This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
For VMS, a belated entry into the $^O jungle that is File::Find::_find_dir.
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
7272f7c1 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137 32#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 33/* if adding more checks watch out for the following tests:
e23c8137
JH
34 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
35 * lib/utf8.t lib/Unicode/Collate/t/index.t
36 * --jhi
37 */
6f207bd3 38# define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
39 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
40 assert((cache)[2] <= (cache)[3]); \
41 assert((cache)[3] <= (cache)[1]);} \
42 } STMT_END
e23c8137 43#else
6f207bd3 44# define ASSERT_UTF8_CACHE(cache) NOOP
e23c8137
JH
45#endif
46
f8c7b90f 47#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 48#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 49#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 50/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 51 on-write. */
765f542d 52#endif
645c22ef
DM
53
54/* ============================================================================
55
56=head1 Allocation and deallocation of SVs.
57
d2a0f284
JC
58An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
59sv, av, hv...) contains type and reference count information, and for
60many types, a pointer to the body (struct xrv, xpv, xpviv...), which
61contains fields specific to each type. Some types store all they need
62in the head, so don't have a body.
63
64In all but the most memory-paranoid configuations (ex: PURIFY), heads
65and bodies are allocated out of arenas, which by default are
66approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
67Sv-bodies are allocated by their sv-type, guaranteeing size
68consistency needed to allocate safely from arrays.
69
d2a0f284
JC
70For SV-heads, the first slot in each arena is reserved, and holds a
71link to the next arena, some flags, and a note of the number of slots.
72Snaked through each arena chain is a linked list of free items; when
73this becomes empty, an extra arena is allocated and divided up into N
74items which are threaded into the free list.
75
76SV-bodies are similar, but they use arena-sets by default, which
77separate the link and info from the arena itself, and reclaim the 1st
78slot in the arena. SV-bodies are further described later.
645c22ef
DM
79
80The following global variables are associated with arenas:
81
82 PL_sv_arenaroot pointer to list of SV arenas
83 PL_sv_root pointer to list of free SV structures
84
d2a0f284
JC
85 PL_body_arenas head of linked-list of body arenas
86 PL_body_roots[] array of pointers to list of free bodies of svtype
87 arrays are indexed by the svtype needed
93e68bfb 88
d2a0f284
JC
89A few special SV heads are not allocated from an arena, but are
90instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
91The size of arenas can be changed from the default by setting
92PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
93
94The SV arena serves the secondary purpose of allowing still-live SVs
95to be located and destroyed during final cleanup.
96
97At the lowest level, the macros new_SV() and del_SV() grab and free
98an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
99to return the SV to the free list with error checking.) new_SV() calls
100more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
101SVs in the free list have their SvTYPE field set to all ones.
102
ff276b08 103At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 104perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 105start of the interpreter.
645c22ef
DM
106
107Manipulation of any of the PL_*root pointers is protected by enclosing
108LOCK_SV_MUTEX; ... UNLOCK_SV_MUTEX calls which should Do the Right Thing
109if threads are enabled.
110
111The function visit() scans the SV arenas list, and calls a specified
112function for each SV it finds which is still live - ie which has an SvTYPE
113other than all 1's, and a non-zero SvREFCNT. visit() is used by the
114following functions (specified as [function that calls visit()] / [function
115called by visit() for each SV]):
116
117 sv_report_used() / do_report_used()
f2524eef 118 dump all remaining SVs (debugging aid)
645c22ef
DM
119
120 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
121 Attempt to free all objects pointed to by RVs,
122 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
123 try to do the same for all objects indirectly
124 referenced by typeglobs too. Called once from
125 perl_destruct(), prior to calling sv_clean_all()
126 below.
127
128 sv_clean_all() / do_clean_all()
129 SvREFCNT_dec(sv) each remaining SV, possibly
130 triggering an sv_free(). It also sets the
131 SVf_BREAK flag on the SV to indicate that the
132 refcnt has been artificially lowered, and thus
133 stopping sv_free() from giving spurious warnings
134 about SVs which unexpectedly have a refcnt
135 of zero. called repeatedly from perl_destruct()
136 until there are no SVs left.
137
93e68bfb 138=head2 Arena allocator API Summary
645c22ef
DM
139
140Private API to rest of sv.c
141
142 new_SV(), del_SV(),
143
144 new_XIV(), del_XIV(),
145 new_XNV(), del_XNV(),
146 etc
147
148Public API:
149
8cf8f3d1 150 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 151
645c22ef
DM
152=cut
153
154============================================================================ */
155
4561caa4
CS
156/*
157 * "A time to plant, and a time to uproot what was planted..."
158 */
159
77354fb4
NC
160/*
161 * nice_chunk and nice_chunk size need to be set
162 * and queried under the protection of sv_mutex
163 */
164void
165Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
166{
97aff369 167 dVAR;
77354fb4
NC
168 void *new_chunk;
169 U32 new_chunk_size;
170 LOCK_SV_MUTEX;
171 new_chunk = (void *)(chunk);
172 new_chunk_size = (chunk_size);
173 if (new_chunk_size > PL_nice_chunk_size) {
174 Safefree(PL_nice_chunk);
175 PL_nice_chunk = (char *) new_chunk;
176 PL_nice_chunk_size = new_chunk_size;
177 } else {
178 Safefree(chunk);
179 }
180 UNLOCK_SV_MUTEX;
181}
cac9b346 182
fd0854ff 183#ifdef DEBUG_LEAKING_SCALARS
22162ca8 184# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
185#else
186# define FREE_SV_DEBUG_FILE(sv)
187#endif
188
48614a46
NC
189#ifdef PERL_POISON
190# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
191/* Whilst I'd love to do this, it seems that things like to check on
192 unreferenced scalars
7e337ee0 193# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 194*/
7e337ee0
JH
195# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
196 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
197#else
198# define SvARENA_CHAIN(sv) SvANY(sv)
199# define POSION_SV_HEAD(sv)
200#endif
201
053fc874
GS
202#define plant_SV(p) \
203 STMT_START { \
fd0854ff 204 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
205 POSION_SV_HEAD(p); \
206 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
207 SvFLAGS(p) = SVTYPEMASK; \
208 PL_sv_root = (p); \
209 --PL_sv_count; \
210 } STMT_END
a0d0e21e 211
fba3b22e 212/* sv_mutex must be held while calling uproot_SV() */
053fc874
GS
213#define uproot_SV(p) \
214 STMT_START { \
215 (p) = PL_sv_root; \
bb7bbd9c 216 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
217 ++PL_sv_count; \
218 } STMT_END
219
645c22ef 220
cac9b346
NC
221/* make some more SVs by adding another arena */
222
223/* sv_mutex must be held while calling more_sv() */
224STATIC SV*
225S_more_sv(pTHX)
226{
97aff369 227 dVAR;
cac9b346
NC
228 SV* sv;
229
230 if (PL_nice_chunk) {
231 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 232 PL_nice_chunk = NULL;
cac9b346
NC
233 PL_nice_chunk_size = 0;
234 }
235 else {
236 char *chunk; /* must use New here to match call to */
d2a0f284 237 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 238 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
239 }
240 uproot_SV(sv);
241 return sv;
242}
243
645c22ef
DM
244/* new_SV(): return a new, empty SV head */
245
eba0f806
DM
246#ifdef DEBUG_LEAKING_SCALARS
247/* provide a real function for a debugger to play with */
248STATIC SV*
249S_new_SV(pTHX)
250{
251 SV* sv;
252
253 LOCK_SV_MUTEX;
254 if (PL_sv_root)
255 uproot_SV(sv);
256 else
cac9b346 257 sv = S_more_sv(aTHX);
eba0f806
DM
258 UNLOCK_SV_MUTEX;
259 SvANY(sv) = 0;
260 SvREFCNT(sv) = 1;
261 SvFLAGS(sv) = 0;
fd0854ff
DM
262 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
263 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
264 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
265 sv->sv_debug_inpad = 0;
266 sv->sv_debug_cloned = 0;
fd0854ff 267 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 268
eba0f806
DM
269 return sv;
270}
271# define new_SV(p) (p)=S_new_SV(aTHX)
272
273#else
274# define new_SV(p) \
053fc874
GS
275 STMT_START { \
276 LOCK_SV_MUTEX; \
277 if (PL_sv_root) \
278 uproot_SV(p); \
279 else \
cac9b346 280 (p) = S_more_sv(aTHX); \
053fc874
GS
281 UNLOCK_SV_MUTEX; \
282 SvANY(p) = 0; \
283 SvREFCNT(p) = 1; \
284 SvFLAGS(p) = 0; \
285 } STMT_END
eba0f806 286#endif
463ee0b2 287
645c22ef
DM
288
289/* del_SV(): return an empty SV head to the free list */
290
a0d0e21e 291#ifdef DEBUGGING
4561caa4 292
053fc874
GS
293#define del_SV(p) \
294 STMT_START { \
295 LOCK_SV_MUTEX; \
aea4f609 296 if (DEBUG_D_TEST) \
053fc874
GS
297 del_sv(p); \
298 else \
299 plant_SV(p); \
300 UNLOCK_SV_MUTEX; \
301 } STMT_END
a0d0e21e 302
76e3520e 303STATIC void
cea2e8a9 304S_del_sv(pTHX_ SV *p)
463ee0b2 305{
97aff369 306 dVAR;
aea4f609 307 if (DEBUG_D_TEST) {
4633a7c4 308 SV* sva;
a3b680e6 309 bool ok = 0;
3280af22 310 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
311 const SV * const sv = sva + 1;
312 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 313 if (p >= sv && p < svend) {
a0d0e21e 314 ok = 1;
c0ff570e
NC
315 break;
316 }
a0d0e21e
LW
317 }
318 if (!ok) {
0453d815 319 if (ckWARN_d(WARN_INTERNAL))
9014280d 320 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
321 "Attempt to free non-arena SV: 0x%"UVxf
322 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
323 return;
324 }
325 }
4561caa4 326 plant_SV(p);
463ee0b2 327}
a0d0e21e 328
4561caa4
CS
329#else /* ! DEBUGGING */
330
331#define del_SV(p) plant_SV(p)
332
333#endif /* DEBUGGING */
463ee0b2 334
645c22ef
DM
335
336/*
ccfc67b7
JH
337=head1 SV Manipulation Functions
338
645c22ef
DM
339=for apidoc sv_add_arena
340
341Given a chunk of memory, link it to the head of the list of arenas,
342and split it into a list of free SVs.
343
344=cut
345*/
346
4633a7c4 347void
864dbfa3 348Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 349{
97aff369 350 dVAR;
0bd48802 351 SV* const sva = (SV*)ptr;
463ee0b2
LW
352 register SV* sv;
353 register SV* svend;
4633a7c4
LW
354
355 /* The first SV in an arena isn't an SV. */
3280af22 356 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
357 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
358 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
359
3280af22
NIS
360 PL_sv_arenaroot = sva;
361 PL_sv_root = sva + 1;
4633a7c4
LW
362
363 svend = &sva[SvREFCNT(sva) - 1];
364 sv = sva + 1;
463ee0b2 365 while (sv < svend) {
48614a46 366 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 367#ifdef DEBUGGING
978b032e 368 SvREFCNT(sv) = 0;
03e36789
NC
369#endif
370 /* Must always set typemask because it's awlays checked in on cleanup
371 when the arenas are walked looking for objects. */
8990e307 372 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
373 sv++;
374 }
48614a46 375 SvARENA_CHAIN(sv) = 0;
03e36789
NC
376#ifdef DEBUGGING
377 SvREFCNT(sv) = 0;
378#endif
4633a7c4
LW
379 SvFLAGS(sv) = SVTYPEMASK;
380}
381
055972dc
DM
382/* visit(): call the named function for each non-free SV in the arenas
383 * whose flags field matches the flags/mask args. */
645c22ef 384
5226ed68 385STATIC I32
055972dc 386S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 387{
97aff369 388 dVAR;
4633a7c4 389 SV* sva;
5226ed68 390 I32 visited = 0;
8990e307 391
3280af22 392 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 393 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 394 register SV* sv;
4561caa4 395 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
396 if (SvTYPE(sv) != SVTYPEMASK
397 && (sv->sv_flags & mask) == flags
398 && SvREFCNT(sv))
399 {
acfe0abc 400 (FCALL)(aTHX_ sv);
5226ed68
JH
401 ++visited;
402 }
8990e307
LW
403 }
404 }
5226ed68 405 return visited;
8990e307
LW
406}
407
758a08c3
JH
408#ifdef DEBUGGING
409
645c22ef
DM
410/* called by sv_report_used() for each live SV */
411
412static void
acfe0abc 413do_report_used(pTHX_ SV *sv)
645c22ef
DM
414{
415 if (SvTYPE(sv) != SVTYPEMASK) {
416 PerlIO_printf(Perl_debug_log, "****\n");
417 sv_dump(sv);
418 }
419}
758a08c3 420#endif
645c22ef
DM
421
422/*
423=for apidoc sv_report_used
424
425Dump the contents of all SVs not yet freed. (Debugging aid).
426
427=cut
428*/
429
8990e307 430void
864dbfa3 431Perl_sv_report_used(pTHX)
4561caa4 432{
ff270d3a 433#ifdef DEBUGGING
055972dc 434 visit(do_report_used, 0, 0);
96a5add6
AL
435#else
436 PERL_UNUSED_CONTEXT;
ff270d3a 437#endif
4561caa4
CS
438}
439
645c22ef
DM
440/* called by sv_clean_objs() for each live SV */
441
442static void
e15faf7d 443do_clean_objs(pTHX_ SV *ref)
645c22ef 444{
97aff369 445 dVAR;
ea724faa
NC
446 assert (SvROK(ref));
447 {
823a54a3
AL
448 SV * const target = SvRV(ref);
449 if (SvOBJECT(target)) {
450 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
451 if (SvWEAKREF(ref)) {
452 sv_del_backref(target, ref);
453 SvWEAKREF_off(ref);
454 SvRV_set(ref, NULL);
455 } else {
456 SvROK_off(ref);
457 SvRV_set(ref, NULL);
458 SvREFCNT_dec(target);
459 }
645c22ef
DM
460 }
461 }
462
463 /* XXX Might want to check arrays, etc. */
464}
465
466/* called by sv_clean_objs() for each live SV */
467
468#ifndef DISABLE_DESTRUCTOR_KLUDGE
469static void
acfe0abc 470do_clean_named_objs(pTHX_ SV *sv)
645c22ef 471{
97aff369 472 dVAR;
ea724faa 473 assert(SvTYPE(sv) == SVt_PVGV);
d011219a
NC
474 assert(isGV_with_GP(sv));
475 if (GvGP(sv)) {
c69033f2
NC
476 if ((
477#ifdef PERL_DONT_CREATE_GVSV
478 GvSV(sv) &&
479#endif
480 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
481 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
482 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
483 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
484 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
485 {
486 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 487 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
488 SvREFCNT_dec(sv);
489 }
490 }
491}
492#endif
493
494/*
495=for apidoc sv_clean_objs
496
497Attempt to destroy all objects not yet freed
498
499=cut
500*/
501
4561caa4 502void
864dbfa3 503Perl_sv_clean_objs(pTHX)
4561caa4 504{
97aff369 505 dVAR;
3280af22 506 PL_in_clean_objs = TRUE;
055972dc 507 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 508#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 509 /* some barnacles may yet remain, clinging to typeglobs */
d011219a 510 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
4561caa4 511#endif
3280af22 512 PL_in_clean_objs = FALSE;
4561caa4
CS
513}
514
645c22ef
DM
515/* called by sv_clean_all() for each live SV */
516
517static void
acfe0abc 518do_clean_all(pTHX_ SV *sv)
645c22ef 519{
97aff369 520 dVAR;
645c22ef
DM
521 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
522 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b 523 if (PL_comppad == (AV*)sv) {
7d49f689 524 PL_comppad = NULL;
4608196e 525 PL_curpad = NULL;
0e705b3b 526 }
645c22ef
DM
527 SvREFCNT_dec(sv);
528}
529
530/*
531=for apidoc sv_clean_all
532
533Decrement the refcnt of each remaining SV, possibly triggering a
534cleanup. This function may have to be called multiple times to free
ff276b08 535SVs which are in complex self-referential hierarchies.
645c22ef
DM
536
537=cut
538*/
539
5226ed68 540I32
864dbfa3 541Perl_sv_clean_all(pTHX)
8990e307 542{
97aff369 543 dVAR;
5226ed68 544 I32 cleaned;
3280af22 545 PL_in_clean_all = TRUE;
055972dc 546 cleaned = visit(do_clean_all, 0,0);
3280af22 547 PL_in_clean_all = FALSE;
5226ed68 548 return cleaned;
8990e307 549}
463ee0b2 550
5e258f8c
JC
551/*
552 ARENASETS: a meta-arena implementation which separates arena-info
553 into struct arena_set, which contains an array of struct
554 arena_descs, each holding info for a single arena. By separating
555 the meta-info from the arena, we recover the 1st slot, formerly
556 borrowed for list management. The arena_set is about the size of an
39244528 557 arena, avoiding the needless malloc overhead of a naive linked-list.
5e258f8c
JC
558
559 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
560 memory in the last arena-set (1/2 on average). In trade, we get
561 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284
JC
562 smaller types). The recovery of the wasted space allows use of
563 small arenas for large, rare body types,
5e258f8c 564*/
5e258f8c 565struct arena_desc {
398c677b
NC
566 char *arena; /* the raw storage, allocated aligned */
567 size_t size; /* its size ~4k typ */
0a848332 568 U32 misc; /* type, and in future other things. */
5e258f8c
JC
569};
570
e6148039
NC
571struct arena_set;
572
573/* Get the maximum number of elements in set[] such that struct arena_set
574 will fit within PERL_ARENA_SIZE, which is probabably just under 4K, and
575 therefore likely to be 1 aligned memory page. */
576
577#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
578 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
579
580struct arena_set {
581 struct arena_set* next;
0a848332
NC
582 unsigned int set_size; /* ie ARENAS_PER_SET */
583 unsigned int curr; /* index of next available arena-desc */
5e258f8c
JC
584 struct arena_desc set[ARENAS_PER_SET];
585};
586
645c22ef
DM
587/*
588=for apidoc sv_free_arenas
589
590Deallocate the memory used by all arenas. Note that all the individual SV
591heads and bodies within the arenas must already have been freed.
592
593=cut
594*/
4633a7c4 595void
864dbfa3 596Perl_sv_free_arenas(pTHX)
4633a7c4 597{
97aff369 598 dVAR;
4633a7c4
LW
599 SV* sva;
600 SV* svanext;
0a848332 601 unsigned int i;
4633a7c4
LW
602
603 /* Free arenas here, but be careful about fake ones. (We assume
604 contiguity of the fake ones with the corresponding real ones.) */
605
3280af22 606 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
607 svanext = (SV*) SvANY(sva);
608 while (svanext && SvFAKE(svanext))
609 svanext = (SV*) SvANY(svanext);
610
611 if (!SvFAKE(sva))
1df70142 612 Safefree(sva);
4633a7c4 613 }
93e68bfb 614
5e258f8c 615 {
0a848332
NC
616 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
617
618 while (aroot) {
619 struct arena_set *current = aroot;
620 i = aroot->curr;
621 while (i--) {
5e258f8c
JC
622 assert(aroot->set[i].arena);
623 Safefree(aroot->set[i].arena);
624 }
0a848332
NC
625 aroot = aroot->next;
626 Safefree(current);
5e258f8c
JC
627 }
628 }
dc8220bf 629 PL_body_arenas = 0;
fdda85ca 630
0a848332
NC
631 i = PERL_ARENA_ROOTS_SIZE;
632 while (i--)
93e68bfb 633 PL_body_roots[i] = 0;
93e68bfb 634
43c5f42d 635 Safefree(PL_nice_chunk);
bd61b366 636 PL_nice_chunk = NULL;
3280af22
NIS
637 PL_nice_chunk_size = 0;
638 PL_sv_arenaroot = 0;
639 PL_sv_root = 0;
4633a7c4
LW
640}
641
bd81e77b
NC
642/*
643 Here are mid-level routines that manage the allocation of bodies out
644 of the various arenas. There are 5 kinds of arenas:
29489e7c 645
bd81e77b
NC
646 1. SV-head arenas, which are discussed and handled above
647 2. regular body arenas
648 3. arenas for reduced-size bodies
649 4. Hash-Entry arenas
650 5. pte arenas (thread related)
29489e7c 651
bd81e77b
NC
652 Arena types 2 & 3 are chained by body-type off an array of
653 arena-root pointers, which is indexed by svtype. Some of the
654 larger/less used body types are malloced singly, since a large
655 unused block of them is wasteful. Also, several svtypes dont have
656 bodies; the data fits into the sv-head itself. The arena-root
657 pointer thus has a few unused root-pointers (which may be hijacked
658 later for arena types 4,5)
29489e7c 659
bd81e77b
NC
660 3 differs from 2 as an optimization; some body types have several
661 unused fields in the front of the structure (which are kept in-place
662 for consistency). These bodies can be allocated in smaller chunks,
663 because the leading fields arent accessed. Pointers to such bodies
664 are decremented to point at the unused 'ghost' memory, knowing that
665 the pointers are used with offsets to the real memory.
29489e7c 666
bd81e77b
NC
667 HE, HEK arenas are managed separately, with separate code, but may
668 be merge-able later..
669
670 PTE arenas are not sv-bodies, but they share these mid-level
671 mechanics, so are considered here. The new mid-level mechanics rely
672 on the sv_type of the body being allocated, so we just reserve one
673 of the unused body-slots for PTEs, then use it in those (2) PTE
674 contexts below (line ~10k)
675*/
676
bd26d9a3 677/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
678 TBD: export properly for hv.c: S_more_he().
679*/
680void*
0a848332 681Perl_get_arena(pTHX_ size_t arena_size, U32 misc)
5e258f8c 682{
7a89be66 683 dVAR;
5e258f8c 684 struct arena_desc* adesc;
39244528 685 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
0a848332 686 unsigned int curr;
5e258f8c 687
476a1e16
JC
688 /* shouldnt need this
689 if (!arena_size) arena_size = PERL_ARENA_SIZE;
690 */
5e258f8c
JC
691
692 /* may need new arena-set to hold new arena */
39244528
NC
693 if (!aroot || aroot->curr >= aroot->set_size) {
694 struct arena_set *newroot;
5e258f8c
JC
695 Newxz(newroot, 1, struct arena_set);
696 newroot->set_size = ARENAS_PER_SET;
39244528
NC
697 newroot->next = aroot;
698 aroot = newroot;
699 PL_body_arenas = (void *) newroot;
52944de8 700 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
5e258f8c
JC
701 }
702
703 /* ok, now have arena-set with at least 1 empty/available arena-desc */
39244528
NC
704 curr = aroot->curr++;
705 adesc = &(aroot->set[curr]);
5e258f8c
JC
706 assert(!adesc->arena);
707
89086707 708 Newx(adesc->arena, arena_size, char);
5e258f8c 709 adesc->size = arena_size;
0a848332 710 adesc->misc = misc;
d2a0f284 711 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
6c9570dc 712 curr, (void*)adesc->arena, arena_size));
5e258f8c
JC
713
714 return adesc->arena;
5e258f8c
JC
715}
716
53c1dcc0 717
bd81e77b 718/* return a thing to the free list */
29489e7c 719
bd81e77b
NC
720#define del_body(thing, root) \
721 STMT_START { \
00b6aa41 722 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
723 LOCK_SV_MUTEX; \
724 *thing_copy = *root; \
725 *root = (void*)thing_copy; \
726 UNLOCK_SV_MUTEX; \
727 } STMT_END
29489e7c 728
bd81e77b 729/*
d2a0f284
JC
730
731=head1 SV-Body Allocation
732
733Allocation of SV-bodies is similar to SV-heads, differing as follows;
734the allocation mechanism is used for many body types, so is somewhat
735more complicated, it uses arena-sets, and has no need for still-live
736SV detection.
737
738At the outermost level, (new|del)_X*V macros return bodies of the
739appropriate type. These macros call either (new|del)_body_type or
740(new|del)_body_allocated macro pairs, depending on specifics of the
741type. Most body types use the former pair, the latter pair is used to
742allocate body types with "ghost fields".
743
744"ghost fields" are fields that are unused in certain types, and
745consequently dont need to actually exist. They are declared because
746they're part of a "base type", which allows use of functions as
747methods. The simplest examples are AVs and HVs, 2 aggregate types
748which don't use the fields which support SCALAR semantics.
749
750For these types, the arenas are carved up into *_allocated size
751chunks, we thus avoid wasted memory for those unaccessed members.
752When bodies are allocated, we adjust the pointer back in memory by the
753size of the bit not allocated, so it's as if we allocated the full
754structure. (But things will all go boom if you write to the part that
755is "not there", because you'll be overwriting the last members of the
756preceding structure in memory.)
757
758We calculate the correction using the STRUCT_OFFSET macro. For
759example, if xpv_allocated is the same structure as XPV then the two
760OFFSETs sum to zero, and the pointer is unchanged. If the allocated
761structure is smaller (no initial NV actually allocated) then the net
762effect is to subtract the size of the NV from the pointer, to return a
763new pointer as if an initial NV were actually allocated.
764
765This is the same trick as was used for NV and IV bodies. Ironically it
766doesn't need to be used for NV bodies any more, because NV is now at
767the start of the structure. IV bodies don't need it either, because
768they are no longer allocated.
769
770In turn, the new_body_* allocators call S_new_body(), which invokes
771new_body_inline macro, which takes a lock, and takes a body off the
772linked list at PL_body_roots[sv_type], calling S_more_bodies() if
773necessary to refresh an empty list. Then the lock is released, and
774the body is returned.
775
776S_more_bodies calls get_arena(), and carves it up into an array of N
777bodies, which it strings into a linked list. It looks up arena-size
778and body-size from the body_details table described below, thus
779supporting the multiple body-types.
780
781If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
782the (new|del)_X*V macros are mapped directly to malloc/free.
783
784*/
785
786/*
787
788For each sv-type, struct body_details bodies_by_type[] carries
789parameters which control these aspects of SV handling:
790
791Arena_size determines whether arenas are used for this body type, and if
792so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
793zero, forcing individual mallocs and frees.
794
795Body_size determines how big a body is, and therefore how many fit into
796each arena. Offset carries the body-pointer adjustment needed for
797*_allocated body types, and is used in *_allocated macros.
798
799But its main purpose is to parameterize info needed in
800Perl_sv_upgrade(). The info here dramatically simplifies the function
801vs the implementation in 5.8.7, making it table-driven. All fields
802are used for this, except for arena_size.
803
804For the sv-types that have no bodies, arenas are not used, so those
805PL_body_roots[sv_type] are unused, and can be overloaded. In
806something of a special case, SVt_NULL is borrowed for HE arenas;
807PL_body_roots[SVt_NULL] is filled by S_more_he, but the
808bodies_by_type[SVt_NULL] slot is not used, as the table is not
809available in hv.c,
810
811PTEs also use arenas, but are never seen in Perl_sv_upgrade.
812Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
813they can just use the same allocation semantics. At first, PTEs were
814also overloaded to a non-body sv-type, but this yielded hard-to-find
815malloc bugs, so was simplified by claiming a new slot. This choice
816has no consequence at this time.
817
29489e7c
DM
818*/
819
bd81e77b 820struct body_details {
0fb58b32 821 U8 body_size; /* Size to allocate */
10666ae3 822 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 823 U8 offset;
10666ae3
NC
824 unsigned int type : 4; /* We have space for a sanity check. */
825 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
826 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
827 unsigned int arena : 1; /* Allocated from an arena */
828 size_t arena_size; /* Size of arena to allocate */
bd81e77b 829};
29489e7c 830
bd81e77b
NC
831#define HADNV FALSE
832#define NONV TRUE
29489e7c 833
d2a0f284 834
bd81e77b
NC
835#ifdef PURIFY
836/* With -DPURFIY we allocate everything directly, and don't use arenas.
837 This seems a rather elegant way to simplify some of the code below. */
838#define HASARENA FALSE
839#else
840#define HASARENA TRUE
841#endif
842#define NOARENA FALSE
29489e7c 843
d2a0f284
JC
844/* Size the arenas to exactly fit a given number of bodies. A count
845 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
846 simplifying the default. If count > 0, the arena is sized to fit
847 only that many bodies, allowing arenas to be used for large, rare
848 bodies (XPVFM, XPVIO) without undue waste. The arena size is
849 limited by PERL_ARENA_SIZE, so we can safely oversize the
850 declarations.
851 */
95db5f15
MB
852#define FIT_ARENA0(body_size) \
853 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
854#define FIT_ARENAn(count,body_size) \
855 ( count * body_size <= PERL_ARENA_SIZE) \
856 ? count * body_size \
857 : FIT_ARENA0 (body_size)
858#define FIT_ARENA(count,body_size) \
859 count \
860 ? FIT_ARENAn (count, body_size) \
861 : FIT_ARENA0 (body_size)
d2a0f284 862
bd81e77b 863/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 864
bd81e77b
NC
865typedef struct {
866 STRLEN xpv_cur;
867 STRLEN xpv_len;
868} xpv_allocated;
29489e7c 869
bd81e77b 870to make its members accessible via a pointer to (say)
29489e7c 871
bd81e77b
NC
872struct xpv {
873 NV xnv_nv;
874 STRLEN xpv_cur;
875 STRLEN xpv_len;
876};
29489e7c 877
bd81e77b 878*/
29489e7c 879
bd81e77b
NC
880#define relative_STRUCT_OFFSET(longer, shorter, member) \
881 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 882
bd81e77b
NC
883/* Calculate the length to copy. Specifically work out the length less any
884 final padding the compiler needed to add. See the comment in sv_upgrade
885 for why copying the padding proved to be a bug. */
29489e7c 886
bd81e77b
NC
887#define copy_length(type, last_member) \
888 STRUCT_OFFSET(type, last_member) \
889 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 890
bd81e77b 891static const struct body_details bodies_by_type[] = {
10666ae3
NC
892 { sizeof(HE), 0, 0, SVt_NULL,
893 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284 894
1cb9cd50
NC
895 /* The bind placeholder pretends to be an RV for now.
896 Also it's marked as "can't upgrade" top stop anyone using it before it's
897 implemented. */
898 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
899
d2a0f284
JC
900 /* IVs are in the head, so the allocation size is 0.
901 However, the slot is overloaded for PTEs. */
902 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
903 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 904 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
905 NOARENA /* IVS don't need an arena */,
906 /* But PTEs need to know the size of their arena */
907 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
908 },
909
bd81e77b 910 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 911 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
912 FIT_ARENA(0, sizeof(NV)) },
913
914 /* RVs are in the head now. */
10666ae3 915 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 916
bd81e77b 917 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
918 { sizeof(xpv_allocated),
919 copy_length(XPV, xpv_len)
920 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
921 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 922 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 923
bd81e77b 924 /* 12 */
d2a0f284
JC
925 { sizeof(xpviv_allocated),
926 copy_length(XPVIV, xiv_u)
927 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
928 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 929 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 930
bd81e77b 931 /* 20 */
10666ae3 932 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
933 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
934
bd81e77b 935 /* 28 */
10666ae3 936 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
937 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
938
bd81e77b 939 /* 48 */
10666ae3 940 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
941 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
942
bd81e77b 943 /* 64 */
10666ae3 944 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
945 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
946
947 { sizeof(xpvav_allocated),
948 copy_length(XPVAV, xmg_stash)
949 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
950 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 951 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
952
953 { sizeof(xpvhv_allocated),
954 copy_length(XPVHV, xmg_stash)
955 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
956 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 957 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 958
c84c4652 959 /* 56 */
4115f141 960 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 961 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 962 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 963
4115f141 964 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 965 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 966 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
967
968 /* XPVIO is 84 bytes, fits 48x */
10666ae3 969 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 970 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 971};
29489e7c 972
d2a0f284
JC
973#define new_body_type(sv_type) \
974 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 975
bd81e77b
NC
976#define del_body_type(p, sv_type) \
977 del_body(p, &PL_body_roots[sv_type])
29489e7c 978
29489e7c 979
bd81e77b 980#define new_body_allocated(sv_type) \
d2a0f284 981 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 982 - bodies_by_type[sv_type].offset)
29489e7c 983
bd81e77b
NC
984#define del_body_allocated(p, sv_type) \
985 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 986
29489e7c 987
bd81e77b
NC
988#define my_safemalloc(s) (void*)safemalloc(s)
989#define my_safecalloc(s) (void*)safecalloc(s, 1)
990#define my_safefree(p) safefree((char*)p)
29489e7c 991
bd81e77b 992#ifdef PURIFY
29489e7c 993
bd81e77b
NC
994#define new_XNV() my_safemalloc(sizeof(XPVNV))
995#define del_XNV(p) my_safefree(p)
29489e7c 996
bd81e77b
NC
997#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
998#define del_XPVNV(p) my_safefree(p)
29489e7c 999
bd81e77b
NC
1000#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1001#define del_XPVAV(p) my_safefree(p)
29489e7c 1002
bd81e77b
NC
1003#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1004#define del_XPVHV(p) my_safefree(p)
29489e7c 1005
bd81e77b
NC
1006#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1007#define del_XPVMG(p) my_safefree(p)
29489e7c 1008
bd81e77b
NC
1009#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1010#define del_XPVGV(p) my_safefree(p)
29489e7c 1011
bd81e77b 1012#else /* !PURIFY */
29489e7c 1013
bd81e77b
NC
1014#define new_XNV() new_body_type(SVt_NV)
1015#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1016
bd81e77b
NC
1017#define new_XPVNV() new_body_type(SVt_PVNV)
1018#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1019
bd81e77b
NC
1020#define new_XPVAV() new_body_allocated(SVt_PVAV)
1021#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1022
bd81e77b
NC
1023#define new_XPVHV() new_body_allocated(SVt_PVHV)
1024#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1025
bd81e77b
NC
1026#define new_XPVMG() new_body_type(SVt_PVMG)
1027#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1028
bd81e77b
NC
1029#define new_XPVGV() new_body_type(SVt_PVGV)
1030#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1031
bd81e77b 1032#endif /* PURIFY */
93e68bfb 1033
bd81e77b 1034/* no arena for you! */
93e68bfb 1035
bd81e77b 1036#define new_NOARENA(details) \
d2a0f284 1037 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1038#define new_NOARENAZ(details) \
d2a0f284
JC
1039 my_safecalloc((details)->body_size + (details)->offset)
1040
1041STATIC void *
1042S_more_bodies (pTHX_ svtype sv_type)
1043{
1044 dVAR;
1045 void ** const root = &PL_body_roots[sv_type];
96a5add6 1046 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1047 const size_t body_size = bdp->body_size;
1048 char *start;
1049 const char *end;
0b2d3faa 1050#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
23e9d66c
NC
1051 static bool done_sanity_check;
1052
0b2d3faa
JH
1053 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1054 * variables like done_sanity_check. */
10666ae3 1055 if (!done_sanity_check) {
ea471437 1056 unsigned int i = SVt_LAST;
10666ae3
NC
1057
1058 done_sanity_check = TRUE;
1059
1060 while (i--)
1061 assert (bodies_by_type[i].type == i);
1062 }
1063#endif
1064
23e9d66c
NC
1065 assert(bdp->arena_size);
1066
0a848332 1067 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size, sv_type);
d2a0f284
JC
1068
1069 end = start + bdp->arena_size - body_size;
1070
d2a0f284
JC
1071 /* computed count doesnt reflect the 1st slot reservation */
1072 DEBUG_m(PerlIO_printf(Perl_debug_log,
1073 "arena %p end %p arena-size %d type %d size %d ct %d\n",
6c9570dc 1074 (void*)start, (void*)end,
0e84aef4
JH
1075 (int)bdp->arena_size, sv_type, (int)body_size,
1076 (int)bdp->arena_size / (int)body_size));
d2a0f284
JC
1077
1078 *root = (void *)start;
1079
1080 while (start < end) {
1081 char * const next = start + body_size;
1082 *(void**) start = (void *)next;
1083 start = next;
1084 }
1085 *(void **)start = 0;
1086
1087 return *root;
1088}
1089
1090/* grab a new thing from the free list, allocating more if necessary.
1091 The inline version is used for speed in hot routines, and the
1092 function using it serves the rest (unless PURIFY).
1093*/
1094#define new_body_inline(xpv, sv_type) \
1095 STMT_START { \
1096 void ** const r3wt = &PL_body_roots[sv_type]; \
1097 LOCK_SV_MUTEX; \
11b79775
DD
1098 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1099 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284
JC
1100 *(r3wt) = *(void**)(xpv); \
1101 UNLOCK_SV_MUTEX; \
1102 } STMT_END
1103
1104#ifndef PURIFY
1105
1106STATIC void *
1107S_new_body(pTHX_ svtype sv_type)
1108{
1109 dVAR;
1110 void *xpv;
1111 new_body_inline(xpv, sv_type);
1112 return xpv;
1113}
1114
1115#endif
93e68bfb 1116
bd81e77b
NC
1117/*
1118=for apidoc sv_upgrade
93e68bfb 1119
bd81e77b
NC
1120Upgrade an SV to a more complex form. Generally adds a new body type to the
1121SV, then copies across as much information as possible from the old body.
1122You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1123
bd81e77b 1124=cut
93e68bfb 1125*/
93e68bfb 1126
bd81e77b 1127void
42d0e0b7 1128Perl_sv_upgrade(pTHX_ register SV *sv, svtype new_type)
cac9b346 1129{
97aff369 1130 dVAR;
bd81e77b
NC
1131 void* old_body;
1132 void* new_body;
42d0e0b7 1133 const svtype old_type = SvTYPE(sv);
d2a0f284 1134 const struct body_details *new_type_details;
bd81e77b
NC
1135 const struct body_details *const old_type_details
1136 = bodies_by_type + old_type;
cac9b346 1137
bd81e77b
NC
1138 if (new_type != SVt_PV && SvIsCOW(sv)) {
1139 sv_force_normal_flags(sv, 0);
1140 }
cac9b346 1141
bd81e77b
NC
1142 if (old_type == new_type)
1143 return;
cac9b346 1144
bd81e77b
NC
1145 if (old_type > new_type)
1146 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1147 (int)old_type, (int)new_type);
cac9b346 1148
cac9b346 1149
bd81e77b 1150 old_body = SvANY(sv);
de042e1d 1151
bd81e77b
NC
1152 /* Copying structures onto other structures that have been neatly zeroed
1153 has a subtle gotcha. Consider XPVMG
cac9b346 1154
bd81e77b
NC
1155 +------+------+------+------+------+-------+-------+
1156 | NV | CUR | LEN | IV | MAGIC | STASH |
1157 +------+------+------+------+------+-------+-------+
1158 0 4 8 12 16 20 24 28
645c22ef 1159
bd81e77b
NC
1160 where NVs are aligned to 8 bytes, so that sizeof that structure is
1161 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1162
bd81e77b
NC
1163 +------+------+------+------+------+-------+-------+------+
1164 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1165 +------+------+------+------+------+-------+-------+------+
1166 0 4 8 12 16 20 24 28 32
08742458 1167
bd81e77b 1168 so what happens if you allocate memory for this structure:
30f9da9e 1169
bd81e77b
NC
1170 +------+------+------+------+------+-------+-------+------+------+...
1171 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1172 +------+------+------+------+------+-------+-------+------+------+...
1173 0 4 8 12 16 20 24 28 32 36
bfc44f79 1174
bd81e77b
NC
1175 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1176 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1177 started out as zero once, but it's quite possible that it isn't. So now,
1178 rather than a nicely zeroed GP, you have it pointing somewhere random.
1179 Bugs ensue.
bfc44f79 1180
bd81e77b
NC
1181 (In fact, GP ends up pointing at a previous GP structure, because the
1182 principle cause of the padding in XPVMG getting garbage is a copy of
6c9e42f7
NC
1183 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1184 this happens to be moot because XPVGV has been re-ordered, with GP
1185 no longer after STASH)
30f9da9e 1186
bd81e77b
NC
1187 So we are careful and work out the size of used parts of all the
1188 structures. */
bfc44f79 1189
bd81e77b
NC
1190 switch (old_type) {
1191 case SVt_NULL:
1192 break;
1193 case SVt_IV:
1194 if (new_type < SVt_PVIV) {
1195 new_type = (new_type == SVt_NV)
1196 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1197 }
1198 break;
1199 case SVt_NV:
1200 if (new_type < SVt_PVNV) {
1201 new_type = SVt_PVNV;
bd81e77b
NC
1202 }
1203 break;
1204 case SVt_RV:
1205 break;
1206 case SVt_PV:
1207 assert(new_type > SVt_PV);
1208 assert(SVt_IV < SVt_PV);
1209 assert(SVt_NV < SVt_PV);
1210 break;
1211 case SVt_PVIV:
1212 break;
1213 case SVt_PVNV:
1214 break;
1215 case SVt_PVMG:
1216 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1217 there's no way that it can be safely upgraded, because perl.c
1218 expects to Safefree(SvANY(PL_mess_sv)) */
1219 assert(sv != PL_mess_sv);
1220 /* This flag bit is used to mean other things in other scalar types.
1221 Given that it only has meaning inside the pad, it shouldn't be set
1222 on anything that can get upgraded. */
00b1698f 1223 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1224 break;
1225 default:
1226 if (old_type_details->cant_upgrade)
c81225bc
NC
1227 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1228 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1229 }
2fa1109b 1230 new_type_details = bodies_by_type + new_type;
645c22ef 1231
bd81e77b
NC
1232 SvFLAGS(sv) &= ~SVTYPEMASK;
1233 SvFLAGS(sv) |= new_type;
932e9ff9 1234
ab4416c0
NC
1235 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1236 the return statements above will have triggered. */
1237 assert (new_type != SVt_NULL);
bd81e77b 1238 switch (new_type) {
bd81e77b
NC
1239 case SVt_IV:
1240 assert(old_type == SVt_NULL);
1241 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1242 SvIV_set(sv, 0);
1243 return;
1244 case SVt_NV:
1245 assert(old_type == SVt_NULL);
1246 SvANY(sv) = new_XNV();
1247 SvNV_set(sv, 0);
1248 return;
1249 case SVt_RV:
1250 assert(old_type == SVt_NULL);
1251 SvANY(sv) = &sv->sv_u.svu_rv;
1252 SvRV_set(sv, 0);
1253 return;
1254 case SVt_PVHV:
bd81e77b 1255 case SVt_PVAV:
d2a0f284 1256 assert(new_type_details->body_size);
c1ae03ae
NC
1257
1258#ifndef PURIFY
1259 assert(new_type_details->arena);
d2a0f284 1260 assert(new_type_details->arena_size);
c1ae03ae 1261 /* This points to the start of the allocated area. */
d2a0f284
JC
1262 new_body_inline(new_body, new_type);
1263 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1264 new_body = ((char *)new_body) - new_type_details->offset;
1265#else
1266 /* We always allocated the full length item with PURIFY. To do this
1267 we fake things so that arena is false for all 16 types.. */
1268 new_body = new_NOARENAZ(new_type_details);
1269#endif
1270 SvANY(sv) = new_body;
1271 if (new_type == SVt_PVAV) {
1272 AvMAX(sv) = -1;
1273 AvFILLp(sv) = -1;
1274 AvREAL_only(sv);
1275 }
aeb18a1e 1276
bd81e77b
NC
1277 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1278 The target created by newSVrv also is, and it can have magic.
1279 However, it never has SvPVX set.
1280 */
1281 if (old_type >= SVt_RV) {
1282 assert(SvPVX_const(sv) == 0);
1283 }
aeb18a1e 1284
bd81e77b 1285 if (old_type >= SVt_PVMG) {
e736a858 1286 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1287 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1288 } else {
1289 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1290 }
1291 break;
93e68bfb 1292
93e68bfb 1293
bd81e77b
NC
1294 case SVt_PVIV:
1295 /* XXX Is this still needed? Was it ever needed? Surely as there is
1296 no route from NV to PVIV, NOK can never be true */
1297 assert(!SvNOKp(sv));
1298 assert(!SvNOK(sv));
1299 case SVt_PVIO:
1300 case SVt_PVFM:
bd81e77b
NC
1301 case SVt_PVGV:
1302 case SVt_PVCV:
1303 case SVt_PVLV:
1304 case SVt_PVMG:
1305 case SVt_PVNV:
1306 case SVt_PV:
93e68bfb 1307
d2a0f284 1308 assert(new_type_details->body_size);
bd81e77b
NC
1309 /* We always allocated the full length item with PURIFY. To do this
1310 we fake things so that arena is false for all 16 types.. */
1311 if(new_type_details->arena) {
1312 /* This points to the start of the allocated area. */
d2a0f284
JC
1313 new_body_inline(new_body, new_type);
1314 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1315 new_body = ((char *)new_body) - new_type_details->offset;
1316 } else {
1317 new_body = new_NOARENAZ(new_type_details);
1318 }
1319 SvANY(sv) = new_body;
5e2fc214 1320
bd81e77b 1321 if (old_type_details->copy) {
f9ba3d20
NC
1322 /* There is now the potential for an upgrade from something without
1323 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1324 int offset = old_type_details->offset;
1325 int length = old_type_details->copy;
1326
1327 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1328 const int difference
f9ba3d20
NC
1329 = new_type_details->offset - old_type_details->offset;
1330 offset += difference;
1331 length -= difference;
1332 }
1333 assert (length >= 0);
1334
1335 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1336 char);
bd81e77b
NC
1337 }
1338
1339#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1340 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1341 * correct 0.0 for us. Otherwise, if the old body didn't have an
1342 * NV slot, but the new one does, then we need to initialise the
1343 * freshly created NV slot with whatever the correct bit pattern is
1344 * for 0.0 */
1345 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1346 SvNV_set(sv, 0);
82048762 1347#endif
5e2fc214 1348
bd81e77b 1349 if (new_type == SVt_PVIO)
f2524eef 1350 IoPAGE_LEN(sv) = 60;
bd81e77b 1351 if (old_type < SVt_RV)
6136c704 1352 SvPV_set(sv, NULL);
bd81e77b
NC
1353 break;
1354 default:
afd78fd5
JH
1355 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1356 (unsigned long)new_type);
bd81e77b 1357 }
73171d91 1358
d2a0f284
JC
1359 if (old_type_details->arena) {
1360 /* If there was an old body, then we need to free it.
1361 Note that there is an assumption that all bodies of types that
1362 can be upgraded came from arenas. Only the more complex non-
1363 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1364#ifdef PURIFY
1365 my_safefree(old_body);
1366#else
1367 del_body((void*)((char*)old_body + old_type_details->offset),
1368 &PL_body_roots[old_type]);
1369#endif
1370 }
1371}
73171d91 1372
bd81e77b
NC
1373/*
1374=for apidoc sv_backoff
73171d91 1375
bd81e77b
NC
1376Remove any string offset. You should normally use the C<SvOOK_off> macro
1377wrapper instead.
73171d91 1378
bd81e77b 1379=cut
73171d91
NC
1380*/
1381
bd81e77b
NC
1382int
1383Perl_sv_backoff(pTHX_ register SV *sv)
1384{
96a5add6 1385 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1386 assert(SvOOK(sv));
1387 assert(SvTYPE(sv) != SVt_PVHV);
1388 assert(SvTYPE(sv) != SVt_PVAV);
1389 if (SvIVX(sv)) {
1390 const char * const s = SvPVX_const(sv);
1391 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1392 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1393 SvIV_set(sv, 0);
1394 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1395 }
1396 SvFLAGS(sv) &= ~SVf_OOK;
1397 return 0;
1398}
73171d91 1399
bd81e77b
NC
1400/*
1401=for apidoc sv_grow
73171d91 1402
bd81e77b
NC
1403Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1404upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1405Use the C<SvGROW> wrapper instead.
93e68bfb 1406
bd81e77b
NC
1407=cut
1408*/
93e68bfb 1409
bd81e77b
NC
1410char *
1411Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1412{
1413 register char *s;
93e68bfb 1414
5db06880
NC
1415 if (PL_madskills && newlen >= 0x100000) {
1416 PerlIO_printf(Perl_debug_log,
1417 "Allocation too large: %"UVxf"\n", (UV)newlen);
1418 }
bd81e77b
NC
1419#ifdef HAS_64K_LIMIT
1420 if (newlen >= 0x10000) {
1421 PerlIO_printf(Perl_debug_log,
1422 "Allocation too large: %"UVxf"\n", (UV)newlen);
1423 my_exit(1);
1424 }
1425#endif /* HAS_64K_LIMIT */
1426 if (SvROK(sv))
1427 sv_unref(sv);
1428 if (SvTYPE(sv) < SVt_PV) {
1429 sv_upgrade(sv, SVt_PV);
1430 s = SvPVX_mutable(sv);
1431 }
1432 else if (SvOOK(sv)) { /* pv is offset? */
1433 sv_backoff(sv);
1434 s = SvPVX_mutable(sv);
1435 if (newlen > SvLEN(sv))
1436 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1437#ifdef HAS_64K_LIMIT
1438 if (newlen >= 0x10000)
1439 newlen = 0xFFFF;
1440#endif
1441 }
1442 else
1443 s = SvPVX_mutable(sv);
aeb18a1e 1444
bd81e77b
NC
1445 if (newlen > SvLEN(sv)) { /* need more room? */
1446 newlen = PERL_STRLEN_ROUNDUP(newlen);
1447 if (SvLEN(sv) && s) {
1448#ifdef MYMALLOC
1449 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1450 if (newlen <= l) {
1451 SvLEN_set(sv, l);
1452 return s;
1453 } else
1454#endif
10edeb5d 1455 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1456 }
1457 else {
10edeb5d 1458 s = (char*)safemalloc(newlen);
bd81e77b
NC
1459 if (SvPVX_const(sv) && SvCUR(sv)) {
1460 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1461 }
1462 }
1463 SvPV_set(sv, s);
1464 SvLEN_set(sv, newlen);
1465 }
1466 return s;
1467}
aeb18a1e 1468
bd81e77b
NC
1469/*
1470=for apidoc sv_setiv
932e9ff9 1471
bd81e77b
NC
1472Copies an integer into the given SV, upgrading first if necessary.
1473Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1474
bd81e77b
NC
1475=cut
1476*/
463ee0b2 1477
bd81e77b
NC
1478void
1479Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1480{
97aff369 1481 dVAR;
bd81e77b
NC
1482 SV_CHECK_THINKFIRST_COW_DROP(sv);
1483 switch (SvTYPE(sv)) {
1484 case SVt_NULL:
1485 sv_upgrade(sv, SVt_IV);
1486 break;
1487 case SVt_NV:
1488 sv_upgrade(sv, SVt_PVNV);
1489 break;
1490 case SVt_RV:
1491 case SVt_PV:
1492 sv_upgrade(sv, SVt_PVIV);
1493 break;
463ee0b2 1494
bd81e77b
NC
1495 case SVt_PVGV:
1496 case SVt_PVAV:
1497 case SVt_PVHV:
1498 case SVt_PVCV:
1499 case SVt_PVFM:
1500 case SVt_PVIO:
1501 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1502 OP_DESC(PL_op));
42d0e0b7 1503 default: NOOP;
bd81e77b
NC
1504 }
1505 (void)SvIOK_only(sv); /* validate number */
1506 SvIV_set(sv, i);
1507 SvTAINT(sv);
1508}
932e9ff9 1509
bd81e77b
NC
1510/*
1511=for apidoc sv_setiv_mg
d33b2eba 1512
bd81e77b 1513Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1514
bd81e77b
NC
1515=cut
1516*/
d33b2eba 1517
bd81e77b
NC
1518void
1519Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1520{
1521 sv_setiv(sv,i);
1522 SvSETMAGIC(sv);
1523}
727879eb 1524
bd81e77b
NC
1525/*
1526=for apidoc sv_setuv
d33b2eba 1527
bd81e77b
NC
1528Copies an unsigned integer into the given SV, upgrading first if necessary.
1529Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1530
bd81e77b
NC
1531=cut
1532*/
d33b2eba 1533
bd81e77b
NC
1534void
1535Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1536{
1537 /* With these two if statements:
1538 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1539
bd81e77b
NC
1540 without
1541 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1542
bd81e77b
NC
1543 If you wish to remove them, please benchmark to see what the effect is
1544 */
1545 if (u <= (UV)IV_MAX) {
1546 sv_setiv(sv, (IV)u);
1547 return;
1548 }
1549 sv_setiv(sv, 0);
1550 SvIsUV_on(sv);
1551 SvUV_set(sv, u);
1552}
d33b2eba 1553
bd81e77b
NC
1554/*
1555=for apidoc sv_setuv_mg
727879eb 1556
bd81e77b 1557Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1558
bd81e77b
NC
1559=cut
1560*/
5e2fc214 1561
bd81e77b
NC
1562void
1563Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1564{
bd81e77b
NC
1565 sv_setuv(sv,u);
1566 SvSETMAGIC(sv);
1567}
5e2fc214 1568
954c1994 1569/*
bd81e77b 1570=for apidoc sv_setnv
954c1994 1571
bd81e77b
NC
1572Copies a double into the given SV, upgrading first if necessary.
1573Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1574
1575=cut
1576*/
1577
63f97190 1578void
bd81e77b 1579Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1580{
97aff369 1581 dVAR;
bd81e77b
NC
1582 SV_CHECK_THINKFIRST_COW_DROP(sv);
1583 switch (SvTYPE(sv)) {
79072805 1584 case SVt_NULL:
79072805 1585 case SVt_IV:
bd81e77b 1586 sv_upgrade(sv, SVt_NV);
79072805 1587 break;
ed6116ce 1588 case SVt_RV:
79072805 1589 case SVt_PV:
79072805 1590 case SVt_PVIV:
bd81e77b 1591 sv_upgrade(sv, SVt_PVNV);
79072805 1592 break;
bd4b1eb5 1593
bd4b1eb5 1594 case SVt_PVGV:
bd81e77b
NC
1595 case SVt_PVAV:
1596 case SVt_PVHV:
79072805 1597 case SVt_PVCV:
bd81e77b
NC
1598 case SVt_PVFM:
1599 case SVt_PVIO:
1600 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1601 OP_NAME(PL_op));
42d0e0b7 1602 default: NOOP;
2068cd4d 1603 }
bd81e77b
NC
1604 SvNV_set(sv, num);
1605 (void)SvNOK_only(sv); /* validate number */
1606 SvTAINT(sv);
79072805
LW
1607}
1608
645c22ef 1609/*
bd81e77b 1610=for apidoc sv_setnv_mg
645c22ef 1611
bd81e77b 1612Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1613
1614=cut
1615*/
1616
bd81e77b
NC
1617void
1618Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1619{
bd81e77b
NC
1620 sv_setnv(sv,num);
1621 SvSETMAGIC(sv);
79072805
LW
1622}
1623
bd81e77b
NC
1624/* Print an "isn't numeric" warning, using a cleaned-up,
1625 * printable version of the offending string
1626 */
954c1994 1627
bd81e77b
NC
1628STATIC void
1629S_not_a_number(pTHX_ SV *sv)
79072805 1630{
97aff369 1631 dVAR;
bd81e77b
NC
1632 SV *dsv;
1633 char tmpbuf[64];
1634 const char *pv;
94463019
JH
1635
1636 if (DO_UTF8(sv)) {
396482e1 1637 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1638 pv = sv_uni_display(dsv, sv, 10, 0);
1639 } else {
1640 char *d = tmpbuf;
551405c4 1641 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1642 /* each *s can expand to 4 chars + "...\0",
1643 i.e. need room for 8 chars */
ecdeb87c 1644
00b6aa41
AL
1645 const char *s = SvPVX_const(sv);
1646 const char * const end = s + SvCUR(sv);
1647 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1648 int ch = *s & 0xFF;
1649 if (ch & 128 && !isPRINT_LC(ch)) {
1650 *d++ = 'M';
1651 *d++ = '-';
1652 ch &= 127;
1653 }
1654 if (ch == '\n') {
1655 *d++ = '\\';
1656 *d++ = 'n';
1657 }
1658 else if (ch == '\r') {
1659 *d++ = '\\';
1660 *d++ = 'r';
1661 }
1662 else if (ch == '\f') {
1663 *d++ = '\\';
1664 *d++ = 'f';
1665 }
1666 else if (ch == '\\') {
1667 *d++ = '\\';
1668 *d++ = '\\';
1669 }
1670 else if (ch == '\0') {
1671 *d++ = '\\';
1672 *d++ = '0';
1673 }
1674 else if (isPRINT_LC(ch))
1675 *d++ = ch;
1676 else {
1677 *d++ = '^';
1678 *d++ = toCTRL(ch);
1679 }
1680 }
1681 if (s < end) {
1682 *d++ = '.';
1683 *d++ = '.';
1684 *d++ = '.';
1685 }
1686 *d = '\0';
1687 pv = tmpbuf;
a0d0e21e 1688 }
a0d0e21e 1689
533c011a 1690 if (PL_op)
9014280d 1691 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1692 "Argument \"%s\" isn't numeric in %s", pv,
1693 OP_DESC(PL_op));
a0d0e21e 1694 else
9014280d 1695 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1696 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1697}
1698
c2988b20
NC
1699/*
1700=for apidoc looks_like_number
1701
645c22ef
DM
1702Test if the content of an SV looks like a number (or is a number).
1703C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1704non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1705
1706=cut
1707*/
1708
1709I32
1710Perl_looks_like_number(pTHX_ SV *sv)
1711{
a3b680e6 1712 register const char *sbegin;
c2988b20
NC
1713 STRLEN len;
1714
1715 if (SvPOK(sv)) {
3f7c398e 1716 sbegin = SvPVX_const(sv);
c2988b20
NC
1717 len = SvCUR(sv);
1718 }
1719 else if (SvPOKp(sv))
83003860 1720 sbegin = SvPV_const(sv, len);
c2988b20 1721 else
e0ab1c0e 1722 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1723 return grok_number(sbegin, len, NULL);
1724}
25da4f38 1725
19f6321d
NC
1726STATIC bool
1727S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1728{
1729 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1730 SV *const buffer = sv_newmortal();
1731
1732 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1733 is on. */
1734 SvFAKE_off(gv);
1735 gv_efullname3(buffer, gv, "*");
1736 SvFLAGS(gv) |= wasfake;
1737
675c862f
AL
1738 /* We know that all GVs stringify to something that is not-a-number,
1739 so no need to test that. */
1740 if (ckWARN(WARN_NUMERIC))
1741 not_a_number(buffer);
1742 /* We just want something true to return, so that S_sv_2iuv_common
1743 can tail call us and return true. */
19f6321d 1744 return TRUE;
675c862f
AL
1745}
1746
1747STATIC char *
19f6321d 1748S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1749{
1750 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1751 SV *const buffer = sv_newmortal();
1752
1753 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1754 is on. */
1755 SvFAKE_off(gv);
1756 gv_efullname3(buffer, gv, "*");
1757 SvFLAGS(gv) |= wasfake;
1758
1759 assert(SvPOK(buffer));
a6d61a6c
NC
1760 if (len) {
1761 *len = SvCUR(buffer);
1762 }
675c862f 1763 return SvPVX(buffer);
180488f8
NC
1764}
1765
25da4f38
IZ
1766/* Actually, ISO C leaves conversion of UV to IV undefined, but
1767 until proven guilty, assume that things are not that bad... */
1768
645c22ef
DM
1769/*
1770 NV_PRESERVES_UV:
1771
1772 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1773 an IV (an assumption perl has been based on to date) it becomes necessary
1774 to remove the assumption that the NV always carries enough precision to
1775 recreate the IV whenever needed, and that the NV is the canonical form.
1776 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1777 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1778 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1779 1) to distinguish between IV/UV/NV slots that have cached a valid
1780 conversion where precision was lost and IV/UV/NV slots that have a
1781 valid conversion which has lost no precision
645c22ef 1782 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1783 would lose precision, the precise conversion (or differently
1784 imprecise conversion) is also performed and cached, to prevent
1785 requests for different numeric formats on the same SV causing
1786 lossy conversion chains. (lossless conversion chains are perfectly
1787 acceptable (still))
1788
1789
1790 flags are used:
1791 SvIOKp is true if the IV slot contains a valid value
1792 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1793 SvNOKp is true if the NV slot contains a valid value
1794 SvNOK is true only if the NV value is accurate
1795
1796 so
645c22ef 1797 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1798 IV(or UV) would lose accuracy over a direct conversion from PV to
1799 IV(or UV). If it would, cache both conversions, return NV, but mark
1800 SV as IOK NOKp (ie not NOK).
1801
645c22ef 1802 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1803 NV would lose accuracy over a direct conversion from PV to NV. If it
1804 would, cache both conversions, flag similarly.
1805
1806 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1807 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1808 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1809 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1810 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1811
645c22ef
DM
1812 The benefit of this is that operations such as pp_add know that if
1813 SvIOK is true for both left and right operands, then integer addition
1814 can be used instead of floating point (for cases where the result won't
1815 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1816 loss of precision compared with integer addition.
1817
1818 * making IV and NV equal status should make maths accurate on 64 bit
1819 platforms
1820 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1821 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1822 looking for SvIOK and checking for overflow will not outweigh the
1823 fp to integer speedup)
1824 * will slow down integer operations (callers of SvIV) on "inaccurate"
1825 values, as the change from SvIOK to SvIOKp will cause a call into
1826 sv_2iv each time rather than a macro access direct to the IV slot
1827 * should speed up number->string conversion on integers as IV is
645c22ef 1828 favoured when IV and NV are equally accurate
28e5dec8
JH
1829
1830 ####################################################################
645c22ef
DM
1831 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1832 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1833 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1834 ####################################################################
1835
645c22ef 1836 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1837 performance ratio.
1838*/
1839
1840#ifndef NV_PRESERVES_UV
645c22ef
DM
1841# define IS_NUMBER_UNDERFLOW_IV 1
1842# define IS_NUMBER_UNDERFLOW_UV 2
1843# define IS_NUMBER_IV_AND_UV 2
1844# define IS_NUMBER_OVERFLOW_IV 4
1845# define IS_NUMBER_OVERFLOW_UV 5
1846
1847/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1848
1849/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1850STATIC int
645c22ef 1851S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1852{
97aff369 1853 dVAR;
b57a0404 1854 PERL_UNUSED_ARG(numtype); /* Used only under DEBUGGING? */
3f7c398e 1855 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
28e5dec8
JH
1856 if (SvNVX(sv) < (NV)IV_MIN) {
1857 (void)SvIOKp_on(sv);
1858 (void)SvNOK_on(sv);
45977657 1859 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1860 return IS_NUMBER_UNDERFLOW_IV;
1861 }
1862 if (SvNVX(sv) > (NV)UV_MAX) {
1863 (void)SvIOKp_on(sv);
1864 (void)SvNOK_on(sv);
1865 SvIsUV_on(sv);
607fa7f2 1866 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1867 return IS_NUMBER_OVERFLOW_UV;
1868 }
c2988b20
NC
1869 (void)SvIOKp_on(sv);
1870 (void)SvNOK_on(sv);
1871 /* Can't use strtol etc to convert this string. (See truth table in
1872 sv_2iv */
1873 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1874 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1875 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1876 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1877 } else {
1878 /* Integer is imprecise. NOK, IOKp */
1879 }
1880 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1881 }
1882 SvIsUV_on(sv);
607fa7f2 1883 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1884 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1885 if (SvUVX(sv) == UV_MAX) {
1886 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1887 possibly be preserved by NV. Hence, it must be overflow.
1888 NOK, IOKp */
1889 return IS_NUMBER_OVERFLOW_UV;
1890 }
1891 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1892 } else {
1893 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1894 }
c2988b20 1895 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1896}
645c22ef
DM
1897#endif /* !NV_PRESERVES_UV*/
1898
af359546
NC
1899STATIC bool
1900S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1901 dVAR;
af359546 1902 if (SvNOKp(sv)) {
28e5dec8
JH
1903 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1904 * without also getting a cached IV/UV from it at the same time
1905 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1906 * IV or UV at same time to avoid this. */
1907 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1908
1909 if (SvTYPE(sv) == SVt_NV)
1910 sv_upgrade(sv, SVt_PVNV);
1911
28e5dec8
JH
1912 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1913 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1914 certainly cast into the IV range at IV_MAX, whereas the correct
1915 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1916 cases go to UV */
cab190d4
JD
1917#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1918 if (Perl_isnan(SvNVX(sv))) {
1919 SvUV_set(sv, 0);
1920 SvIsUV_on(sv);
fdbe6d7c 1921 return FALSE;
cab190d4 1922 }
cab190d4 1923#endif
28e5dec8 1924 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1925 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1926 if (SvNVX(sv) == (NV) SvIVX(sv)
1927#ifndef NV_PRESERVES_UV
1928 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1929 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1930 /* Don't flag it as "accurately an integer" if the number
1931 came from a (by definition imprecise) NV operation, and
1932 we're outside the range of NV integer precision */
1933#endif
1934 ) {
1935 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1936 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1937 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1938 PTR2UV(sv),
1939 SvNVX(sv),
1940 SvIVX(sv)));
1941
1942 } else {
1943 /* IV not precise. No need to convert from PV, as NV
1944 conversion would already have cached IV if it detected
1945 that PV->IV would be better than PV->NV->IV
1946 flags already correct - don't set public IOK. */
1947 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1948 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1949 PTR2UV(sv),
1950 SvNVX(sv),
1951 SvIVX(sv)));
1952 }
1953 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1954 but the cast (NV)IV_MIN rounds to a the value less (more
1955 negative) than IV_MIN which happens to be equal to SvNVX ??
1956 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1957 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1958 (NV)UVX == NVX are both true, but the values differ. :-(
1959 Hopefully for 2s complement IV_MIN is something like
1960 0x8000000000000000 which will be exact. NWC */
d460ef45 1961 }
25da4f38 1962 else {
607fa7f2 1963 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1964 if (
1965 (SvNVX(sv) == (NV) SvUVX(sv))
1966#ifndef NV_PRESERVES_UV
1967 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1968 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1969 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1970 /* Don't flag it as "accurately an integer" if the number
1971 came from a (by definition imprecise) NV operation, and
1972 we're outside the range of NV integer precision */
1973#endif
1974 )
1975 SvIOK_on(sv);
25da4f38 1976 SvIsUV_on(sv);
1c846c1f 1977 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1978 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1979 PTR2UV(sv),
57def98f
JH
1980 SvUVX(sv),
1981 SvUVX(sv)));
25da4f38 1982 }
748a9306
LW
1983 }
1984 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1985 UV value;
504618e9 1986 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1987 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1988 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1989 the same as the direct translation of the initial string
1990 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1991 be careful to ensure that the value with the .456 is around if the
1992 NV value is requested in the future).
1c846c1f 1993
af359546 1994 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1995 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1996 cache the NV if we are sure it's not needed.
25da4f38 1997 */
16b7a9a4 1998
c2988b20
NC
1999 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2000 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2001 == IS_NUMBER_IN_UV) {
5e045b90 2002 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
2003 if (SvTYPE(sv) < SVt_PVIV)
2004 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 2005 (void)SvIOK_on(sv);
c2988b20
NC
2006 } else if (SvTYPE(sv) < SVt_PVNV)
2007 sv_upgrade(sv, SVt_PVNV);
28e5dec8 2008
f2524eef 2009 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2010 we aren't going to call atof() below. If NVs don't preserve UVs
2011 then the value returned may have more precision than atof() will
2012 return, even though value isn't perfectly accurate. */
2013 if ((numtype & (IS_NUMBER_IN_UV
2014#ifdef NV_PRESERVES_UV
2015 | IS_NUMBER_NOT_INT
2016#endif
2017 )) == IS_NUMBER_IN_UV) {
2018 /* This won't turn off the public IOK flag if it was set above */
2019 (void)SvIOKp_on(sv);
2020
2021 if (!(numtype & IS_NUMBER_NEG)) {
2022 /* positive */;
2023 if (value <= (UV)IV_MAX) {
45977657 2024 SvIV_set(sv, (IV)value);
c2988b20 2025 } else {
af359546 2026 /* it didn't overflow, and it was positive. */
607fa7f2 2027 SvUV_set(sv, value);
c2988b20
NC
2028 SvIsUV_on(sv);
2029 }
2030 } else {
2031 /* 2s complement assumption */
2032 if (value <= (UV)IV_MIN) {
45977657 2033 SvIV_set(sv, -(IV)value);
c2988b20
NC
2034 } else {
2035 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2036 I'm assuming it will be rare. */
c2988b20
NC
2037 if (SvTYPE(sv) < SVt_PVNV)
2038 sv_upgrade(sv, SVt_PVNV);
2039 SvNOK_on(sv);
2040 SvIOK_off(sv);
2041 SvIOKp_on(sv);
9d6ce603 2042 SvNV_set(sv, -(NV)value);
45977657 2043 SvIV_set(sv, IV_MIN);
c2988b20
NC
2044 }
2045 }
2046 }
2047 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2048 will be in the previous block to set the IV slot, and the next
2049 block to set the NV slot. So no else here. */
2050
2051 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2052 != IS_NUMBER_IN_UV) {
2053 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2054 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2055
c2988b20
NC
2056 if (! numtype && ckWARN(WARN_NUMERIC))
2057 not_a_number(sv);
28e5dec8 2058
65202027 2059#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2060 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2061 PTR2UV(sv), SvNVX(sv)));
65202027 2062#else
1779d84d 2063 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2064 PTR2UV(sv), SvNVX(sv)));
65202027 2065#endif
28e5dec8 2066
28e5dec8 2067#ifdef NV_PRESERVES_UV
af359546
NC
2068 (void)SvIOKp_on(sv);
2069 (void)SvNOK_on(sv);
2070 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2071 SvIV_set(sv, I_V(SvNVX(sv)));
2072 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2073 SvIOK_on(sv);
2074 } else {
6f207bd3 2075 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2076 }
2077 /* UV will not work better than IV */
2078 } else {
2079 if (SvNVX(sv) > (NV)UV_MAX) {
2080 SvIsUV_on(sv);
2081 /* Integer is inaccurate. NOK, IOKp, is UV */
2082 SvUV_set(sv, UV_MAX);
af359546
NC
2083 } else {
2084 SvUV_set(sv, U_V(SvNVX(sv)));
2085 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2086 NV preservse UV so can do correct comparison. */
2087 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2088 SvIOK_on(sv);
af359546 2089 } else {
6f207bd3 2090 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2091 }
2092 }
4b0c9573 2093 SvIsUV_on(sv);
af359546 2094 }
28e5dec8 2095#else /* NV_PRESERVES_UV */
c2988b20
NC
2096 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2097 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2098 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2099 grok_number above. The NV slot has just been set using
2100 Atof. */
560b0c46 2101 SvNOK_on(sv);
c2988b20
NC
2102 assert (SvIOKp(sv));
2103 } else {
2104 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2105 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2106 /* Small enough to preserve all bits. */
2107 (void)SvIOKp_on(sv);
2108 SvNOK_on(sv);
45977657 2109 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2110 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2111 SvIOK_on(sv);
2112 /* Assumption: first non-preserved integer is < IV_MAX,
2113 this NV is in the preserved range, therefore: */
2114 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2115 < (UV)IV_MAX)) {
32fdb065 2116 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
c2988b20
NC
2117 }
2118 } else {
2119 /* IN_UV NOT_INT
2120 0 0 already failed to read UV.
2121 0 1 already failed to read UV.
2122 1 0 you won't get here in this case. IV/UV
2123 slot set, public IOK, Atof() unneeded.
2124 1 1 already read UV.
2125 so there's no point in sv_2iuv_non_preserve() attempting
2126 to use atol, strtol, strtoul etc. */
40a17c4c 2127 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2128 }
2129 }
28e5dec8 2130#endif /* NV_PRESERVES_UV */
25da4f38 2131 }
af359546
NC
2132 }
2133 else {
675c862f 2134 if (isGV_with_GP(sv))
a0933d07 2135 return glob_2number((GV *)sv);
180488f8 2136
af359546
NC
2137 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2138 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2139 report_uninit(sv);
2140 }
25da4f38
IZ
2141 if (SvTYPE(sv) < SVt_IV)
2142 /* Typically the caller expects that sv_any is not NULL now. */
2143 sv_upgrade(sv, SVt_IV);
af359546
NC
2144 /* Return 0 from the caller. */
2145 return TRUE;
2146 }
2147 return FALSE;
2148}
2149
2150/*
2151=for apidoc sv_2iv_flags
2152
2153Return the integer value of an SV, doing any necessary string
2154conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2155Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2156
2157=cut
2158*/
2159
2160IV
2161Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2162{
97aff369 2163 dVAR;
af359546 2164 if (!sv)
a0d0e21e 2165 return 0;
cecf5685
NC
2166 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2167 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e
NC
2168 cache IVs just in case. In practice it seems that they never
2169 actually anywhere accessible by user Perl code, let alone get used
2170 in anything other than a string context. */
af359546
NC
2171 if (flags & SV_GMAGIC)
2172 mg_get(sv);
2173 if (SvIOKp(sv))
2174 return SvIVX(sv);
2175 if (SvNOKp(sv)) {
2176 return I_V(SvNVX(sv));
2177 }
71c558c3
NC
2178 if (SvPOKp(sv) && SvLEN(sv)) {
2179 UV value;
2180 const int numtype
2181 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2182
2183 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2184 == IS_NUMBER_IN_UV) {
2185 /* It's definitely an integer */
2186 if (numtype & IS_NUMBER_NEG) {
2187 if (value < (UV)IV_MIN)
2188 return -(IV)value;
2189 } else {
2190 if (value < (UV)IV_MAX)
2191 return (IV)value;
2192 }
2193 }
2194 if (!numtype) {
2195 if (ckWARN(WARN_NUMERIC))
2196 not_a_number(sv);
2197 }
2198 return I_V(Atof(SvPVX_const(sv)));
2199 }
1c7ff15e
NC
2200 if (SvROK(sv)) {
2201 goto return_rok;
af359546 2202 }
1c7ff15e
NC
2203 assert(SvTYPE(sv) >= SVt_PVMG);
2204 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2205 } else if (SvTHINKFIRST(sv)) {
af359546 2206 if (SvROK(sv)) {
1c7ff15e 2207 return_rok:
af359546
NC
2208 if (SvAMAGIC(sv)) {
2209 SV * const tmpstr=AMG_CALLun(sv,numer);
2210 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2211 return SvIV(tmpstr);
2212 }
2213 }
2214 return PTR2IV(SvRV(sv));
2215 }
2216 if (SvIsCOW(sv)) {
2217 sv_force_normal_flags(sv, 0);
2218 }
2219 if (SvREADONLY(sv) && !SvOK(sv)) {
2220 if (ckWARN(WARN_UNINITIALIZED))
2221 report_uninit(sv);
2222 return 0;
2223 }
2224 }
2225 if (!SvIOKp(sv)) {
2226 if (S_sv_2iuv_common(aTHX_ sv))
2227 return 0;
79072805 2228 }
1d7c1841
GS
2229 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2230 PTR2UV(sv),SvIVX(sv)));
25da4f38 2231 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2232}
2233
645c22ef 2234/*
891f9566 2235=for apidoc sv_2uv_flags
645c22ef
DM
2236
2237Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2238conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2239Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2240
2241=cut
2242*/
2243
ff68c719 2244UV
891f9566 2245Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2246{
97aff369 2247 dVAR;
ff68c719 2248 if (!sv)
2249 return 0;
cecf5685
NC
2250 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2251 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2252 cache IVs just in case. */
891f9566
YST
2253 if (flags & SV_GMAGIC)
2254 mg_get(sv);
ff68c719 2255 if (SvIOKp(sv))
2256 return SvUVX(sv);
2257 if (SvNOKp(sv))
2258 return U_V(SvNVX(sv));
71c558c3
NC
2259 if (SvPOKp(sv) && SvLEN(sv)) {
2260 UV value;
2261 const int numtype
2262 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2263
2264 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2265 == IS_NUMBER_IN_UV) {
2266 /* It's definitely an integer */
2267 if (!(numtype & IS_NUMBER_NEG))
2268 return value;
2269 }
2270 if (!numtype) {
2271 if (ckWARN(WARN_NUMERIC))
2272 not_a_number(sv);
2273 }
2274 return U_V(Atof(SvPVX_const(sv)));
2275 }
1c7ff15e
NC
2276 if (SvROK(sv)) {
2277 goto return_rok;
3fe9a6f1 2278 }
1c7ff15e
NC
2279 assert(SvTYPE(sv) >= SVt_PVMG);
2280 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2281 } else if (SvTHINKFIRST(sv)) {
ff68c719 2282 if (SvROK(sv)) {
1c7ff15e 2283 return_rok:
deb46114
NC
2284 if (SvAMAGIC(sv)) {
2285 SV *const tmpstr = AMG_CALLun(sv,numer);
2286 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2287 return SvUV(tmpstr);
2288 }
2289 }
2290 return PTR2UV(SvRV(sv));
ff68c719 2291 }
765f542d
NC
2292 if (SvIsCOW(sv)) {
2293 sv_force_normal_flags(sv, 0);
8a818333 2294 }
0336b60e 2295 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2296 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2297 report_uninit(sv);
ff68c719 2298 return 0;
2299 }
2300 }
af359546
NC
2301 if (!SvIOKp(sv)) {
2302 if (S_sv_2iuv_common(aTHX_ sv))
2303 return 0;
ff68c719 2304 }
25da4f38 2305
1d7c1841
GS
2306 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2307 PTR2UV(sv),SvUVX(sv)));
25da4f38 2308 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2309}
2310
645c22ef
DM
2311/*
2312=for apidoc sv_2nv
2313
2314Return the num value of an SV, doing any necessary string or integer
2315conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2316macros.
2317
2318=cut
2319*/
2320
65202027 2321NV
864dbfa3 2322Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2323{
97aff369 2324 dVAR;
79072805
LW
2325 if (!sv)
2326 return 0.0;
cecf5685
NC
2327 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2328 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2329 cache IVs just in case. */
463ee0b2
LW
2330 mg_get(sv);
2331 if (SvNOKp(sv))
2332 return SvNVX(sv);
0aa395f8 2333 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2334 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2335 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2336 not_a_number(sv);
3f7c398e 2337 return Atof(SvPVX_const(sv));
a0d0e21e 2338 }
25da4f38 2339 if (SvIOKp(sv)) {
1c846c1f 2340 if (SvIsUV(sv))
65202027 2341 return (NV)SvUVX(sv);
25da4f38 2342 else
65202027 2343 return (NV)SvIVX(sv);
47a72cb8
NC
2344 }
2345 if (SvROK(sv)) {
2346 goto return_rok;
2347 }
2348 assert(SvTYPE(sv) >= SVt_PVMG);
2349 /* This falls through to the report_uninit near the end of the
2350 function. */
2351 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2352 if (SvROK(sv)) {
47a72cb8 2353 return_rok:
deb46114
NC
2354 if (SvAMAGIC(sv)) {
2355 SV *const tmpstr = AMG_CALLun(sv,numer);
2356 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2357 return SvNV(tmpstr);
2358 }
2359 }
2360 return PTR2NV(SvRV(sv));
a0d0e21e 2361 }
765f542d
NC
2362 if (SvIsCOW(sv)) {
2363 sv_force_normal_flags(sv, 0);
8a818333 2364 }
0336b60e 2365 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2366 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2367 report_uninit(sv);
ed6116ce
LW
2368 return 0.0;
2369 }
79072805
LW
2370 }
2371 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2372 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2373 sv_upgrade(sv, SVt_NV);
906f284f 2374#ifdef USE_LONG_DOUBLE
097ee67d 2375 DEBUG_c({
f93f4e46 2376 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2377 PerlIO_printf(Perl_debug_log,
2378 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2379 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2380 RESTORE_NUMERIC_LOCAL();
2381 });
65202027 2382#else
572bbb43 2383 DEBUG_c({
f93f4e46 2384 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2385 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2386 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2387 RESTORE_NUMERIC_LOCAL();
2388 });
572bbb43 2389#endif
79072805
LW
2390 }
2391 else if (SvTYPE(sv) < SVt_PVNV)
2392 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2393 if (SvNOKp(sv)) {
2394 return SvNVX(sv);
61604483 2395 }
59d8ce62 2396 if (SvIOKp(sv)) {
9d6ce603 2397 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2398#ifdef NV_PRESERVES_UV
2399 SvNOK_on(sv);
2400#else
2401 /* Only set the public NV OK flag if this NV preserves the IV */
2402 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2403 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2404 : (SvIVX(sv) == I_V(SvNVX(sv))))
2405 SvNOK_on(sv);
2406 else
2407 SvNOKp_on(sv);
2408#endif
93a17b20 2409 }
748a9306 2410 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2411 UV value;
3f7c398e 2412 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2413 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2414 not_a_number(sv);
28e5dec8 2415#ifdef NV_PRESERVES_UV
c2988b20
NC
2416 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2417 == IS_NUMBER_IN_UV) {
5e045b90 2418 /* It's definitely an integer */
9d6ce603 2419 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2420 } else
3f7c398e 2421 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2422 SvNOK_on(sv);
2423#else
3f7c398e 2424 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2425 /* Only set the public NV OK flag if this NV preserves the value in
2426 the PV at least as well as an IV/UV would.
2427 Not sure how to do this 100% reliably. */
2428 /* if that shift count is out of range then Configure's test is
2429 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2430 UV_BITS */
2431 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2432 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2433 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2434 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2435 /* Can't use strtol etc to convert this string, so don't try.
2436 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2437 SvNOK_on(sv);
2438 } else {
2439 /* value has been set. It may not be precise. */
2440 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2441 /* 2s complement assumption for (UV)IV_MIN */
2442 SvNOK_on(sv); /* Integer is too negative. */
2443 } else {
2444 SvNOKp_on(sv);
2445 SvIOKp_on(sv);
6fa402ec 2446
c2988b20 2447 if (numtype & IS_NUMBER_NEG) {
45977657 2448 SvIV_set(sv, -(IV)value);
c2988b20 2449 } else if (value <= (UV)IV_MAX) {
45977657 2450 SvIV_set(sv, (IV)value);
c2988b20 2451 } else {
607fa7f2 2452 SvUV_set(sv, value);
c2988b20
NC
2453 SvIsUV_on(sv);
2454 }
2455
2456 if (numtype & IS_NUMBER_NOT_INT) {
2457 /* I believe that even if the original PV had decimals,
2458 they are lost beyond the limit of the FP precision.
2459 However, neither is canonical, so both only get p
2460 flags. NWC, 2000/11/25 */
2461 /* Both already have p flags, so do nothing */
2462 } else {
66a1b24b 2463 const NV nv = SvNVX(sv);
c2988b20
NC
2464 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2465 if (SvIVX(sv) == I_V(nv)) {
2466 SvNOK_on(sv);
c2988b20 2467 } else {
c2988b20
NC
2468 /* It had no "." so it must be integer. */
2469 }
00b6aa41 2470 SvIOK_on(sv);
c2988b20
NC
2471 } else {
2472 /* between IV_MAX and NV(UV_MAX).
2473 Could be slightly > UV_MAX */
6fa402ec 2474
c2988b20
NC
2475 if (numtype & IS_NUMBER_NOT_INT) {
2476 /* UV and NV both imprecise. */
2477 } else {
66a1b24b 2478 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2479
2480 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2481 SvNOK_on(sv);
c2988b20 2482 }
00b6aa41 2483 SvIOK_on(sv);
c2988b20
NC
2484 }
2485 }
2486 }
2487 }
2488 }
28e5dec8 2489#endif /* NV_PRESERVES_UV */
93a17b20 2490 }
79072805 2491 else {
f7877b28 2492 if (isGV_with_GP(sv)) {
19f6321d 2493 glob_2number((GV *)sv);
180488f8
NC
2494 return 0.0;
2495 }
2496
041457d9 2497 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2498 report_uninit(sv);
7e25a7e9
NC
2499 assert (SvTYPE(sv) >= SVt_NV);
2500 /* Typically the caller expects that sv_any is not NULL now. */
2501 /* XXX Ilya implies that this is a bug in callers that assume this
2502 and ideally should be fixed. */
a0d0e21e 2503 return 0.0;
79072805 2504 }
572bbb43 2505#if defined(USE_LONG_DOUBLE)
097ee67d 2506 DEBUG_c({
f93f4e46 2507 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2508 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2509 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2510 RESTORE_NUMERIC_LOCAL();
2511 });
65202027 2512#else
572bbb43 2513 DEBUG_c({
f93f4e46 2514 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2515 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2516 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2517 RESTORE_NUMERIC_LOCAL();
2518 });
572bbb43 2519#endif
463ee0b2 2520 return SvNVX(sv);
79072805
LW
2521}
2522
645c22ef
DM
2523/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2524 * UV as a string towards the end of buf, and return pointers to start and
2525 * end of it.
2526 *
2527 * We assume that buf is at least TYPE_CHARS(UV) long.
2528 */
2529
864dbfa3 2530static char *
aec46f14 2531S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2532{
25da4f38 2533 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2534 char * const ebuf = ptr;
25da4f38 2535 int sign;
25da4f38
IZ
2536
2537 if (is_uv)
2538 sign = 0;
2539 else if (iv >= 0) {
2540 uv = iv;
2541 sign = 0;
2542 } else {
2543 uv = -iv;
2544 sign = 1;
2545 }
2546 do {
eb160463 2547 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2548 } while (uv /= 10);
2549 if (sign)
2550 *--ptr = '-';
2551 *peob = ebuf;
2552 return ptr;
2553}
2554
645c22ef
DM
2555/*
2556=for apidoc sv_2pv_flags
2557
ff276b08 2558Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2559If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2560if necessary.
2561Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2562usually end up here too.
2563
2564=cut
2565*/
2566
8d6d96c1
HS
2567char *
2568Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2569{
97aff369 2570 dVAR;
79072805 2571 register char *s;
79072805 2572
463ee0b2 2573 if (!sv) {
cdb061a3
NC
2574 if (lp)
2575 *lp = 0;
73d840c0 2576 return (char *)"";
463ee0b2 2577 }
8990e307 2578 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2579 if (flags & SV_GMAGIC)
2580 mg_get(sv);
463ee0b2 2581 if (SvPOKp(sv)) {
cdb061a3
NC
2582 if (lp)
2583 *lp = SvCUR(sv);
10516c54
NC
2584 if (flags & SV_MUTABLE_RETURN)
2585 return SvPVX_mutable(sv);
4d84ee25
NC
2586 if (flags & SV_CONST_RETURN)
2587 return (char *)SvPVX_const(sv);
463ee0b2
LW
2588 return SvPVX(sv);
2589 }
75dfc8ec
NC
2590 if (SvIOKp(sv) || SvNOKp(sv)) {
2591 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2592 STRLEN len;
2593
2594 if (SvIOKp(sv)) {
e80fed9d 2595 len = SvIsUV(sv)
d9fad198
JH
2596 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2597 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2598 } else {
e8ada2d0
NC
2599 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2600 len = strlen(tbuf);
75dfc8ec 2601 }
b5b886f0
NC
2602 assert(!SvROK(sv));
2603 {
75dfc8ec
NC
2604 dVAR;
2605
2606#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2607 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2608 tbuf[0] = '0';
2609 tbuf[1] = 0;
75dfc8ec
NC
2610 len = 1;
2611 }
2612#endif
2613 SvUPGRADE(sv, SVt_PV);
2614 if (lp)
2615 *lp = len;
2616 s = SvGROW_mutable(sv, len + 1);
2617 SvCUR_set(sv, len);
2618 SvPOKp_on(sv);
10edeb5d 2619 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2620 }
463ee0b2 2621 }
1c7ff15e
NC
2622 if (SvROK(sv)) {
2623 goto return_rok;
2624 }
2625 assert(SvTYPE(sv) >= SVt_PVMG);
2626 /* This falls through to the report_uninit near the end of the
2627 function. */
2628 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2629 if (SvROK(sv)) {
1c7ff15e 2630 return_rok:
deb46114
NC
2631 if (SvAMAGIC(sv)) {
2632 SV *const tmpstr = AMG_CALLun(sv,string);
2633 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2634 /* Unwrap this: */
2635 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2636 */
2637
2638 char *pv;
2639 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2640 if (flags & SV_CONST_RETURN) {
2641 pv = (char *) SvPVX_const(tmpstr);
2642 } else {
2643 pv = (flags & SV_MUTABLE_RETURN)
2644 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2645 }
2646 if (lp)
2647 *lp = SvCUR(tmpstr);
50adf7d2 2648 } else {
deb46114 2649 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2650 }
deb46114
NC
2651 if (SvUTF8(tmpstr))
2652 SvUTF8_on(sv);
2653 else
2654 SvUTF8_off(sv);
2655 return pv;
50adf7d2 2656 }
deb46114
NC
2657 }
2658 {
fafee734
NC
2659 STRLEN len;
2660 char *retval;
2661 char *buffer;
f9277f47 2662 MAGIC *mg;
d8eae41e
NC
2663 const SV *const referent = (SV*)SvRV(sv);
2664
2665 if (!referent) {
fafee734
NC
2666 len = 7;
2667 retval = buffer = savepvn("NULLREF", len);
042dae7a
NC
2668 } else if (SvTYPE(referent) == SVt_PVMG
2669 && ((SvFLAGS(referent) &
2670 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2671 == (SVs_OBJECT|SVs_SMG))
de8c5301
YO
2672 && (mg = mg_find(referent, PERL_MAGIC_qr)))
2673 {
2674 char *str = NULL;
2675 I32 haseval = 0;
60df1e07 2676 U32 flags = 0;
de8c5301
YO
2677 (str) = CALLREG_AS_STR(mg,lp,&flags,&haseval);
2678 if (flags & 1)
2679 SvUTF8_on(sv);
2680 else
2681 SvUTF8_off(sv);
2682 PL_reginterp_cnt += haseval;
2683 return str;
d8eae41e
NC
2684 } else {
2685 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2686 const STRLEN typelen = strlen(typestr);
2687 UV addr = PTR2UV(referent);
2688 const char *stashname = NULL;
2689 STRLEN stashnamelen = 0; /* hush, gcc */
2690 const char *buffer_end;
d8eae41e 2691
d8eae41e 2692 if (SvOBJECT(referent)) {
fafee734
NC
2693 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2694
2695 if (name) {
2696 stashname = HEK_KEY(name);
2697 stashnamelen = HEK_LEN(name);
2698
2699 if (HEK_UTF8(name)) {
2700 SvUTF8_on(sv);
2701 } else {
2702 SvUTF8_off(sv);
2703 }
2704 } else {
2705 stashname = "__ANON__";
2706 stashnamelen = 8;
2707 }
2708 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2709 + 2 * sizeof(UV) + 2 /* )\0 */;
2710 } else {
2711 len = typelen + 3 /* (0x */
2712 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2713 }
fafee734
NC
2714
2715 Newx(buffer, len, char);
2716 buffer_end = retval = buffer + len;
2717
2718 /* Working backwards */
2719 *--retval = '\0';
2720 *--retval = ')';
2721 do {
2722 *--retval = PL_hexdigit[addr & 15];
2723 } while (addr >>= 4);
2724 *--retval = 'x';
2725 *--retval = '0';
2726 *--retval = '(';
2727
2728 retval -= typelen;
2729 memcpy(retval, typestr, typelen);
2730
2731 if (stashname) {
2732 *--retval = '=';
2733 retval -= stashnamelen;
2734 memcpy(retval, stashname, stashnamelen);
2735 }
2736 /* retval may not neccesarily have reached the start of the
2737 buffer here. */
2738 assert (retval >= buffer);
2739
2740 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2741 }
042dae7a 2742 if (lp)
fafee734
NC
2743 *lp = len;
2744 SAVEFREEPV(buffer);
2745 return retval;
463ee0b2 2746 }
79072805 2747 }
0336b60e 2748 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2749 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2750 report_uninit(sv);
cdb061a3
NC
2751 if (lp)
2752 *lp = 0;
73d840c0 2753 return (char *)"";
79072805 2754 }
79072805 2755 }
28e5dec8
JH
2756 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2757 /* I'm assuming that if both IV and NV are equally valid then
2758 converting the IV is going to be more efficient */
e1ec3a88 2759 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2760 char buf[TYPE_CHARS(UV)];
2761 char *ebuf, *ptr;
2762
2763 if (SvTYPE(sv) < SVt_PVIV)
2764 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2765 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2766 /* inlined from sv_setpvn */
2767 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2768 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2769 SvCUR_set(sv, ebuf - ptr);
2770 s = SvEND(sv);
2771 *s = '\0';
28e5dec8
JH
2772 }
2773 else if (SvNOKp(sv)) {
c81271c3 2774 const int olderrno = errno;
79072805
LW
2775 if (SvTYPE(sv) < SVt_PVNV)
2776 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2777 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2778 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2779 /* some Xenix systems wipe out errno here */
79072805 2780#ifdef apollo
463ee0b2 2781 if (SvNVX(sv) == 0.0)
d1307786 2782 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2783 else
2784#endif /*apollo*/
bbce6d69 2785 {
2d4389e4 2786 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2787 }
79072805 2788 errno = olderrno;
a0d0e21e
LW
2789#ifdef FIXNEGATIVEZERO
2790 if (*s == '-' && s[1] == '0' && !s[2])
d1307786 2791 my_strlcpy(s, "0", SvLEN(s));
a0d0e21e 2792#endif
79072805
LW
2793 while (*s) s++;
2794#ifdef hcx
2795 if (s[-1] == '.')
46fc3d4c 2796 *--s = '\0';
79072805
LW
2797#endif
2798 }
79072805 2799 else {
675c862f 2800 if (isGV_with_GP(sv))
19f6321d 2801 return glob_2pv((GV *)sv, lp);
180488f8 2802
041457d9 2803 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2804 report_uninit(sv);
cdb061a3 2805 if (lp)
00b6aa41 2806 *lp = 0;
25da4f38
IZ
2807 if (SvTYPE(sv) < SVt_PV)
2808 /* Typically the caller expects that sv_any is not NULL now. */
2809 sv_upgrade(sv, SVt_PV);
73d840c0 2810 return (char *)"";
79072805 2811 }
cdb061a3 2812 {
823a54a3 2813 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2814 if (lp)
2815 *lp = len;
2816 SvCUR_set(sv, len);
2817 }
79072805 2818 SvPOK_on(sv);
1d7c1841 2819 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2820 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2821 if (flags & SV_CONST_RETURN)
2822 return (char *)SvPVX_const(sv);
10516c54
NC
2823 if (flags & SV_MUTABLE_RETURN)
2824 return SvPVX_mutable(sv);
463ee0b2
LW
2825 return SvPVX(sv);
2826}
2827
645c22ef 2828/*
6050d10e
JP
2829=for apidoc sv_copypv
2830
2831Copies a stringified representation of the source SV into the
2832destination SV. Automatically performs any necessary mg_get and
54f0641b 2833coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2834UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2835sv_2pv[_flags] but operates directly on an SV instead of just the
2836string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2837would lose the UTF-8'ness of the PV.
2838
2839=cut
2840*/
2841
2842void
2843Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2844{
446eaa42 2845 STRLEN len;
53c1dcc0 2846 const char * const s = SvPV_const(ssv,len);
cb50f42d 2847 sv_setpvn(dsv,s,len);
446eaa42 2848 if (SvUTF8(ssv))
cb50f42d 2849 SvUTF8_on(dsv);
446eaa42 2850 else
cb50f42d 2851 SvUTF8_off(dsv);
6050d10e
JP
2852}
2853
2854/*
645c22ef
DM
2855=for apidoc sv_2pvbyte
2856
2857Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2858to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2859side-effect.
2860
2861Usually accessed via the C<SvPVbyte> macro.
2862
2863=cut
2864*/
2865
7340a771
GS
2866char *
2867Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2868{
0875d2fe 2869 sv_utf8_downgrade(sv,0);
97972285 2870 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2871}
2872
645c22ef 2873/*
035cbb0e
RGS
2874=for apidoc sv_2pvutf8
2875
2876Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2877to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2878
2879Usually accessed via the C<SvPVutf8> macro.
2880
2881=cut
2882*/
645c22ef 2883
7340a771
GS
2884char *
2885Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2886{
035cbb0e
RGS
2887 sv_utf8_upgrade(sv);
2888 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2889}
1c846c1f 2890
7ee2227d 2891
645c22ef
DM
2892/*
2893=for apidoc sv_2bool
2894
2895This function is only called on magical items, and is only used by
8cf8f3d1 2896sv_true() or its macro equivalent.
645c22ef
DM
2897
2898=cut
2899*/
2900
463ee0b2 2901bool
864dbfa3 2902Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2903{
97aff369 2904 dVAR;
5b295bef 2905 SvGETMAGIC(sv);
463ee0b2 2906
a0d0e21e
LW
2907 if (!SvOK(sv))
2908 return 0;
2909 if (SvROK(sv)) {
fabdb6c0
AL
2910 if (SvAMAGIC(sv)) {
2911 SV * const tmpsv = AMG_CALLun(sv,bool_);
2912 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2913 return (bool)SvTRUE(tmpsv);
2914 }
2915 return SvRV(sv) != 0;
a0d0e21e 2916 }
463ee0b2 2917 if (SvPOKp(sv)) {
53c1dcc0
AL
2918 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2919 if (Xpvtmp &&
339049b0 2920 (*sv->sv_u.svu_pv > '0' ||
11343788 2921 Xpvtmp->xpv_cur > 1 ||
339049b0 2922 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2923 return 1;
2924 else
2925 return 0;
2926 }
2927 else {
2928 if (SvIOKp(sv))
2929 return SvIVX(sv) != 0;
2930 else {
2931 if (SvNOKp(sv))
2932 return SvNVX(sv) != 0.0;
180488f8 2933 else {
f7877b28 2934 if (isGV_with_GP(sv))
180488f8
NC
2935 return TRUE;
2936 else
2937 return FALSE;
2938 }
463ee0b2
LW
2939 }
2940 }
79072805
LW
2941}
2942
c461cf8f
JH
2943/*
2944=for apidoc sv_utf8_upgrade
2945
78ea37eb 2946Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2947Forces the SV to string form if it is not already.
4411f3b6
NIS
2948Always sets the SvUTF8 flag to avoid future validity checks even
2949if all the bytes have hibit clear.
c461cf8f 2950
13a6c0e0
JH
2951This is not as a general purpose byte encoding to Unicode interface:
2952use the Encode extension for that.
2953
8d6d96c1
HS
2954=for apidoc sv_utf8_upgrade_flags
2955
78ea37eb 2956Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2957Forces the SV to string form if it is not already.
8d6d96c1
HS
2958Always sets the SvUTF8 flag to avoid future validity checks even
2959if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2960will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2961C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2962
13a6c0e0
JH
2963This is not as a general purpose byte encoding to Unicode interface:
2964use the Encode extension for that.
2965
8d6d96c1
HS
2966=cut
2967*/
2968
2969STRLEN
2970Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2971{
97aff369 2972 dVAR;
808c356f
RGS
2973 if (sv == &PL_sv_undef)
2974 return 0;
e0e62c2a
NIS
2975 if (!SvPOK(sv)) {
2976 STRLEN len = 0;
d52b7888
NC
2977 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2978 (void) sv_2pv_flags(sv,&len, flags);
2979 if (SvUTF8(sv))
2980 return len;
2981 } else {
2982 (void) SvPV_force(sv,len);
2983 }
e0e62c2a 2984 }
4411f3b6 2985
f5cee72b 2986 if (SvUTF8(sv)) {
5fec3b1d 2987 return SvCUR(sv);
f5cee72b 2988 }
5fec3b1d 2989
765f542d
NC
2990 if (SvIsCOW(sv)) {
2991 sv_force_normal_flags(sv, 0);
db42d148
NIS
2992 }
2993
88632417 2994 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2995 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2996 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2997 /* This function could be much more efficient if we
2998 * had a FLAG in SVs to signal if there are any hibit
2999 * chars in the PV. Given that there isn't such a flag
3000 * make the loop as fast as possible. */
00b6aa41 3001 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3002 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3003 const U8 *t = s;
c4e7c712
NC
3004
3005 while (t < e) {
53c1dcc0 3006 const U8 ch = *t++;
00b6aa41
AL
3007 /* Check for hi bit */
3008 if (!NATIVE_IS_INVARIANT(ch)) {
3009 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3010 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3011
3012 SvPV_free(sv); /* No longer using what was there before. */
3013 SvPV_set(sv, (char*)recoded);
3014 SvCUR_set(sv, len - 1);
3015 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3016 break;
00b6aa41 3017 }
c4e7c712
NC
3018 }
3019 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3020 SvUTF8_on(sv);
560a288e 3021 }
4411f3b6 3022 return SvCUR(sv);
560a288e
GS
3023}
3024
c461cf8f
JH
3025/*
3026=for apidoc sv_utf8_downgrade
3027
78ea37eb
TS
3028Attempts to convert the PV of an SV from characters to bytes.
3029If the PV contains a character beyond byte, this conversion will fail;
3030in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3031true, croaks.
3032
13a6c0e0
JH
3033This is not as a general purpose Unicode to byte encoding interface:
3034use the Encode extension for that.
3035
c461cf8f
JH
3036=cut
3037*/
3038
560a288e
GS
3039bool
3040Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3041{
97aff369 3042 dVAR;
78ea37eb 3043 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3044 if (SvCUR(sv)) {
03cfe0ae 3045 U8 *s;
652088fc 3046 STRLEN len;
fa301091 3047
765f542d
NC
3048 if (SvIsCOW(sv)) {
3049 sv_force_normal_flags(sv, 0);
3050 }
03cfe0ae
NIS
3051 s = (U8 *) SvPV(sv, len);
3052 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3053 if (fail_ok)
3054 return FALSE;
3055 else {
3056 if (PL_op)
3057 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3058 OP_DESC(PL_op));
fa301091
JH
3059 else
3060 Perl_croak(aTHX_ "Wide character");
3061 }
4b3603a4 3062 }
b162af07 3063 SvCUR_set(sv, len);
67e989fb 3064 }
560a288e 3065 }
ffebcc3e 3066 SvUTF8_off(sv);
560a288e
GS
3067 return TRUE;
3068}
3069
c461cf8f
JH
3070/*
3071=for apidoc sv_utf8_encode
3072
78ea37eb
TS
3073Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3074flag off so that it looks like octets again.
c461cf8f
JH
3075
3076=cut
3077*/
3078
560a288e
GS
3079void
3080Perl_sv_utf8_encode(pTHX_ register SV *sv)
3081{
4c94c214
NC
3082 if (SvIsCOW(sv)) {
3083 sv_force_normal_flags(sv, 0);
3084 }
3085 if (SvREADONLY(sv)) {
3086 Perl_croak(aTHX_ PL_no_modify);
3087 }
a5f5288a 3088 (void) sv_utf8_upgrade(sv);
560a288e
GS
3089 SvUTF8_off(sv);
3090}
3091
4411f3b6
NIS
3092/*
3093=for apidoc sv_utf8_decode
3094
78ea37eb
TS
3095If the PV of the SV is an octet sequence in UTF-8
3096and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3097so that it looks like a character. If the PV contains only single-byte
3098characters, the C<SvUTF8> flag stays being off.
3099Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3100
3101=cut
3102*/
3103
560a288e
GS
3104bool
3105Perl_sv_utf8_decode(pTHX_ register SV *sv)
3106{
78ea37eb 3107 if (SvPOKp(sv)) {
93524f2b
NC
3108 const U8 *c;
3109 const U8 *e;
9cbac4c7 3110
645c22ef
DM
3111 /* The octets may have got themselves encoded - get them back as
3112 * bytes
3113 */
3114 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3115 return FALSE;
3116
3117 /* it is actually just a matter of turning the utf8 flag on, but
3118 * we want to make sure everything inside is valid utf8 first.
3119 */
93524f2b 3120 c = (const U8 *) SvPVX_const(sv);
63cd0674 3121 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3122 return FALSE;
93524f2b 3123 e = (const U8 *) SvEND(sv);
511c2ff0 3124 while (c < e) {
b64e5050 3125 const U8 ch = *c++;
c4d5f83a 3126 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3127 SvUTF8_on(sv);
3128 break;
3129 }
560a288e 3130 }
560a288e
GS
3131 }
3132 return TRUE;
3133}
3134
954c1994
GS
3135/*
3136=for apidoc sv_setsv
3137
645c22ef
DM
3138Copies the contents of the source SV C<ssv> into the destination SV
3139C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3140function if the source SV needs to be reused. Does not handle 'set' magic.
3141Loosely speaking, it performs a copy-by-value, obliterating any previous
3142content of the destination.
3143
3144You probably want to use one of the assortment of wrappers, such as
3145C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3146C<SvSetMagicSV_nosteal>.
3147
8d6d96c1
HS
3148=for apidoc sv_setsv_flags
3149
645c22ef
DM
3150Copies the contents of the source SV C<ssv> into the destination SV
3151C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3152function if the source SV needs to be reused. Does not handle 'set' magic.
3153Loosely speaking, it performs a copy-by-value, obliterating any previous
3154content of the destination.
3155If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3156C<ssv> if appropriate, else not. If the C<flags> parameter has the
3157C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3158and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3159
3160You probably want to use one of the assortment of wrappers, such as
3161C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3162C<SvSetMagicSV_nosteal>.
3163
3164This is the primary function for copying scalars, and most other
3165copy-ish functions and macros use this underneath.
8d6d96c1
HS
3166
3167=cut
3168*/
3169
5d0301b7 3170static void
2eb42952 3171S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3172{
3173 if (dtype != SVt_PVGV) {
3174 const char * const name = GvNAME(sstr);
3175 const STRLEN len = GvNAMELEN(sstr);
0d092c36 3176 {
f7877b28
NC
3177 if (dtype >= SVt_PV) {
3178 SvPV_free(dstr);
3179 SvPV_set(dstr, 0);
3180 SvLEN_set(dstr, 0);
3181 SvCUR_set(dstr, 0);
3182 }
0d092c36 3183 SvUPGRADE(dstr, SVt_PVGV);
dedf8e73 3184 (void)SvOK_off(dstr);
2e5b91de
NC
3185 /* FIXME - why are we doing this, then turning it off and on again
3186 below? */
3187 isGV_with_GP_on(dstr);
f7877b28 3188 }
5d0301b7
NC
3189 GvSTASH(dstr) = GvSTASH(sstr);
3190 if (GvSTASH(dstr))
3191 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3192 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3193 SvFAKE_on(dstr); /* can coerce to non-glob */
3194 }
3195
3196#ifdef GV_UNIQUE_CHECK
3197 if (GvUNIQUE((GV*)dstr)) {
3198 Perl_croak(aTHX_ PL_no_modify);
3199 }
3200#endif
3201
f7877b28 3202 gp_free((GV*)dstr);
2e5b91de 3203 isGV_with_GP_off(dstr);
5d0301b7 3204 (void)SvOK_off(dstr);
2e5b91de 3205 isGV_with_GP_on(dstr);
dedf8e73 3206 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3207 GvGP(dstr) = gp_ref(GvGP(sstr));
3208 if (SvTAINTED(sstr))
3209 SvTAINT(dstr);
3210 if (GvIMPORTED(dstr) != GVf_IMPORTED
3211 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3212 {
3213 GvIMPORTED_on(dstr);
3214 }
3215 GvMULTI_on(dstr);
3216 return;
3217}
3218
b8473700 3219static void
2eb42952 3220S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3221 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3222 SV *dref = NULL;
3223 const int intro = GvINTRO(dstr);
2440974c 3224 SV **location;
3386d083 3225 U8 import_flag = 0;
27242d61
NC
3226 const U32 stype = SvTYPE(sref);
3227
b8473700
NC
3228
3229#ifdef GV_UNIQUE_CHECK
3230 if (GvUNIQUE((GV*)dstr)) {
3231 Perl_croak(aTHX_ PL_no_modify);
3232 }
3233#endif
3234
3235 if (intro) {
3236 GvINTRO_off(dstr); /* one-shot flag */
3237 GvLINE(dstr) = CopLINE(PL_curcop);
3238 GvEGV(dstr) = (GV*)dstr;
3239 }
3240 GvMULTI_on(dstr);
27242d61 3241 switch (stype) {
b8473700 3242 case SVt_PVCV:
27242d61
NC
3243 location = (SV **) &GvCV(dstr);
3244 import_flag = GVf_IMPORTED_CV;
3245 goto common;
3246 case SVt_PVHV:
3247 location = (SV **) &GvHV(dstr);
3248 import_flag = GVf_IMPORTED_HV;
3249 goto common;
3250 case SVt_PVAV:
3251 location = (SV **) &GvAV(dstr);
3252 import_flag = GVf_IMPORTED_AV;
3253 goto common;
3254 case SVt_PVIO:
3255 location = (SV **) &GvIOp(dstr);
3256 goto common;
3257 case SVt_PVFM:
3258 location = (SV **) &GvFORM(dstr);
3259 default:
3260 location = &GvSV(dstr);
3261 import_flag = GVf_IMPORTED_SV;
3262 common:
b8473700 3263 if (intro) {
27242d61
NC
3264 if (stype == SVt_PVCV) {
3265 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3266 SvREFCNT_dec(GvCV(dstr));
3267 GvCV(dstr) = NULL;
3268 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3269 PL_sub_generation++;
3270 }
b8473700 3271 }
27242d61 3272 SAVEGENERICSV(*location);
b8473700
NC
3273 }
3274 else
27242d61
NC
3275 dref = *location;
3276 if (stype == SVt_PVCV && *location != sref) {
3277 CV* const cv = (CV*)*location;
b8473700
NC
3278 if (cv) {
3279 if (!GvCVGEN((GV*)dstr) &&
3280 (CvROOT(cv) || CvXSUB(cv)))
3281 {
3282 /* Redefining a sub - warning is mandatory if
3283 it was a const and its value changed. */
3284 if (CvCONST(cv) && CvCONST((CV*)sref)
3285 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3286 NOOP;
b8473700
NC
3287 /* They are 2 constant subroutines generated from
3288 the same constant. This probably means that
3289 they are really the "same" proxy subroutine
3290 instantiated in 2 places. Most likely this is
3291 when a constant is exported twice. Don't warn.
3292 */
3293 }
3294 else if (ckWARN(WARN_REDEFINE)
3295 || (CvCONST(cv)
3296 && (!CvCONST((CV*)sref)
3297 || sv_cmp(cv_const_sv(cv),
3298 cv_const_sv((CV*)sref))))) {
3299 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3300 (const char *)
3301 (CvCONST(cv)
3302 ? "Constant subroutine %s::%s redefined"
3303 : "Subroutine %s::%s redefined"),
b8473700
NC
3304 HvNAME_get(GvSTASH((GV*)dstr)),
3305 GvENAME((GV*)dstr));
3306 }
3307 }
3308 if (!intro)
cbf82dd0
NC
3309 cv_ckproto_len(cv, (GV*)dstr,
3310 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3311 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3312 }
b8473700
NC
3313 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3314 GvASSUMECV_on(dstr);
3315 PL_sub_generation++;
3316 }
2440974c 3317 *location = sref;
3386d083
NC
3318 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3319 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3320 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3321 }
3322 break;
3323 }
b37c2d43 3324 SvREFCNT_dec(dref);
b8473700
NC
3325 if (SvTAINTED(sstr))
3326 SvTAINT(dstr);
3327 return;
3328}
3329
8d6d96c1
HS
3330void
3331Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3332{
97aff369 3333 dVAR;
8990e307
LW
3334 register U32 sflags;
3335 register int dtype;
42d0e0b7 3336 register svtype stype;
463ee0b2 3337
79072805
LW
3338 if (sstr == dstr)
3339 return;
29f4f0ab
NC
3340
3341 if (SvIS_FREED(dstr)) {
3342 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
be2597df 3343 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
29f4f0ab 3344 }
765f542d 3345 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3346 if (!sstr)
3280af22 3347 sstr = &PL_sv_undef;
29f4f0ab 3348 if (SvIS_FREED(sstr)) {
6c9570dc
MHM
3349 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3350 (void*)sstr, (void*)dstr);
29f4f0ab 3351 }
8990e307
LW
3352 stype = SvTYPE(sstr);
3353 dtype = SvTYPE(dstr);
79072805 3354
52944de8 3355 (void)SvAMAGIC_off(dstr);
7a5fa8a2 3356 if ( SvVOK(dstr) )
ece467f9
JP
3357 {
3358 /* need to nuke the magic */
3359 mg_free(dstr);
3360 SvRMAGICAL_off(dstr);
3361 }
9e7bc3e8 3362
463ee0b2 3363 /* There's a lot of redundancy below but we're going for speed here */
79072805 3364
8990e307 3365 switch (stype) {
79072805 3366 case SVt_NULL:
aece5585 3367 undef_sstr:
20408e3c
GS
3368 if (dtype != SVt_PVGV) {
3369 (void)SvOK_off(dstr);
3370 return;
3371 }
3372 break;
463ee0b2 3373 case SVt_IV:
aece5585
GA
3374 if (SvIOK(sstr)) {
3375 switch (dtype) {
3376 case SVt_NULL:
8990e307 3377 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3378 break;
3379 case SVt_NV:
aece5585
GA
3380 case SVt_RV:
3381 case SVt_PV:
a0d0e21e 3382 sv_upgrade(dstr, SVt_PVIV);
aece5585 3383 break;
010be86b
NC
3384 case SVt_PVGV:
3385 goto end_of_first_switch;
aece5585
GA
3386 }
3387 (void)SvIOK_only(dstr);
45977657 3388 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3389 if (SvIsUV(sstr))
3390 SvIsUV_on(dstr);
37c25af0
NC
3391 /* SvTAINTED can only be true if the SV has taint magic, which in
3392 turn means that the SV type is PVMG (or greater). This is the
3393 case statement for SVt_IV, so this cannot be true (whatever gcov
3394 may say). */
3395 assert(!SvTAINTED(sstr));
aece5585 3396 return;
8990e307 3397 }
aece5585
GA
3398 goto undef_sstr;
3399
463ee0b2 3400 case SVt_NV:
aece5585
GA
3401 if (SvNOK(sstr)) {
3402 switch (dtype) {
3403 case SVt_NULL:
3404 case SVt_IV:
8990e307 3405 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3406 break;
3407 case SVt_RV:
3408 case SVt_PV:
3409 case SVt_PVIV:
a0d0e21e 3410 sv_upgrade(dstr, SVt_PVNV);
aece5585 3411 break;
010be86b
NC
3412 case SVt_PVGV:
3413 goto end_of_first_switch;
aece5585 3414 }
9d6ce603 3415 SvNV_set(dstr, SvNVX(sstr));
aece5585 3416 (void)SvNOK_only(dstr);
37c25af0
NC
3417 /* SvTAINTED can only be true if the SV has taint magic, which in
3418 turn means that the SV type is PVMG (or greater). This is the
3419 case statement for SVt_NV, so this cannot be true (whatever gcov
3420 may say). */
3421 assert(!SvTAINTED(sstr));
aece5585 3422 return;
8990e307 3423 }
aece5585
GA
3424 goto undef_sstr;
3425
ed6116ce 3426 case SVt_RV:
8990e307 3427 if (dtype < SVt_RV)
ed6116ce 3428 sv_upgrade(dstr, SVt_RV);
ed6116ce 3429 break;
fc36a67e 3430 case SVt_PVFM:
f8c7b90f 3431#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3432 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3433 if (dtype < SVt_PVIV)
3434 sv_upgrade(dstr, SVt_PVIV);
3435 break;
3436 }
3437 /* Fall through */
3438#endif
3439 case SVt_PV:
8990e307 3440 if (dtype < SVt_PV)
463ee0b2 3441 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3442 break;
3443 case SVt_PVIV:
8990e307 3444 if (dtype < SVt_PVIV)
463ee0b2 3445 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3446 break;
3447 case SVt_PVNV:
8990e307 3448 if (dtype < SVt_PVNV)
463ee0b2 3449 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3450 break;
489f7bfe 3451 default:
a3b680e6
AL
3452 {
3453 const char * const type = sv_reftype(sstr,0);
533c011a 3454 if (PL_op)
a3b680e6 3455 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3456 else
a3b680e6
AL
3457 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3458 }
4633a7c4
LW
3459 break;
3460
cecf5685 3461 /* case SVt_BIND: */
39cb70dc 3462 case SVt_PVLV:
79072805 3463 case SVt_PVGV:
cecf5685 3464 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
d4c19fe8 3465 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3466 return;
79072805 3467 }
cecf5685 3468 /* SvVALID means that this PVGV is playing at being an FBM. */
5f66b61c 3469 /*FALLTHROUGH*/
79072805 3470
489f7bfe 3471 case SVt_PVMG:
8d6d96c1 3472 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3473 mg_get(sstr);
1d9c78c6 3474 if (SvTYPE(sstr) != stype) {
973f89ab 3475 stype = SvTYPE(sstr);
cecf5685 3476 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3477 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3478 return;
3479 }
973f89ab
CS
3480 }
3481 }
ded42b9f 3482 if (stype == SVt_PVLV)
862a34c6 3483 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3484 else
42d0e0b7 3485 SvUPGRADE(dstr, (svtype)stype);
79072805 3486 }
010be86b 3487 end_of_first_switch:
79072805 3488
ff920335
NC
3489 /* dstr may have been upgraded. */
3490 dtype = SvTYPE(dstr);
8990e307
LW
3491 sflags = SvFLAGS(sstr);
3492
ba2fdce6 3493 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
85324b4d
NC
3494 /* Assigning to a subroutine sets the prototype. */
3495 if (SvOK(sstr)) {
3496 STRLEN len;
3497 const char *const ptr = SvPV_const(sstr, len);
3498
3499 SvGROW(dstr, len + 1);
3500 Copy(ptr, SvPVX(dstr), len + 1, char);
3501 SvCUR_set(dstr, len);
fcddd32e 3502 SvPOK_only(dstr);
ba2fdce6 3503 SvFLAGS(dstr) |= sflags & SVf_UTF8;
85324b4d
NC
3504 } else {
3505 SvOK_off(dstr);
3506 }
ba2fdce6
NC
3507 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3508 const char * const type = sv_reftype(dstr,0);
3509 if (PL_op)
3510 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3511 else
3512 Perl_croak(aTHX_ "Cannot copy to %s", type);
85324b4d 3513 } else if (sflags & SVf_ROK) {
cecf5685
NC
3514 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3515 && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
acaa9288
NC
3516 sstr = SvRV(sstr);
3517 if (sstr == dstr) {
3518 if (GvIMPORTED(dstr) != GVf_IMPORTED
3519 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3520 {
3521 GvIMPORTED_on(dstr);
3522 }
3523 GvMULTI_on(dstr);
3524 return;
3525 }
d4c19fe8 3526 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3527 return;
3528 }
3529
8990e307 3530 if (dtype >= SVt_PV) {
b8c701c1 3531 if (dtype == SVt_PVGV) {
d4c19fe8 3532 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3533 return;
3534 }
3f7c398e 3535 if (SvPVX_const(dstr)) {
8bd4d4c5 3536 SvPV_free(dstr);
b162af07
SP
3537 SvLEN_set(dstr, 0);
3538 SvCUR_set(dstr, 0);
a0d0e21e 3539 }
8990e307 3540 }
a0d0e21e 3541 (void)SvOK_off(dstr);
b162af07 3542 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
96d4b0ee 3543 SvFLAGS(dstr) |= sflags & SVf_ROK;
dfd48732
NC
3544 assert(!(sflags & SVp_NOK));
3545 assert(!(sflags & SVp_IOK));
3546 assert(!(sflags & SVf_NOK));
3547 assert(!(sflags & SVf_IOK));
ed6116ce 3548 }
cecf5685 3549 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
c0c44674
NC
3550 if (!(sflags & SVf_OK)) {
3551 if (ckWARN(WARN_MISC))
3552 Perl_warner(aTHX_ packWARN(WARN_MISC),
3553 "Undefined value assigned to typeglob");
3554 }
3555 else {
3556 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3557 if (dstr != (SV*)gv) {
3558 if (GvGP(dstr))
3559 gp_free((GV*)dstr);
3560 GvGP(dstr) = gp_ref(GvGP(gv));
3561 }
3562 }
3563 }
8990e307 3564 else if (sflags & SVp_POK) {
765f542d 3565 bool isSwipe = 0;
79072805
LW
3566
3567 /*
3568 * Check to see if we can just swipe the string. If so, it's a
3569 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3570 * It might even be a win on short strings if SvPVX_const(dstr)
3571 * has to be allocated and SvPVX_const(sstr) has to be freed.
34482cd6
NC
3572 * Likewise if we can set up COW rather than doing an actual copy, we
3573 * drop to the else clause, as the swipe code and the COW setup code
3574 * have much in common.
79072805
LW
3575 */
3576
120fac95
NC
3577 /* Whichever path we take through the next code, we want this true,
3578 and doing it now facilitates the COW check. */
3579 (void)SvPOK_only(dstr);
3580
765f542d 3581 if (
34482cd6
NC
3582 /* If we're already COW then this clause is not true, and if COW
3583 is allowed then we drop down to the else and make dest COW
3584 with us. If caller hasn't said that we're allowed to COW
3585 shared hash keys then we don't do the COW setup, even if the
3586 source scalar is a shared hash key scalar. */
3587 (((flags & SV_COW_SHARED_HASH_KEYS)
3588 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
3589 : 1 /* If making a COW copy is forbidden then the behaviour we
3590 desire is as if the source SV isn't actually already
3591 COW, even if it is. So we act as if the source flags
3592 are not COW, rather than actually testing them. */
3593 )
f8c7b90f 3594#ifndef PERL_OLD_COPY_ON_WRITE
34482cd6
NC
3595 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
3596 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
3597 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
3598 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
3599 but in turn, it's somewhat dead code, never expected to go
3600 live, but more kept as a placeholder on how to do it better
3601 in a newer implementation. */
3602 /* If we are COW and dstr is a suitable target then we drop down
3603 into the else and make dest a COW of us. */
b8f9541a
NC
3604 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3605#endif
3606 )
765f542d 3607 &&
765f542d
NC
3608 !(isSwipe =
3609 (sflags & SVs_TEMP) && /* slated for free anyway? */
3610 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3611 (!(flags & SV_NOSTEAL)) &&
3612 /* and we're allowed to steal temps */
765f542d
NC
3613 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3614 SvLEN(sstr) && /* and really is a string */
645c22ef 3615 /* and won't be needed again, potentially */
765f542d 3616 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3617#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3618 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3619 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3620 && SvTYPE(sstr) >= SVt_PVIV)
3621#endif
3622 ) {
3623 /* Failed the swipe test, and it's not a shared hash key either.
3624 Have to copy the string. */
3625 STRLEN len = SvCUR(sstr);
3626 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3627 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3628 SvCUR_set(dstr, len);
3629 *SvEND(dstr) = '\0';
765f542d 3630 } else {
f8c7b90f 3631 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3632 be true in here. */
765f542d
NC
3633 /* Either it's a shared hash key, or it's suitable for
3634 copy-on-write or we can swipe the string. */
46187eeb 3635 if (DEBUG_C_TEST) {
ed252734 3636 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3637 sv_dump(sstr);
3638 sv_dump(dstr);
46187eeb 3639 }
f8c7b90f 3640#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3641 if (!isSwipe) {
3642 /* I believe I should acquire a global SV mutex if
3643 it's a COW sv (not a shared hash key) to stop
3644 it going un copy-on-write.
3645 If the source SV has gone un copy on write between up there
3646 and down here, then (assert() that) it is of the correct
3647 form to make it copy on write again */
3648 if ((sflags & (SVf_FAKE | SVf_READONLY))
3649 != (SVf_FAKE | SVf_READONLY)) {
3650 SvREADONLY_on(sstr);
3651 SvFAKE_on(sstr);
3652 /* Make the source SV into a loop of 1.
3653 (about to become 2) */
a29f6d03 3654 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3655 }
3656 }
3657#endif
3658 /* Initial code is common. */
94010e71
NC
3659 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3660 SvPV_free(dstr);
79072805 3661 }
765f542d 3662
765f542d
NC
3663 if (!isSwipe) {
3664 /* making another shared SV. */
3665 STRLEN cur = SvCUR(sstr);
3666 STRLEN len = SvLEN(sstr);
f8c7b90f 3667#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3668 if (len) {
b8f9541a 3669 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3670 /* SvIsCOW_normal */
3671 /* splice us in between source and next-after-source. */
a29f6d03
NC
3672 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3673 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3674 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3675 } else
3676#endif
3677 {
765f542d 3678 /* SvIsCOW_shared_hash */
46187eeb
NC
3679 DEBUG_C(PerlIO_printf(Perl_debug_log,
3680 "Copy on write: Sharing hash\n"));
b8f9541a 3681
bdd68bc3 3682 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3683 SvPV_set(dstr,
d1db91c6 3684 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3685 }
87a1ef3d
SP
3686 SvLEN_set(dstr, len);
3687 SvCUR_set(dstr, cur);
765f542d
NC
3688 SvREADONLY_on(dstr);
3689 SvFAKE_on(dstr);
3690 /* Relesase a global SV mutex. */
3691 }
3692 else
765f542d 3693 { /* Passes the swipe test. */
78d1e721 3694 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3695 SvLEN_set(dstr, SvLEN(sstr));
3696 SvCUR_set(dstr, SvCUR(sstr));
3697
3698 SvTEMP_off(dstr);
3699 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3700 SvPV_set(sstr, NULL);
765f542d
NC
3701 SvLEN_set(sstr, 0);
3702 SvCUR_set(sstr, 0);
3703 SvTEMP_off(sstr);
3704 }
3705 }
8990e307 3706 if (sflags & SVp_NOK) {
9d6ce603 3707 SvNV_set(dstr, SvNVX(sstr));
79072805 3708 }
8990e307 3709 if (sflags & SVp_IOK) {
88555484 3710 SvOOK_off(dstr);
23525414
NC
3711 SvIV_set(dstr, SvIVX(sstr));
3712 /* Must do this otherwise some other overloaded use of 0x80000000
3713 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3714 if (sflags & SVf_IVisUV)
25da4f38 3715 SvIsUV_on(dstr);
79072805 3716 }
96d4b0ee 3717 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4f2da183 3718 {
b0a11fe1 3719 const MAGIC * const smg = SvVSTRING_mg(sstr);
4f2da183
NC
3720 if (smg) {
3721 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3722 smg->mg_ptr, smg->mg_len);
3723 SvRMAGICAL_on(dstr);
3724 }
7a5fa8a2 3725 }
79072805 3726 }
5d581361 3727 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3728 (void)SvOK_off(dstr);
96d4b0ee 3729 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
5d581361
NC
3730 if (sflags & SVp_IOK) {
3731 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3732 SvIV_set(dstr, SvIVX(sstr));
3733 }
3332b3c1 3734 if (sflags & SVp_NOK) {
9d6ce603 3735 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3736 }
3737 }
79072805 3738 else {
f7877b28 3739 if (isGV_with_GP(sstr)) {
180488f8
NC
3740 /* This stringification rule for globs is spread in 3 places.
3741 This feels bad. FIXME. */
3742 const U32 wasfake = sflags & SVf_FAKE;
3743
3744 /* FAKE globs can get coerced, so need to turn this off
3745 temporarily if it is on. */
3746 SvFAKE_off(sstr);
3747 gv_efullname3(dstr, (GV *)sstr, "*");
3748 SvFLAGS(sstr) |= wasfake;
3749 }
20408e3c
GS
3750 else
3751 (void)SvOK_off(dstr);
a0d0e21e 3752 }
27c9684d
AP
3753 if (SvTAINTED(sstr))
3754 SvTAINT(dstr);
79072805
LW
3755}
3756
954c1994
GS
3757/*
3758=for apidoc sv_setsv_mg
3759
3760Like C<sv_setsv>, but also handles 'set' magic.
3761
3762=cut
3763*/
3764
79072805 3765void
864dbfa3 3766Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3767{
3768 sv_setsv(dstr,sstr);
3769 SvSETMAGIC(dstr);
3770}
3771
f8c7b90f 3772#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3773SV *
3774Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3775{
3776 STRLEN cur = SvCUR(sstr);
3777 STRLEN len = SvLEN(sstr);
3778 register char *new_pv;
3779
3780 if (DEBUG_C_TEST) {
3781 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
6c9570dc 3782 (void*)sstr, (void*)dstr);
ed252734
NC
3783 sv_dump(sstr);
3784 if (dstr)
3785 sv_dump(dstr);
3786 }
3787
3788 if (dstr) {
3789 if (SvTHINKFIRST(dstr))
3790 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3791 else if (SvPVX_const(dstr))
3792 Safefree(SvPVX_const(dstr));
ed252734
NC
3793 }
3794 else
3795 new_SV(dstr);
862a34c6 3796 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3797
3798 assert (SvPOK(sstr));
3799 assert (SvPOKp(sstr));
3800 assert (!SvIOK(sstr));
3801 assert (!SvIOKp(sstr));
3802 assert (!SvNOK(sstr));
3803 assert (!SvNOKp(sstr));
3804
3805 if (SvIsCOW(sstr)) {
3806
3807 if (SvLEN(sstr) == 0) {
3808 /* source is a COW shared hash key. */
ed252734
NC
3809 DEBUG_C(PerlIO_printf(Perl_debug_log,
3810 "Fast copy on write: Sharing hash\n"));
d1db91c6 3811 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3812 goto common_exit;
3813 }
3814 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3815 } else {
3816 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3817 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3818 SvREADONLY_on(sstr);
3819 SvFAKE_on(sstr);
3820 DEBUG_C(PerlIO_printf(Perl_debug_log,
3821 "Fast copy on write: Converting sstr to COW\n"));
3822 SV_COW_NEXT_SV_SET(dstr, sstr);
3823 }
3824 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3825 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3826
3827 common_exit:
3828 SvPV_set(dstr, new_pv);
3829 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3830 if (SvUTF8(sstr))
3831 SvUTF8_on(dstr);
87a1ef3d
SP
3832 SvLEN_set(dstr, len);
3833 SvCUR_set(dstr, cur);
ed252734
NC
3834 if (DEBUG_C_TEST) {
3835 sv_dump(dstr);
3836 }
3837 return dstr;
3838}
3839#endif
3840
954c1994
GS
3841/*
3842=for apidoc sv_setpvn
3843
3844Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3845bytes to be copied. If the C<ptr> argument is NULL the SV will become
3846undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3847
3848=cut
3849*/
3850
ef50df4b 3851void
864dbfa3 3852Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3853{
97aff369 3854 dVAR;
c6f8c383 3855 register char *dptr;
22c522df 3856
765f542d 3857 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3858 if (!ptr) {
a0d0e21e 3859 (void)SvOK_off(sv);
463ee0b2
LW
3860 return;
3861 }
22c522df
JH
3862 else {
3863 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3864 const IV iv = len;
9c5ffd7c
JH
3865 if (iv < 0)
3866 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3867 }
862a34c6 3868 SvUPGRADE(sv, SVt_PV);
c6f8c383 3869
5902b6a9 3870 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3871 Move(ptr,dptr,len,char);
3872 dptr[len] = '\0';
79072805 3873 SvCUR_set(sv, len);
1aa99e6b 3874 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3875 SvTAINT(sv);
79072805
LW
3876}
3877
954c1994
GS
3878/*
3879=for apidoc sv_setpvn_mg
3880
3881Like C<sv_setpvn>, but also handles 'set' magic.
3882
3883=cut
3884*/
3885
79072805 3886void
864dbfa3 3887Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3888{
3889 sv_setpvn(sv,ptr,len);
3890 SvSETMAGIC(sv);
3891}
3892
954c1994
GS
3893/*
3894=for apidoc sv_setpv
3895
3896Copies a string into an SV. The string must be null-terminated. Does not
3897handle 'set' magic. See C<sv_setpv_mg>.
3898
3899=cut
3900*/
3901
ef50df4b 3902void
864dbfa3 3903Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3904{
97aff369 3905 dVAR;
79072805
LW
3906 register STRLEN len;
3907
765f542d 3908 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3909 if (!ptr) {
a0d0e21e 3910 (void)SvOK_off(sv);
463ee0b2
LW
3911 return;
3912 }
79072805 3913 len = strlen(ptr);
862a34c6 3914 SvUPGRADE(sv, SVt_PV);
c6f8c383 3915
79072805 3916 SvGROW(sv, len + 1);
463ee0b2 3917 Move(ptr,SvPVX(sv),len+1,char);
79072805 3918 SvCUR_set(sv, len);
1aa99e6b 3919 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3920 SvTAINT(sv);
3921}
3922
954c1994
GS
3923/*
3924=for apidoc sv_setpv_mg
3925
3926Like C<sv_setpv>, but also handles 'set' magic.
3927
3928=cut
3929*/
3930
463ee0b2 3931void
864dbfa3 3932Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3933{
3934 sv_setpv(sv,ptr);
3935 SvSETMAGIC(sv);
3936}
3937
954c1994 3938/*
47518d95 3939=for apidoc sv_usepvn_flags
954c1994 3940
794a0d33
JH
3941Tells an SV to use C<ptr> to find its string value. Normally the
3942string is stored inside the SV but sv_usepvn allows the SV to use an
3943outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
3944by C<malloc>. The string length, C<len>, must be supplied. By default
3945this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
3946so that pointer should not be freed or used by the programmer after
3947giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
3948that pointer (e.g. ptr + 1) be used.
3949
3950If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3951SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 3952will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 3953C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3954
3955=cut
3956*/
3957
ef50df4b 3958void
47518d95 3959Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3960{
97aff369 3961 dVAR;
1936d2a7 3962 STRLEN allocate;
765f542d 3963 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3964 SvUPGRADE(sv, SVt_PV);
463ee0b2 3965 if (!ptr) {
a0d0e21e 3966 (void)SvOK_off(sv);
47518d95
NC
3967 if (flags & SV_SMAGIC)
3968 SvSETMAGIC(sv);
463ee0b2
LW
3969 return;
3970 }
3f7c398e 3971 if (SvPVX_const(sv))
8bd4d4c5 3972 SvPV_free(sv);
1936d2a7 3973
0b7042f9 3974#ifdef DEBUGGING
2e90b4cd
NC
3975 if (flags & SV_HAS_TRAILING_NUL)
3976 assert(ptr[len] == '\0');
0b7042f9 3977#endif
2e90b4cd 3978
c1c21316 3979 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3980 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
cbf82dd0
NC
3981 if (flags & SV_HAS_TRAILING_NUL) {
3982 /* It's long enough - do nothing.
3983 Specfically Perl_newCONSTSUB is relying on this. */
3984 } else {
69d25b4f 3985#ifdef DEBUGGING
69d25b4f 3986 /* Force a move to shake out bugs in callers. */
10edeb5d 3987 char *new_ptr = (char*)safemalloc(allocate);
69d25b4f
NC
3988 Copy(ptr, new_ptr, len, char);
3989 PoisonFree(ptr,len,char);
3990 Safefree(ptr);
3991 ptr = new_ptr;
69d25b4f 3992#else
10edeb5d 3993 ptr = (char*) saferealloc (ptr, allocate);
69d25b4f 3994#endif
cbf82dd0 3995 }
f880fe2f 3996 SvPV_set(sv, ptr);
463ee0b2 3997 SvCUR_set(sv, len);
1936d2a7 3998 SvLEN_set(sv, allocate);
c1c21316
NC
3999 if (!(flags & SV_HAS_TRAILING_NUL)) {
4000 *SvEND(sv) = '\0';
4001 }
1aa99e6b 4002 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4003 SvTAINT(sv);
47518d95
NC
4004 if (flags & SV_SMAGIC)
4005 SvSETMAGIC(sv);
ef50df4b
GS
4006}
4007
f8c7b90f 4008#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4009/* Need to do this *after* making the SV normal, as we need the buffer
4010 pointer to remain valid until after we've copied it. If we let go too early,
4011 another thread could invalidate it by unsharing last of the same hash key
4012 (which it can do by means other than releasing copy-on-write Svs)
4013 or by changing the other copy-on-write SVs in the loop. */
4014STATIC void
5302ffd4 4015S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
765f542d 4016{
5302ffd4 4017 { /* this SV was SvIsCOW_normal(sv) */
765f542d 4018 /* we need to find the SV pointing to us. */
cf5629ad 4019 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 4020
765f542d
NC
4021 if (current == sv) {
4022 /* The SV we point to points back to us (there were only two of us
4023 in the loop.)
4024 Hence other SV is no longer copy on write either. */
4025 SvFAKE_off(after);
4026 SvREADONLY_off(after);
4027 } else {
4028 /* We need to follow the pointers around the loop. */
4029 SV *next;
4030 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4031 assert (next);
4032 current = next;
4033 /* don't loop forever if the structure is bust, and we have
4034 a pointer into a closed loop. */
4035 assert (current != after);
3f7c398e 4036 assert (SvPVX_const(current) == pvx);
765f542d
NC
4037 }
4038 /* Make the SV before us point to the SV after us. */
a29f6d03 4039 SV_COW_NEXT_SV_SET(current, after);
765f542d 4040 }
765f542d
NC
4041 }
4042}
765f542d 4043#endif
645c22ef
DM
4044/*
4045=for apidoc sv_force_normal_flags
4046
4047Undo various types of fakery on an SV: if the PV is a shared string, make
4048a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4049an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4050we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4051then a copy-on-write scalar drops its PV buffer (if any) and becomes
4052SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4053set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4054C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4055with flags set to 0.
645c22ef
DM
4056
4057=cut
4058*/
4059
6fc92669 4060void
840a7b70 4061Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4062{
97aff369 4063 dVAR;
f8c7b90f 4064#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4065 if (SvREADONLY(sv)) {
4066 /* At this point I believe I should acquire a global SV mutex. */
4067 if (SvFAKE(sv)) {
b64e5050 4068 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4069 const STRLEN len = SvLEN(sv);
4070 const STRLEN cur = SvCUR(sv);
5302ffd4
NC
4071 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4072 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4073 we'll fail an assertion. */
4074 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4075
46187eeb
NC
4076 if (DEBUG_C_TEST) {
4077 PerlIO_printf(Perl_debug_log,
4078 "Copy on write: Force normal %ld\n",
4079 (long) flags);
e419cbc5 4080 sv_dump(sv);
46187eeb 4081 }
765f542d
NC
4082 SvFAKE_off(sv);
4083 SvREADONLY_off(sv);
9f653bb5 4084 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4085 SvPV_set(sv, NULL);
87a1ef3d 4086 SvLEN_set(sv, 0);
765f542d
NC
4087 if (flags & SV_COW_DROP_PV) {
4088 /* OK, so we don't need to copy our buffer. */
4089 SvPOK_off(sv);
4090 } else {
4091 SvGROW(sv, cur + 1);
4092 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4093 SvCUR_set(sv, cur);
765f542d
NC
4094 *SvEND(sv) = '\0';
4095 }
5302ffd4
NC
4096 if (len) {
4097 sv_release_COW(sv, pvx, next);
4098 } else {
4099 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4100 }
46187eeb 4101 if (DEBUG_C_TEST) {
e419cbc5 4102 sv_dump(sv);
46187eeb 4103 }
765f542d 4104 }
923e4eb5 4105 else if (IN_PERL_RUNTIME)
765f542d
NC
4106 Perl_croak(aTHX_ PL_no_modify);
4107 /* At this point I believe that I can drop the global SV mutex. */
4108 }
4109#else
2213622d 4110 if (SvREADONLY(sv)) {
1c846c1f 4111 if (SvFAKE(sv)) {
b64e5050 4112 const char * const pvx = SvPVX_const(sv);
66a1b24b 4113 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4114 SvFAKE_off(sv);
4115 SvREADONLY_off(sv);
bd61b366 4116 SvPV_set(sv, NULL);
66a1b24b 4117 SvLEN_set(sv, 0);
1c846c1f 4118 SvGROW(sv, len + 1);
706aa1c9 4119 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4120 *SvEND(sv) = '\0';
bdd68bc3 4121 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4122 }
923e4eb5 4123 else if (IN_PERL_RUNTIME)
cea2e8a9 4124 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4125 }
765f542d 4126#endif
2213622d 4127 if (SvROK(sv))
840a7b70 4128 sv_unref_flags(sv, flags);
6fc92669
GS
4129 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4130 sv_unglob(sv);
0f15f207 4131}
1c846c1f 4132
645c22ef 4133/*
954c1994
GS
4134=for apidoc sv_chop
4135
1c846c1f 4136Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4137SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4138the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4139string. Uses the "OOK hack".
3f7c398e 4140Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4141refer to the same chunk of data.
954c1994
GS
4142
4143=cut
4144*/
4145
79072805 4146void
f54cb97a 4147Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4148{
4149 register STRLEN delta;
a0d0e21e 4150 if (!ptr || !SvPOKp(sv))
79072805 4151 return;
3f7c398e 4152 delta = ptr - SvPVX_const(sv);
2213622d 4153 SV_CHECK_THINKFIRST(sv);
79072805
LW
4154 if (SvTYPE(sv) < SVt_PVIV)
4155 sv_upgrade(sv,SVt_PVIV);
4156
4157 if (!SvOOK(sv)) {
50483b2c 4158 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4159 const char *pvx = SvPVX_const(sv);
a28509cc 4160 const STRLEN len = SvCUR(sv);
50483b2c 4161 SvGROW(sv, len + 1);
706aa1c9 4162 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4163 *SvEND(sv) = '\0';
4164 }
45977657 4165 SvIV_set(sv, 0);
a4bfb290
AB
4166 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4167 and we do that anyway inside the SvNIOK_off
4168 */
7a5fa8a2 4169 SvFLAGS(sv) |= SVf_OOK;
79072805 4170 }
a4bfb290 4171 SvNIOK_off(sv);
b162af07
SP
4172 SvLEN_set(sv, SvLEN(sv) - delta);
4173 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4174 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4175 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4176}
4177
954c1994
GS
4178/*
4179=for apidoc sv_catpvn
4180
4181Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4182C<len> indicates number of bytes to copy. If the SV has the UTF-8
4183status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4184Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4185
8d6d96c1
HS
4186=for apidoc sv_catpvn_flags
4187
4188Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4189C<len> indicates number of bytes to copy. If the SV has the UTF-8
4190status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4191If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4192appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4193in terms of this function.
4194
4195=cut
4196*/
4197
4198void
4199Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4200{
97aff369 4201 dVAR;
8d6d96c1 4202 STRLEN dlen;
fabdb6c0 4203 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4204
8d6d96c1
HS
4205 SvGROW(dsv, dlen + slen + 1);
4206 if (sstr == dstr)
3f7c398e 4207 sstr = SvPVX_const(dsv);
8d6d96c1 4208 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4209 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4210 *SvEND(dsv) = '\0';
4211 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4212 SvTAINT(dsv);
bddd5118
NC
4213 if (flags & SV_SMAGIC)
4214 SvSETMAGIC(dsv);
79072805
LW
4215}
4216
954c1994 4217/*
954c1994
GS
4218=for apidoc sv_catsv
4219
13e8c8e3
JH
4220Concatenates the string from SV C<ssv> onto the end of the string in
4221SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4222not 'set' magic. See C<sv_catsv_mg>.
954c1994 4223
8d6d96c1
HS
4224=for apidoc sv_catsv_flags
4225
4226Concatenates the string from SV C<ssv> onto the end of the string in
4227SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4228bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4229and C<sv_catsv_nomg> are implemented in terms of this function.
4230
4231=cut */
4232
ef50df4b 4233void
8d6d96c1 4234Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4235{
97aff369 4236 dVAR;
bddd5118 4237 if (ssv) {
00b6aa41
AL
4238 STRLEN slen;
4239 const char *spv = SvPV_const(ssv, slen);
4240 if (spv) {
bddd5118
NC
4241 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4242 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4243 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4244 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4245 dsv->sv_flags doesn't have that bit set.
4fd84b44 4246 Andy Dougherty 12 Oct 2001
bddd5118
NC
4247 */
4248 const I32 sutf8 = DO_UTF8(ssv);
4249 I32 dutf8;
13e8c8e3 4250
bddd5118
NC
4251 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4252 mg_get(dsv);
4253 dutf8 = DO_UTF8(dsv);
8d6d96c1 4254
bddd5118
NC
4255 if (dutf8 != sutf8) {
4256 if (dutf8) {
4257 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4258 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4259
bddd5118
NC
4260 sv_utf8_upgrade(csv);
4261 spv = SvPV_const(csv, slen);
4262 }
4263 else
4264 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4265 }
bddd5118 4266 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4267 }
560a288e 4268 }
bddd5118
NC
4269 if (flags & SV_SMAGIC)
4270 SvSETMAGIC(dsv);
79072805
LW
4271}
4272
954c1994 4273/*
954c1994
GS
4274=for apidoc sv_catpv
4275
4276Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4277If the SV has the UTF-8 status set, then the bytes appended should be
4278valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4279
d5ce4a7c 4280=cut */
954c1994 4281
ef50df4b 4282void
0c981600 4283Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4284{
97aff369 4285 dVAR;
79072805 4286 register STRLEN len;
463ee0b2 4287 STRLEN tlen;
748a9306 4288 char *junk;
79072805 4289
0c981600 4290 if (!ptr)
79072805 4291 return;
748a9306 4292 junk = SvPV_force(sv, tlen);
0c981600 4293 len = strlen(ptr);
463ee0b2 4294 SvGROW(sv, tlen + len + 1);
0c981600 4295 if (ptr == junk)
3f7c398e 4296 ptr = SvPVX_const(sv);
0c981600 4297 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4298 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4299 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4300 SvTAINT(sv);
79072805
LW
4301}
4302
954c1994
GS
4303/*
4304=for apidoc sv_catpv_mg
4305
4306Like C<sv_catpv>, but also handles 'set' magic.
4307
4308=cut
4309*/
4310
ef50df4b 4311void
0c981600 4312Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4313{
0c981600 4314 sv_catpv(sv,ptr);
ef50df4b
GS
4315 SvSETMAGIC(sv);
4316}
4317
645c22ef
DM
4318/*
4319=for apidoc newSV
4320
561b68a9
SH
4321Creates a new SV. A non-zero C<len> parameter indicates the number of
4322bytes of preallocated string space the SV should have. An extra byte for a
4323trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4324space is allocated.) The reference count for the new SV is set to 1.
4325
4326In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4327parameter, I<x>, a debug aid which allowed callers to identify themselves.
4328This aid has been superseded by a new build option, PERL_MEM_LOG (see
4329L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4330modules supporting older perls.
645c22ef
DM
4331
4332=cut
4333*/
4334
79072805 4335SV *
864dbfa3 4336Perl_newSV(pTHX_ STRLEN len)
79072805 4337{
97aff369 4338 dVAR;
79072805 4339 register SV *sv;
1c846c1f 4340
4561caa4 4341 new_SV(sv);
79072805
LW
4342 if (len) {
4343 sv_upgrade(sv, SVt_PV);
4344 SvGROW(sv, len + 1);
4345 }
4346 return sv;
4347}
954c1994 4348/*
92110913 4349=for apidoc sv_magicext
954c1994 4350
68795e93 4351Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4352supplied vtable and returns a pointer to the magic added.
92110913 4353
2d8d5d5a
SH
4354Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4355In particular, you can add magic to SvREADONLY SVs, and add more than
4356one instance of the same 'how'.
645c22ef 4357
2d8d5d5a
SH
4358If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4359stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4360special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4361to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4362
2d8d5d5a 4363(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4364
4365=cut
4366*/
92110913 4367MAGIC *
53d44271 4368Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, const MGVTBL *vtable,
92110913 4369 const char* name, I32 namlen)
79072805 4370{
97aff369 4371 dVAR;
79072805 4372 MAGIC* mg;
68795e93 4373
92110913 4374 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4375 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4376 }
a02a5408 4377 Newxz(mg, 1, MAGIC);
79072805 4378 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4379 SvMAGIC_set(sv, mg);
75f9d97a 4380
05f95b08
SB
4381 /* Sometimes a magic contains a reference loop, where the sv and
4382 object refer to each other. To prevent a reference loop that
4383 would prevent such objects being freed, we look for such loops
4384 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4385
4386 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4387 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4388
4389 */
14befaf4
DM
4390 if (!obj || obj == sv ||
4391 how == PERL_MAGIC_arylen ||
4392 how == PERL_MAGIC_qr ||
8d2f4536 4393 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4394 (SvTYPE(obj) == SVt_PVGV &&
4395 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4396 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4397 GvFORM(obj) == (CV*)sv)))
75f9d97a 4398 {
8990e307 4399 mg->mg_obj = obj;
75f9d97a 4400 }
85e6fe83 4401 else {
b37c2d43 4402 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4403 mg->mg_flags |= MGf_REFCOUNTED;
4404 }
b5ccf5f2
YST
4405
4406 /* Normal self-ties simply pass a null object, and instead of
4407 using mg_obj directly, use the SvTIED_obj macro to produce a
4408 new RV as needed. For glob "self-ties", we are tieing the PVIO
4409 with an RV obj pointing to the glob containing the PVIO. In
4410 this case, to avoid a reference loop, we need to weaken the
4411 reference.
4412 */
4413
4414 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4415 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4416 {
4417 sv_rvweaken(obj);
4418 }
4419
79072805 4420 mg->mg_type = how;
565764a8 4421 mg->mg_len = namlen;
9cbac4c7 4422 if (name) {
92110913 4423 if (namlen > 0)
1edc1566 4424 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4425 else if (namlen == HEf_SVKEY)
b37c2d43 4426 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4427 else
92110913 4428 mg->mg_ptr = (char *) name;
9cbac4c7 4429 }
53d44271 4430 mg->mg_virtual = (MGVTBL *) vtable;
68795e93 4431
92110913
NIS
4432 mg_magical(sv);
4433 if (SvGMAGICAL(sv))
4434 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4435 return mg;
4436}
4437
4438/*
4439=for apidoc sv_magic
1c846c1f 4440
92110913
NIS
4441Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4442then adds a new magic item of type C<how> to the head of the magic list.
4443
2d8d5d5a
SH
4444See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4445handling of the C<name> and C<namlen> arguments.
4446
4509d3fb
SB
4447You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4448to add more than one instance of the same 'how'.
4449
92110913
NIS
4450=cut
4451*/
4452
4453void
4454Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4455{
97aff369 4456 dVAR;
53d44271 4457 const MGVTBL *vtable;
92110913 4458 MAGIC* mg;
92110913 4459
f8c7b90f 4460#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4461 if (SvIsCOW(sv))
4462 sv_force_normal_flags(sv, 0);
4463#endif
92110913 4464 if (SvREADONLY(sv)) {
d8084ca5
DM
4465 if (
4466 /* its okay to attach magic to shared strings; the subsequent
4467 * upgrade to PVMG will unshare the string */
4468 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4469
4470 && IN_PERL_RUNTIME
92110913
NIS
4471 && how != PERL_MAGIC_regex_global
4472 && how != PERL_MAGIC_bm
4473 && how != PERL_MAGIC_fm
4474 && how != PERL_MAGIC_sv
e6469971 4475 && how != PERL_MAGIC_backref
92110913
NIS
4476 )
4477 {
4478 Perl_croak(aTHX_ PL_no_modify);
4479 }
4480 }
4481 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4482 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4483 /* sv_magic() refuses to add a magic of the same 'how' as an
4484 existing one
92110913 4485 */
2a509ed3 4486 if (how == PERL_MAGIC_taint) {
92110913 4487 mg->mg_len |= 1;
2a509ed3
NC
4488 /* Any scalar which already had taint magic on which someone
4489 (erroneously?) did SvIOK_on() or similar will now be
4490 incorrectly sporting public "OK" flags. */
4491 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4492 }
92110913
NIS
4493 return;
4494 }
4495 }
68795e93 4496
79072805 4497 switch (how) {
14befaf4 4498 case PERL_MAGIC_sv:
92110913 4499 vtable = &PL_vtbl_sv;
79072805 4500 break;
14befaf4 4501 case PERL_MAGIC_overload:
92110913 4502 vtable = &PL_vtbl_amagic;
a0d0e21e 4503 break;
14befaf4 4504 case PERL_MAGIC_overload_elem:
92110913 4505 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4506 break;
14befaf4 4507 case PERL_MAGIC_overload_table:
92110913 4508 vtable = &PL_vtbl_ovrld;
a0d0e21e 4509 break;
14befaf4 4510 case PERL_MAGIC_bm:
92110913 4511 vtable = &PL_vtbl_bm;
79072805 4512 break;
14befaf4 4513 case PERL_MAGIC_regdata:
92110913 4514 vtable = &PL_vtbl_regdata;
6cef1e77 4515 break;
14befaf4 4516 case PERL_MAGIC_regdatum:
92110913 4517 vtable = &PL_vtbl_regdatum;
6cef1e77 4518 break;
14befaf4 4519 case PERL_MAGIC_env:
92110913 4520 vtable = &PL_vtbl_env;
79072805 4521 break;
14befaf4 4522 case PERL_MAGIC_fm:
92110913 4523 vtable = &PL_vtbl_fm;
55497cff 4524 break;
14befaf4 4525 case PERL_MAGIC_envelem:
92110913 4526 vtable = &PL_vtbl_envelem;
79072805 4527 break;
14befaf4 4528 case PERL_MAGIC_regex_global:
92110913 4529 vtable = &PL_vtbl_mglob;
93a17b20 4530 break;
14befaf4 4531 case PERL_MAGIC_isa:
92110913 4532 vtable = &PL_vtbl_isa;
463ee0b2 4533 break;
14befaf4 4534 case PERL_MAGIC_isaelem:
92110913 4535 vtable = &PL_vtbl_isaelem;
463ee0b2 4536 break;
14befaf4 4537 case PERL_MAGIC_nkeys:
92110913 4538 vtable = &PL_vtbl_nkeys;
16660edb 4539 break;
14befaf4 4540 case PERL_MAGIC_dbfile:
aec46f14 4541 vtable = NULL;
93a17b20 4542 break;
14befaf4 4543 case PERL_MAGIC_dbline:
92110913 4544 vtable = &PL_vtbl_dbline;
79072805 4545 break;
36477c24 4546#ifdef USE_LOCALE_COLLATE
14befaf4 4547 case PERL_MAGIC_collxfrm:
92110913 4548 vtable = &PL_vtbl_collxfrm;
bbce6d69 4549 break;
36477c24 4550#endif /* USE_LOCALE_COLLATE */
14befaf4 4551 case PERL_MAGIC_tied:
92110913 4552 vtable = &PL_vtbl_pack;
463ee0b2 4553 break;
14befaf4
DM
4554 case PERL_MAGIC_tiedelem:
4555 case PERL_MAGIC_tiedscalar:
92110913 4556 vtable = &PL_vtbl_packelem;
463ee0b2 4557 break;
14befaf4 4558 case PERL_MAGIC_qr:
92110913 4559 vtable = &PL_vtbl_regexp;
c277df42 4560 break;
b3ca2e83
NC
4561 case PERL_MAGIC_hints:
4562 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4563 case PERL_MAGIC_sig:
92110913 4564 vtable = &PL_vtbl_sig;
79072805 4565 break;
14befaf4 4566 case PERL_MAGIC_sigelem:
92110913 4567 vtable = &PL_vtbl_sigelem;
79072805 4568 break;
14befaf4 4569 case PERL_MAGIC_taint:
92110913 4570 vtable = &PL_vtbl_taint;
463ee0b2 4571 break;
14befaf4 4572 case PERL_MAGIC_uvar:
92110913 4573 vtable = &PL_vtbl_uvar;
79072805 4574 break;
14befaf4 4575 case PERL_MAGIC_vec:
92110913 4576 vtable = &PL_vtbl_vec;
79072805 4577 break;
a3874608 4578 case PERL_MAGIC_arylen_p:
bfcb3514 4579 case PERL_MAGIC_rhash:
8d2f4536 4580 case PERL_MAGIC_symtab:
ece467f9 4581 case PERL_MAGIC_vstring:
aec46f14 4582 vtable = NULL;
ece467f9 4583 break;
7e8c5dac
HS
4584 case PERL_MAGIC_utf8:
4585 vtable = &PL_vtbl_utf8;
4586 break;
14befaf4 4587 case PERL_MAGIC_substr:
92110913 4588 vtable = &PL_vtbl_substr;
79072805 4589 break;
14befaf4 4590 case PERL_MAGIC_defelem:
92110913 4591 vtable = &PL_vtbl_defelem;
5f05dabc 4592 break;
14befaf4 4593 case PERL_MAGIC_arylen:
92110913 4594 vtable = &PL_vtbl_arylen;
79072805 4595 break;
14befaf4 4596 case PERL_MAGIC_pos:
92110913 4597 vtable = &PL_vtbl_pos;
a0d0e21e 4598 break;
14befaf4 4599 case PERL_MAGIC_backref:
92110913 4600 vtable = &PL_vtbl_backref;
810b8aa5 4601 break;
b3ca2e83
NC
4602 case PERL_MAGIC_hintselem:
4603 vtable = &PL_vtbl_hintselem;
4604 break;
14befaf4
DM
4605 case PERL_MAGIC_ext:
4606 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4607 /* Useful for attaching extension internal data to perl vars. */
4608 /* Note that multiple extensions may clash if magical scalars */
4609 /* etc holding private data from one are passed to another. */
aec46f14 4610 vtable = NULL;
a0d0e21e 4611 break;
79072805 4612 default:
14befaf4 4613 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4614 }
68795e93 4615
92110913 4616 /* Rest of work is done else where */
aec46f14 4617 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4618
92110913
NIS
4619 switch (how) {
4620 case PERL_MAGIC_taint:
4621 mg->mg_len = 1;
4622 break;
4623 case PERL_MAGIC_ext:
4624 case PERL_MAGIC_dbfile:
4625 SvRMAGICAL_on(sv);
4626 break;
4627 }
463ee0b2
LW
4628}
4629
c461cf8f
JH
4630/*
4631=for apidoc sv_unmagic
4632
645c22ef 4633Removes all magic of type C<type> from an SV.
c461cf8f
JH
4634
4635=cut
4636*/
4637
463ee0b2 4638int
864dbfa3 4639Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4640{
4641 MAGIC* mg;
4642 MAGIC** mgp;
91bba347 4643 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4644 return 0;
064cf529 4645 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4646 for (mg = *mgp; mg; mg = *mgp) {
4647 if (mg->mg_type == type) {
e1ec3a88 4648 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4649 *mgp = mg->mg_moremagic;
1d7c1841 4650 if (vtbl && vtbl->svt_free)
fc0dc3b3 4651 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4652 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4653 if (mg->mg_len > 0)
1edc1566 4654 Safefree(mg->mg_ptr);
565764a8 4655 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4656 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4657 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4658 Safefree(mg->mg_ptr);
9cbac4c7 4659 }
a0d0e21e
LW
4660 if (mg->mg_flags & MGf_REFCOUNTED)
4661 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4662 Safefree(mg);
4663 }
4664 else
4665 mgp = &mg->mg_moremagic;
79072805 4666 }
91bba347 4667 if (!SvMAGIC(sv)) {
463ee0b2 4668 SvMAGICAL_off(sv);
c268c2a6 4669 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4670 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4671 }
4672
4673 return 0;
79072805
LW
4674}
4675
c461cf8f
JH
4676/*
4677=for apidoc sv_rvweaken
4678
645c22ef
DM
4679Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4680referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4681push a back-reference to this RV onto the array of backreferences
1e73acc8
AS
4682associated with that magic. If the RV is magical, set magic will be
4683called after the RV is cleared.
c461cf8f
JH
4684
4685=cut
4686*/
4687
810b8aa5 4688SV *
864dbfa3 4689Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4690{
4691 SV *tsv;
4692 if (!SvOK(sv)) /* let undefs pass */
4693 return sv;
4694 if (!SvROK(sv))
cea2e8a9 4695 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4696 else if (SvWEAKREF(sv)) {
810b8aa5 4697 if (ckWARN(WARN_MISC))
9014280d 4698 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4699 return sv;
4700 }
4701 tsv = SvRV(sv);
e15faf7d 4702 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4703 SvWEAKREF_on(sv);
1c846c1f 4704 SvREFCNT_dec(tsv);
810b8aa5
GS
4705 return sv;
4706}
4707
645c22ef
DM
4708/* Give tsv backref magic if it hasn't already got it, then push a
4709 * back-reference to sv onto the array associated with the backref magic.
4710 */
4711
e15faf7d
NC
4712void
4713Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4714{
97aff369 4715 dVAR;
810b8aa5 4716 AV *av;
86f55936
NC
4717
4718 if (SvTYPE(tsv) == SVt_PVHV) {
4719 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4720
4721 av = *avp;
4722 if (!av) {
4723 /* There is no AV in the offical place - try a fixup. */
4724 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4725
4726 if (mg) {
4727 /* Aha. They've got it stowed in magic. Bring it back. */
4728 av = (AV*)mg->mg_obj;
4729 /* Stop mg_free decreasing the refernce count. */
4730 mg->mg_obj = NULL;
4731 /* Stop mg_free even calling the destructor, given that
4732 there's no AV to free up. */
4733 mg->mg_virtual = 0;
4734 sv_unmagic(tsv, PERL_MAGIC_backref);
4735 } else {
4736 av = newAV();
4737 AvREAL_off(av);
b37c2d43 4738 SvREFCNT_inc_simple_void(av);
86f55936
NC
4739 }
4740 *avp = av;
4741 }
4742 } else {
4743 const MAGIC *const mg
4744 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4745 if (mg)
4746 av = (AV*)mg->mg_obj;
4747 else {
4748 av = newAV();
4749 AvREAL_off(av);
4750 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4751 /* av now has a refcnt of 2, which avoids it getting freed
4752 * before us during global cleanup. The extra ref is removed
4753 * by magic_killbackrefs() when tsv is being freed */
4754 }
810b8aa5 4755 }
d91d49e8 4756 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4757 av_extend(av, AvFILLp(av)+1);
4758 }
4759 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4760}
4761
645c22ef
DM
4762/* delete a back-reference to ourselves from the backref magic associated
4763 * with the SV we point to.
4764 */
4765
1c846c1f 4766STATIC void
e15faf7d 4767S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4768{
97aff369 4769 dVAR;
86f55936 4770 AV *av = NULL;
810b8aa5
GS
4771 SV **svp;
4772 I32 i;
86f55936
NC
4773
4774 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4775 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4776 /* We mustn't attempt to "fix up" the hash here by moving the
4777 backreference array back to the hv_aux structure, as that is stored
4778 in the main HvARRAY(), and hfreentries assumes that no-one
4779 reallocates HvARRAY() while it is running. */
86f55936
NC
4780 }
4781 if (!av) {
4782 const MAGIC *const mg
4783 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4784 if (mg)
4785 av = (AV *)mg->mg_obj;
4786 }
4787 if (!av) {
e15faf7d
NC
4788 if (PL_in_clean_all)
4789 return;
cea2e8a9 4790 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4791 }
4792
4793 if (SvIS_FREED(av))
4794 return;
4795
810b8aa5 4796 svp = AvARRAY(av);
6a76db8b
NC
4797 /* We shouldn't be in here more than once, but for paranoia reasons lets
4798 not assume this. */
4799 for (i = AvFILLp(av); i >= 0; i--) {
4800 if (svp[i] == sv) {
4801 const SSize_t fill = AvFILLp(av);
4802 if (i != fill) {
4803 /* We weren't the last entry.
4804 An unordered list has this property that you can take the
4805 last element off the end to fill the hole, and it's still
4806 an unordered list :-)
4807 */
4808 svp[i] = svp[fill];
4809 }
a0714e2c 4810 svp[fill] = NULL;
6a76db8b
NC
4811 AvFILLp(av) = fill - 1;
4812 }
4813 }
810b8aa5
GS
4814}
4815
86f55936
NC
4816int
4817Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4818{
4819 SV **svp = AvARRAY(av);
4820
4821 PERL_UNUSED_ARG(sv);
4822
4823 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4824 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4825 if (svp && !SvIS_FREED(av)) {
4826 SV *const *const last = svp + AvFILLp(av);
4827
4828 while (svp <= last) {
4829 if (*svp) {
4830 SV *const referrer = *svp;
4831 if (SvWEAKREF(referrer)) {
4832 /* XXX Should we check that it hasn't changed? */
4833 SvRV_set(referrer, 0);
4834 SvOK_off(referrer);
4835 SvWEAKREF_off(referrer);
1e73acc8 4836 SvSETMAGIC(referrer);
86f55936
NC
4837 } else if (SvTYPE(referrer) == SVt_PVGV ||
4838 SvTYPE(referrer) == SVt_PVLV) {
4839 /* You lookin' at me? */
4840 assert(GvSTASH(referrer));
4841 assert(GvSTASH(referrer) == (HV*)sv);
4842 GvSTASH(referrer) = 0;
4843 } else {
4844 Perl_croak(aTHX_
4845 "panic: magic_killbackrefs (flags=%"UVxf")",
4846 (UV)SvFLAGS(referrer));
4847 }
4848
a0714e2c 4849 *svp = NULL;
86f55936
NC
4850 }
4851 svp++;
4852 }
4853 }
4854 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4855 return 0;
4856}
4857
954c1994
GS
4858/*
4859=for apidoc sv_insert
4860
4861Inserts a string at the specified offset/length within the SV. Similar to
4862the Perl substr() function.
4863
4864=cut
4865*/
4866
79072805 4867void
e1ec3a88 4868Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4869{
97aff369 4870 dVAR;
79072805
LW
4871 register char *big;
4872 register char *mid;
4873 register char *midend;
4874 register char *bigend;
4875 register I32 i;
6ff81951 4876 STRLEN curlen;
1c846c1f 4877
79072805 4878
8990e307 4879 if (!bigstr)
cea2e8a9 4880 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4881 SvPV_force(bigstr, curlen);
60fa28ff 4882 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4883 if (offset + len > curlen) {
4884 SvGROW(bigstr, offset+len+1);
93524f2b 4885 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4886 SvCUR_set(bigstr, offset+len);
4887 }
79072805 4888
69b47968 4889 SvTAINT(bigstr);
79072805
LW
4890 i = littlelen - len;
4891 if (i > 0) { /* string might grow */
a0d0e21e 4892 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4893 mid = big + offset + len;
4894 midend = bigend = big + SvCUR(bigstr);
4895 bigend += i;
4896 *bigend = '\0';
4897 while (midend > mid) /* shove everything down */
4898 *--bigend = *--midend;
4899 Move(little,big+offset,littlelen,char);
b162af07 4900 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4901 SvSETMAGIC(bigstr);
4902 return;
4903 }
4904 else if (i == 0) {
463ee0b2 4905 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4906 SvSETMAGIC(bigstr);
4907 return;
4908 }
4909
463ee0b2 4910 big = SvPVX(bigstr);
79072805
LW
4911 mid = big + offset;
4912 midend = mid + len;
4913 bigend = big + SvCUR(bigstr);
4914
4915 if (midend > bigend)
cea2e8a9 4916 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4917
4918 if (mid - big > bigend - midend) { /* faster to shorten from end */
4919 if (littlelen) {
4920 Move(little, mid, littlelen,char);
4921 mid += littlelen;
4922 }
4923 i = bigend - midend;
4924 if (i > 0) {
4925 Move(midend, mid, i,char);
4926 mid += i;
4927 }
4928 *mid = '\0';
4929 SvCUR_set(bigstr, mid - big);
4930 }
155aba94 4931 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4932 midend -= littlelen;
4933 mid = midend;
4934 sv_chop(bigstr,midend-i);
4935 big += i;
4936 while (i--)
4937 *--midend = *--big;
4938 if (littlelen)
4939 Move(little, mid, littlelen,char);
4940 }
4941 else if (littlelen) {
4942 midend -= littlelen;
4943 sv_chop(bigstr,midend);
4944 Move(little,midend,littlelen,char);
4945 }
4946 else {
4947 sv_chop(bigstr,midend);
4948 }
4949 SvSETMAGIC(bigstr);
4950}
4951
c461cf8f
JH
4952/*
4953=for apidoc sv_replace
4954
4955Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4956The target SV physically takes over ownership of the body of the source SV
4957and inherits its flags; however, the target keeps any magic it owns,
4958and any magic in the source is discarded.
ff276b08 4959Note that this is a rather specialist SV copying operation; most of the
645c22ef 4960time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4961
4962=cut
4963*/
79072805
LW
4964
4965void
864dbfa3 4966Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4967{
97aff369 4968 dVAR;
a3b680e6 4969 const U32 refcnt = SvREFCNT(sv);
765f542d 4970 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4971 if (SvREFCNT(nsv) != 1) {
7437becc 4972 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4973 UVuf " != 1)", (UV) SvREFCNT(nsv));
4974 }
93a17b20 4975 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4976 if (SvMAGICAL(nsv))
4977 mg_free(nsv);
4978 else
4979 sv_upgrade(nsv, SVt_PVMG);
b162af07 4980 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4981 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4982 SvMAGICAL_off(sv);
b162af07 4983 SvMAGIC_set(sv, NULL);
93a17b20 4984 }
79072805
LW
4985 SvREFCNT(sv) = 0;
4986 sv_clear(sv);
477f5d66 4987 assert(!SvREFCNT(sv));
fd0854ff
DM
4988#ifdef DEBUG_LEAKING_SCALARS
4989 sv->sv_flags = nsv->sv_flags;
4990 sv->sv_any = nsv->sv_any;
4991 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4992 sv->sv_u = nsv->sv_u;
fd0854ff 4993#else
79072805 4994 StructCopy(nsv,sv,SV);
fd0854ff 4995#endif
7b2c381c
NC
4996 /* Currently could join these into one piece of pointer arithmetic, but
4997 it would be unclear. */
4998 if(SvTYPE(sv) == SVt_IV)
4999 SvANY(sv)
339049b0 5000 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 5001 else if (SvTYPE(sv) == SVt_RV) {
339049b0 5002 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
5003 }
5004
fd0854ff 5005
f8c7b90f 5006#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
5007 if (SvIsCOW_normal(nsv)) {
5008 /* We need to follow the pointers around the loop to make the
5009 previous SV point to sv, rather than nsv. */
5010 SV *next;
5011 SV *current = nsv;
5012 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5013 assert(next);
5014 current = next;
3f7c398e 5015 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
5016 }
5017 /* Make the SV before us point to the SV after us. */
5018 if (DEBUG_C_TEST) {
5019 PerlIO_printf(Perl_debug_log, "previous is\n");
5020 sv_dump(current);
a29f6d03
NC
5021 PerlIO_printf(Perl_debug_log,
5022 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
5023 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5024 }
a29f6d03 5025 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
5026 }
5027#endif
79072805 5028 SvREFCNT(sv) = refcnt;
1edc1566 5029 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 5030 SvREFCNT(nsv) = 0;
463ee0b2 5031 del_SV(nsv);
79072805
LW
5032}
5033
c461cf8f
JH
5034/*
5035=for apidoc sv_clear
5036
645c22ef
DM
5037Clear an SV: call any destructors, free up any memory used by the body,
5038and free the body itself. The SV's head is I<not> freed, although
5039its type is set to all 1's so that it won't inadvertently be assumed
5040to be live during global destruction etc.
5041This function should only be called when REFCNT is zero. Most of the time
5042you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5043instead.
c461cf8f
JH
5044
5045=cut
5046*/
5047
79072805 5048void
864dbfa3 5049Perl_sv_clear(pTHX_ register SV *sv)
79072805 5050{
27da23d5 5051 dVAR;
82bb6deb 5052 const U32 type = SvTYPE(sv);
8edfc514
NC
5053 const struct body_details *const sv_type_details
5054 = bodies_by_type + type;
82bb6deb 5055
79072805
LW
5056 assert(sv);
5057 assert(SvREFCNT(sv) == 0);
5058
d2a0f284
JC
5059 if (type <= SVt_IV) {
5060 /* See the comment in sv.h about the collusion between this early
5061 return and the overloading of the NULL and IV slots in the size
5062 table. */
82bb6deb 5063 return;
d2a0f284 5064 }
82bb6deb 5065
ed6116ce 5066 if (SvOBJECT(sv)) {
3280af22 5067 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5068 dSP;
893645bd 5069 HV* stash;
d460ef45 5070 do {
b464bac0 5071 CV* destructor;
4e8e7886 5072 stash = SvSTASH(sv);
32251b26 5073 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5074 if (destructor) {
1b6737cc 5075 SV* const tmpref = newRV(sv);
5cc433a6 5076 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5077 ENTER;
e788e7d3 5078 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5079 EXTEND(SP, 2);
5080 PUSHMARK(SP);
5cc433a6 5081 PUSHs(tmpref);
4e8e7886 5082 PUTBACK;
44389ee9 5083 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5084
5085
d3acc0f7 5086 POPSTACK;
3095d977 5087 SPAGAIN;
4e8e7886 5088 LEAVE;
5cc433a6
AB
5089 if(SvREFCNT(tmpref) < 2) {
5090 /* tmpref is not kept alive! */
5091 SvREFCNT(sv)--;
b162af07 5092 SvRV_set(tmpref, NULL);
5cc433a6
AB
5093 SvROK_off(tmpref);
5094 }
5095 SvREFCNT_dec(tmpref);
4e8e7886
GS
5096 }
5097 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5098
6f44e0a4
JP
5099
5100 if (SvREFCNT(sv)) {
5101 if (PL_in_clean_objs)
cea2e8a9 5102 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5103 HvNAME_get(stash));
6f44e0a4
JP
5104 /* DESTROY gave object new lease on life */
5105 return;
5106 }
a0d0e21e 5107 }
4e8e7886 5108
a0d0e21e 5109 if (SvOBJECT(sv)) {
4e8e7886 5110 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5111 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5112 if (type != SVt_PVIO)
3280af22 5113 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5114 }
463ee0b2 5115 }
82bb6deb 5116 if (type >= SVt_PVMG) {
cecf5685 5117 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
73d95100 5118 SvREFCNT_dec(SvOURSTASH(sv));
e736a858 5119 } else if (SvMAGIC(sv))
524189f1 5120 mg_free(sv);
00b1698f 5121 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5122 SvREFCNT_dec(SvSTASH(sv));
5123 }
82bb6deb 5124 switch (type) {
cecf5685 5125 /* case SVt_BIND: */
8990e307 5126 case SVt_PVIO:
df0bd2f4
GS
5127 if (IoIFP(sv) &&
5128 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5129 IoIFP(sv) != PerlIO_stdout() &&
5130 IoIFP(sv) != PerlIO_stderr())
93578b34 5131 {
f2b5be74 5132 io_close((IO*)sv, FALSE);
93578b34 5133 }
1d7c1841 5134 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5135 PerlDir_close(IoDIRP(sv));
1d7c1841 5136 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5137 Safefree(IoTOP_NAME(sv));
5138 Safefree(IoFMT_NAME(sv));
5139 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5140 goto freescalar;
79072805 5141 case SVt_PVCV:
748a9306 5142 case SVt_PVFM:
85e6fe83 5143 cv_undef((CV*)sv);
a0d0e21e 5144 goto freescalar;
79072805 5145 case SVt_PVHV:
86f55936 5146 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5147 hv_undef((HV*)sv);
a0d0e21e 5148 break;
79072805 5149 case SVt_PVAV:
85e6fe83 5150 av_undef((AV*)sv);
a0d0e21e 5151 break;
02270b4e 5152 case SVt_PVLV:
dd28f7bb
DM
5153 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5154 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5155 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5156 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5157 }
5158 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5159 SvREFCNT_dec(LvTARG(sv));
a0d0e21e 5160 case SVt_PVGV:
cecf5685
NC
5161 if (isGV_with_GP(sv)) {
5162 gp_free((GV*)sv);
5163 if (GvNAME_HEK(sv))
5164 unshare_hek(GvNAME_HEK(sv));
893645bd
NC
5165 /* If we're in a stash, we don't own a reference to it. However it does
5166 have a back reference to us, which needs to be cleared. */
cecf5685
NC
5167 if (!SvVALID(sv) && GvSTASH(sv))
5168 sv_del_backref((SV*)GvSTASH(sv), sv);
5169 }
79072805 5170 case SVt_PVMG:
79072805
LW
5171 case SVt_PVNV:
5172 case SVt_PVIV:
a0d0e21e 5173 freescalar:
5228ca4e
NC
5174 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5175 if (SvOOK(sv)) {
93524f2b 5176 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5177 /* Don't even bother with turning off the OOK flag. */
5178 }
79072805 5179 case SVt_PV:
a0d0e21e 5180 case SVt_RV:
810b8aa5 5181 if (SvROK(sv)) {
b37c2d43 5182 SV * const target = SvRV(sv);
810b8aa5 5183 if (SvWEAKREF(sv))
e15faf7d 5184 sv_del_backref(target, sv);
810b8aa5 5185 else
e15faf7d 5186 SvREFCNT_dec(target);
810b8aa5 5187 }
f8c7b90f 5188#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5189 else if (SvPVX_const(sv)) {
765f542d
NC
5190 if (SvIsCOW(sv)) {
5191 /* I believe I need to grab the global SV mutex here and
5192 then recheck the COW status. */
46187eeb
NC
5193 if (DEBUG_C_TEST) {
5194 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5195 sv_dump(sv);
46187eeb 5196 }
5302ffd4
NC
5197 if (SvLEN(sv)) {
5198 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5199 } else {
5200 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5201 }
5202
765f542d
NC
5203 /* And drop it here. */
5204 SvFAKE_off(sv);
5205 } else if (SvLEN(sv)) {
3f7c398e 5206 Safefree(SvPVX_const(sv));
765f542d
NC
5207 }
5208 }
5209#else
3f7c398e 5210 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5211 Safefree(SvPVX_mutable(sv));
3f7c398e 5212 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5213 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5214 SvFAKE_off(sv);
5215 }
765f542d 5216#endif
79072805
LW
5217 break;
5218 case SVt_NV:
79072805
LW
5219 break;
5220 }
5221
893645bd
NC
5222 SvFLAGS(sv) &= SVf_BREAK;
5223 SvFLAGS(sv) |= SVTYPEMASK;
5224
8edfc514 5225 if (sv_type_details->arena) {
b9502f15 5226 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5227 &PL_body_roots[type]);
5228 }
d2a0f284 5229 else if (sv_type_details->body_size) {
8edfc514
NC
5230 my_safefree(SvANY(sv));
5231 }
79072805
LW
5232}
5233
645c22ef
DM
5234/*
5235=for apidoc sv_newref
5236
5237Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5238instead.
5239
5240=cut
5241*/
5242
79072805 5243SV *
864dbfa3 5244Perl_sv_newref(pTHX_ SV *sv)
79072805 5245{
96a5add6 5246 PERL_UNUSED_CONTEXT;
463ee0b2 5247 if (sv)
4db098f4 5248 (SvREFCNT(sv))++;
79072805
LW
5249 return sv;
5250}
5251
c461cf8f
JH
5252/*
5253=for apidoc sv_free
5254
645c22ef
DM
5255Decrement an SV's reference count, and if it drops to zero, call
5256C<sv_clear> to invoke destructors and free up any memory used by
5257the body; finally, deallocate the SV's head itself.
5258Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5259
5260=cut
5261*/
5262
79072805 5263void
864dbfa3 5264Perl_sv_free(pTHX_ SV *sv)
79072805 5265{
27da23d5 5266 dVAR;
79072805
LW
5267 if (!sv)
5268 return;
a0d0e21e
LW
5269 if (SvREFCNT(sv) == 0) {
5270 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5271 /* this SV's refcnt has been artificially decremented to
5272 * trigger cleanup */
a0d0e21e 5273 return;
3280af22 5274 if (PL_in_clean_all) /* All is fair */
1edc1566 5275 return;
d689ffdd
JP
5276 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5277 /* make sure SvREFCNT(sv)==0 happens very seldom */
5278 SvREFCNT(sv) = (~(U32)0)/2;
5279 return;
5280 }
41e4abd8 5281 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5282 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5283 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5284 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5285#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5286 Perl_dump_sv_child(aTHX_ sv);
5287#endif
5288 }
79072805
LW
5289 return;
5290 }
4db098f4 5291 if (--(SvREFCNT(sv)) > 0)
8990e307 5292 return;
8c4d3c90
NC
5293 Perl_sv_free2(aTHX_ sv);
5294}
5295
5296void
5297Perl_sv_free2(pTHX_ SV *sv)
5298{
27da23d5 5299 dVAR;
463ee0b2
LW
5300#ifdef DEBUGGING
5301 if (SvTEMP(sv)) {
0453d815 5302 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5303 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5304 "Attempt to free temp prematurely: SV 0x%"UVxf
5305 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5306 return;
79072805 5307 }
463ee0b2 5308#endif
d689ffdd
JP
5309 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5310 /* make sure SvREFCNT(sv)==0 happens very seldom */
5311 SvREFCNT(sv) = (~(U32)0)/2;
5312 return;
5313 }
79072805 5314 sv_clear(sv);
477f5d66
CS
5315 if (! SvREFCNT(sv))
5316 del_SV(sv);
79072805
LW
5317}
5318
954c1994
GS
5319/*
5320=for apidoc sv_len
5321
645c22ef
DM
5322Returns the length of the string in the SV. Handles magic and type
5323coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5324
5325=cut
5326*/
5327
79072805 5328STRLEN
864dbfa3 5329Perl_sv_len(pTHX_ register SV *sv)
79072805 5330{
463ee0b2 5331 STRLEN len;
79072805
LW
5332
5333 if (!sv)
5334 return 0;
5335
8990e307 5336 if (SvGMAGICAL(sv))
565764a8 5337 len = mg_length(sv);
8990e307 5338 else
4d84ee25 5339 (void)SvPV_const(sv, len);
463ee0b2 5340 return len;
79072805
LW
5341}
5342
c461cf8f
JH
5343/*
5344=for apidoc sv_len_utf8
5345
5346Returns the number of characters in the string in an SV, counting wide
1e54db1a 5347UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5348
5349=cut
5350*/
5351
7e8c5dac
HS
5352/*
5353 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5354 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5355 * (Note that the mg_len is not the length of the mg_ptr field.
5356 * This allows the cache to store the character length of the string without
5357 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5358 *
7e8c5dac
HS
5359 */
5360
a0ed51b3 5361STRLEN
864dbfa3 5362Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5363{
a0ed51b3
LW
5364 if (!sv)
5365 return 0;
5366
a0ed51b3 5367 if (SvGMAGICAL(sv))
b76347f2 5368 return mg_length(sv);
a0ed51b3 5369 else
b76347f2 5370 {
26346457 5371 STRLEN len;
e62f0680 5372 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5373
26346457
NC
5374 if (PL_utf8cache) {
5375 STRLEN ulen;
5376 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5377
5378 if (mg && mg->mg_len != -1) {
5379 ulen = mg->mg_len;
5380 if (PL_utf8cache < 0) {
5381 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5382 if (real != ulen) {
5383 /* Need to turn the assertions off otherwise we may
5384 recurse infinitely while printing error messages.
5385 */
5386 SAVEI8(PL_utf8cache);
5387 PL_utf8cache = 0;
f5992bc4
RB
5388 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
5389 " real %"UVuf" for %"SVf,
be2597df 5390 (UV) ulen, (UV) real, SVfARG(sv));
26346457
NC
5391 }
5392 }
5393 }
5394 else {
5395 ulen = Perl_utf8_length(aTHX_ s, s + len);
5396 if (!SvREADONLY(sv)) {
5397 if (!mg) {
5398 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5399 &PL_vtbl_utf8, 0, 0);
5400 }
cb9e20bb 5401 assert(mg);
26346457 5402 mg->mg_len = ulen;
cb9e20bb 5403 }
cb9e20bb 5404 }
26346457 5405 return ulen;
7e8c5dac 5406 }
26346457 5407 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5408 }
5409}
5410
9564a3bd
NC
5411/* Walk forwards to find the byte corresponding to the passed in UTF-8
5412 offset. */
bdf30dd6 5413static STRLEN
721e86b6 5414S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
bdf30dd6
NC
5415 STRLEN uoffset)
5416{
5417 const U8 *s = start;
5418
5419 while (s < send && uoffset--)
5420 s += UTF8SKIP(s);
5421 if (s > send) {
5422 /* This is the existing behaviour. Possibly it should be a croak, as
5423 it's actually a bounds error */
5424 s = send;
5425 }
5426 return s - start;
5427}
5428
9564a3bd
NC
5429/* Given the length of the string in both bytes and UTF-8 characters, decide
5430 whether to walk forwards or backwards to find the byte corresponding to
5431 the passed in UTF-8 offset. */
c336ad0b 5432static STRLEN
721e86b6 5433S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
c336ad0b
NC
5434 STRLEN uoffset, STRLEN uend)
5435{
5436 STRLEN backw = uend - uoffset;
5437 if (uoffset < 2 * backw) {
25a8a4ef 5438 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5439 forward (that's where the 2 * backw comes from).
5440 (The real figure of course depends on the UTF-8 data.) */
721e86b6 5441 return sv_pos_u2b_forwards(start, send, uoffset);
c336ad0b
NC
5442 }
5443
5444 while (backw--) {
5445 send--;
5446 while (UTF8_IS_CONTINUATION(*send))
5447 send--;
5448 }
5449 return send - start;
5450}
5451
9564a3bd
NC
5452/* For the string representation of the given scalar, find the byte
5453 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5454 give another position in the string, *before* the sought offset, which
5455 (which is always true, as 0, 0 is a valid pair of positions), which should
5456 help reduce the amount of linear searching.
5457 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5458 will be used to reduce the amount of linear searching. The cache will be
5459 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5460static STRLEN
5461S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5462 const U8 *const send, STRLEN uoffset,
5463 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5464 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5465 bool found = FALSE;
5466
75c33c12
NC
5467 assert (uoffset >= uoffset0);
5468
c336ad0b 5469 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5470 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5471 if ((*mgp)->mg_ptr) {
5472 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5473 if (cache[0] == uoffset) {
5474 /* An exact match. */
5475 return cache[1];
5476 }
ab455f60
NC
5477 if (cache[2] == uoffset) {
5478 /* An exact match. */
5479 return cache[3];
5480 }
668af93f
NC
5481
5482 if (cache[0] < uoffset) {
d8b2e1f9
NC
5483 /* The cache already knows part of the way. */
5484 if (cache[0] > uoffset0) {
5485 /* The cache knows more than the passed in pair */
5486 uoffset0 = cache[0];
5487 boffset0 = cache[1];
5488 }
5489 if ((*mgp)->mg_len != -1) {
5490 /* And we know the end too. */
5491 boffset = boffset0
721e86b6 5492 + sv_pos_u2b_midway(start + boffset0, send,
d8b2e1f9
NC
5493 uoffset - uoffset0,
5494 (*mgp)->mg_len - uoffset0);
5495 } else {
5496 boffset = boffset0
721e86b6 5497 + sv_pos_u2b_forwards(start + boffset0,
d8b2e1f9
NC
5498 send, uoffset - uoffset0);
5499 }
dd7c5fd3
NC
5500 }
5501 else if (cache[2] < uoffset) {
5502 /* We're between the two cache entries. */
5503 if (cache[2] > uoffset0) {
5504 /* and the cache knows more than the passed in pair */
5505 uoffset0 = cache[2];
5506 boffset0 = cache[3];
5507 }
5508
668af93f 5509 boffset = boffset0
721e86b6 5510 + sv_pos_u2b_midway(start + boffset0,
668af93f
NC
5511 start + cache[1],
5512 uoffset - uoffset0,
5513 cache[0] - uoffset0);
dd7c5fd3
NC
5514 } else {
5515 boffset = boffset0
721e86b6 5516 + sv_pos_u2b_midway(start + boffset0,
dd7c5fd3
NC
5517 start + cache[3],
5518 uoffset - uoffset0,
5519 cache[2] - uoffset0);
d8b2e1f9 5520 }
668af93f 5521 found = TRUE;
d8b2e1f9
NC
5522 }
5523 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5524 /* If we can take advantage of a passed in offset, do so. */
5525 /* In fact, offset0 is either 0, or less than offset, so don't
5526 need to worry about the other possibility. */
5527 boffset = boffset0
721e86b6 5528 + sv_pos_u2b_midway(start + boffset0, send,
75c33c12
NC
5529 uoffset - uoffset0,
5530 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5531 found = TRUE;
5532 }
28ccbf94 5533 }
c336ad0b
NC
5534
5535 if (!found || PL_utf8cache < 0) {
75c33c12 5536 const STRLEN real_boffset
721e86b6 5537 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
75c33c12
NC
5538 send, uoffset - uoffset0);
5539
c336ad0b
NC
5540 if (found && PL_utf8cache < 0) {
5541 if (real_boffset != boffset) {
5542 /* Need to turn the assertions off otherwise we may recurse
5543 infinitely while printing error messages. */
5544 SAVEI8(PL_utf8cache);
5545 PL_utf8cache = 0;
f5992bc4
RB
5546 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
5547 " real %"UVuf" for %"SVf,
be2597df 5548 (UV) boffset, (UV) real_boffset, SVfARG(sv));
c336ad0b
NC
5549 }
5550 }
5551 boffset = real_boffset;
28ccbf94 5552 }
0905937d 5553
ab455f60 5554 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5555 return boffset;
5556}
5557
9564a3bd
NC
5558
5559/*
5560=for apidoc sv_pos_u2b
5561
5562Converts the value pointed to by offsetp from a count of UTF-8 chars from
5563the start of the string, to a count of the equivalent number of bytes; if
5564lenp is non-zero, it does the same to lenp, but this time starting from
5565the offset, rather than from the start of the string. Handles magic and
5566type coercion.
5567
5568=cut
5569*/
5570
5571/*
5572 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5573 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5574 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5575 *
5576 */
5577
a0ed51b3 5578void
864dbfa3 5579Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5580{
245d4a47 5581 const U8 *start;
a0ed51b3
LW
5582 STRLEN len;
5583
5584 if (!sv)
5585 return;
5586
245d4a47 5587 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5588 if (len) {
bdf30dd6
NC
5589 STRLEN uoffset = (STRLEN) *offsetp;
5590 const U8 * const send = start + len;
0905937d 5591 MAGIC *mg = NULL;
721e86b6 5592 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
28ccbf94 5593 uoffset, 0, 0);
bdf30dd6
NC
5594
5595 *offsetp = (I32) boffset;
5596
5597 if (lenp) {
28ccbf94 5598 /* Convert the relative offset to absolute. */
721e86b6
AL
5599 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5600 const STRLEN boffset2
5601 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
28ccbf94 5602 uoffset, boffset) - boffset;
bdf30dd6 5603
28ccbf94 5604 *lenp = boffset2;
bdf30dd6 5605 }
7e8c5dac
HS
5606 }
5607 else {
5608 *offsetp = 0;
5609 if (lenp)
5610 *lenp = 0;
a0ed51b3 5611 }
e23c8137 5612
a0ed51b3
LW
5613 return;
5614}
5615
9564a3bd
NC
5616/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5617 byte length pairing. The (byte) length of the total SV is passed in too,
5618 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5619 may not have updated SvCUR, so we can't rely on reading it directly.
5620
5621 The proffered utf8/byte length pairing isn't used if the cache already has
5622 two pairs, and swapping either for the proffered pair would increase the
5623 RMS of the intervals between known byte offsets.
5624
5625 The cache itself consists of 4 STRLEN values
5626 0: larger UTF-8 offset
5627 1: corresponding byte offset
5628 2: smaller UTF-8 offset
5629 3: corresponding byte offset
5630
5631 Unused cache pairs have the value 0, 0.
5632 Keeping the cache "backwards" means that the invariant of
5633 cache[0] >= cache[2] is maintained even with empty slots, which means that
5634 the code that uses it doesn't need to worry if only 1 entry has actually
5635 been set to non-zero. It also makes the "position beyond the end of the
5636 cache" logic much simpler, as the first slot is always the one to start
5637 from.
645c22ef 5638*/
ec07b5e0 5639static void
ab455f60
NC
5640S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5641 STRLEN blen)
ec07b5e0
NC
5642{
5643 STRLEN *cache;
5644 if (SvREADONLY(sv))
5645 return;
5646
5647 if (!*mgp) {
5648 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5649 0);
5650 (*mgp)->mg_len = -1;
5651 }
5652 assert(*mgp);
5653
5654 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5655 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5656 (*mgp)->mg_ptr = (char *) cache;
5657 }
5658 assert(cache);
5659
5660 if (PL_utf8cache < 0) {
ef816a78 5661 const U8 *start = (const U8 *) SvPVX_const(sv);
6448472a 5662 const STRLEN realutf8 = utf8_length(start, start + byte);
ec07b5e0
NC
5663
5664 if (realutf8 != utf8) {
5665 /* Need to turn the assertions off otherwise we may recurse
5666 infinitely while printing error messages. */
5667 SAVEI8(PL_utf8cache);
5668 PL_utf8cache = 0;
f5992bc4 5669 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
be2597df 5670 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
ec07b5e0
NC
5671 }
5672 }
ab455f60
NC
5673
5674 /* Cache is held with the later position first, to simplify the code
5675 that deals with unbounded ends. */
5676
5677 ASSERT_UTF8_CACHE(cache);
5678 if (cache[1] == 0) {
5679 /* Cache is totally empty */
5680 cache[0] = utf8;
5681 cache[1] = byte;
5682 } else if (cache[3] == 0) {
5683 if (byte > cache[1]) {
5684 /* New one is larger, so goes first. */
5685 cache[2] = cache[0];
5686 cache[3] = cache[1];
5687 cache[0] = utf8;
5688 cache[1] = byte;
5689 } else {
5690 cache[2] = utf8;
5691 cache[3] = byte;
5692 }
5693 } else {
5694#define THREEWAY_SQUARE(a,b,c,d) \
5695 ((float)((d) - (c))) * ((float)((d) - (c))) \
5696 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5697 + ((float)((b) - (a))) * ((float)((b) - (a)))
5698
5699 /* Cache has 2 slots in use, and we know three potential pairs.
5700 Keep the two that give the lowest RMS distance. Do the
5701 calcualation in bytes simply because we always know the byte
5702 length. squareroot has the same ordering as the positive value,
5703 so don't bother with the actual square root. */
5704 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5705 if (byte > cache[1]) {
5706 /* New position is after the existing pair of pairs. */
5707 const float keep_earlier
5708 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5709 const float keep_later
5710 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5711
5712 if (keep_later < keep_earlier) {
5713 if (keep_later < existing) {
5714 cache[2] = cache[0];
5715 cache[3] = cache[1];
5716 cache[0] = utf8;
5717 cache[1] = byte;
5718 }
5719 }
5720 else {
5721 if (keep_earlier < existing) {
5722 cache[0] = utf8;
5723 cache[1] = byte;
5724 }
5725 }
5726 }
57d7fbf1
NC
5727 else if (byte > cache[3]) {
5728 /* New position is between the existing pair of pairs. */
5729 const float keep_earlier
5730 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5731 const float keep_later
5732 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5733
5734 if (keep_later < keep_earlier) {
5735 if (keep_later < existing) {
5736 cache[2] = utf8;
5737 cache[3] = byte;
5738 }
5739 }
5740 else {
5741 if (keep_earlier < existing) {
5742 cache[0] = utf8;
5743 cache[1] = byte;
5744 }
5745 }
5746 }
5747 else {
5748 /* New position is before the existing pair of pairs. */
5749 const float keep_earlier
5750 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5751 const float keep_later
5752 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5753
5754 if (keep_later < keep_earlier) {
5755 if (keep_later < existing) {
5756 cache[2] = utf8;
5757 cache[3] = byte;
5758 }
5759 }
5760 else {
5761 if (keep_earlier < existing) {
5762 cache[0] = cache[2];
5763 cache[1] = cache[3];
5764 cache[2] = utf8;
5765 cache[3] = byte;
5766 }
5767 }
5768 }
ab455f60 5769 }
0905937d 5770 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5771}
5772
ec07b5e0 5773/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5774 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5775 backward is half the speed of walking forward. */
ec07b5e0
NC
5776static STRLEN
5777S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5778 STRLEN endu)
5779{
5780 const STRLEN forw = target - s;
5781 STRLEN backw = end - target;
5782
5783 if (forw < 2 * backw) {
6448472a 5784 return utf8_length(s, target);
ec07b5e0
NC
5785 }
5786
5787 while (end > target) {
5788 end--;
5789 while (UTF8_IS_CONTINUATION(*end)) {
5790 end--;
5791 }
5792 endu--;
5793 }
5794 return endu;
5795}
5796
9564a3bd
NC
5797/*
5798=for apidoc sv_pos_b2u
5799
5800Converts the value pointed to by offsetp from a count of bytes from the
5801start of the string, to a count of the equivalent number of UTF-8 chars.
5802Handles magic and type coercion.
5803
5804=cut
5805*/
5806
5807/*
5808 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5809 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5810 * byte offsets.
5811 *
5812 */
a0ed51b3 5813void
7e8c5dac 5814Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5815{
83003860 5816 const U8* s;
ec07b5e0 5817 const STRLEN byte = *offsetp;
7087a21c 5818 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5819 STRLEN blen;
ec07b5e0
NC
5820 MAGIC* mg = NULL;
5821 const U8* send;
a922f900 5822 bool found = FALSE;
a0ed51b3
LW
5823
5824 if (!sv)
5825 return;
5826
ab455f60 5827 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5828
ab455f60 5829 if (blen < byte)
ec07b5e0 5830 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5831
ec07b5e0 5832 send = s + byte;
a67d7df9 5833
ffca234a
NC
5834 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5835 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5836 if (mg->mg_ptr) {
d4c19fe8 5837 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5838 if (cache[1] == byte) {
ec07b5e0
NC
5839 /* An exact match. */
5840 *offsetp = cache[0];
ec07b5e0 5841 return;
7e8c5dac 5842 }
ab455f60
NC
5843 if (cache[3] == byte) {
5844 /* An exact match. */
5845 *offsetp = cache[2];
5846 return;
5847 }
668af93f
NC
5848
5849 if (cache[1] < byte) {
ec07b5e0 5850 /* We already know part of the way. */
b9f984a5
NC
5851 if (mg->mg_len != -1) {
5852 /* Actually, we know the end too. */
5853 len = cache[0]
5854 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5855 s + blen, mg->mg_len - cache[0]);
b9f984a5 5856 } else {
6448472a 5857 len = cache[0] + utf8_length(s + cache[1], send);
b9f984a5 5858 }
7e8c5dac 5859 }
9f985e4c
NC
5860 else if (cache[3] < byte) {
5861 /* We're between the two cached pairs, so we do the calculation
5862 offset by the byte/utf-8 positions for the earlier pair,
5863 then add the utf-8 characters from the string start to
5864 there. */
5865 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5866 s + cache[1], cache[0] - cache[2])
5867 + cache[2];
5868
5869 }
5870 else { /* cache[3] > byte */
5871 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5872 cache[2]);
7e8c5dac 5873
7e8c5dac 5874 }
ec07b5e0 5875 ASSERT_UTF8_CACHE(cache);
a922f900 5876 found = TRUE;
ffca234a 5877 } else if (mg->mg_len != -1) {
ab455f60 5878 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5879 found = TRUE;
7e8c5dac 5880 }
a0ed51b3 5881 }
a922f900 5882 if (!found || PL_utf8cache < 0) {
6448472a 5883 const STRLEN real_len = utf8_length(s, send);
a922f900
NC
5884
5885 if (found && PL_utf8cache < 0) {
5886 if (len != real_len) {
5887 /* Need to turn the assertions off otherwise we may recurse
5888 infinitely while printing error messages. */
5889 SAVEI8(PL_utf8cache);
5890 PL_utf8cache = 0;
f5992bc4
RB
5891 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
5892 " real %"UVuf" for %"SVf,
be2597df 5893 (UV) len, (UV) real_len, SVfARG(sv));
a922f900
NC
5894 }
5895 }
5896 len = real_len;
ec07b5e0
NC
5897 }
5898 *offsetp = len;
5899
ab455f60 5900 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5901}
5902
954c1994
GS
5903/*
5904=for apidoc sv_eq
5905
5906Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5907identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5908coerce its args to strings if necessary.
954c1994
GS
5909
5910=cut
5911*/
5912
79072805 5913I32
e01b9e88 5914Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5915{
97aff369 5916 dVAR;
e1ec3a88 5917 const char *pv1;
463ee0b2 5918 STRLEN cur1;
e1ec3a88 5919 const char *pv2;
463ee0b2 5920 STRLEN cur2;
e01b9e88 5921 I32 eq = 0;
bd61b366 5922 char *tpv = NULL;
a0714e2c 5923 SV* svrecode = NULL;
79072805 5924
e01b9e88 5925 if (!sv1) {
79072805
LW
5926 pv1 = "";
5927 cur1 = 0;
5928 }
ced497e2
YST
5929 else {
5930 /* if pv1 and pv2 are the same, second SvPV_const call may
5931 * invalidate pv1, so we may need to make a copy */
5932 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
5933 pv1 = SvPV_const(sv1, cur1);
5934 sv1 = sv_2mortal(newSVpvn(pv1, cur1));
5935 if (SvUTF8(sv2)) SvUTF8_on(sv1);
5936 }
4d84ee25 5937 pv1 = SvPV_const(sv1, cur1);
ced497e2 5938 }
79072805 5939
e01b9e88
SC
5940 if (!sv2){
5941 pv2 = "";
5942 cur2 = 0;
92d29cee 5943 }
e01b9e88 5944 else
4d84ee25 5945 pv2 = SvPV_const(sv2, cur2);
79072805 5946
cf48d248 5947 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5948 /* Differing utf8ness.
5949 * Do not UTF8size the comparands as a side-effect. */
5950 if (PL_encoding) {
5951 if (SvUTF8(sv1)) {
553e1bcc
AT
5952 svrecode = newSVpvn(pv2, cur2);
5953 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5954 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5955 }
5956 else {
553e1bcc
AT
5957 svrecode = newSVpvn(pv1, cur1);
5958 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5959 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5960 }
5961 /* Now both are in UTF-8. */
0a1bd7ac
DM
5962 if (cur1 != cur2) {
5963 SvREFCNT_dec(svrecode);
799ef3cb 5964 return FALSE;
0a1bd7ac 5965 }
799ef3cb
JH
5966 }
5967 else {
5968 bool is_utf8 = TRUE;
5969
5970 if (SvUTF8(sv1)) {
5971 /* sv1 is the UTF-8 one,
5972 * if is equal it must be downgrade-able */
9d4ba2ae 5973 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5974 &cur1, &is_utf8);
5975 if (pv != pv1)
553e1bcc 5976 pv1 = tpv = pv;
799ef3cb
JH
5977 }
5978 else {
5979 /* sv2 is the UTF-8 one,
5980 * if is equal it must be downgrade-able */
9d4ba2ae 5981 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5982 &cur2, &is_utf8);
5983 if (pv != pv2)
553e1bcc 5984 pv2 = tpv = pv;
799ef3cb
JH
5985 }
5986 if (is_utf8) {
5987 /* Downgrade not possible - cannot be eq */
bf694877 5988 assert (tpv == 0);
799ef3cb
JH
5989 return FALSE;
5990 }
5991 }
cf48d248
JH
5992 }
5993
5994 if (cur1 == cur2)
765f542d 5995 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5996
b37c2d43 5997 SvREFCNT_dec(svrecode);
553e1bcc
AT
5998 if (tpv)
5999 Safefree(tpv);
cf48d248 6000
e01b9e88 6001 return eq;
79072805
LW
6002}
6003
954c1994
GS
6004/*
6005=for apidoc sv_cmp
6006
6007Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6008string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
6009C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6010coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
6011
6012=cut
6013*/
6014
79072805 6015I32
e01b9e88 6016Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 6017{
97aff369 6018 dVAR;
560a288e 6019 STRLEN cur1, cur2;
e1ec3a88 6020 const char *pv1, *pv2;
bd61b366 6021 char *tpv = NULL;
cf48d248 6022 I32 cmp;
a0714e2c 6023 SV *svrecode = NULL;
560a288e 6024
e01b9e88
SC
6025 if (!sv1) {
6026 pv1 = "";
560a288e
GS
6027 cur1 = 0;
6028 }
e01b9e88 6029 else
4d84ee25 6030 pv1 = SvPV_const(sv1, cur1);
560a288e 6031
553e1bcc 6032 if (!sv2) {
e01b9e88 6033 pv2 = "";
560a288e
GS
6034 cur2 = 0;
6035 }
e01b9e88 6036 else
4d84ee25 6037 pv2 = SvPV_const(sv2, cur2);
79072805 6038
cf48d248 6039 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6040 /* Differing utf8ness.
6041 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6042 if (SvUTF8(sv1)) {
799ef3cb 6043 if (PL_encoding) {
553e1bcc
AT
6044 svrecode = newSVpvn(pv2, cur2);
6045 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6046 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6047 }
6048 else {
e1ec3a88 6049 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6050 }
cf48d248
JH
6051 }
6052 else {
799ef3cb 6053 if (PL_encoding) {
553e1bcc
AT
6054 svrecode = newSVpvn(pv1, cur1);
6055 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6056 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6057 }
6058 else {
e1ec3a88 6059 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6060 }
cf48d248
JH
6061 }
6062 }
6063
e01b9e88 6064 if (!cur1) {
cf48d248 6065 cmp = cur2 ? -1 : 0;
e01b9e88 6066 } else if (!cur2) {
cf48d248
JH
6067 cmp = 1;
6068 } else {
e1ec3a88 6069 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6070
6071 if (retval) {
cf48d248 6072 cmp = retval < 0 ? -1 : 1;
e01b9e88 6073 } else if (cur1 == cur2) {
cf48d248
JH
6074 cmp = 0;
6075 } else {
6076 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6077 }
cf48d248 6078 }
16660edb 6079
b37c2d43 6080 SvREFCNT_dec(svrecode);
553e1bcc
AT
6081 if (tpv)
6082 Safefree(tpv);
cf48d248
JH
6083
6084 return cmp;
bbce6d69 6085}
16660edb 6086
c461cf8f
JH
6087/*
6088=for apidoc sv_cmp_locale
6089
645c22ef
DM
6090Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6091'use bytes' aware, handles get magic, and will coerce its args to strings
6092if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6093
6094=cut
6095*/
6096
bbce6d69 6097I32
864dbfa3 6098Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6099{
97aff369 6100 dVAR;
36477c24 6101#ifdef USE_LOCALE_COLLATE
16660edb 6102
bbce6d69 6103 char *pv1, *pv2;
6104 STRLEN len1, len2;
6105 I32 retval;
16660edb 6106
3280af22 6107 if (PL_collation_standard)
bbce6d69 6108 goto raw_compare;
16660edb 6109
bbce6d69 6110 len1 = 0;
8ac85365 6111 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6112 len2 = 0;
8ac85365 6113 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6114
bbce6d69 6115 if (!pv1 || !len1) {
6116 if (pv2 && len2)
6117 return -1;
6118 else
6119 goto raw_compare;
6120 }
6121 else {
6122 if (!pv2 || !len2)
6123 return 1;
6124 }
16660edb 6125
bbce6d69 6126 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6127
bbce6d69 6128 if (retval)
16660edb 6129 return retval < 0 ? -1 : 1;
6130
bbce6d69 6131 /*
6132 * When the result of collation is equality, that doesn't mean
6133 * that there are no differences -- some locales exclude some
6134 * characters from consideration. So to avoid false equalities,
6135 * we use the raw string as a tiebreaker.
6136 */
16660edb 6137
bbce6d69 6138 raw_compare:
5f66b61c 6139 /*FALLTHROUGH*/
16660edb 6140
36477c24 6141#endif /* USE_LOCALE_COLLATE */
16660edb 6142
bbce6d69 6143 return sv_cmp(sv1, sv2);
6144}
79072805 6145
645c22ef 6146
36477c24 6147#ifdef USE_LOCALE_COLLATE
645c22ef 6148
7a4c00b4 6149/*
645c22ef
DM
6150=for apidoc sv_collxfrm
6151
6152Add Collate Transform magic to an SV if it doesn't already have it.
6153
6154Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6155scalar data of the variable, but transformed to such a format that a normal
6156memory comparison can be used to compare the data according to the locale
6157settings.
6158
6159=cut
6160*/
6161
bbce6d69 6162char *
864dbfa3 6163Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6164{
97aff369 6165 dVAR;
7a4c00b4 6166 MAGIC *mg;
16660edb 6167
14befaf4 6168 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6169 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6170 const char *s;
6171 char *xf;
bbce6d69 6172 STRLEN len, xlen;
6173
7a4c00b4 6174 if (mg)
6175 Safefree(mg->mg_ptr);
93524f2b 6176 s = SvPV_const(sv, len);
bbce6d69 6177 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6178 if (SvREADONLY(sv)) {
6179 SAVEFREEPV(xf);
6180 *nxp = xlen;
3280af22 6181 return xf + sizeof(PL_collation_ix);
ff0cee69 6182 }
7a4c00b4 6183 if (! mg) {
d83f0a82
NC
6184#ifdef PERL_OLD_COPY_ON_WRITE
6185 if (SvIsCOW(sv))
6186 sv_force_normal_flags(sv, 0);
6187#endif
6188 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6189 0, 0);
7a4c00b4 6190 assert(mg);
bbce6d69 6191 }
7a4c00b4 6192 mg->mg_ptr = xf;
565764a8 6193 mg->mg_len = xlen;
7a4c00b4 6194 }
6195 else {
ff0cee69 6196 if (mg) {
6197 mg->mg_ptr = NULL;
565764a8 6198 mg->mg_len = -1;
ff0cee69 6199 }
bbce6d69 6200 }
6201 }
7a4c00b4 6202 if (mg && mg->mg_ptr) {
565764a8 6203 *nxp = mg->mg_len;
3280af22 6204 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6205 }
6206 else {
6207 *nxp = 0;
6208 return NULL;
16660edb 6209 }
79072805
LW
6210}
6211
36477c24 6212#endif /* USE_LOCALE_COLLATE */
bbce6d69 6213
c461cf8f
JH
6214/*
6215=for apidoc sv_gets
6216
6217Get a line from the filehandle and store it into the SV, optionally
6218appending to the currently-stored string.
6219
6220=cut
6221*/
6222
79072805 6223char *
864dbfa3 6224Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6225{
97aff369 6226 dVAR;
e1ec3a88 6227 const char *rsptr;
c07a80fd 6228 STRLEN rslen;
6229 register STDCHAR rslast;
6230 register STDCHAR *bp;
6231 register I32 cnt;
9c5ffd7c 6232 I32 i = 0;
8bfdd7d9 6233 I32 rspara = 0;
c07a80fd 6234
bc44a8a2
NC
6235 if (SvTHINKFIRST(sv))
6236 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6237 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6238 from <>.
6239 However, perlbench says it's slower, because the existing swipe code
6240 is faster than copy on write.
6241 Swings and roundabouts. */
862a34c6 6242 SvUPGRADE(sv, SVt_PV);
99491443 6243
ff68c719 6244 SvSCREAM_off(sv);
efd8b2ba
AE
6245
6246 if (append) {
6247 if (PerlIO_isutf8(fp)) {
6248 if (!SvUTF8(sv)) {
6249 sv_utf8_upgrade_nomg(sv);
6250 sv_pos_u2b(sv,&append,0);
6251 }
6252 } else if (SvUTF8(sv)) {
561b68a9 6253 SV * const tsv = newSV(0);
efd8b2ba
AE
6254 sv_gets(tsv, fp, 0);
6255 sv_utf8_upgrade_nomg(tsv);
6256 SvCUR_set(sv,append);
6257 sv_catsv(sv,tsv);
6258 sv_free(tsv);
6259 goto return_string_or_null;
6260 }
6261 }
6262
6263 SvPOK_only(sv);
6264 if (PerlIO_isutf8(fp))
6265 SvUTF8_on(sv);
c07a80fd 6266
923e4eb5 6267 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6268 /* we always read code in line mode */
6269 rsptr = "\n";
6270 rslen = 1;
6271 }
6272 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6273 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6274 of amount we are going to read -- may result in mallocing
6275 more memory than we really need if the layers below reduce
6276 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6277 */
e311fd51 6278 Stat_t st;
e468d35b 6279 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6280 const Off_t offset = PerlIO_tell(fp);
58f1856e 6281 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6282 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6283 }
6284 }
c07a80fd 6285 rsptr = NULL;
6286 rslen = 0;
6287 }
3280af22 6288 else if (RsRECORD(PL_rs)) {
e311fd51 6289 I32 bytesread;
5b2b9c68 6290 char *buffer;
acbd132f 6291 U32 recsize;
5b2b9c68
HM
6292
6293 /* Grab the size of the record we're getting */
acbd132f 6294 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6295 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6296 /* Go yank in */
6297#ifdef VMS
6298 /* VMS wants read instead of fread, because fread doesn't respect */
6299 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6300 /* doing, but we've got no other real choice - except avoid stdio
6301 as implementation - perhaps write a :vms layer ?
6302 */
5b2b9c68
HM
6303 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6304#else
6305 bytesread = PerlIO_read(fp, buffer, recsize);
6306#endif
27e6ca2d
AE
6307 if (bytesread < 0)
6308 bytesread = 0;
e311fd51 6309 SvCUR_set(sv, bytesread += append);
e670df4e 6310 buffer[bytesread] = '\0';
efd8b2ba 6311 goto return_string_or_null;
5b2b9c68 6312 }
3280af22 6313 else if (RsPARA(PL_rs)) {
c07a80fd 6314 rsptr = "\n\n";
6315 rslen = 2;
8bfdd7d9 6316 rspara = 1;
c07a80fd 6317 }
7d59b7e4
NIS
6318 else {
6319 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6320 if (PerlIO_isutf8(fp)) {
6321 rsptr = SvPVutf8(PL_rs, rslen);
6322 }
6323 else {
6324 if (SvUTF8(PL_rs)) {
6325 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6326 Perl_croak(aTHX_ "Wide character in $/");
6327 }
6328 }
93524f2b 6329 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6330 }
6331 }
6332
c07a80fd 6333 rslast = rslen ? rsptr[rslen - 1] : '\0';
6334
8bfdd7d9 6335 if (rspara) { /* have to do this both before and after */
79072805 6336 do { /* to make sure file boundaries work right */
760ac839 6337 if (PerlIO_eof(fp))
a0d0e21e 6338 return 0;
760ac839 6339 i = PerlIO_getc(fp);
79072805 6340 if (i != '\n') {
a0d0e21e
LW
6341 if (i == -1)
6342 return 0;
760ac839 6343 PerlIO_ungetc(fp,i);
79072805
LW
6344 break;
6345 }
6346 } while (i != EOF);
6347 }
c07a80fd 6348
760ac839
LW
6349 /* See if we know enough about I/O mechanism to cheat it ! */
6350
6351 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6352 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6353 enough here - and may even be a macro allowing compile
6354 time optimization.
6355 */
6356
6357 if (PerlIO_fast_gets(fp)) {
6358
6359 /*
6360 * We're going to steal some values from the stdio struct
6361 * and put EVERYTHING in the innermost loop into registers.
6362 */
6363 register STDCHAR *ptr;
6364 STRLEN bpx;
6365 I32 shortbuffered;
6366
16660edb 6367#if defined(VMS) && defined(PERLIO_IS_STDIO)
6368 /* An ungetc()d char is handled separately from the regular
6369 * buffer, so we getc() it back out and stuff it in the buffer.
6370 */
6371 i = PerlIO_getc(fp);
6372 if (i == EOF) return 0;
6373 *(--((*fp)->_ptr)) = (unsigned char) i;
6374 (*fp)->_cnt++;
6375#endif
c07a80fd 6376
c2960299 6377 /* Here is some breathtakingly efficient cheating */
c07a80fd 6378
a20bf0c3 6379 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6380 /* make sure we have the room */
7a5fa8a2 6381 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6382 /* Not room for all of it
7a5fa8a2 6383 if we are looking for a separator and room for some
e468d35b
NIS
6384 */
6385 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6386 /* just process what we have room for */
79072805
LW
6387 shortbuffered = cnt - SvLEN(sv) + append + 1;
6388 cnt -= shortbuffered;
6389 }
6390 else {
6391 shortbuffered = 0;
bbce6d69 6392 /* remember that cnt can be negative */
eb160463 6393 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6394 }
6395 }
7a5fa8a2 6396 else
79072805 6397 shortbuffered = 0;
3f7c398e 6398 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6399 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6400 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6401 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6402 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6403 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6404 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6405 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6406 for (;;) {
6407 screamer:
93a17b20 6408 if (cnt > 0) {
c07a80fd 6409 if (rslen) {
760ac839
LW
6410 while (cnt > 0) { /* this | eat */
6411 cnt--;
c07a80fd 6412 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6413 goto thats_all_folks; /* screams | sed :-) */
6414 }
6415 }
6416 else {
1c846c1f
NIS
6417 Copy(ptr, bp, cnt, char); /* this | eat */
6418 bp += cnt; /* screams | dust */
c07a80fd 6419 ptr += cnt; /* louder | sed :-) */
a5f75d66 6420 cnt = 0;
93a17b20 6421 }
79072805
LW
6422 }
6423
748a9306 6424 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6425 cnt = shortbuffered;
6426 shortbuffered = 0;
3f7c398e 6427 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6428 SvCUR_set(sv, bpx);
6429 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6430 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6431 continue;
6432 }
6433
16660edb 6434 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6435 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6436 PTR2UV(ptr),(long)cnt));
cc00df79 6437 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6438#if 0
16660edb 6439 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6440 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6441 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6442 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6443#endif
1c846c1f 6444 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6445 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6446 another abstraction. */
760ac839 6447 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6448#if 0
16660edb 6449 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6450 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6451 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6452 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6453#endif
a20bf0c3
JH
6454 cnt = PerlIO_get_cnt(fp);
6455 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6456 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6457 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6458
748a9306
LW
6459 if (i == EOF) /* all done for ever? */
6460 goto thats_really_all_folks;
6461
3f7c398e 6462 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6463 SvCUR_set(sv, bpx);
6464 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6465 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6466
eb160463 6467 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6468
c07a80fd 6469 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6470 goto thats_all_folks;
79072805
LW
6471 }
6472
6473thats_all_folks:
3f7c398e 6474 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6475 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6476 goto screamer; /* go back to the fray */
79072805
LW
6477thats_really_all_folks:
6478 if (shortbuffered)
6479 cnt += shortbuffered;
16660edb 6480 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6481 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6482 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6483 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6484 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6485 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6486 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6487 *bp = '\0';
3f7c398e 6488 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6489 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6490 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6491 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6492 }
6493 else
79072805 6494 {
6edd2cd5 6495 /*The big, slow, and stupid way. */
27da23d5 6496#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6497 STDCHAR *buf = NULL;
a02a5408 6498 Newx(buf, 8192, STDCHAR);
6edd2cd5 6499 assert(buf);
4d2c4e07 6500#else
6edd2cd5 6501 STDCHAR buf[8192];
4d2c4e07 6502#endif
79072805 6503
760ac839 6504screamer2:
c07a80fd 6505 if (rslen) {
00b6aa41 6506 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6507 bp = buf;
eb160463 6508 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6509 ; /* keep reading */
6510 cnt = bp - buf;
c07a80fd 6511 }
6512 else {
760ac839 6513 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6514 /* Accomodate broken VAXC compiler, which applies U8 cast to
6515 * both args of ?: operator, causing EOF to change into 255
6516 */
37be0adf 6517 if (cnt > 0)
cbe9e203
JH
6518 i = (U8)buf[cnt - 1];
6519 else
37be0adf 6520 i = EOF;
c07a80fd 6521 }
79072805 6522
cbe9e203
JH
6523 if (cnt < 0)
6524 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6525 if (append)
6526 sv_catpvn(sv, (char *) buf, cnt);
6527 else
6528 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6529
6530 if (i != EOF && /* joy */
6531 (!rslen ||
6532 SvCUR(sv) < rslen ||
3f7c398e 6533 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6534 {
6535 append = -1;
63e4d877
CS
6536 /*
6537 * If we're reading from a TTY and we get a short read,
6538 * indicating that the user hit his EOF character, we need
6539 * to notice it now, because if we try to read from the TTY
6540 * again, the EOF condition will disappear.
6541 *
6542 * The comparison of cnt to sizeof(buf) is an optimization
6543 * that prevents unnecessary calls to feof().
6544 *
6545 * - jik 9/25/96
6546 */
bb7a0f54 6547 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 6548 goto screamer2;
79072805 6549 }
6edd2cd5 6550
27da23d5 6551#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6552 Safefree(buf);
6553#endif
79072805
LW
6554 }
6555
8bfdd7d9 6556 if (rspara) { /* have to do this both before and after */
c07a80fd 6557 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6558 i = PerlIO_getc(fp);
79072805 6559 if (i != '\n') {
760ac839 6560 PerlIO_ungetc(fp,i);
79072805
LW
6561 break;
6562 }
6563 }
6564 }
c07a80fd 6565
efd8b2ba 6566return_string_or_null:
bd61b366 6567 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6568}
6569
954c1994
GS
6570/*
6571=for apidoc sv_inc
6572
645c22ef
DM
6573Auto-increment of the value in the SV, doing string to numeric conversion
6574if necessary. Handles 'get' magic.
954c1994
GS
6575
6576=cut
6577*/
6578
79072805 6579void
864dbfa3 6580Perl_sv_inc(pTHX_ register SV *sv)
79072805 6581{
97aff369 6582 dVAR;
79072805 6583 register char *d;
463ee0b2 6584 int flags;
79072805
LW
6585
6586 if (!sv)
6587 return;
5b295bef 6588 SvGETMAGIC(sv);
ed6116ce 6589 if (SvTHINKFIRST(sv)) {
765f542d
NC
6590 if (SvIsCOW(sv))
6591 sv_force_normal_flags(sv, 0);
0f15f207 6592 if (SvREADONLY(sv)) {
923e4eb5 6593 if (IN_PERL_RUNTIME)
cea2e8a9 6594 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6595 }
a0d0e21e 6596 if (SvROK(sv)) {
b5be31e9 6597 IV i;
9e7bc3e8
JD
6598 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6599 return;
56431972 6600 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6601 sv_unref(sv);
6602 sv_setiv(sv, i);
a0d0e21e 6603 }
ed6116ce 6604 }
8990e307 6605 flags = SvFLAGS(sv);
28e5dec8
JH
6606 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6607 /* It's (privately or publicly) a float, but not tested as an
6608 integer, so test it to see. */
d460ef45 6609 (void) SvIV(sv);
28e5dec8
JH
6610 flags = SvFLAGS(sv);
6611 }
6612 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6613 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6614#ifdef PERL_PRESERVE_IVUV
28e5dec8 6615 oops_its_int:
59d8ce62 6616#endif
25da4f38
IZ
6617 if (SvIsUV(sv)) {
6618 if (SvUVX(sv) == UV_MAX)
a1e868e7 6619 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6620 else
6621 (void)SvIOK_only_UV(sv);
607fa7f2 6622 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6623 } else {
6624 if (SvIVX(sv) == IV_MAX)
28e5dec8 6625 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6626 else {
6627 (void)SvIOK_only(sv);
45977657 6628 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6629 }
55497cff 6630 }
79072805
LW
6631 return;
6632 }
28e5dec8
JH
6633 if (flags & SVp_NOK) {
6634 (void)SvNOK_only(sv);
9d6ce603 6635 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6636 return;
6637 }
6638
3f7c398e 6639 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6640 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6641 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6642 (void)SvIOK_only(sv);
45977657 6643 SvIV_set(sv, 1);
79072805
LW
6644 return;
6645 }
463ee0b2 6646 d = SvPVX(sv);
79072805
LW
6647 while (isALPHA(*d)) d++;
6648 while (isDIGIT(*d)) d++;
6649 if (*d) {
28e5dec8 6650#ifdef PERL_PRESERVE_IVUV
d1be9408 6651 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6652 warnings. Probably ought to make the sv_iv_please() that does
6653 the conversion if possible, and silently. */
504618e9 6654 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6655 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6656 /* Need to try really hard to see if it's an integer.
6657 9.22337203685478e+18 is an integer.
6658 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6659 so $a="9.22337203685478e+18"; $a+0; $a++
6660 needs to be the same as $a="9.22337203685478e+18"; $a++
6661 or we go insane. */
d460ef45 6662
28e5dec8
JH
6663 (void) sv_2iv(sv);
6664 if (SvIOK(sv))
6665 goto oops_its_int;
6666
6667 /* sv_2iv *should* have made this an NV */
6668 if (flags & SVp_NOK) {
6669 (void)SvNOK_only(sv);
9d6ce603 6670 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6671 return;
6672 }
6673 /* I don't think we can get here. Maybe I should assert this
6674 And if we do get here I suspect that sv_setnv will croak. NWC
6675 Fall through. */
6676#if defined(USE_LONG_DOUBLE)
6677 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
3f7c398e 6678 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6679#else
1779d84d 6680 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
3f7c398e 6681 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6682#endif
6683 }
6684#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6685 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6686 return;
6687 }
6688 d--;
3f7c398e 6689 while (d >= SvPVX_const(sv)) {
79072805
LW
6690 if (isDIGIT(*d)) {
6691 if (++*d <= '9')
6692 return;
6693 *(d--) = '0';
6694 }
6695 else {
9d116dd7
JH
6696#ifdef EBCDIC
6697 /* MKS: The original code here died if letters weren't consecutive.
6698 * at least it didn't have to worry about non-C locales. The
6699 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6700 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6701 * [A-Za-z] are accepted by isALPHA in the C locale.
6702 */
6703 if (*d != 'z' && *d != 'Z') {
6704 do { ++*d; } while (!isALPHA(*d));
6705 return;
6706 }
6707 *(d--) -= 'z' - 'a';
6708#else
79072805
LW
6709 ++*d;
6710 if (isALPHA(*d))
6711 return;
6712 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6713#endif
79072805
LW
6714 }
6715 }
6716 /* oh,oh, the number grew */
6717 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6718 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6719 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6720 *d = d[-1];
6721 if (isDIGIT(d[1]))
6722 *d = '1';
6723 else
6724 *d = d[1];
6725}
6726
954c1994
GS
6727/*
6728=for apidoc sv_dec
6729
645c22ef
DM
6730Auto-decrement of the value in the SV, doing string to numeric conversion
6731if necessary. Handles 'get' magic.
954c1994
GS
6732
6733=cut
6734*/
6735
79072805 6736void
864dbfa3 6737Perl_sv_dec(pTHX_ register SV *sv)
79072805 6738{
97aff369 6739 dVAR;
463ee0b2
LW
6740 int flags;
6741
79072805
LW
6742 if (!sv)
6743 return;
5b295bef 6744 SvGETMAGIC(sv);
ed6116ce 6745 if (SvTHINKFIRST(sv)) {
765f542d
NC
6746 if (SvIsCOW(sv))
6747 sv_force_normal_flags(sv, 0);
0f15f207 6748 if (SvREADONLY(sv)) {
923e4eb5 6749 if (IN_PERL_RUNTIME)
cea2e8a9 6750 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6751 }
a0d0e21e 6752 if (SvROK(sv)) {
b5be31e9 6753 IV i;
9e7bc3e8
JD
6754 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6755 return;
56431972 6756 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6757 sv_unref(sv);
6758 sv_setiv(sv, i);
a0d0e21e 6759 }
ed6116ce 6760 }
28e5dec8
JH
6761 /* Unlike sv_inc we don't have to worry about string-never-numbers
6762 and keeping them magic. But we mustn't warn on punting */
8990e307 6763 flags = SvFLAGS(sv);
28e5dec8
JH
6764 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6765 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6766#ifdef PERL_PRESERVE_IVUV
28e5dec8 6767 oops_its_int:
59d8ce62 6768#endif
25da4f38
IZ
6769 if (SvIsUV(sv)) {
6770 if (SvUVX(sv) == 0) {
6771 (void)SvIOK_only(sv);
45977657 6772 SvIV_set(sv, -1);
25da4f38
IZ
6773 }
6774 else {
6775 (void)SvIOK_only_UV(sv);
f4eee32f 6776 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6777 }
25da4f38
IZ
6778 } else {
6779 if (SvIVX(sv) == IV_MIN)
65202027 6780 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6781 else {
6782 (void)SvIOK_only(sv);
45977657 6783 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6784 }
55497cff 6785 }
6786 return;
6787 }
28e5dec8 6788 if (flags & SVp_NOK) {
9d6ce603 6789 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6790 (void)SvNOK_only(sv);
6791 return;
6792 }
8990e307 6793 if (!(flags & SVp_POK)) {
ef088171
NC
6794 if ((flags & SVTYPEMASK) < SVt_PVIV)
6795 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6796 SvIV_set(sv, -1);
6797 (void)SvIOK_only(sv);
79072805
LW
6798 return;
6799 }
28e5dec8
JH
6800#ifdef PERL_PRESERVE_IVUV
6801 {
504618e9 6802 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6803 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6804 /* Need to try really hard to see if it's an integer.
6805 9.22337203685478e+18 is an integer.
6806 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6807 so $a="9.22337203685478e+18"; $a+0; $a--
6808 needs to be the same as $a="9.22337203685478e+18"; $a--
6809 or we go insane. */
d460ef45 6810
28e5dec8
JH
6811 (void) sv_2iv(sv);
6812 if (SvIOK(sv))
6813 goto oops_its_int;
6814
6815 /* sv_2iv *should* have made this an NV */
6816 if (flags & SVp_NOK) {
6817 (void)SvNOK_only(sv);
9d6ce603 6818 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6819 return;
6820 }
6821 /* I don't think we can get here. Maybe I should assert this
6822 And if we do get here I suspect that sv_setnv will croak. NWC
6823 Fall through. */
6824#if defined(USE_LONG_DOUBLE)
6825 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
3f7c398e 6826 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6827#else
1779d84d 6828 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
3f7c398e 6829 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6830#endif
6831 }
6832 }
6833#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6834 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6835}
6836
954c1994
GS
6837/*
6838=for apidoc sv_mortalcopy
6839
645c22ef 6840Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6841The new SV is marked as mortal. It will be destroyed "soon", either by an
6842explicit call to FREETMPS, or by an implicit call at places such as
6843statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6844
6845=cut
6846*/
6847
79072805
LW
6848/* Make a string that will exist for the duration of the expression
6849 * evaluation. Actually, it may have to last longer than that, but
6850 * hopefully we won't free it until it has been assigned to a
6851 * permanent location. */
6852
6853SV *
864dbfa3 6854Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6855{
97aff369 6856 dVAR;
463ee0b2 6857 register SV *sv;
b881518d 6858
4561caa4 6859 new_SV(sv);
79072805 6860 sv_setsv(sv,oldstr);
677b06e3
GS
6861 EXTEND_MORTAL(1);
6862 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6863 SvTEMP_on(sv);
6864 return sv;
6865}
6866
954c1994
GS
6867/*
6868=for apidoc sv_newmortal
6869
645c22ef 6870Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6871set to 1. It will be destroyed "soon", either by an explicit call to
6872FREETMPS, or by an implicit call at places such as statement boundaries.
6873See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6874
6875=cut
6876*/
6877
8990e307 6878SV *
864dbfa3 6879Perl_sv_newmortal(pTHX)
8990e307 6880{
97aff369 6881 dVAR;
8990e307
LW
6882 register SV *sv;
6883
4561caa4 6884 new_SV(sv);
8990e307 6885 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6886 EXTEND_MORTAL(1);
6887 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6888 return sv;
6889}
6890
954c1994
GS
6891/*
6892=for apidoc sv_2mortal
6893
d4236ebc
DM
6894Marks an existing SV as mortal. The SV will be destroyed "soon", either
6895by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6896statement boundaries. SvTEMP() is turned on which means that the SV's
6897string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6898and C<sv_mortalcopy>.
954c1994
GS
6899
6900=cut
6901*/
6902
79072805 6903SV *
864dbfa3 6904Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6905{
27da23d5 6906 dVAR;
79072805 6907 if (!sv)
7a5b473e 6908 return NULL;
d689ffdd 6909 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6910 return sv;
677b06e3
GS
6911 EXTEND_MORTAL(1);
6912 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6913 SvTEMP_on(sv);
79072805
LW
6914 return sv;
6915}
6916
954c1994
GS
6917/*
6918=for apidoc newSVpv
6919
6920Creates a new SV and copies a string into it. The reference count for the
6921SV is set to 1. If C<len> is zero, Perl will compute the length using
6922strlen(). For efficiency, consider using C<newSVpvn> instead.
6923
6924=cut
6925*/
6926
79072805 6927SV *
864dbfa3 6928Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6929{
97aff369 6930 dVAR;
463ee0b2 6931 register SV *sv;
79072805 6932
4561caa4 6933 new_SV(sv);
ddfa59c7 6934 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
79072805
LW
6935 return sv;
6936}
6937
954c1994
GS
6938/*
6939=for apidoc newSVpvn
6940
6941Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6942SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6943string. You are responsible for ensuring that the source string is at least
9e09f5f2 6944C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6945
6946=cut
6947*/
6948
9da1e3b5 6949SV *
864dbfa3 6950Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6951{
97aff369 6952 dVAR;
9da1e3b5
MUN
6953 register SV *sv;
6954
6955 new_SV(sv);
9da1e3b5
MUN
6956 sv_setpvn(sv,s,len);
6957 return sv;
6958}
6959
bd08039b
NC
6960
6961/*
926f8064 6962=for apidoc newSVhek
bd08039b
NC
6963
6964Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6965point to the shared string table where possible. Returns a new (undefined)
6966SV if the hek is NULL.
bd08039b
NC
6967
6968=cut
6969*/
6970
6971SV *
c1b02ed8 6972Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6973{
97aff369 6974 dVAR;
5aaec2b4
NC
6975 if (!hek) {
6976 SV *sv;
6977
6978 new_SV(sv);
6979 return sv;
6980 }
6981
bd08039b
NC
6982 if (HEK_LEN(hek) == HEf_SVKEY) {
6983 return newSVsv(*(SV**)HEK_KEY(hek));
6984 } else {
6985 const int flags = HEK_FLAGS(hek);
6986 if (flags & HVhek_WASUTF8) {
6987 /* Trouble :-)
6988 Andreas would like keys he put in as utf8 to come back as utf8
6989 */
6990 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6991 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6992 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6993
6994 SvUTF8_on (sv);
6995 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6996 return sv;
45e34800 6997 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
bd08039b
NC
6998 /* We don't have a pointer to the hv, so we have to replicate the
6999 flag into every HEK. This hv is using custom a hasing
7000 algorithm. Hence we can't return a shared string scalar, as
7001 that would contain the (wrong) hash value, and might get passed
45e34800
NC
7002 into an hv routine with a regular hash.
7003 Similarly, a hash that isn't using shared hash keys has to have
7004 the flag in every key so that we know not to try to call
7005 share_hek_kek on it. */
bd08039b 7006
b64e5050 7007 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
7008 if (HEK_UTF8(hek))
7009 SvUTF8_on (sv);
7010 return sv;
7011 }
7012 /* This will be overwhelminly the most common case. */
409dfe77
NC
7013 {
7014 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7015 more efficient than sharepvn(). */
7016 SV *sv;
7017
7018 new_SV(sv);
7019 sv_upgrade(sv, SVt_PV);
7020 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7021 SvCUR_set(sv, HEK_LEN(hek));
7022 SvLEN_set(sv, 0);
7023 SvREADONLY_on(sv);
7024 SvFAKE_on(sv);
7025 SvPOK_on(sv);
7026 if (HEK_UTF8(hek))
7027 SvUTF8_on(sv);
7028 return sv;
7029 }
bd08039b
NC
7030 }
7031}
7032
1c846c1f
NIS
7033/*
7034=for apidoc newSVpvn_share
7035
3f7c398e 7036Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7037table. If the string does not already exist in the table, it is created
7038first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7039slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7040otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7041is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7042hash lookup will avoid string compare.
1c846c1f
NIS
7043
7044=cut
7045*/
7046
7047SV *
c3654f1a 7048Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7049{
97aff369 7050 dVAR;
1c846c1f 7051 register SV *sv;
c3654f1a 7052 bool is_utf8 = FALSE;
a51caccf
NC
7053 const char *const orig_src = src;
7054
c3654f1a 7055 if (len < 0) {
77caf834 7056 STRLEN tmplen = -len;
c3654f1a 7057 is_utf8 = TRUE;
75a54232 7058 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7059 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7060 len = tmplen;
7061 }
1c846c1f 7062 if (!hash)
5afd6d42 7063 PERL_HASH(hash, src, len);
1c846c1f 7064 new_SV(sv);
bdd68bc3 7065 sv_upgrade(sv, SVt_PV);
f880fe2f 7066 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7067 SvCUR_set(sv, len);
b162af07 7068 SvLEN_set(sv, 0);
1c846c1f
NIS
7069 SvREADONLY_on(sv);
7070 SvFAKE_on(sv);
7071 SvPOK_on(sv);
c3654f1a
IH
7072 if (is_utf8)
7073 SvUTF8_on(sv);
a51caccf
NC
7074 if (src != orig_src)
7075 Safefree(src);
1c846c1f
NIS
7076 return sv;
7077}
7078
645c22ef 7079
cea2e8a9 7080#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7081
7082/* pTHX_ magic can't cope with varargs, so this is a no-context
7083 * version of the main function, (which may itself be aliased to us).
7084 * Don't access this version directly.
7085 */
7086
46fc3d4c 7087SV *
cea2e8a9 7088Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7089{
cea2e8a9 7090 dTHX;
46fc3d4c 7091 register SV *sv;
7092 va_list args;
46fc3d4c 7093 va_start(args, pat);
c5be433b 7094 sv = vnewSVpvf(pat, &args);
46fc3d4c 7095 va_end(args);
7096 return sv;
7097}
cea2e8a9 7098#endif
46fc3d4c 7099
954c1994
GS
7100/*
7101=for apidoc newSVpvf
7102
645c22ef 7103Creates a new SV and initializes it with the string formatted like
954c1994
GS
7104C<sprintf>.
7105
7106=cut
7107*/
7108
cea2e8a9
GS
7109SV *
7110Perl_newSVpvf(pTHX_ const char* pat, ...)
7111{
7112 register SV *sv;
7113 va_list args;
cea2e8a9 7114 va_start(args, pat);
c5be433b 7115 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7116 va_end(args);
7117 return sv;
7118}
46fc3d4c 7119
645c22ef
DM
7120/* backend for newSVpvf() and newSVpvf_nocontext() */
7121
79072805 7122SV *
c5be433b
GS
7123Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7124{
97aff369 7125 dVAR;
c5be433b
GS
7126 register SV *sv;
7127 new_SV(sv);
4608196e 7128 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7129 return sv;
7130}
7131
954c1994
GS
7132/*
7133=for apidoc newSVnv
7134
7135Creates a new SV and copies a floating point value into it.
7136The reference count for the SV is set to 1.
7137
7138=cut
7139*/
7140
c5be433b 7141SV *
65202027 7142Perl_newSVnv(pTHX_ NV n)
79072805 7143{
97aff369 7144 dVAR;
463ee0b2 7145 register SV *sv;
79072805 7146
4561caa4 7147 new_SV(sv);
79072805
LW
7148 sv_setnv(sv,n);
7149 return sv;
7150}
7151
954c1994
GS
7152/*
7153=for apidoc newSViv
7154
7155Creates a new SV and copies an integer into it. The reference count for the
7156SV is set to 1.
7157
7158=cut
7159*/
7160
79072805 7161SV *
864dbfa3 7162Perl_newSViv(pTHX_ IV i)
79072805 7163{
97aff369 7164 dVAR;
463ee0b2 7165 register SV *sv;
79072805 7166
4561caa4 7167 new_SV(sv);
79072805
LW
7168 sv_setiv(sv,i);
7169 return sv;
7170}
7171
954c1994 7172/*
1a3327fb
JH
7173=for apidoc newSVuv
7174
7175Creates a new SV and copies an unsigned integer into it.
7176The reference count for the SV is set to 1.
7177
7178=cut
7179*/
7180
7181SV *
7182Perl_newSVuv(pTHX_ UV u)
7183{
97aff369 7184 dVAR;
1a3327fb
JH
7185 register SV *sv;
7186
7187 new_SV(sv);
7188 sv_setuv(sv,u);
7189 return sv;
7190}
7191
7192/*
954c1994
GS
7193=for apidoc newRV_noinc
7194
7195Creates an RV wrapper for an SV. The reference count for the original
7196SV is B<not> incremented.
7197
7198=cut
7199*/
7200
2304df62 7201SV *
864dbfa3 7202Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7203{
97aff369 7204 dVAR;
2304df62
AD
7205 register SV *sv;
7206
4561caa4 7207 new_SV(sv);
2304df62 7208 sv_upgrade(sv, SVt_RV);
76e3520e 7209 SvTEMP_off(tmpRef);
b162af07 7210 SvRV_set(sv, tmpRef);
2304df62 7211 SvROK_on(sv);
2304df62
AD
7212 return sv;
7213}
7214
ff276b08 7215/* newRV_inc is the official function name to use now.
645c22ef
DM
7216 * newRV_inc is in fact #defined to newRV in sv.h
7217 */
7218
5f05dabc 7219SV *
7f466ec7 7220Perl_newRV(pTHX_ SV *sv)
5f05dabc 7221{
97aff369 7222 dVAR;
7f466ec7 7223 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7224}
5f05dabc 7225
954c1994
GS
7226/*
7227=for apidoc newSVsv
7228
7229Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7230(Uses C<sv_setsv>).
954c1994
GS
7231
7232=cut
7233*/
7234
79072805 7235SV *
864dbfa3 7236Perl_newSVsv(pTHX_ register SV *old)
79072805 7237{
97aff369 7238 dVAR;
463ee0b2 7239 register SV *sv;
79072805
LW
7240
7241 if (!old)
7a5b473e 7242 return NULL;
8990e307 7243 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7244 if (ckWARN_d(WARN_INTERNAL))
9014280d 7245 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7246 return NULL;
79072805 7247 }
4561caa4 7248 new_SV(sv);
e90aabeb
NC
7249 /* SV_GMAGIC is the default for sv_setv()
7250 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7251 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7252 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7253 return sv;
79072805
LW
7254}
7255
645c22ef
DM
7256/*
7257=for apidoc sv_reset
7258
7259Underlying implementation for the C<reset> Perl function.
7260Note that the perl-level function is vaguely deprecated.
7261
7262=cut
7263*/
7264
79072805 7265void
e1ec3a88 7266Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7267{
27da23d5 7268 dVAR;
4802d5d7 7269 char todo[PERL_UCHAR_MAX+1];
79072805 7270
49d8d3a1
MB
7271 if (!stash)
7272 return;
7273
79072805 7274 if (!*s) { /* reset ?? searches */
aec46f14 7275 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7276 if (mg) {
7277 PMOP *pm = (PMOP *) mg->mg_obj;
7278 while (pm) {
7279 pm->op_pmdynflags &= ~PMdf_USED;
7280 pm = pm->op_pmnext;
7281 }
79072805
LW
7282 }
7283 return;
7284 }
7285
7286 /* reset variables */
7287
7288 if (!HvARRAY(stash))
7289 return;
463ee0b2
LW
7290
7291 Zero(todo, 256, char);
79072805 7292 while (*s) {
b464bac0
AL
7293 I32 max;
7294 I32 i = (unsigned char)*s;
79072805
LW
7295 if (s[1] == '-') {
7296 s += 2;
7297 }
4802d5d7 7298 max = (unsigned char)*s++;
79072805 7299 for ( ; i <= max; i++) {
463ee0b2
LW
7300 todo[i] = 1;
7301 }
a0d0e21e 7302 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7303 HE *entry;
79072805 7304 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7305 entry;
7306 entry = HeNEXT(entry))
7307 {
b464bac0
AL
7308 register GV *gv;
7309 register SV *sv;
7310
1edc1566 7311 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7312 continue;
1edc1566 7313 gv = (GV*)HeVAL(entry);
79072805 7314 sv = GvSV(gv);
e203899d
NC
7315 if (sv) {
7316 if (SvTHINKFIRST(sv)) {
7317 if (!SvREADONLY(sv) && SvROK(sv))
7318 sv_unref(sv);
7319 /* XXX Is this continue a bug? Why should THINKFIRST
7320 exempt us from resetting arrays and hashes? */
7321 continue;
7322 }
7323 SvOK_off(sv);
7324 if (SvTYPE(sv) >= SVt_PV) {
7325 SvCUR_set(sv, 0);
bd61b366 7326 if (SvPVX_const(sv) != NULL)
e203899d
NC
7327 *SvPVX(sv) = '\0';
7328 SvTAINT(sv);
7329 }
79072805
LW
7330 }
7331 if (GvAV(gv)) {
7332 av_clear(GvAV(gv));
7333 }
bfcb3514 7334 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7335#if defined(VMS)
7336 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7337#else /* ! VMS */
463ee0b2 7338 hv_clear(GvHV(gv));
b0269e46
AB
7339# if defined(USE_ENVIRON_ARRAY)
7340 if (gv == PL_envgv)
7341 my_clearenv();
7342# endif /* USE_ENVIRON_ARRAY */
7343#endif /* VMS */
79072805
LW
7344 }
7345 }
7346 }
7347 }
7348}
7349
645c22ef
DM
7350/*
7351=for apidoc sv_2io
7352
7353Using various gambits, try to get an IO from an SV: the IO slot if its a
7354GV; or the recursive result if we're an RV; or the IO slot of the symbol
7355named after the PV if we're a string.
7356
7357=cut
7358*/
7359
46fc3d4c 7360IO*
864dbfa3 7361Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7362{
7363 IO* io;
7364 GV* gv;
7365
7366 switch (SvTYPE(sv)) {
7367 case SVt_PVIO:
7368 io = (IO*)sv;
7369 break;
7370 case SVt_PVGV:
7371 gv = (GV*)sv;
7372 io = GvIO(gv);
7373 if (!io)
cea2e8a9 7374 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7375 break;
7376 default:
7377 if (!SvOK(sv))
cea2e8a9 7378 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7379 if (SvROK(sv))
7380 return sv_2io(SvRV(sv));
f776e3cd 7381 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7382 if (gv)
7383 io = GvIO(gv);
7384 else
7385 io = 0;
7386 if (!io)
be2597df 7387 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
46fc3d4c 7388 break;
7389 }
7390 return io;
7391}
7392
645c22ef
DM
7393/*
7394=for apidoc sv_2cv
7395
7396Using various gambits, try to get a CV from an SV; in addition, try if
7397possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7398The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7399
7400=cut
7401*/
7402
79072805 7403CV *
864dbfa3 7404Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7405{
27da23d5 7406 dVAR;
a0714e2c 7407 GV *gv = NULL;
601f1833 7408 CV *cv = NULL;
79072805 7409
85dec29a
NC
7410 if (!sv) {
7411 *st = NULL;
7412 *gvp = NULL;
7413 return NULL;
7414 }
79072805 7415 switch (SvTYPE(sv)) {
79072805
LW
7416 case SVt_PVCV:
7417 *st = CvSTASH(sv);
a0714e2c 7418 *gvp = NULL;
79072805
LW
7419 return (CV*)sv;
7420 case SVt_PVHV:
7421 case SVt_PVAV:
ef58ba18 7422 *st = NULL;
a0714e2c 7423 *gvp = NULL;
601f1833 7424 return NULL;
8990e307
LW
7425 case SVt_PVGV:
7426 gv = (GV*)sv;
a0d0e21e 7427 *gvp = gv;
8990e307
LW
7428 *st = GvESTASH(gv);
7429 goto fix_gv;
7430
79072805 7431 default:
5b295bef 7432 SvGETMAGIC(sv);
a0d0e21e 7433 if (SvROK(sv)) {
823a54a3 7434 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7435 tryAMAGICunDEREF(to_cv);
7436
62f274bf
GS
7437 sv = SvRV(sv);
7438 if (SvTYPE(sv) == SVt_PVCV) {
7439 cv = (CV*)sv;
a0714e2c 7440 *gvp = NULL;
62f274bf
GS
7441 *st = CvSTASH(cv);
7442 return cv;
7443 }
7444 else if(isGV(sv))
7445 gv = (GV*)sv;
7446 else
cea2e8a9 7447 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7448 }
62f274bf 7449 else if (isGV(sv))
79072805
LW
7450 gv = (GV*)sv;
7451 else
7a5fd60d 7452 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7453 *gvp = gv;
ef58ba18
NC
7454 if (!gv) {
7455 *st = NULL;
601f1833 7456 return NULL;
ef58ba18 7457 }
e26df76a
NC
7458 /* Some flags to gv_fetchsv mean don't really create the GV */
7459 if (SvTYPE(gv) != SVt_PVGV) {
7460 *st = NULL;
7461 return NULL;
7462 }
79072805 7463 *st = GvESTASH(gv);
8990e307 7464 fix_gv:
8ebc5c01 7465 if (lref && !GvCVu(gv)) {
4633a7c4 7466 SV *tmpsv;
748a9306 7467 ENTER;
561b68a9 7468 tmpsv = newSV(0);
bd61b366 7469 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7470 /* XXX this is probably not what they think they're getting.
7471 * It has the same effect as "sub name;", i.e. just a forward
7472 * declaration! */
774d564b 7473 newSUB(start_subparse(FALSE, 0),
4633a7c4 7474 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7475 NULL, NULL);
748a9306 7476 LEAVE;
8ebc5c01 7477 if (!GvCVu(gv))
35c1215d 7478 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
be2597df 7479 SVfARG(sv));
8990e307 7480 }
8ebc5c01 7481 return GvCVu(gv);
79072805
LW
7482 }
7483}
7484
c461cf8f
JH
7485/*
7486=for apidoc sv_true
7487
7488Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7489Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7490instead use an in-line version.
c461cf8f
JH
7491
7492=cut
7493*/
7494
79072805 7495I32
864dbfa3 7496Perl_sv_true(pTHX_ register SV *sv)
79072805 7497{
8990e307
LW
7498 if (!sv)
7499 return 0;
79072805 7500 if (SvPOK(sv)) {
823a54a3
AL
7501 register const XPV* const tXpv = (XPV*)SvANY(sv);
7502 if (tXpv &&
c2f1de04 7503 (tXpv->xpv_cur > 1 ||
339049b0 7504 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7505 return 1;
7506 else
7507 return 0;
7508 }
7509 else {
7510 if (SvIOK(sv))
463ee0b2 7511 return SvIVX(sv) != 0;
79072805
LW
7512 else {
7513 if (SvNOK(sv))
463ee0b2 7514 return SvNVX(sv) != 0.0;
79072805 7515 else
463ee0b2 7516 return sv_2bool(sv);
79072805
LW
7517 }
7518 }
7519}
79072805 7520
645c22ef 7521/*
c461cf8f
JH
7522=for apidoc sv_pvn_force
7523
7524Get a sensible string out of the SV somehow.
645c22ef
DM
7525A private implementation of the C<SvPV_force> macro for compilers which
7526can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7527
8d6d96c1
HS
7528=for apidoc sv_pvn_force_flags
7529
7530Get a sensible string out of the SV somehow.
7531If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7532appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7533implemented in terms of this function.
645c22ef
DM
7534You normally want to use the various wrapper macros instead: see
7535C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7536
7537=cut
7538*/
7539
7540char *
7541Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7542{
97aff369 7543 dVAR;
6fc92669 7544 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7545 sv_force_normal_flags(sv, 0);
1c846c1f 7546
a0d0e21e 7547 if (SvPOK(sv)) {
13c5b33c
NC
7548 if (lp)
7549 *lp = SvCUR(sv);
a0d0e21e
LW
7550 }
7551 else {
a3b680e6 7552 char *s;
13c5b33c
NC
7553 STRLEN len;
7554
4d84ee25 7555 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7556 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7557 if (PL_op)
7558 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7559 ref, OP_NAME(PL_op));
4d84ee25 7560 else
b64e5050 7561 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7562 }
b64e5050 7563 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7564 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7565 OP_NAME(PL_op));
b64e5050 7566 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7567 if (lp)
7568 *lp = len;
7569
3f7c398e 7570 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7571 if (SvROK(sv))
7572 sv_unref(sv);
862a34c6 7573 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7574 SvGROW(sv, len + 1);
706aa1c9 7575 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7576 SvCUR_set(sv, len);
7577 *SvEND(sv) = '\0';
7578 }
7579 if (!SvPOK(sv)) {
7580 SvPOK_on(sv); /* validate pointer */
7581 SvTAINT(sv);
1d7c1841 7582 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7583 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7584 }
7585 }
4d84ee25 7586 return SvPVX_mutable(sv);
a0d0e21e
LW
7587}
7588
645c22ef 7589/*
645c22ef
DM
7590=for apidoc sv_pvbyten_force
7591
0feed65a 7592The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7593
7594=cut
7595*/
7596
7340a771
GS
7597char *
7598Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7599{
46ec2f14 7600 sv_pvn_force(sv,lp);
ffebcc3e 7601 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7602 *lp = SvCUR(sv);
7603 return SvPVX(sv);
7340a771
GS
7604}
7605
645c22ef 7606/*
c461cf8f
JH
7607=for apidoc sv_pvutf8n_force
7608
0feed65a 7609The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7610
7611=cut
7612*/
7613
7340a771
GS
7614char *
7615Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7616{
46ec2f14 7617 sv_pvn_force(sv,lp);
560a288e 7618 sv_utf8_upgrade(sv);
46ec2f14
TS
7619 *lp = SvCUR(sv);
7620 return SvPVX(sv);
7340a771
GS
7621}
7622
c461cf8f
JH
7623/*
7624=for apidoc sv_reftype
7625
7626Returns a string describing what the SV is a reference to.
7627
7628=cut
7629*/
7630
2b388283 7631const char *
bfed75c6 7632Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7633{
07409e01
NC
7634 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7635 inside return suggests a const propagation bug in g++. */
c86bf373 7636 if (ob && SvOBJECT(sv)) {
1b6737cc 7637 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7638 return name ? name : (char *) "__ANON__";
c86bf373 7639 }
a0d0e21e
LW
7640 else {
7641 switch (SvTYPE(sv)) {
7642 case SVt_NULL:
7643 case SVt_IV:
7644 case SVt_NV:
7645 case SVt_RV:
7646 case SVt_PV:
7647 case SVt_PVIV:
7648 case SVt_PVNV:
7649 case SVt_PVMG:
1cb0ed9b 7650 if (SvVOK(sv))
439cb1c4 7651 return "VSTRING";
a0d0e21e
LW
7652 if (SvROK(sv))
7653 return "REF";
7654 else
7655 return "SCALAR";
1cb0ed9b 7656
07409e01 7657 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7658 /* tied lvalues should appear to be
7659 * scalars for backwards compatitbility */
7660 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7661 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7662 case SVt_PVAV: return "ARRAY";
7663 case SVt_PVHV: return "HASH";
7664 case SVt_PVCV: return "CODE";
7665 case SVt_PVGV: return "GLOB";
1d2dff63 7666 case SVt_PVFM: return "FORMAT";
27f9d8f3 7667 case SVt_PVIO: return "IO";
cecf5685 7668 case SVt_BIND: return "BIND";
a0d0e21e
LW
7669 default: return "UNKNOWN";
7670 }
7671 }
7672}
7673
954c1994
GS
7674/*
7675=for apidoc sv_isobject
7676
7677Returns a boolean indicating whether the SV is an RV pointing to a blessed
7678object. If the SV is not an RV, or if the object is not blessed, then this
7679will return false.
7680
7681=cut
7682*/
7683
463ee0b2 7684int
864dbfa3 7685Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7686{
68dc0745 7687 if (!sv)
7688 return 0;
5b295bef 7689 SvGETMAGIC(sv);
85e6fe83
LW
7690 if (!SvROK(sv))
7691 return 0;
7692 sv = (SV*)SvRV(sv);
7693 if (!SvOBJECT(sv))
7694 return 0;
7695 return 1;
7696}
7697
954c1994
GS
7698/*
7699=for apidoc sv_isa
7700
7701Returns a boolean indicating whether the SV is blessed into the specified
7702class. This does not check for subtypes; use C<sv_derived_from> to verify
7703an inheritance relationship.
7704
7705=cut
7706*/
7707
85e6fe83 7708int
864dbfa3 7709Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7710{
bfcb3514 7711 const char *hvname;
68dc0745 7712 if (!sv)
7713 return 0;
5b295bef 7714 SvGETMAGIC(sv);
ed6116ce 7715 if (!SvROK(sv))
463ee0b2 7716 return 0;
ed6116ce
LW
7717 sv = (SV*)SvRV(sv);
7718 if (!SvOBJECT(sv))
463ee0b2 7719 return 0;
bfcb3514
NC
7720 hvname = HvNAME_get(SvSTASH(sv));
7721 if (!hvname)
e27ad1f2 7722 return 0;
463ee0b2 7723
bfcb3514 7724 return strEQ(hvname, name);
463ee0b2
LW
7725}
7726
954c1994
GS
7727/*
7728=for apidoc newSVrv
7729
7730Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7731it will be upgraded to one. If C<classname> is non-null then the new SV will
7732be blessed in the specified package. The new SV is returned and its
7733reference count is 1.
7734
7735=cut
7736*/
7737
463ee0b2 7738SV*
864dbfa3 7739Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7740{
97aff369 7741 dVAR;
463ee0b2
LW
7742 SV *sv;
7743
4561caa4 7744 new_SV(sv);
51cf62d8 7745
765f542d 7746 SV_CHECK_THINKFIRST_COW_DROP(rv);
52944de8 7747 (void)SvAMAGIC_off(rv);
51cf62d8 7748
0199fce9 7749 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7750 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7751 SvREFCNT(rv) = 0;
7752 sv_clear(rv);
7753 SvFLAGS(rv) = 0;
7754 SvREFCNT(rv) = refcnt;
0199fce9 7755
dc5494d2
NC
7756 sv_upgrade(rv, SVt_RV);
7757 } else if (SvROK(rv)) {
7758 SvREFCNT_dec(SvRV(rv));
7759 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7760 sv_upgrade(rv, SVt_RV);
7761 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7762 SvPV_free(rv);
0199fce9
JD
7763 SvCUR_set(rv, 0);
7764 SvLEN_set(rv, 0);
7765 }
51cf62d8 7766
0c34ef67 7767 SvOK_off(rv);
b162af07 7768 SvRV_set(rv, sv);
ed6116ce 7769 SvROK_on(rv);
463ee0b2 7770
a0d0e21e 7771 if (classname) {
da51bb9b 7772 HV* const stash = gv_stashpv(classname, GV_ADD);
a0d0e21e
LW
7773 (void)sv_bless(rv, stash);
7774 }
7775 return sv;
7776}
7777
954c1994
GS
7778/*
7779=for apidoc sv_setref_pv
7780
7781Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7782argument will be upgraded to an RV. That RV will be modified to point to
7783the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7784into the SV. The C<classname> argument indicates the package for the
bd61b366 7785blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7786will have a reference count of 1, and the RV will be returned.
954c1994
GS
7787
7788Do not use with other Perl types such as HV, AV, SV, CV, because those
7789objects will become corrupted by the pointer copy process.
7790
7791Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7792
7793=cut
7794*/
7795
a0d0e21e 7796SV*
864dbfa3 7797Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7798{
97aff369 7799 dVAR;
189b2af5 7800 if (!pv) {
3280af22 7801 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7802 SvSETMAGIC(rv);
7803 }
a0d0e21e 7804 else
56431972 7805 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7806 return rv;
7807}
7808
954c1994
GS
7809/*
7810=for apidoc sv_setref_iv
7811
7812Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7813argument will be upgraded to an RV. That RV will be modified to point to
7814the new SV. The C<classname> argument indicates the package for the
bd61b366 7815blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7816will have a reference count of 1, and the RV will be returned.
954c1994
GS
7817
7818=cut
7819*/
7820
a0d0e21e 7821SV*
864dbfa3 7822Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7823{
7824 sv_setiv(newSVrv(rv,classname), iv);
7825 return rv;
7826}
7827
954c1994 7828/*
e1c57cef
JH
7829=for apidoc sv_setref_uv
7830
7831Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7832argument will be upgraded to an RV. That RV will be modified to point to
7833the new SV. The C<classname> argument indicates the package for the
bd61b366 7834blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7835will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7836
7837=cut
7838*/
7839
7840SV*
7841Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7842{
7843 sv_setuv(newSVrv(rv,classname), uv);
7844 return rv;
7845}
7846
7847/*
954c1994
GS
7848=for apidoc sv_setref_nv
7849
7850Copies a double into a new SV, optionally blessing the SV. The C<rv>
7851argument will be upgraded to an RV. That RV will be modified to point to
7852the new SV. The C<classname> argument indicates the package for the
bd61b366 7853blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7854will have a reference count of 1, and the RV will be returned.
954c1994
GS
7855
7856=cut
7857*/
7858
a0d0e21e 7859SV*
65202027 7860Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7861{
7862 sv_setnv(newSVrv(rv,classname), nv);
7863 return rv;
7864}
463ee0b2 7865
954c1994
GS
7866/*
7867=for apidoc sv_setref_pvn
7868
7869Copies a string into a new SV, optionally blessing the SV. The length of the
7870string must be specified with C<n>. The C<rv> argument will be upgraded to
7871an RV. That RV will be modified to point to the new SV. The C<classname>
7872argument indicates the package for the blessing. Set C<classname> to
bd61b366 7873C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7874of 1, and the RV will be returned.
954c1994
GS
7875
7876Note that C<sv_setref_pv> copies the pointer while this copies the string.
7877
7878=cut
7879*/
7880
a0d0e21e 7881SV*
1b6737cc 7882Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7883{
7884 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7885 return rv;
7886}
7887
954c1994
GS
7888/*
7889=for apidoc sv_bless
7890
7891Blesses an SV into a specified package. The SV must be an RV. The package
7892must be designated by its stash (see C<gv_stashpv()>). The reference count
7893of the SV is unaffected.
7894
7895=cut
7896*/
7897
a0d0e21e 7898SV*
864dbfa3 7899Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7900{
97aff369 7901 dVAR;
76e3520e 7902 SV *tmpRef;
a0d0e21e 7903 if (!SvROK(sv))
cea2e8a9 7904 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7905 tmpRef = SvRV(sv);
7906 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7907 if (SvREADONLY(tmpRef))
cea2e8a9 7908 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7909 if (SvOBJECT(tmpRef)) {
7910 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7911 --PL_sv_objcount;
76e3520e 7912 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7913 }
a0d0e21e 7914 }
76e3520e
GS
7915 SvOBJECT_on(tmpRef);
7916 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7917 ++PL_sv_objcount;
862a34c6 7918 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7919 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7920
2e3febc6
CS
7921 if (Gv_AMG(stash))
7922 SvAMAGIC_on(sv);
7923 else
52944de8 7924 (void)SvAMAGIC_off(sv);
a0d0e21e 7925
1edbfb88
AB
7926 if(SvSMAGICAL(tmpRef))
7927 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7928 mg_set(tmpRef);
7929
7930
ecdeb87c 7931
a0d0e21e
LW
7932 return sv;
7933}
7934
645c22ef 7935/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7936 */
7937
76e3520e 7938STATIC void
cea2e8a9 7939S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7940{
97aff369 7941 dVAR;
850fabdf 7942 void *xpvmg;
b37c2d43 7943 SV * const temp = sv_newmortal();
850fabdf 7944
a0d0e21e
LW
7945 assert(SvTYPE(sv) == SVt_PVGV);
7946 SvFAKE_off(sv);
180488f8
NC
7947 gv_efullname3(temp, (GV *) sv, "*");
7948
f7877b28 7949 if (GvGP(sv)) {
1edc1566 7950 gp_free((GV*)sv);
f7877b28 7951 }
e826b3c7 7952 if (GvSTASH(sv)) {
e15faf7d 7953 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7954 GvSTASH(sv) = NULL;
e826b3c7 7955 }
a5f75d66 7956 GvMULTI_off(sv);
acda4c6a
NC
7957 if (GvNAME_HEK(sv)) {
7958 unshare_hek(GvNAME_HEK(sv));
7959 }
2e5b91de 7960 isGV_with_GP_off(sv);
850fabdf
GS
7961
7962 /* need to keep SvANY(sv) in the right arena */
7963 xpvmg = new_XPVMG();
7964 StructCopy(SvANY(sv), xpvmg, XPVMG);
7965 del_XPVGV(SvANY(sv));
7966 SvANY(sv) = xpvmg;
7967
a0d0e21e
LW
7968 SvFLAGS(sv) &= ~SVTYPEMASK;
7969 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7970
7971 /* Intentionally not calling any local SET magic, as this isn't so much a
7972 set operation as merely an internal storage change. */
7973 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7974}
7975
954c1994 7976/*
840a7b70 7977=for apidoc sv_unref_flags
954c1994
GS
7978
7979Unsets the RV status of the SV, and decrements the reference count of
7980whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7981as a reversal of C<newSVrv>. The C<cflags> argument can contain
7982C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7983(otherwise the decrementing is conditional on the reference count being
7984different from one or the reference being a readonly SV).
7889fe52 7985See C<SvROK_off>.
954c1994
GS
7986
7987=cut
7988*/
7989
ed6116ce 7990void
e15faf7d 7991Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7992{
b64e5050 7993 SV* const target = SvRV(ref);
810b8aa5 7994
e15faf7d
NC
7995 if (SvWEAKREF(ref)) {
7996 sv_del_backref(target, ref);
7997 SvWEAKREF_off(ref);
7998 SvRV_set(ref, NULL);
810b8aa5
GS
7999 return;
8000 }
e15faf7d
NC
8001 SvRV_set(ref, NULL);
8002 SvROK_off(ref);
8003 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 8004 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
8005 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8006 SvREFCNT_dec(target);
840a7b70 8007 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 8008 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 8009}
8990e307 8010
840a7b70 8011/*
645c22ef
DM
8012=for apidoc sv_untaint
8013
8014Untaint an SV. Use C<SvTAINTED_off> instead.
8015=cut
8016*/
8017
bbce6d69 8018void
864dbfa3 8019Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 8020{
13f57bf8 8021 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 8022 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 8023 if (mg)
565764a8 8024 mg->mg_len &= ~1;
36477c24 8025 }
bbce6d69 8026}
8027
645c22ef
DM
8028/*
8029=for apidoc sv_tainted
8030
8031Test an SV for taintedness. Use C<SvTAINTED> instead.
8032=cut
8033*/
8034
bbce6d69 8035bool
864dbfa3 8036Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8037{
13f57bf8 8038 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8039 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8040 if (mg && (mg->mg_len & 1) )
36477c24 8041 return TRUE;
8042 }
8043 return FALSE;
bbce6d69 8044}
8045
09540bc3
JH
8046/*
8047=for apidoc sv_setpviv
8048
8049Copies an integer into the given SV, also updating its string value.
8050Does not handle 'set' magic. See C<sv_setpviv_mg>.
8051
8052=cut
8053*/
8054
8055void
8056Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8057{
8058 char buf[TYPE_CHARS(UV)];
8059 char *ebuf;
b64e5050 8060 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8061
8062 sv_setpvn(sv, ptr, ebuf - ptr);
8063}
8064
8065/*
8066=for apidoc sv_setpviv_mg
8067
8068Like C<sv_setpviv>, but also handles 'set' magic.
8069
8070=cut
8071*/
8072
8073void
8074Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8075{
df7eb254 8076 sv_setpviv(sv, iv);
09540bc3
JH
8077 SvSETMAGIC(sv);
8078}
8079
cea2e8a9 8080#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8081
8082/* pTHX_ magic can't cope with varargs, so this is a no-context
8083 * version of the main function, (which may itself be aliased to us).
8084 * Don't access this version directly.
8085 */
8086
cea2e8a9
GS
8087void
8088Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8089{
8090 dTHX;
8091 va_list args;
8092 va_start(args, pat);
c5be433b 8093 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8094 va_end(args);
8095}
8096
645c22ef
DM
8097/* pTHX_ magic can't cope with varargs, so this is a no-context
8098 * version of the main function, (which may itself be aliased to us).
8099 * Don't access this version directly.
8100 */
cea2e8a9
GS
8101
8102void
8103Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8104{
8105 dTHX;
8106 va_list args;
8107 va_start(args, pat);
c5be433b 8108 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8109 va_end(args);
cea2e8a9
GS
8110}
8111#endif
8112
954c1994
GS
8113/*
8114=for apidoc sv_setpvf
8115
bffc3d17
SH
8116Works like C<sv_catpvf> but copies the text into the SV instead of
8117appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8118
8119=cut
8120*/
8121
46fc3d4c 8122void
864dbfa3 8123Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8124{
8125 va_list args;
46fc3d4c 8126 va_start(args, pat);
c5be433b 8127 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8128 va_end(args);
8129}
8130
bffc3d17
SH
8131/*
8132=for apidoc sv_vsetpvf
8133
8134Works like C<sv_vcatpvf> but copies the text into the SV instead of
8135appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8136
8137Usually used via its frontend C<sv_setpvf>.
8138
8139=cut
8140*/
645c22ef 8141
c5be433b
GS
8142void
8143Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8144{
4608196e 8145 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8146}
ef50df4b 8147
954c1994
GS
8148/*
8149=for apidoc sv_setpvf_mg
8150
8151Like C<sv_setpvf>, but also handles 'set' magic.
8152
8153=cut
8154*/
8155
ef50df4b 8156void
864dbfa3 8157Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8158{
8159 va_list args;
ef50df4b 8160 va_start(args, pat);
c5be433b 8161 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8162 va_end(args);
c5be433b
GS
8163}
8164
bffc3d17
SH
8165/*
8166=for apidoc sv_vsetpvf_mg
8167
8168Like C<sv_vsetpvf>, but also handles 'set' magic.
8169
8170Usually used via its frontend C<sv_setpvf_mg>.
8171
8172=cut
8173*/
645c22ef 8174
c5be433b
GS
8175void
8176Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8177{
4608196e 8178 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8179 SvSETMAGIC(sv);
8180}
8181
cea2e8a9 8182#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8183
8184/* pTHX_ magic can't cope with varargs, so this is a no-context
8185 * version of the main function, (which may itself be aliased to us).
8186 * Don't access this version directly.
8187 */
8188
cea2e8a9
GS
8189void
8190Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8191{
8192 dTHX;
8193 va_list args;
8194 va_start(args, pat);
c5be433b 8195 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8196 va_end(args);
8197}
8198
645c22ef
DM
8199/* pTHX_ magic can't cope with varargs, so this is a no-context
8200 * version of the main function, (which may itself be aliased to us).
8201 * Don't access this version directly.
8202 */
8203
cea2e8a9
GS
8204void
8205Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8206{
8207 dTHX;
8208 va_list args;
8209 va_start(args, pat);
c5be433b 8210 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8211 va_end(args);
cea2e8a9
GS
8212}
8213#endif
8214
954c1994
GS
8215/*
8216=for apidoc sv_catpvf
8217
d5ce4a7c
GA
8218Processes its arguments like C<sprintf> and appends the formatted
8219output to an SV. If the appended data contains "wide" characters
8220(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8221and characters >255 formatted with %c), the original SV might get
bffc3d17 8222upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8223C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8224valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8225
d5ce4a7c 8226=cut */
954c1994 8227
46fc3d4c 8228void
864dbfa3 8229Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8230{
8231 va_list args;
46fc3d4c 8232 va_start(args, pat);
c5be433b 8233 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8234 va_end(args);
8235}
8236
bffc3d17
SH
8237/*
8238=for apidoc sv_vcatpvf
8239
8240Processes its arguments like C<vsprintf> and appends the formatted output
8241to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8242
8243Usually used via its frontend C<sv_catpvf>.
8244
8245=cut
8246*/
645c22ef 8247
ef50df4b 8248void
c5be433b
GS
8249Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8250{
4608196e 8251 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8252}
8253
954c1994
GS
8254/*
8255=for apidoc sv_catpvf_mg
8256
8257Like C<sv_catpvf>, but also handles 'set' magic.
8258
8259=cut
8260*/
8261
c5be433b 8262void
864dbfa3 8263Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8264{
8265 va_list args;
ef50df4b 8266 va_start(args, pat);
c5be433b 8267 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8268 va_end(args);
c5be433b
GS
8269}
8270
bffc3d17
SH
8271/*
8272=for apidoc sv_vcatpvf_mg
8273
8274Like C<sv_vcatpvf>, but also handles 'set' magic.
8275
8276Usually used via its frontend C<sv_catpvf_mg>.
8277
8278=cut
8279*/
645c22ef 8280
c5be433b
GS
8281void
8282Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8283{
4608196e 8284 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8285 SvSETMAGIC(sv);
8286}
8287
954c1994
GS
8288/*
8289=for apidoc sv_vsetpvfn
8290
bffc3d17 8291Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8292appending it.
8293
bffc3d17 8294Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8295
954c1994
GS
8296=cut
8297*/
8298
46fc3d4c 8299void
7d5ea4e7 8300Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8301{
8302 sv_setpvn(sv, "", 0);
7d5ea4e7 8303 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8304}
8305
2d00ba3b 8306STATIC I32
9dd79c3f 8307S_expect_number(pTHX_ char** pattern)
211dfcf1 8308{
97aff369 8309 dVAR;
211dfcf1
HS
8310 I32 var = 0;
8311 switch (**pattern) {
8312 case '1': case '2': case '3':
8313 case '4': case '5': case '6':
8314 case '7': case '8': case '9':
2fba7546
GA
8315 var = *(*pattern)++ - '0';
8316 while (isDIGIT(**pattern)) {
5f66b61c 8317 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8318 if (tmp < var)
8319 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8320 var = tmp;
8321 }
211dfcf1
HS
8322 }
8323 return var;
8324}
211dfcf1 8325
c445ea15
AL
8326STATIC char *
8327S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8328{
a3b680e6 8329 const int neg = nv < 0;
4151a5fe 8330 UV uv;
4151a5fe
IZ
8331
8332 if (neg)
8333 nv = -nv;
8334 if (nv < UV_MAX) {
b464bac0 8335 char *p = endbuf;
4151a5fe 8336 nv += 0.5;
028f8eaa 8337 uv = (UV)nv;
4151a5fe
IZ
8338 if (uv & 1 && uv == nv)
8339 uv--; /* Round to even */
8340 do {
a3b680e6 8341 const unsigned dig = uv % 10;
4151a5fe
IZ
8342 *--p = '0' + dig;
8343 } while (uv /= 10);
8344 if (neg)
8345 *--p = '-';
8346 *len = endbuf - p;
8347 return p;
8348 }
bd61b366 8349 return NULL;
4151a5fe
IZ
8350}
8351
8352
954c1994
GS
8353/*
8354=for apidoc sv_vcatpvfn
8355
8356Processes its arguments like C<vsprintf> and appends the formatted output
8357to an SV. Uses an array of SVs if the C style variable argument list is
8358missing (NULL). When running with taint checks enabled, indicates via
8359C<maybe_tainted> if results are untrustworthy (often due to the use of
8360locales).
8361
bffc3d17 8362Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8363
954c1994
GS
8364=cut
8365*/
8366
8896765a
RB
8367
8368#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8369 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8370 vec_utf8 = DO_UTF8(vecsv);
8371
1ef29b0e
RGS
8372/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8373
46fc3d4c 8374void
7d5ea4e7 8375Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8376{
97aff369 8377 dVAR;
46fc3d4c 8378 char *p;
8379 char *q;
a3b680e6 8380 const char *patend;
fc36a67e 8381 STRLEN origlen;
46fc3d4c 8382 I32 svix = 0;
27da23d5 8383 static const char nullstr[] = "(null)";
a0714e2c 8384 SV *argsv = NULL;
b464bac0
AL
8385 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8386 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8387 SV *nsv = NULL;
4151a5fe
IZ
8388 /* Times 4: a decimal digit takes more than 3 binary digits.
8389 * NV_DIG: mantissa takes than many decimal digits.
8390 * Plus 32: Playing safe. */
8391 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8392 /* large enough for "%#.#f" --chip */
8393 /* what about long double NVs? --jhi */
db79b45b 8394
53c1dcc0
AL
8395 PERL_UNUSED_ARG(maybe_tainted);
8396
46fc3d4c 8397 /* no matter what, this is a string now */
fc36a67e 8398 (void)SvPV_force(sv, origlen);
46fc3d4c 8399
8896765a 8400 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8401 if (patlen == 0)
8402 return;
0dbb1585 8403 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8404 if (args) {
8405 const char * const s = va_arg(*args, char*);
8406 sv_catpv(sv, s ? s : nullstr);
8407 }
8408 else if (svix < svmax) {
8409 sv_catsv(sv, *svargs);
2d03de9c
AL
8410 }
8411 return;
0dbb1585 8412 }
8896765a
RB
8413 if (args && patlen == 3 && pat[0] == '%' &&
8414 pat[1] == '-' && pat[2] == 'p') {
6c9570dc 8415 argsv = (SV*)va_arg(*args, void*);
8896765a 8416 sv_catsv(sv, argsv);
8896765a 8417 return;
46fc3d4c 8418 }
8419
1d917b39 8420#ifndef USE_LONG_DOUBLE
4151a5fe 8421 /* special-case "%.<number>[gf]" */
7af36d83 8422 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8423 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8424 unsigned digits = 0;
8425 const char *pp;
8426
8427 pp = pat + 2;
8428 while (*pp >= '0' && *pp <= '9')
8429 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8430 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8431 NV nv;
8432
7af36d83 8433 if (svix < svmax)
4151a5fe
IZ
8434 nv = SvNV(*svargs);
8435 else
8436 return;
8437 if (*pp == 'g') {
2873255c
NC
8438 /* Add check for digits != 0 because it seems that some
8439 gconverts are buggy in this case, and we don't yet have
8440 a Configure test for this. */
8441 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8442 /* 0, point, slack */
2e59c212 8443 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8444 sv_catpv(sv, ebuf);
8445 if (*ebuf) /* May return an empty string for digits==0 */
8446 return;
8447 }
8448 } else if (!digits) {
8449 STRLEN l;
8450
8451 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8452 sv_catpvn(sv, p, l);
8453 return;
8454 }
8455 }
8456 }
8457 }
1d917b39 8458#endif /* !USE_LONG_DOUBLE */
4151a5fe 8459
2cf2cfc6 8460 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8461 has_utf8 = TRUE;
2cf2cfc6 8462
46fc3d4c 8463 patend = (char*)pat + patlen;
8464 for (p = (char*)pat; p < patend; p = q) {
8465 bool alt = FALSE;
8466 bool left = FALSE;
b22c7a20 8467 bool vectorize = FALSE;
211dfcf1 8468 bool vectorarg = FALSE;
2cf2cfc6 8469 bool vec_utf8 = FALSE;
46fc3d4c 8470 char fill = ' ';
8471 char plus = 0;
8472 char intsize = 0;
8473 STRLEN width = 0;
fc36a67e 8474 STRLEN zeros = 0;
46fc3d4c 8475 bool has_precis = FALSE;
8476 STRLEN precis = 0;
c445ea15 8477 const I32 osvix = svix;
2cf2cfc6 8478 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8479#ifdef HAS_LDBL_SPRINTF_BUG
8480 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8481 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8482 bool fix_ldbl_sprintf_bug = FALSE;
8483#endif
205f51d8 8484
46fc3d4c 8485 char esignbuf[4];
89ebb4a3 8486 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8487 STRLEN esignlen = 0;
8488
bd61b366 8489 const char *eptr = NULL;
fc36a67e 8490 STRLEN elen = 0;
a0714e2c 8491 SV *vecsv = NULL;
4608196e 8492 const U8 *vecstr = NULL;
b22c7a20 8493 STRLEN veclen = 0;
934abaf1 8494 char c = 0;
46fc3d4c 8495 int i;
9c5ffd7c 8496 unsigned base = 0;
8c8eb53c
RB
8497 IV iv = 0;
8498 UV uv = 0;
9e5b023a
JH
8499 /* we need a long double target in case HAS_LONG_DOUBLE but
8500 not USE_LONG_DOUBLE
8501 */
35fff930 8502#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8503 long double nv;
8504#else
65202027 8505 NV nv;
9e5b023a 8506#endif
46fc3d4c 8507 STRLEN have;
8508 STRLEN need;
8509 STRLEN gap;
7af36d83 8510 const char *dotstr = ".";
b22c7a20 8511 STRLEN dotstrlen = 1;
211dfcf1 8512 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8513 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8514 I32 epix = 0; /* explicit precision index */
8515 I32 evix = 0; /* explicit vector index */
eb3fce90 8516 bool asterisk = FALSE;
46fc3d4c 8517
211dfcf1 8518 /* echo everything up to the next format specification */
46fc3d4c 8519 for (q = p; q < patend && *q != '%'; ++q) ;
8520 if (q > p) {
db79b45b
JH
8521 if (has_utf8 && !pat_utf8)
8522 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8523 else
8524 sv_catpvn(sv, p, q - p);
46fc3d4c 8525 p = q;
8526 }
8527 if (q++ >= patend)
8528 break;
8529
211dfcf1
HS
8530/*
8531 We allow format specification elements in this order:
8532 \d+\$ explicit format parameter index
8533 [-+ 0#]+ flags
a472f209 8534 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8535 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8536 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8537 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8538 [hlqLV] size
8896765a
RB
8539 [%bcdefginopsuxDFOUX] format (mandatory)
8540*/
8541
8542 if (args) {
8543/*
8544 As of perl5.9.3, printf format checking is on by default.
8545 Internally, perl uses %p formats to provide an escape to
8546 some extended formatting. This block deals with those
8547 extensions: if it does not match, (char*)q is reset and
8548 the normal format processing code is used.
8549
8550 Currently defined extensions are:
8551 %p include pointer address (standard)
8552 %-p (SVf) include an SV (previously %_)
8553 %-<num>p include an SV with precision <num>
8554 %1p (VDf) include a v-string (as %vd)
8555 %<num>p reserved for future extensions
8556
8557 Robin Barker 2005-07-14
211dfcf1 8558*/
8896765a
RB
8559 char* r = q;
8560 bool sv = FALSE;
8561 STRLEN n = 0;
8562 if (*q == '-')
8563 sv = *q++;
c445ea15 8564 n = expect_number(&q);
8896765a
RB
8565 if (*q++ == 'p') {
8566 if (sv) { /* SVf */
8567 if (n) {
8568 precis = n;
8569 has_precis = TRUE;
8570 }
6c9570dc 8571 argsv = (SV*)va_arg(*args, void*);
8896765a
RB
8572 eptr = SvPVx_const(argsv, elen);
8573 if (DO_UTF8(argsv))
8574 is_utf8 = TRUE;
8575 goto string;
8576 }
8577#if vdNUMBER
8578 else if (n == vdNUMBER) { /* VDf */
8579 vectorize = TRUE;
8580 VECTORIZE_ARGS
8581 goto format_vd;
8582 }
8583#endif
8584 else if (n) {
8585 if (ckWARN_d(WARN_INTERNAL))
8586 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8587 "internal %%<num>p might conflict with future printf extensions");
8588 }
8589 }
8590 q = r;
8591 }
8592
c445ea15 8593 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8594 if (*q == '$') {
8595 ++q;
8596 efix = width;
8597 } else {
8598 goto gotwidth;
8599 }
8600 }
8601
fc36a67e 8602 /* FLAGS */
8603
46fc3d4c 8604 while (*q) {
8605 switch (*q) {
8606 case ' ':
8607 case '+':
9911cee9
TS
8608 if (plus == '+' && *q == ' ') /* '+' over ' ' */
8609 q++;
8610 else
8611 plus = *q++;
46fc3d4c 8612 continue;
8613
8614 case '-':
8615 left = TRUE;
8616 q++;
8617 continue;
8618
8619 case '0':
8620 fill = *q++;
8621 continue;
8622
8623 case '#':
8624 alt = TRUE;
8625 q++;
8626 continue;
8627
fc36a67e 8628 default:
8629 break;
8630 }
8631 break;
8632 }
46fc3d4c 8633
211dfcf1 8634 tryasterisk:
eb3fce90 8635 if (*q == '*') {
211dfcf1 8636 q++;
c445ea15 8637 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8638 if (*q++ != '$')
8639 goto unknown;
eb3fce90 8640 asterisk = TRUE;
211dfcf1
HS
8641 }
8642 if (*q == 'v') {
eb3fce90 8643 q++;
211dfcf1
HS
8644 if (vectorize)
8645 goto unknown;
9cbac4c7 8646 if ((vectorarg = asterisk)) {
211dfcf1
HS
8647 evix = ewix;
8648 ewix = 0;
8649 asterisk = FALSE;
8650 }
8651 vectorize = TRUE;
8652 goto tryasterisk;
eb3fce90
JH
8653 }
8654
211dfcf1 8655 if (!asterisk)
858a90f9 8656 {
7a5fa8a2 8657 if( *q == '0' )
f3583277 8658 fill = *q++;
c445ea15 8659 width = expect_number(&q);
858a90f9 8660 }
211dfcf1
HS
8661
8662 if (vectorize) {
8663 if (vectorarg) {
8664 if (args)
8665 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8666 else if (evix) {
8667 vecsv = (evix > 0 && evix <= svmax)
8668 ? svargs[evix-1] : &PL_sv_undef;
8669 } else {
8670 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8671 }
245d4a47 8672 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8673 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8674 bad with tied or overloaded values that return UTF8. */
211dfcf1 8675 if (DO_UTF8(vecsv))
2cf2cfc6 8676 is_utf8 = TRUE;
640283f5
NC
8677 else if (has_utf8) {
8678 vecsv = sv_mortalcopy(vecsv);
8679 sv_utf8_upgrade(vecsv);
8680 dotstr = SvPV_const(vecsv, dotstrlen);
8681 is_utf8 = TRUE;
8682 }
211dfcf1
HS
8683 }
8684 if (args) {
8896765a 8685 VECTORIZE_ARGS
eb3fce90 8686 }
7ad96abb 8687 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8688 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8689 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8690 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8691
8692 /* if this is a version object, we need to convert
8693 * back into v-string notation and then let the
8694 * vectorize happen normally
d7aa5382 8695 */
96b8f7ce
JP
8696 if (sv_derived_from(vecsv, "version")) {
8697 char *version = savesvpv(vecsv);
34ba6322
SP
8698 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8699 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8700 "vector argument not supported with alpha versions");
8701 goto unknown;
8702 }
96b8f7ce
JP
8703 vecsv = sv_newmortal();
8704 /* scan_vstring is expected to be called during
8705 * tokenization, so we need to fake up the end
8706 * of the buffer for it
8707 */
8708 PL_bufend = version + veclen;
8709 scan_vstring(version, vecsv);
8710 vecstr = (U8*)SvPV_const(vecsv, veclen);
8711 vec_utf8 = DO_UTF8(vecsv);
8712 Safefree(version);
d7aa5382 8713 }
211dfcf1
HS
8714 }
8715 else {
8716 vecstr = (U8*)"";
8717 veclen = 0;
8718 }
eb3fce90 8719 }
fc36a67e 8720
eb3fce90 8721 if (asterisk) {
fc36a67e 8722 if (args)
8723 i = va_arg(*args, int);
8724 else
eb3fce90
JH
8725 i = (ewix ? ewix <= svmax : svix < svmax) ?
8726 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8727 left |= (i < 0);
8728 width = (i < 0) ? -i : i;
fc36a67e 8729 }
211dfcf1 8730 gotwidth:
fc36a67e 8731
8732 /* PRECISION */
46fc3d4c 8733
fc36a67e 8734 if (*q == '.') {
8735 q++;
8736 if (*q == '*') {
211dfcf1 8737 q++;
c445ea15 8738 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8739 goto unknown;
8740 /* XXX: todo, support specified precision parameter */
8741 if (epix)
211dfcf1 8742 goto unknown;
46fc3d4c 8743 if (args)
8744 i = va_arg(*args, int);
8745 else
eb3fce90
JH
8746 i = (ewix ? ewix <= svmax : svix < svmax)
8747 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9911cee9
TS
8748 precis = i;
8749 has_precis = !(i < 0);
fc36a67e 8750 }
8751 else {
8752 precis = 0;
8753 while (isDIGIT(*q))
8754 precis = precis * 10 + (*q++ - '0');
9911cee9 8755 has_precis = TRUE;
fc36a67e 8756 }
fc36a67e 8757 }
46fc3d4c 8758
fc36a67e 8759 /* SIZE */
46fc3d4c 8760
fc36a67e 8761 switch (*q) {
c623ac67
GS
8762#ifdef WIN32
8763 case 'I': /* Ix, I32x, and I64x */
8764# ifdef WIN64
8765 if (q[1] == '6' && q[2] == '4') {
8766 q += 3;
8767 intsize = 'q';
8768 break;
8769 }
8770# endif
8771 if (q[1] == '3' && q[2] == '2') {
8772 q += 3;
8773 break;
8774 }
8775# ifdef WIN64
8776 intsize = 'q';
8777# endif
8778 q++;
8779 break;
8780#endif
9e5b023a 8781#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8782 case 'L': /* Ld */
5f66b61c 8783 /*FALLTHROUGH*/
e5c81feb 8784#ifdef HAS_QUAD
6f9bb7fd 8785 case 'q': /* qd */
9e5b023a 8786#endif
6f9bb7fd
GS
8787 intsize = 'q';
8788 q++;
8789 break;
8790#endif
fc36a67e 8791 case 'l':
9e5b023a 8792#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8793 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8794 intsize = 'q';
8795 q += 2;
46fc3d4c 8796 break;
cf2093f6 8797 }
fc36a67e 8798#endif
5f66b61c 8799 /*FALLTHROUGH*/
fc36a67e 8800 case 'h':
5f66b61c 8801 /*FALLTHROUGH*/
fc36a67e 8802 case 'V':
8803 intsize = *q++;
46fc3d4c 8804 break;
8805 }
8806
fc36a67e 8807 /* CONVERSION */
8808
211dfcf1
HS
8809 if (*q == '%') {
8810 eptr = q++;
8811 elen = 1;
26372e71
GA
8812 if (vectorize) {
8813 c = '%';
8814 goto unknown;
8815 }
211dfcf1
HS
8816 goto string;
8817 }
8818
26372e71 8819 if (!vectorize && !args) {
86c51f8b
NC
8820 if (efix) {
8821 const I32 i = efix-1;
8822 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8823 } else {
8824 argsv = (svix >= 0 && svix < svmax)
8825 ? svargs[svix++] : &PL_sv_undef;
8826 }
863811b2 8827 }
211dfcf1 8828
46fc3d4c 8829 switch (c = *q++) {
8830
8831 /* STRINGS */
8832
46fc3d4c 8833 case 'c':
26372e71
GA
8834 if (vectorize)
8835 goto unknown;
8836 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8837 if ((uv > 255 ||
8838 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8839 && !IN_BYTES) {
dfe13c55 8840 eptr = (char*)utf8buf;
9041c2e3 8841 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8842 is_utf8 = TRUE;
7e2040f0
GS
8843 }
8844 else {
8845 c = (char)uv;
8846 eptr = &c;
8847 elen = 1;
a0ed51b3 8848 }
46fc3d4c 8849 goto string;
8850
46fc3d4c 8851 case 's':
26372e71
GA
8852 if (vectorize)
8853 goto unknown;
8854 if (args) {
fc36a67e 8855 eptr = va_arg(*args, char*);
c635e13b 8856 if (eptr)
1d7c1841
GS
8857#ifdef MACOS_TRADITIONAL
8858 /* On MacOS, %#s format is used for Pascal strings */
8859 if (alt)
8860 elen = *eptr++;
8861 else
8862#endif
c635e13b 8863 elen = strlen(eptr);
8864 else {
27da23d5 8865 eptr = (char *)nullstr;
c635e13b 8866 elen = sizeof nullstr - 1;
8867 }
46fc3d4c 8868 }
211dfcf1 8869 else {
4d84ee25 8870 eptr = SvPVx_const(argsv, elen);
7e2040f0 8871 if (DO_UTF8(argsv)) {
59b61096 8872 I32 old_precis = precis;
a0ed51b3
LW
8873 if (has_precis && precis < elen) {
8874 I32 p = precis;
7e2040f0 8875 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8876 precis = p;
8877 }
8878 if (width) { /* fudge width (can't fudge elen) */
59b61096
AV
8879 if (has_precis && precis < elen)
8880 width += precis - old_precis;
8881 else
8882 width += elen - sv_len_utf8(argsv);
a0ed51b3 8883 }
2cf2cfc6 8884 is_utf8 = TRUE;
a0ed51b3
LW
8885 }
8886 }
fc36a67e 8887
46fc3d4c 8888 string:
8889 if (has_precis && elen > precis)
8890 elen = precis;
8891 break;
8892
8893 /* INTEGERS */
8894
fc36a67e 8895 case 'p':
be75b157 8896 if (alt || vectorize)
c2e66d9e 8897 goto unknown;
211dfcf1 8898 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8899 base = 16;
8900 goto integer;
8901
46fc3d4c 8902 case 'D':
29fe7a80 8903#ifdef IV_IS_QUAD
22f3ae8c 8904 intsize = 'q';
29fe7a80 8905#else
46fc3d4c 8906 intsize = 'l';
29fe7a80 8907#endif
5f66b61c 8908 /*FALLTHROUGH*/
46fc3d4c 8909 case 'd':
8910 case 'i':
8896765a
RB
8911#if vdNUMBER
8912 format_vd:
8913#endif
b22c7a20 8914 if (vectorize) {
ba210ebe 8915 STRLEN ulen;
211dfcf1
HS
8916 if (!veclen)
8917 continue;
2cf2cfc6
A
8918 if (vec_utf8)
8919 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8920 UTF8_ALLOW_ANYUV);
b22c7a20 8921 else {
e83d50c9 8922 uv = *vecstr;
b22c7a20
GS
8923 ulen = 1;
8924 }
8925 vecstr += ulen;
8926 veclen -= ulen;
e83d50c9
JP
8927 if (plus)
8928 esignbuf[esignlen++] = plus;
b22c7a20
GS
8929 }
8930 else if (args) {
46fc3d4c 8931 switch (intsize) {
8932 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8933 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8934 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8935 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8936#ifdef HAS_QUAD
8937 case 'q': iv = va_arg(*args, Quad_t); break;
8938#endif
46fc3d4c 8939 }
8940 }
8941 else {
b10c0dba 8942 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8943 switch (intsize) {
b10c0dba
MHM
8944 case 'h': iv = (short)tiv; break;
8945 case 'l': iv = (long)tiv; break;
8946 case 'V':
8947 default: iv = tiv; break;
cf2093f6 8948#ifdef HAS_QUAD
b10c0dba 8949 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8950#endif
46fc3d4c 8951 }
8952 }
e83d50c9
JP
8953 if ( !vectorize ) /* we already set uv above */
8954 {
8955 if (iv >= 0) {
8956 uv = iv;
8957 if (plus)
8958 esignbuf[esignlen++] = plus;
8959 }
8960 else {
8961 uv = -iv;
8962 esignbuf[esignlen++] = '-';
8963 }
46fc3d4c 8964 }
8965 base = 10;
8966 goto integer;
8967
fc36a67e 8968 case 'U':
29fe7a80 8969#ifdef IV_IS_QUAD
22f3ae8c 8970 intsize = 'q';
29fe7a80 8971#else
fc36a67e 8972 intsize = 'l';
29fe7a80 8973#endif
5f66b61c 8974 /*FALLTHROUGH*/
fc36a67e 8975 case 'u':
8976 base = 10;
8977 goto uns_integer;
8978
7ff06cc7 8979 case 'B':
4f19785b
WSI
8980 case 'b':
8981 base = 2;
8982 goto uns_integer;
8983
46fc3d4c 8984 case 'O':
29fe7a80 8985#ifdef IV_IS_QUAD
22f3ae8c 8986 intsize = 'q';
29fe7a80 8987#else
46fc3d4c 8988 intsize = 'l';
29fe7a80 8989#endif
5f66b61c 8990 /*FALLTHROUGH*/
46fc3d4c 8991 case 'o':
8992 base = 8;
8993 goto uns_integer;
8994
8995 case 'X':
46fc3d4c 8996 case 'x':
8997 base = 16;
46fc3d4c 8998
8999 uns_integer:
b22c7a20 9000 if (vectorize) {
ba210ebe 9001 STRLEN ulen;
b22c7a20 9002 vector:
211dfcf1
HS
9003 if (!veclen)
9004 continue;
2cf2cfc6
A
9005 if (vec_utf8)
9006 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9007 UTF8_ALLOW_ANYUV);
b22c7a20 9008 else {
a05b299f 9009 uv = *vecstr;
b22c7a20
GS
9010 ulen = 1;
9011 }
9012 vecstr += ulen;
9013 veclen -= ulen;
9014 }
9015 else if (args) {
46fc3d4c 9016 switch (intsize) {
9017 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 9018 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 9019 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 9020 default: uv = va_arg(*args, unsigned); break;
cf2093f6 9021#ifdef HAS_QUAD
9e3321a5 9022 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 9023#endif
46fc3d4c 9024 }
9025 }
9026 else {
b10c0dba 9027 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 9028 switch (intsize) {
b10c0dba
MHM
9029 case 'h': uv = (unsigned short)tuv; break;
9030 case 'l': uv = (unsigned long)tuv; break;
9031 case 'V':
9032 default: uv = tuv; break;
cf2093f6 9033#ifdef HAS_QUAD
b10c0dba 9034 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 9035#endif
46fc3d4c 9036 }
9037 }
9038
9039 integer:
4d84ee25
NC
9040 {
9041 char *ptr = ebuf + sizeof ebuf;
1387f30c
DD
9042 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9043 zeros = 0;
9044
4d84ee25
NC
9045 switch (base) {
9046 unsigned dig;
9047 case 16:
14eb61ab 9048 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
4d84ee25
NC
9049 do {
9050 dig = uv & 15;
9051 *--ptr = p[dig];
9052 } while (uv >>= 4);
1387f30c 9053 if (tempalt) {
4d84ee25
NC
9054 esignbuf[esignlen++] = '0';
9055 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9056 }
9057 break;
9058 case 8:
9059 do {
9060 dig = uv & 7;
9061 *--ptr = '0' + dig;
9062 } while (uv >>= 3);
9063 if (alt && *ptr != '0')
9064 *--ptr = '0';
9065 break;
9066 case 2:
9067 do {
9068 dig = uv & 1;
9069 *--ptr = '0' + dig;
9070 } while (uv >>= 1);
1387f30c 9071 if (tempalt) {
4d84ee25 9072 esignbuf[esignlen++] = '0';
7ff06cc7 9073 esignbuf[esignlen++] = c;
4d84ee25
NC
9074 }
9075 break;
9076 default: /* it had better be ten or less */
9077 do {
9078 dig = uv % base;
9079 *--ptr = '0' + dig;
9080 } while (uv /= base);
9081 break;
46fc3d4c 9082 }
4d84ee25
NC
9083 elen = (ebuf + sizeof ebuf) - ptr;
9084 eptr = ptr;
9085 if (has_precis) {
9086 if (precis > elen)
9087 zeros = precis - elen;
e6bb52fd
TS
9088 else if (precis == 0 && elen == 1 && *eptr == '0'
9089 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
4d84ee25 9090 elen = 0;
9911cee9
TS
9091
9092 /* a precision nullifies the 0 flag. */
9093 if (fill == '0')
9094 fill = ' ';
eda88b6d 9095 }
c10ed8b9 9096 }
46fc3d4c 9097 break;
9098
9099 /* FLOATING POINT */
9100
fc36a67e 9101 case 'F':
9102 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9103 /*FALLTHROUGH*/
46fc3d4c 9104 case 'e': case 'E':
fc36a67e 9105 case 'f':
46fc3d4c 9106 case 'g': case 'G':
26372e71
GA
9107 if (vectorize)
9108 goto unknown;
46fc3d4c 9109
9110 /* This is evil, but floating point is even more evil */
9111
9e5b023a
JH
9112 /* for SV-style calling, we can only get NV
9113 for C-style calling, we assume %f is double;
9114 for simplicity we allow any of %Lf, %llf, %qf for long double
9115 */
9116 switch (intsize) {
9117 case 'V':
9118#if defined(USE_LONG_DOUBLE)
9119 intsize = 'q';
9120#endif
9121 break;
8a2e3f14 9122/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9123 case 'l':
5f66b61c 9124 /*FALLTHROUGH*/
9e5b023a
JH
9125 default:
9126#if defined(USE_LONG_DOUBLE)
9127 intsize = args ? 0 : 'q';
9128#endif
9129 break;
9130 case 'q':
9131#if defined(HAS_LONG_DOUBLE)
9132 break;
9133#else
5f66b61c 9134 /*FALLTHROUGH*/
9e5b023a
JH
9135#endif
9136 case 'h':
9e5b023a
JH
9137 goto unknown;
9138 }
9139
9140 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9141 nv = (args) ?
35fff930
JH
9142#if LONG_DOUBLESIZE > DOUBLESIZE
9143 intsize == 'q' ?
205f51d8
AS
9144 va_arg(*args, long double) :
9145 va_arg(*args, double)
35fff930 9146#else
205f51d8 9147 va_arg(*args, double)
35fff930 9148#endif
9e5b023a 9149 : SvNVx(argsv);
fc36a67e 9150
9151 need = 0;
9152 if (c != 'e' && c != 'E') {
9153 i = PERL_INT_MIN;
9e5b023a
JH
9154 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9155 will cast our (long double) to (double) */
73b309ea 9156 (void)Perl_frexp(nv, &i);
fc36a67e 9157 if (i == PERL_INT_MIN)
cea2e8a9 9158 Perl_die(aTHX_ "panic: frexp");
c635e13b 9159 if (i > 0)
fc36a67e 9160 need = BIT_DIGITS(i);
9161 }
9162 need += has_precis ? precis : 6; /* known default */
20f6aaab 9163
fc36a67e 9164 if (need < width)
9165 need = width;
9166
20f6aaab
AS
9167#ifdef HAS_LDBL_SPRINTF_BUG
9168 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9169 with sfio - Allen <allens@cpan.org> */
9170
9171# ifdef DBL_MAX
9172# define MY_DBL_MAX DBL_MAX
9173# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9174# if DOUBLESIZE >= 8
9175# define MY_DBL_MAX 1.7976931348623157E+308L
9176# else
9177# define MY_DBL_MAX 3.40282347E+38L
9178# endif
9179# endif
9180
9181# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9182# define MY_DBL_MAX_BUG 1L
20f6aaab 9183# else
205f51d8 9184# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9185# endif
20f6aaab 9186
205f51d8
AS
9187# ifdef DBL_MIN
9188# define MY_DBL_MIN DBL_MIN
9189# else /* XXX guessing! -Allen */
9190# if DOUBLESIZE >= 8
9191# define MY_DBL_MIN 2.2250738585072014E-308L
9192# else
9193# define MY_DBL_MIN 1.17549435E-38L
9194# endif
9195# endif
20f6aaab 9196
205f51d8
AS
9197 if ((intsize == 'q') && (c == 'f') &&
9198 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9199 (need < DBL_DIG)) {
9200 /* it's going to be short enough that
9201 * long double precision is not needed */
9202
9203 if ((nv <= 0L) && (nv >= -0L))
9204 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9205 else {
9206 /* would use Perl_fp_class as a double-check but not
9207 * functional on IRIX - see perl.h comments */
9208
9209 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9210 /* It's within the range that a double can represent */
9211#if defined(DBL_MAX) && !defined(DBL_MIN)
9212 if ((nv >= ((long double)1/DBL_MAX)) ||
9213 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9214#endif
205f51d8 9215 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9216 }
205f51d8
AS
9217 }
9218 if (fix_ldbl_sprintf_bug == TRUE) {
9219 double temp;
9220
9221 intsize = 0;
9222 temp = (double)nv;
9223 nv = (NV)temp;
9224 }
20f6aaab 9225 }
205f51d8
AS
9226
9227# undef MY_DBL_MAX
9228# undef MY_DBL_MAX_BUG
9229# undef MY_DBL_MIN
9230
20f6aaab
AS
9231#endif /* HAS_LDBL_SPRINTF_BUG */
9232
46fc3d4c 9233 need += 20; /* fudge factor */
80252599
GS
9234 if (PL_efloatsize < need) {
9235 Safefree(PL_efloatbuf);
9236 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9237 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9238 PL_efloatbuf[0] = '\0';
46fc3d4c 9239 }
9240
4151a5fe
IZ
9241 if ( !(width || left || plus || alt) && fill != '0'
9242 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9243 /* See earlier comment about buggy Gconvert when digits,
9244 aka precis is 0 */
9245 if ( c == 'g' && precis) {
2e59c212 9246 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9247 /* May return an empty string for digits==0 */
9248 if (*PL_efloatbuf) {
9249 elen = strlen(PL_efloatbuf);
4151a5fe 9250 goto float_converted;
4150c189 9251 }
4151a5fe
IZ
9252 } else if ( c == 'f' && !precis) {
9253 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9254 break;
9255 }
9256 }
4d84ee25
NC
9257 {
9258 char *ptr = ebuf + sizeof ebuf;
9259 *--ptr = '\0';
9260 *--ptr = c;
9261 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9262#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9263 if (intsize == 'q') {
9264 /* Copy the one or more characters in a long double
9265 * format before the 'base' ([efgEFG]) character to
9266 * the format string. */
9267 static char const prifldbl[] = PERL_PRIfldbl;
9268 char const *p = prifldbl + sizeof(prifldbl) - 3;
9269 while (p >= prifldbl) { *--ptr = *p--; }
9270 }
65202027 9271#endif
4d84ee25
NC
9272 if (has_precis) {
9273 base = precis;
9274 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9275 *--ptr = '.';
9276 }
9277 if (width) {
9278 base = width;
9279 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9280 }
9281 if (fill == '0')
9282 *--ptr = fill;
9283 if (left)
9284 *--ptr = '-';
9285 if (plus)
9286 *--ptr = plus;
9287 if (alt)
9288 *--ptr = '#';
9289 *--ptr = '%';
9290
9291 /* No taint. Otherwise we are in the strange situation
9292 * where printf() taints but print($float) doesn't.
9293 * --jhi */
9e5b023a 9294#if defined(HAS_LONG_DOUBLE)
4150c189 9295 elen = ((intsize == 'q')
d9fad198
JH
9296 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
9297 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
9e5b023a 9298#else
4150c189 9299 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9300#endif
4d84ee25 9301 }
4151a5fe 9302 float_converted:
80252599 9303 eptr = PL_efloatbuf;
46fc3d4c 9304 break;
9305
fc36a67e 9306 /* SPECIAL */
9307
9308 case 'n':
26372e71
GA
9309 if (vectorize)
9310 goto unknown;
fc36a67e 9311 i = SvCUR(sv) - origlen;
26372e71 9312 if (args) {
c635e13b 9313 switch (intsize) {
9314 case 'h': *(va_arg(*args, short*)) = i; break;
9315 default: *(va_arg(*args, int*)) = i; break;
9316 case 'l': *(va_arg(*args, long*)) = i; break;
9317 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9318#ifdef HAS_QUAD
9319 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9320#endif
c635e13b 9321 }
fc36a67e 9322 }
9dd79c3f 9323 else
211dfcf1 9324 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9325 continue; /* not "break" */
9326
9327 /* UNKNOWN */
9328
46fc3d4c 9329 default:
fc36a67e 9330 unknown:
041457d9
DM
9331 if (!args
9332 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9333 && ckWARN(WARN_PRINTF))
9334 {
c4420975 9335 SV * const msg = sv_newmortal();
35c1215d
NC
9336 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9337 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9338 if (c) {
0f4b6630 9339 if (isPRINT(c))
1c846c1f 9340 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9341 "\"%%%c\"", c & 0xFF);
9342 else
9343 Perl_sv_catpvf(aTHX_ msg,
57def98f 9344 "\"%%\\%03"UVof"\"",
0f4b6630 9345 (UV)c & 0xFF);
0f4b6630 9346 } else
396482e1 9347 sv_catpvs(msg, "end of string");
be2597df 9348 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
c635e13b 9349 }
fb73857a 9350
9351 /* output mangled stuff ... */
9352 if (c == '\0')
9353 --q;
46fc3d4c 9354 eptr = p;
9355 elen = q - p;
fb73857a 9356
9357 /* ... right here, because formatting flags should not apply */
9358 SvGROW(sv, SvCUR(sv) + elen + 1);
9359 p = SvEND(sv);
4459522c 9360 Copy(eptr, p, elen, char);
fb73857a 9361 p += elen;
9362 *p = '\0';
3f7c398e 9363 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9364 svix = osvix;
fb73857a 9365 continue; /* not "break" */
46fc3d4c 9366 }
9367
cc61b222
TS
9368 if (is_utf8 != has_utf8) {
9369 if (is_utf8) {
9370 if (SvCUR(sv))
9371 sv_utf8_upgrade(sv);
9372 }
9373 else {
9374 const STRLEN old_elen = elen;
9375 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
9376 sv_utf8_upgrade(nsv);
9377 eptr = SvPVX_const(nsv);
9378 elen = SvCUR(nsv);
9379
9380 if (width) { /* fudge width (can't fudge elen) */
9381 width += elen - old_elen;
9382 }
9383 is_utf8 = TRUE;
9384 }
9385 }
9386
6c94ec8b 9387 have = esignlen + zeros + elen;
ed2b91d2
GA
9388 if (have < zeros)
9389 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9390
46fc3d4c 9391 need = (have > width ? have : width);
9392 gap = need - have;
9393
d2641cbd
PC
9394 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9395 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9396 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9397 p = SvEND(sv);
9398 if (esignlen && fill == '0') {
53c1dcc0 9399 int i;
eb160463 9400 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9401 *p++ = esignbuf[i];
9402 }
9403 if (gap && !left) {
9404 memset(p, fill, gap);
9405 p += gap;
9406 }
9407 if (esignlen && fill != '0') {
53c1dcc0 9408 int i;
eb160463 9409 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9410 *p++ = esignbuf[i];
9411 }
fc36a67e 9412 if (zeros) {
53c1dcc0 9413 int i;
fc36a67e 9414 for (i = zeros; i; i--)
9415 *p++ = '0';
9416 }
46fc3d4c 9417 if (elen) {
4459522c 9418 Copy(eptr, p, elen, char);
46fc3d4c 9419 p += elen;
9420 }
9421 if (gap && left) {
9422 memset(p, ' ', gap);
9423 p += gap;
9424 }
b22c7a20
GS
9425 if (vectorize) {
9426 if (veclen) {
4459522c 9427 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9428 p += dotstrlen;
9429 }
9430 else
9431 vectorize = FALSE; /* done iterating over vecstr */
9432 }
2cf2cfc6
A
9433 if (is_utf8)
9434 has_utf8 = TRUE;
9435 if (has_utf8)
7e2040f0 9436 SvUTF8_on(sv);
46fc3d4c 9437 *p = '\0';
3f7c398e 9438 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9439 if (vectorize) {
9440 esignlen = 0;
9441 goto vector;
9442 }
46fc3d4c 9443 }
9444}
51371543 9445
645c22ef
DM
9446/* =========================================================================
9447
9448=head1 Cloning an interpreter
9449
9450All the macros and functions in this section are for the private use of
9451the main function, perl_clone().
9452
9453The foo_dup() functions make an exact copy of an existing foo thinngy.
9454During the course of a cloning, a hash table is used to map old addresses
9455to new addresses. The table is created and manipulated with the
9456ptr_table_* functions.
9457
9458=cut
9459
9460============================================================================*/
9461
9462
1d7c1841
GS
9463#if defined(USE_ITHREADS)
9464
d4c19fe8 9465/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9466#ifndef GpREFCNT_inc
9467# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9468#endif
9469
9470
a41cc44e 9471/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
3e07292d
NC
9472 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
9473 If this changes, please unmerge ss_dup. */
d2d73c3e 9474#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9475#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9476#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9477#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9478#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9479#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9480#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9481#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9482#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9483#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9484#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9485#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9486#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9487#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9488
199e78b7
DM
9489/* clone a parser */
9490
9491yy_parser *
9492Perl_parser_dup(pTHX_ const yy_parser *proto, CLONE_PARAMS* param)
9493{
9494 yy_parser *parser;
9495
9496 if (!proto)
9497 return NULL;
9498
7c197c94
DM
9499 /* look for it in the table first */
9500 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
9501 if (parser)
9502 return parser;
9503
9504 /* create anew and remember what it is */
199e78b7 9505 Newxz(parser, 1, yy_parser);
7c197c94 9506 ptr_table_store(PL_ptr_table, proto, parser);
199e78b7
DM
9507
9508 parser->yyerrstatus = 0;
9509 parser->yychar = YYEMPTY; /* Cause a token to be read. */
9510
9511 /* XXX these not yet duped */
9512 parser->old_parser = NULL;
9513 parser->stack = NULL;
9514 parser->ps = NULL;
9515 parser->stack_size = 0;
9516 /* XXX parser->stack->state = 0; */
9517
9518 /* XXX eventually, just Copy() most of the parser struct ? */
9519
9520 parser->lex_brackets = proto->lex_brackets;
9521 parser->lex_casemods = proto->lex_casemods;
9522 parser->lex_brackstack = savepvn(proto->lex_brackstack,
9523 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
9524 parser->lex_casestack = savepvn(proto->lex_casestack,
9525 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
9526 parser->lex_defer = proto->lex_defer;
9527 parser->lex_dojoin = proto->lex_dojoin;
9528 parser->lex_expect = proto->lex_expect;
9529 parser->lex_formbrack = proto->lex_formbrack;
9530 parser->lex_inpat = proto->lex_inpat;
9531 parser->lex_inwhat = proto->lex_inwhat;
9532 parser->lex_op = proto->lex_op;
9533 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
9534 parser->lex_starts = proto->lex_starts;
9535 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
9536 parser->multi_close = proto->multi_close;
9537 parser->multi_open = proto->multi_open;
9538 parser->multi_start = proto->multi_start;
9539 parser->pending_ident = proto->pending_ident;
9540 parser->preambled = proto->preambled;
9541 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
9542
9543#ifdef PERL_MAD
9544 parser->endwhite = proto->endwhite;
9545 parser->faketokens = proto->faketokens;
9546 parser->lasttoke = proto->lasttoke;
9547 parser->nextwhite = proto->nextwhite;
9548 parser->realtokenstart = proto->realtokenstart;
9549 parser->skipwhite = proto->skipwhite;
9550 parser->thisclose = proto->thisclose;
9551 parser->thismad = proto->thismad;
9552 parser->thisopen = proto->thisopen;
9553 parser->thisstuff = proto->thisstuff;
9554 parser->thistoken = proto->thistoken;
9555 parser->thiswhite = proto->thiswhite;
9556#endif
9557 return parser;
9558}
9559
d2d73c3e 9560
d2d73c3e 9561/* duplicate a file handle */
645c22ef 9562
1d7c1841 9563PerlIO *
a8fc9800 9564Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9565{
9566 PerlIO *ret;
53c1dcc0
AL
9567
9568 PERL_UNUSED_ARG(type);
73d840c0 9569
1d7c1841
GS
9570 if (!fp)
9571 return (PerlIO*)NULL;
9572
9573 /* look for it in the table first */
9574 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9575 if (ret)
9576 return ret;
9577
9578 /* create anew and remember what it is */
ecdeb87c 9579 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9580 ptr_table_store(PL_ptr_table, fp, ret);
9581 return ret;
9582}
9583
645c22ef
DM
9584/* duplicate a directory handle */
9585
1d7c1841
GS
9586DIR *
9587Perl_dirp_dup(pTHX_ DIR *dp)
9588{
96a5add6 9589 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9590 if (!dp)
9591 return (DIR*)NULL;
9592 /* XXX TODO */
9593 return dp;
9594}
9595
ff276b08 9596/* duplicate a typeglob */
645c22ef 9597
1d7c1841 9598GP *
a8fc9800 9599Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9600{
9601 GP *ret;
b37c2d43 9602
1d7c1841
GS
9603 if (!gp)
9604 return (GP*)NULL;
9605 /* look for it in the table first */
9606 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9607 if (ret)
9608 return ret;
9609
9610 /* create anew and remember what it is */
a02a5408 9611 Newxz(ret, 1, GP);
1d7c1841
GS
9612 ptr_table_store(PL_ptr_table, gp, ret);
9613
9614 /* clone */
9615 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9616 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9617 ret->gp_io = io_dup_inc(gp->gp_io, param);
9618 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9619 ret->gp_av = av_dup_inc(gp->gp_av, param);
9620 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9621 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9622 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9623 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841 9624 ret->gp_line = gp->gp_line;
f4890806 9625 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
1d7c1841
GS
9626 return ret;
9627}
9628
645c22ef
DM
9629/* duplicate a chain of magic */
9630
1d7c1841 9631MAGIC *
a8fc9800 9632Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9633{
cb359b41
JH
9634 MAGIC *mgprev = (MAGIC*)NULL;
9635 MAGIC *mgret;
1d7c1841
GS
9636 if (!mg)
9637 return (MAGIC*)NULL;
9638 /* look for it in the table first */
9639 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9640 if (mgret)
9641 return mgret;
9642
9643 for (; mg; mg = mg->mg_moremagic) {
9644 MAGIC *nmg;
a02a5408 9645 Newxz(nmg, 1, MAGIC);
cb359b41 9646 if (mgprev)
1d7c1841 9647 mgprev->mg_moremagic = nmg;
cb359b41
JH
9648 else
9649 mgret = nmg;
1d7c1841
GS
9650 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9651 nmg->mg_private = mg->mg_private;
9652 nmg->mg_type = mg->mg_type;
9653 nmg->mg_flags = mg->mg_flags;
14befaf4 9654 if (mg->mg_type == PERL_MAGIC_qr) {
f8149455 9655 nmg->mg_obj = (SV*)CALLREGDUPE((REGEXP*)mg->mg_obj, param);
1d7c1841 9656 }
05bd4103 9657 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9658 /* The backref AV has its reference count deliberately bumped by
9659 1. */
9660 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9661 }
8d2f4536
NC
9662 else if (mg->mg_type == PERL_MAGIC_symtab) {
9663 nmg->mg_obj = mg->mg_obj;
9664 }
1d7c1841
GS
9665 else {
9666 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9667 ? sv_dup_inc(mg->mg_obj, param)
9668 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9669 }
9670 nmg->mg_len = mg->mg_len;
9671 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9672 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9673 if (mg->mg_len > 0) {
1d7c1841 9674 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9675 if (mg->mg_type == PERL_MAGIC_overload_table &&
9676 AMT_AMAGIC((AMT*)mg->mg_ptr))
9677 {
c445ea15 9678 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9679 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9680 I32 i;
9681 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9682 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9683 }
9684 }
9685 }
9686 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9687 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9688 }
68795e93
NIS
9689 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9690 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9691 }
1d7c1841
GS
9692 mgprev = nmg;
9693 }
9694 return mgret;
9695}
9696
4674ade5
NC
9697#endif /* USE_ITHREADS */
9698
645c22ef
DM
9699/* create a new pointer-mapping table */
9700
1d7c1841
GS
9701PTR_TBL_t *
9702Perl_ptr_table_new(pTHX)
9703{
9704 PTR_TBL_t *tbl;
96a5add6
AL
9705 PERL_UNUSED_CONTEXT;
9706
a02a5408 9707 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9708 tbl->tbl_max = 511;
9709 tbl->tbl_items = 0;
a02a5408 9710 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9711 return tbl;
9712}
9713
7119fd33
NC
9714#define PTR_TABLE_HASH(ptr) \
9715 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9716
93e68bfb
JC
9717/*
9718 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9719 following define) and at call to new_body_inline made below in
9720 Perl_ptr_table_store()
9721 */
9722
9723#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9724
645c22ef
DM
9725/* map an existing pointer using a table */
9726
7bf61b54 9727STATIC PTR_TBL_ENT_t *
b0e6ae5b 9728S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9729 PTR_TBL_ENT_t *tblent;
4373e329 9730 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9731 assert(tbl);
9732 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9733 for (; tblent; tblent = tblent->next) {
9734 if (tblent->oldval == sv)
7bf61b54 9735 return tblent;
1d7c1841 9736 }
d4c19fe8 9737 return NULL;
7bf61b54
NC
9738}
9739
9740void *
9741Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9742{
b0e6ae5b 9743 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9744 PERL_UNUSED_CONTEXT;
d4c19fe8 9745 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9746}
9747
645c22ef
DM
9748/* add a new entry to a pointer-mapping table */
9749
1d7c1841 9750void
44f8325f 9751Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9752{
0c9fdfe0 9753 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9754 PERL_UNUSED_CONTEXT;
1d7c1841 9755
7bf61b54
NC
9756 if (tblent) {
9757 tblent->newval = newsv;
9758 } else {
9759 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9760
d2a0f284
JC
9761 new_body_inline(tblent, PTE_SVSLOT);
9762
7bf61b54
NC
9763 tblent->oldval = oldsv;
9764 tblent->newval = newsv;
9765 tblent->next = tbl->tbl_ary[entry];
9766 tbl->tbl_ary[entry] = tblent;
9767 tbl->tbl_items++;
9768 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9769 ptr_table_split(tbl);
1d7c1841 9770 }
1d7c1841
GS
9771}
9772
645c22ef
DM
9773/* double the hash bucket size of an existing ptr table */
9774
1d7c1841
GS
9775void
9776Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9777{
9778 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9779 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9780 UV newsize = oldsize * 2;
9781 UV i;
96a5add6 9782 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9783
9784 Renew(ary, newsize, PTR_TBL_ENT_t*);
9785 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9786 tbl->tbl_max = --newsize;
9787 tbl->tbl_ary = ary;
9788 for (i=0; i < oldsize; i++, ary++) {
9789 PTR_TBL_ENT_t **curentp, **entp, *ent;
9790 if (!*ary)
9791 continue;
9792 curentp = ary + oldsize;
9793 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9794 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9795 *entp = ent->next;
9796 ent->next = *curentp;
9797 *curentp = ent;
9798 continue;
9799 }
9800 else
9801 entp = &ent->next;
9802 }
9803 }
9804}
9805
645c22ef
DM
9806/* remove all the entries from a ptr table */
9807
a0739874
DM
9808void
9809Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9810{
d5cefff9 9811 if (tbl && tbl->tbl_items) {
c445ea15 9812 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9813 UV riter = tbl->tbl_max;
a0739874 9814
d5cefff9
NC
9815 do {
9816 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9817
d5cefff9 9818 while (entry) {
00b6aa41 9819 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9820 entry = entry->next;
9821 del_pte(oentry);
9822 }
9823 } while (riter--);
a0739874 9824
d5cefff9
NC
9825 tbl->tbl_items = 0;
9826 }
a0739874
DM
9827}
9828
645c22ef
DM
9829/* clear and free a ptr table */
9830
a0739874
DM
9831void
9832Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9833{
9834 if (!tbl) {
9835 return;
9836 }
9837 ptr_table_clear(tbl);
9838 Safefree(tbl->tbl_ary);
9839 Safefree(tbl);
9840}
9841
4674ade5 9842#if defined(USE_ITHREADS)
5bd07a3d 9843
83841fad 9844void
eb86f8b3 9845Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9846{
9847 if (SvROK(sstr)) {
b162af07
SP
9848 SvRV_set(dstr, SvWEAKREF(sstr)
9849 ? sv_dup(SvRV(sstr), param)
9850 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9851
83841fad 9852 }
3f7c398e 9853 else if (SvPVX_const(sstr)) {
83841fad
NIS
9854 /* Has something there */
9855 if (SvLEN(sstr)) {
68795e93 9856 /* Normal PV - clone whole allocated space */
3f7c398e 9857 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9858 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9859 /* Not that normal - actually sstr is copy on write.
9860 But we are a true, independant SV, so: */
9861 SvREADONLY_off(dstr);
9862 SvFAKE_off(dstr);
9863 }
68795e93 9864 }
83841fad
NIS
9865 else {
9866 /* Special case - not normally malloced for some reason */
f7877b28
NC
9867 if (isGV_with_GP(sstr)) {
9868 /* Don't need to do anything here. */
9869 }
9870 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9871 /* A "shared" PV - clone it as "shared" PV */
9872 SvPV_set(dstr,
9873 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9874 param)));
83841fad
NIS
9875 }
9876 else {
9877 /* Some other special case - random pointer */
f880fe2f 9878 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9879 }
83841fad
NIS
9880 }
9881 }
9882 else {
4608196e 9883 /* Copy the NULL */
f880fe2f 9884 if (SvTYPE(dstr) == SVt_RV)
b162af07 9885 SvRV_set(dstr, NULL);
f880fe2f 9886 else
6136c704 9887 SvPV_set(dstr, NULL);
83841fad
NIS
9888 }
9889}
9890
662fb8b2
NC
9891/* duplicate an SV of any type (including AV, HV etc) */
9892
1d7c1841 9893SV *
eb86f8b3 9894Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9895{
27da23d5 9896 dVAR;
1d7c1841
GS
9897 SV *dstr;
9898
9899 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9900 return NULL;
1d7c1841
GS
9901 /* look for it in the table first */
9902 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9903 if (dstr)
9904 return dstr;
9905
0405e91e
AB
9906 if(param->flags & CLONEf_JOIN_IN) {
9907 /** We are joining here so we don't want do clone
9908 something that is bad **/
eb86f8b3
AL
9909 if (SvTYPE(sstr) == SVt_PVHV) {
9910 const char * const hvname = HvNAME_get(sstr);
9911 if (hvname)
9912 /** don't clone stashes if they already exist **/
9913 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9914 }
9915 }
9916
1d7c1841
GS
9917 /* create anew and remember what it is */
9918 new_SV(dstr);
fd0854ff
DM
9919
9920#ifdef DEBUG_LEAKING_SCALARS
9921 dstr->sv_debug_optype = sstr->sv_debug_optype;
9922 dstr->sv_debug_line = sstr->sv_debug_line;
9923 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9924 dstr->sv_debug_cloned = 1;
fd0854ff 9925 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9926#endif
9927
1d7c1841
GS
9928 ptr_table_store(PL_ptr_table, sstr, dstr);
9929
9930 /* clone */
9931 SvFLAGS(dstr) = SvFLAGS(sstr);
9932 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9933 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9934
9935#ifdef DEBUGGING
3f7c398e 9936 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9937 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
6c9570dc 9938 (void*)PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9939#endif
9940
9660f481
DM
9941 /* don't clone objects whose class has asked us not to */
9942 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9943 SvFLAGS(dstr) &= ~SVTYPEMASK;
9944 SvOBJECT_off(dstr);
9945 return dstr;
9946 }
9947
1d7c1841
GS
9948 switch (SvTYPE(sstr)) {
9949 case SVt_NULL:
9950 SvANY(dstr) = NULL;
9951 break;
9952 case SVt_IV:
339049b0 9953 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9954 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9955 break;
9956 case SVt_NV:
9957 SvANY(dstr) = new_XNV();
9d6ce603 9958 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9959 break;
9960 case SVt_RV:
339049b0 9961 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9962 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9963 break;
cecf5685 9964 /* case SVt_BIND: */
662fb8b2
NC
9965 default:
9966 {
9967 /* These are all the types that need complex bodies allocating. */
662fb8b2 9968 void *new_body;
2bcc16b3
NC
9969 const svtype sv_type = SvTYPE(sstr);
9970 const struct body_details *const sv_type_details
9971 = bodies_by_type + sv_type;
662fb8b2 9972
93e68bfb 9973 switch (sv_type) {
662fb8b2 9974 default:
bb263b4e 9975 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9976 break;
9977
662fb8b2
NC
9978 case SVt_PVGV:
9979 if (GvUNIQUE((GV*)sstr)) {
6f207bd3 9980 NOOP; /* Do sharing here, and fall through */
662fb8b2 9981 }
c22188b4
NC
9982 case SVt_PVIO:
9983 case SVt_PVFM:
9984 case SVt_PVHV:
9985 case SVt_PVAV:
662fb8b2 9986 case SVt_PVCV:
662fb8b2 9987 case SVt_PVLV:
662fb8b2 9988 case SVt_PVMG:
662fb8b2 9989 case SVt_PVNV:
662fb8b2 9990 case SVt_PVIV:
662fb8b2 9991 case SVt_PV:
d2a0f284 9992 assert(sv_type_details->body_size);
c22188b4 9993 if (sv_type_details->arena) {
d2a0f284 9994 new_body_inline(new_body, sv_type);
c22188b4 9995 new_body
b9502f15 9996 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9997 } else {
9998 new_body = new_NOARENA(sv_type_details);
9999 }
1d7c1841 10000 }
662fb8b2
NC
10001 assert(new_body);
10002 SvANY(dstr) = new_body;
10003
2bcc16b3 10004#ifndef PURIFY
b9502f15
NC
10005 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10006 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 10007 sv_type_details->copy, char);
2bcc16b3
NC
10008#else
10009 Copy(((char*)SvANY(sstr)),
10010 ((char*)SvANY(dstr)),
d2a0f284 10011 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10012#endif
662fb8b2 10013
f7877b28
NC
10014 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10015 && !isGV_with_GP(dstr))
662fb8b2
NC
10016 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10017
10018 /* The Copy above means that all the source (unduplicated) pointers
10019 are now in the destination. We can check the flags and the
10020 pointers in either, but it's possible that there's less cache
10021 missing by always going for the destination.
10022 FIXME - instrument and check that assumption */
f32993d6 10023 if (sv_type >= SVt_PVMG) {
885ffcb3 10024 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
73d95100 10025 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
e736a858 10026 } else if (SvMAGIC(dstr))
662fb8b2
NC
10027 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10028 if (SvSTASH(dstr))
10029 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10030 }
662fb8b2 10031
f32993d6
NC
10032 /* The cast silences a GCC warning about unhandled types. */
10033 switch ((int)sv_type) {
662fb8b2
NC
10034 case SVt_PV:
10035 break;
10036 case SVt_PVIV:
10037 break;
10038 case SVt_PVNV:
10039 break;
10040 case SVt_PVMG:
10041 break;
662fb8b2
NC
10042 case SVt_PVLV:
10043 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10044 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10045 LvTARG(dstr) = dstr;
10046 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10047 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10048 else
10049 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
662fb8b2 10050 case SVt_PVGV:
cecf5685
NC
10051 if(isGV_with_GP(sstr)) {
10052 if (GvNAME_HEK(dstr))
10053 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
39cb70dc
NC
10054 /* Don't call sv_add_backref here as it's going to be
10055 created as part of the magic cloning of the symbol
10056 table. */
f7877b28
NC
10057 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10058 at the point of this comment. */
39cb70dc 10059 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
f7877b28
NC
10060 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10061 (void)GpREFCNT_inc(GvGP(dstr));
10062 } else
10063 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10064 break;
10065 case SVt_PVIO:
10066 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10067 if (IoOFP(dstr) == IoIFP(sstr))
10068 IoOFP(dstr) = IoIFP(dstr);
10069 else
10070 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10071 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10072 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10073 /* I have no idea why fake dirp (rsfps)
10074 should be treated differently but otherwise
10075 we end up with leaks -- sky*/
10076 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10077 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10078 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10079 } else {
10080 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10081 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10082 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10083 if (IoDIRP(dstr)) {
10084 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10085 } else {
6f207bd3 10086 NOOP;
100ce7e1
NC
10087 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10088 }
662fb8b2
NC
10089 }
10090 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10091 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10092 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10093 break;
10094 case SVt_PVAV:
10095 if (AvARRAY((AV*)sstr)) {
10096 SV **dst_ary, **src_ary;
10097 SSize_t items = AvFILLp((AV*)sstr) + 1;
10098
10099 src_ary = AvARRAY((AV*)sstr);
a02a5408 10100 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2 10101 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
9c6bc640 10102 AvARRAY((AV*)dstr) = dst_ary;
662fb8b2
NC
10103 AvALLOC((AV*)dstr) = dst_ary;
10104 if (AvREAL((AV*)sstr)) {
10105 while (items-- > 0)
10106 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10107 }
10108 else {
10109 while (items-- > 0)
10110 *dst_ary++ = sv_dup(*src_ary++, param);
10111 }
10112 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10113 while (items-- > 0) {
10114 *dst_ary++ = &PL_sv_undef;
10115 }
bfcb3514 10116 }
662fb8b2 10117 else {
9c6bc640 10118 AvARRAY((AV*)dstr) = NULL;
662fb8b2 10119 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10120 }
662fb8b2
NC
10121 break;
10122 case SVt_PVHV:
7e265ef3
AL
10123 if (HvARRAY((HV*)sstr)) {
10124 STRLEN i = 0;
10125 const bool sharekeys = !!HvSHAREKEYS(sstr);
10126 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10127 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10128 char *darray;
10129 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
10130 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10131 char);
10132 HvARRAY(dstr) = (HE**)darray;
10133 while (i <= sxhv->xhv_max) {
10134 const HE * const source = HvARRAY(sstr)[i];
10135 HvARRAY(dstr)[i] = source
10136 ? he_dup(source, sharekeys, param) : 0;
10137 ++i;
10138 }
10139 if (SvOOK(sstr)) {
10140 HEK *hvname;
10141 const struct xpvhv_aux * const saux = HvAUX(sstr);
10142 struct xpvhv_aux * const daux = HvAUX(dstr);
10143 /* This flag isn't copied. */
10144 /* SvOOK_on(hv) attacks the IV flags. */
10145 SvFLAGS(dstr) |= SVf_OOK;
10146
10147 hvname = saux->xhv_name;
10148 daux->xhv_name = hvname ? hek_dup(hvname, param) : hvname;
10149
10150 daux->xhv_riter = saux->xhv_riter;
10151 daux->xhv_eiter = saux->xhv_eiter
10152 ? he_dup(saux->xhv_eiter,
10153 (bool)!!HvSHAREKEYS(sstr), param) : 0;
10154 daux->xhv_backreferences =
10155 saux->xhv_backreferences
86f55936 10156 ? (AV*) SvREFCNT_inc(
7e265ef3 10157 sv_dup((SV*)saux->xhv_backreferences, param))
86f55936 10158 : 0;
7e265ef3
AL
10159 /* Record stashes for possible cloning in Perl_clone(). */
10160 if (hvname)
10161 av_push(param->stashes, dstr);
662fb8b2 10162 }
662fb8b2 10163 }
7e265ef3 10164 else
797c7171 10165 HvARRAY((HV*)dstr) = NULL;
662fb8b2 10166 break;
662fb8b2 10167 case SVt_PVCV:
bb172083
NC
10168 if (!(param->flags & CLONEf_COPY_STACKS)) {
10169 CvDEPTH(dstr) = 0;
10170 }
10171 case SVt_PVFM:
662fb8b2
NC
10172 /* NOTE: not refcounted */
10173 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10174 OP_REFCNT_LOCK;
d04ba589
NC
10175 if (!CvISXSUB(dstr))
10176 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10177 OP_REFCNT_UNLOCK;
cfae286e 10178 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10179 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10180 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10181 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10182 }
10183 /* don't dup if copying back - CvGV isn't refcounted, so the
10184 * duped GV may never be freed. A bit of a hack! DAPM */
10185 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10186 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10187 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10188 CvOUTSIDE(dstr) =
10189 CvWEAKOUTSIDE(sstr)
10190 ? cv_dup( CvOUTSIDE(dstr), param)
10191 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10192 if (!CvISXSUB(dstr))
662fb8b2
NC
10193 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10194 break;
bfcb3514 10195 }
1d7c1841 10196 }
1d7c1841
GS
10197 }
10198
10199 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10200 ++PL_sv_objcount;
10201
10202 return dstr;
d2d73c3e 10203 }
1d7c1841 10204
645c22ef
DM
10205/* duplicate a context */
10206
1d7c1841 10207PERL_CONTEXT *
a8fc9800 10208Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10209{
10210 PERL_CONTEXT *ncxs;
10211
10212 if (!cxs)
10213 return (PERL_CONTEXT*)NULL;
10214
10215 /* look for it in the table first */
10216 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10217 if (ncxs)
10218 return ncxs;
10219
10220 /* create anew and remember what it is */
a02a5408 10221 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10222 ptr_table_store(PL_ptr_table, cxs, ncxs);
10223
10224 while (ix >= 0) {
c445ea15
AL
10225 PERL_CONTEXT * const cx = &cxs[ix];
10226 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10227 ncx->cx_type = cx->cx_type;
10228 if (CxTYPE(cx) == CXt_SUBST) {
10229 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10230 }
10231 else {
10232 ncx->blk_oldsp = cx->blk_oldsp;
10233 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10234 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10235 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10236 ncx->blk_oldpm = cx->blk_oldpm;
10237 ncx->blk_gimme = cx->blk_gimme;
10238 switch (CxTYPE(cx)) {
10239 case CXt_SUB:
10240 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10241 ? cv_dup_inc(cx->blk_sub.cv, param)
10242 : cv_dup(cx->blk_sub.cv,param));
cc8d50a7 10243 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10244 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10245 : NULL);
d2d73c3e 10246 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841 10247 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
cc8d50a7
NC
10248 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10249 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10250 ncx->blk_sub.retop = cx->blk_sub.retop;
d8d97e70
DM
10251 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
10252 cx->blk_sub.oldcomppad);
1d7c1841
GS
10253 break;
10254 case CXt_EVAL:
10255 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10256 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10257 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10258 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10259 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10260 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10261 break;
10262 case CXt_LOOP:
10263 ncx->blk_loop.label = cx->blk_loop.label;
10264 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
022eaa24 10265 ncx->blk_loop.my_op = cx->blk_loop.my_op;
1d7c1841
GS
10266 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10267 ? cx->blk_loop.iterdata
d2d73c3e 10268 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10269 ncx->blk_loop.oldcomppad
10270 = (PAD*)ptr_table_fetch(PL_ptr_table,
10271 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10272 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10273 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10274 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10275 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10276 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10277 break;
10278 case CXt_FORMAT:
d2d73c3e
AB
10279 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10280 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10281 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
cc8d50a7 10282 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10283 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10284 break;
10285 case CXt_BLOCK:
10286 case CXt_NULL:
10287 break;
10288 }
10289 }
10290 --ix;
10291 }
10292 return ncxs;
10293}
10294
645c22ef
DM
10295/* duplicate a stack info structure */
10296
1d7c1841 10297PERL_SI *
a8fc9800 10298Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10299{
10300 PERL_SI *nsi;
10301
10302 if (!si)
10303 return (PERL_SI*)NULL;
10304
10305 /* look for it in the table first */
10306 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10307 if (nsi)
10308 return nsi;
10309
10310 /* create anew and remember what it is */
a02a5408 10311 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10312 ptr_table_store(PL_ptr_table, si, nsi);
10313
d2d73c3e 10314 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10315 nsi->si_cxix = si->si_cxix;
10316 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10317 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10318 nsi->si_type = si->si_type;
d2d73c3e
AB
10319 nsi->si_prev = si_dup(si->si_prev, param);
10320 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10321 nsi->si_markoff = si->si_markoff;
10322
10323 return nsi;
10324}
10325
10326#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10327#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10328#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10329#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10330#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10331#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10332#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10333#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10334#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10335#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10336#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10337#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10338#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10339#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10340
10341/* XXXXX todo */
10342#define pv_dup_inc(p) SAVEPV(p)
10343#define pv_dup(p) SAVEPV(p)
10344#define svp_dup_inc(p,pp) any_dup(p,pp)
10345
645c22ef
DM
10346/* map any object to the new equivent - either something in the
10347 * ptr table, or something in the interpreter structure
10348 */
10349
1d7c1841 10350void *
53c1dcc0 10351Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10352{
10353 void *ret;
10354
10355 if (!v)
10356 return (void*)NULL;
10357
10358 /* look for it in the table first */
10359 ret = ptr_table_fetch(PL_ptr_table, v);
10360 if (ret)
10361 return ret;
10362
10363 /* see if it is part of the interpreter structure */
10364 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10365 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10366 else {
1d7c1841 10367 ret = v;
05ec9bb3 10368 }
1d7c1841
GS
10369
10370 return ret;
10371}
10372
645c22ef
DM
10373/* duplicate the save stack */
10374
1d7c1841 10375ANY *
a8fc9800 10376Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10377{
53d44271 10378 dVAR;
53c1dcc0
AL
10379 ANY * const ss = proto_perl->Tsavestack;
10380 const I32 max = proto_perl->Tsavestack_max;
10381 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10382 ANY *nss;
10383 SV *sv;
10384 GV *gv;
10385 AV *av;
10386 HV *hv;
10387 void* ptr;
10388 int intval;
10389 long longval;
10390 GP *gp;
10391 IV iv;
b24356f5 10392 I32 i;
c4e33207 10393 char *c = NULL;
1d7c1841 10394 void (*dptr) (void*);
acfe0abc 10395 void (*dxptr) (pTHX_ void*);
1d7c1841 10396
a02a5408 10397 Newxz(nss, max, ANY);
1d7c1841
GS
10398
10399 while (ix > 0) {
b24356f5
NC
10400 const I32 type = POPINT(ss,ix);
10401 TOPINT(nss,ix) = type;
10402 switch (type) {
3e07292d
NC
10403 case SAVEt_HELEM: /* hash element */
10404 sv = (SV*)POPPTR(ss,ix);
10405 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
10406 /* fall through */
1d7c1841 10407 case SAVEt_ITEM: /* normal string */
a41cc44e 10408 case SAVEt_SV: /* scalar reference */
1d7c1841 10409 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10410 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10411 /* fall through */
10412 case SAVEt_FREESV:
10413 case SAVEt_MORTALIZESV:
1d7c1841 10414 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10415 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10416 break;
05ec9bb3
NIS
10417 case SAVEt_SHARED_PVREF: /* char* in shared space */
10418 c = (char*)POPPTR(ss,ix);
10419 TOPPTR(nss,ix) = savesharedpv(c);
10420 ptr = POPPTR(ss,ix);
10421 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10422 break;
1d7c1841
GS
10423 case SAVEt_GENERIC_SVREF: /* generic sv */
10424 case SAVEt_SVREF: /* scalar reference */
10425 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10426 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10427 ptr = POPPTR(ss,ix);
10428 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10429 break;
a41cc44e 10430 case SAVEt_HV: /* hash reference */
1d7c1841 10431 case SAVEt_AV: /* array reference */
11b79775 10432 sv = (SV*) POPPTR(ss,ix);
337d28f5 10433 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10434 /* fall through */
10435 case SAVEt_COMPPAD:
10436 case SAVEt_NSTAB:
667e2948 10437 sv = (SV*) POPPTR(ss,ix);
3e07292d 10438 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10439 break;
10440 case SAVEt_INT: /* int reference */
10441 ptr = POPPTR(ss,ix);
10442 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10443 intval = (int)POPINT(ss,ix);
10444 TOPINT(nss,ix) = intval;
10445 break;
10446 case SAVEt_LONG: /* long reference */
10447 ptr = POPPTR(ss,ix);
10448 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
3e07292d
NC
10449 /* fall through */
10450 case SAVEt_CLEARSV:
1d7c1841
GS
10451 longval = (long)POPLONG(ss,ix);
10452 TOPLONG(nss,ix) = longval;
10453 break;
10454 case SAVEt_I32: /* I32 reference */
10455 case SAVEt_I16: /* I16 reference */
10456 case SAVEt_I8: /* I8 reference */
88effcc9 10457 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10458 ptr = POPPTR(ss,ix);
10459 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
1ccabee8 10460 i = POPINT(ss,ix);
1d7c1841
GS
10461 TOPINT(nss,ix) = i;
10462 break;
10463 case SAVEt_IV: /* IV reference */
10464 ptr = POPPTR(ss,ix);
10465 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10466 iv = POPIV(ss,ix);
10467 TOPIV(nss,ix) = iv;
10468 break;
a41cc44e
NC
10469 case SAVEt_HPTR: /* HV* reference */
10470 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10471 case SAVEt_SPTR: /* SV* reference */
10472 ptr = POPPTR(ss,ix);
10473 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10474 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10475 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10476 break;
10477 case SAVEt_VPTR: /* random* reference */
10478 ptr = POPPTR(ss,ix);
10479 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10480 ptr = POPPTR(ss,ix);
10481 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10482 break;
b03d03b0 10483 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10484 case SAVEt_PPTR: /* char* reference */
10485 ptr = POPPTR(ss,ix);
10486 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10487 c = (char*)POPPTR(ss,ix);
10488 TOPPTR(nss,ix) = pv_dup(c);
10489 break;
1d7c1841
GS
10490 case SAVEt_GP: /* scalar reference */
10491 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10492 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10493 (void)GpREFCNT_inc(gp);
10494 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10495 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841 10496 break;
1d7c1841
GS
10497 case SAVEt_FREEOP:
10498 ptr = POPPTR(ss,ix);
10499 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10500 /* these are assumed to be refcounted properly */
53c1dcc0 10501 OP *o;
1d7c1841
GS
10502 switch (((OP*)ptr)->op_type) {
10503 case OP_LEAVESUB:
10504 case OP_LEAVESUBLV:
10505 case OP_LEAVEEVAL:
10506 case OP_LEAVE:
10507 case OP_SCOPE:
10508 case OP_LEAVEWRITE:
e977893f
GS
10509 TOPPTR(nss,ix) = ptr;
10510 o = (OP*)ptr;
d3c72c2a 10511 OP_REFCNT_LOCK;
e977893f 10512 OpREFCNT_inc(o);
d3c72c2a 10513 OP_REFCNT_UNLOCK;
1d7c1841
GS
10514 break;
10515 default:
5f66b61c 10516 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10517 break;
10518 }
10519 }
10520 else
5f66b61c 10521 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10522 break;
10523 case SAVEt_FREEPV:
10524 c = (char*)POPPTR(ss,ix);
10525 TOPPTR(nss,ix) = pv_dup_inc(c);
10526 break;
1d7c1841
GS
10527 case SAVEt_DELETE:
10528 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10529 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10530 c = (char*)POPPTR(ss,ix);
10531 TOPPTR(nss,ix) = pv_dup_inc(c);
3e07292d
NC
10532 /* fall through */
10533 case SAVEt_STACK_POS: /* Position on Perl stack */
1d7c1841
GS
10534 i = POPINT(ss,ix);
10535 TOPINT(nss,ix) = i;
10536 break;
10537 case SAVEt_DESTRUCTOR:
10538 ptr = POPPTR(ss,ix);
10539 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10540 dptr = POPDPTR(ss,ix);
8141890a
JH
10541 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10542 any_dup(FPTR2DPTR(void *, dptr),
10543 proto_perl));
1d7c1841
GS
10544 break;
10545 case SAVEt_DESTRUCTOR_X:
10546 ptr = POPPTR(ss,ix);
10547 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10548 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10549 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10550 any_dup(FPTR2DPTR(void *, dxptr),
10551 proto_perl));
1d7c1841
GS
10552 break;
10553 case SAVEt_REGCONTEXT:
10554 case SAVEt_ALLOC:
10555 i = POPINT(ss,ix);
10556 TOPINT(nss,ix) = i;
10557 ix -= i;
10558 break;
1d7c1841
GS
10559 case SAVEt_AELEM: /* array element */
10560 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10561 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10562 i = POPINT(ss,ix);
10563 TOPINT(nss,ix) = i;
10564 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10565 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841 10566 break;
1d7c1841
GS
10567 case SAVEt_OP:
10568 ptr = POPPTR(ss,ix);
10569 TOPPTR(nss,ix) = ptr;
10570 break;
10571 case SAVEt_HINTS:
10572 i = POPINT(ss,ix);
10573 TOPINT(nss,ix) = i;
b3ca2e83 10574 ptr = POPPTR(ss,ix);
080ac856 10575 if (ptr) {
7b6dd8c3 10576 HINTS_REFCNT_LOCK;
080ac856 10577 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10578 HINTS_REFCNT_UNLOCK;
10579 }
cbb1fbea 10580 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10581 if (i & HINT_LOCALIZE_HH) {
10582 hv = (HV*)POPPTR(ss,ix);
10583 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10584 }
1d7c1841 10585 break;
c3564e5c
GS
10586 case SAVEt_PADSV:
10587 longval = (long)POPLONG(ss,ix);
10588 TOPLONG(nss,ix) = longval;
10589 ptr = POPPTR(ss,ix);
10590 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10591 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10592 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10593 break;
a1bb4754 10594 case SAVEt_BOOL:
38d8b13e 10595 ptr = POPPTR(ss,ix);
b9609c01 10596 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10597 longval = (long)POPBOOL(ss,ix);
b9609c01 10598 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10599 break;
8bd2680e
MHM
10600 case SAVEt_SET_SVFLAGS:
10601 i = POPINT(ss,ix);
10602 TOPINT(nss,ix) = i;
10603 i = POPINT(ss,ix);
10604 TOPINT(nss,ix) = i;
10605 sv = (SV*)POPPTR(ss,ix);
10606 TOPPTR(nss,ix) = sv_dup(sv, param);
10607 break;
5bfb7d0e
NC
10608 case SAVEt_RE_STATE:
10609 {
10610 const struct re_save_state *const old_state
10611 = (struct re_save_state *)
10612 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10613 struct re_save_state *const new_state
10614 = (struct re_save_state *)
10615 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10616
10617 Copy(old_state, new_state, 1, struct re_save_state);
10618 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10619
10620 new_state->re_state_bostr
10621 = pv_dup(old_state->re_state_bostr);
10622 new_state->re_state_reginput
10623 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10624 new_state->re_state_regeol
10625 = pv_dup(old_state->re_state_regeol);
10626 new_state->re_state_regstartp
11b79775 10627 = (I32*) any_dup(old_state->re_state_regstartp, proto_perl);
5bfb7d0e 10628 new_state->re_state_regendp
11b79775 10629 = (I32*) any_dup(old_state->re_state_regendp, proto_perl);
5bfb7d0e 10630 new_state->re_state_reglastparen
11b79775
DD
10631 = (U32*) any_dup(old_state->re_state_reglastparen,
10632 proto_perl);
5bfb7d0e 10633 new_state->re_state_reglastcloseparen
11b79775 10634 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
5bfb7d0e 10635 proto_perl);
5bfb7d0e
NC
10636 /* XXX This just has to be broken. The old save_re_context
10637 code did SAVEGENERICPV(PL_reg_start_tmp);
10638 PL_reg_start_tmp is char **.
10639 Look above to what the dup code does for
10640 SAVEt_GENERIC_PVREF
10641 It can never have worked.
10642 So this is merely a faithful copy of the exiting bug: */
10643 new_state->re_state_reg_start_tmp
10644 = (char **) pv_dup((char *)
10645 old_state->re_state_reg_start_tmp);
10646 /* I assume that it only ever "worked" because no-one called
10647 (pseudo)fork while the regexp engine had re-entered itself.
10648 */
5bfb7d0e
NC
10649#ifdef PERL_OLD_COPY_ON_WRITE
10650 new_state->re_state_nrs
10651 = sv_dup(old_state->re_state_nrs, param);
10652#endif
10653 new_state->re_state_reg_magic
11b79775
DD
10654 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
10655 proto_perl);
5bfb7d0e 10656 new_state->re_state_reg_oldcurpm
11b79775
DD
10657 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
10658 proto_perl);
5bfb7d0e 10659 new_state->re_state_reg_curpm
11b79775
DD
10660 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
10661 proto_perl);
5bfb7d0e
NC
10662 new_state->re_state_reg_oldsaved
10663 = pv_dup(old_state->re_state_reg_oldsaved);
10664 new_state->re_state_reg_poscache
10665 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
10666 new_state->re_state_reg_starttry
10667 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
10668 break;
10669 }
68da3b2f
NC
10670 case SAVEt_COMPILE_WARNINGS:
10671 ptr = POPPTR(ss,ix);
10672 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10673 break;
7c197c94
DM
10674 case SAVEt_PARSER:
10675 ptr = POPPTR(ss,ix);
10676 TOPPTR(nss,ix) = parser_dup(ptr, param);
10677 break;
1d7c1841 10678 default:
147bc374
NC
10679 Perl_croak(aTHX_
10680 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
1d7c1841
GS
10681 }
10682 }
10683
bd81e77b
NC
10684 return nss;
10685}
10686
10687
10688/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10689 * flag to the result. This is done for each stash before cloning starts,
10690 * so we know which stashes want their objects cloned */
10691
10692static void
10693do_mark_cloneable_stash(pTHX_ SV *sv)
10694{
10695 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10696 if (hvname) {
10697 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10698 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10699 if (cloner && GvCV(cloner)) {
10700 dSP;
10701 UV status;
10702
10703 ENTER;
10704 SAVETMPS;
10705 PUSHMARK(SP);
10706 XPUSHs(sv_2mortal(newSVhek(hvname)));
10707 PUTBACK;
10708 call_sv((SV*)GvCV(cloner), G_SCALAR);
10709 SPAGAIN;
10710 status = POPu;
10711 PUTBACK;
10712 FREETMPS;
10713 LEAVE;
10714 if (status)
10715 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10716 }
10717 }
10718}
10719
10720
10721
10722/*
10723=for apidoc perl_clone
10724
10725Create and return a new interpreter by cloning the current one.
10726
10727perl_clone takes these flags as parameters:
10728
10729CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10730without it we only clone the data and zero the stacks,
10731with it we copy the stacks and the new perl interpreter is
10732ready to run at the exact same point as the previous one.
10733The pseudo-fork code uses COPY_STACKS while the
10734threads->new doesn't.
10735
10736CLONEf_KEEP_PTR_TABLE
10737perl_clone keeps a ptr_table with the pointer of the old
10738variable as a key and the new variable as a value,
10739this allows it to check if something has been cloned and not
10740clone it again but rather just use the value and increase the
10741refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10742the ptr_table using the function
10743C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10744reason to keep it around is if you want to dup some of your own
10745variable who are outside the graph perl scans, example of this
10746code is in threads.xs create
10747
10748CLONEf_CLONE_HOST
10749This is a win32 thing, it is ignored on unix, it tells perls
10750win32host code (which is c++) to clone itself, this is needed on
10751win32 if you want to run two threads at the same time,
10752if you just want to do some stuff in a separate perl interpreter
10753and then throw it away and return to the original one,
10754you don't need to do anything.
10755
10756=cut
10757*/
10758
10759/* XXX the above needs expanding by someone who actually understands it ! */
10760EXTERN_C PerlInterpreter *
10761perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10762
10763PerlInterpreter *
10764perl_clone(PerlInterpreter *proto_perl, UV flags)
10765{
10766 dVAR;
10767#ifdef PERL_IMPLICIT_SYS
10768
10769 /* perlhost.h so we need to call into it
10770 to clone the host, CPerlHost should have a c interface, sky */
10771
10772 if (flags & CLONEf_CLONE_HOST) {
10773 return perl_clone_host(proto_perl,flags);
10774 }
10775 return perl_clone_using(proto_perl, flags,
10776 proto_perl->IMem,
10777 proto_perl->IMemShared,
10778 proto_perl->IMemParse,
10779 proto_perl->IEnv,
10780 proto_perl->IStdIO,
10781 proto_perl->ILIO,
10782 proto_perl->IDir,
10783 proto_perl->ISock,
10784 proto_perl->IProc);
10785}
10786
10787PerlInterpreter *
10788perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10789 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10790 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10791 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10792 struct IPerlDir* ipD, struct IPerlSock* ipS,
10793 struct IPerlProc* ipP)
10794{
10795 /* XXX many of the string copies here can be optimized if they're
10796 * constants; they need to be allocated as common memory and just
10797 * their pointers copied. */
10798
10799 IV i;
10800 CLONE_PARAMS clone_params;
5f66b61c 10801 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10802
5f66b61c 10803 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10804 /* for each stash, determine whether its objects should be cloned */
10805 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10806 PERL_SET_THX(my_perl);
10807
10808# ifdef DEBUGGING
7e337ee0 10809 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10810 PL_op = NULL;
10811 PL_curcop = NULL;
bd81e77b
NC
10812 PL_markstack = 0;
10813 PL_scopestack = 0;
10814 PL_savestack = 0;
10815 PL_savestack_ix = 0;
10816 PL_savestack_max = -1;
10817 PL_sig_pending = 0;
10818 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10819# else /* !DEBUGGING */
10820 Zero(my_perl, 1, PerlInterpreter);
10821# endif /* DEBUGGING */
10822
10823 /* host pointers */
10824 PL_Mem = ipM;
10825 PL_MemShared = ipMS;
10826 PL_MemParse = ipMP;
10827 PL_Env = ipE;
10828 PL_StdIO = ipStd;
10829 PL_LIO = ipLIO;
10830 PL_Dir = ipD;
10831 PL_Sock = ipS;
10832 PL_Proc = ipP;
10833#else /* !PERL_IMPLICIT_SYS */
10834 IV i;
10835 CLONE_PARAMS clone_params;
10836 CLONE_PARAMS* param = &clone_params;
5f66b61c 10837 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10838 /* for each stash, determine whether its objects should be cloned */
10839 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10840 PERL_SET_THX(my_perl);
10841
10842# ifdef DEBUGGING
7e337ee0 10843 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10844 PL_op = NULL;
10845 PL_curcop = NULL;
bd81e77b
NC
10846 PL_markstack = 0;
10847 PL_scopestack = 0;
10848 PL_savestack = 0;
10849 PL_savestack_ix = 0;
10850 PL_savestack_max = -1;
10851 PL_sig_pending = 0;
10852 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10853# else /* !DEBUGGING */
10854 Zero(my_perl, 1, PerlInterpreter);
10855# endif /* DEBUGGING */
10856#endif /* PERL_IMPLICIT_SYS */
10857 param->flags = flags;
10858 param->proto_perl = proto_perl;
10859
7cb608b5
NC
10860 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10861
fdda85ca 10862 PL_body_arenas = NULL;
bd81e77b
NC
10863 Zero(&PL_body_roots, 1, PL_body_roots);
10864
10865 PL_nice_chunk = NULL;
10866 PL_nice_chunk_size = 0;
10867 PL_sv_count = 0;
10868 PL_sv_objcount = 0;
a0714e2c
SS
10869 PL_sv_root = NULL;
10870 PL_sv_arenaroot = NULL;
bd81e77b
NC
10871
10872 PL_debug = proto_perl->Idebug;
10873
10874 PL_hash_seed = proto_perl->Ihash_seed;
10875 PL_rehash_seed = proto_perl->Irehash_seed;
10876
10877#ifdef USE_REENTRANT_API
10878 /* XXX: things like -Dm will segfault here in perlio, but doing
10879 * PERL_SET_CONTEXT(proto_perl);
10880 * breaks too many other things
10881 */
10882 Perl_reentrant_init(aTHX);
10883#endif
10884
10885 /* create SV map for pointer relocation */
10886 PL_ptr_table = ptr_table_new();
10887
10888 /* initialize these special pointers as early as possible */
10889 SvANY(&PL_sv_undef) = NULL;
10890 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10891 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10892 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10893
10894 SvANY(&PL_sv_no) = new_XPVNV();
10895 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10896 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10897 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10898 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
10899 SvCUR_set(&PL_sv_no, 0);
10900 SvLEN_set(&PL_sv_no, 1);
10901 SvIV_set(&PL_sv_no, 0);
10902 SvNV_set(&PL_sv_no, 0);
10903 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10904
10905 SvANY(&PL_sv_yes) = new_XPVNV();
10906 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10907 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10908 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10909 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
10910 SvCUR_set(&PL_sv_yes, 1);
10911 SvLEN_set(&PL_sv_yes, 2);
10912 SvIV_set(&PL_sv_yes, 1);
10913 SvNV_set(&PL_sv_yes, 1);
10914 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10915
10916 /* create (a non-shared!) shared string table */
10917 PL_strtab = newHV();
10918 HvSHAREKEYS_off(PL_strtab);
10919 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10920 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10921
10922 PL_compiling = proto_perl->Icompiling;
10923
10924 /* These two PVs will be free'd special way so must set them same way op.c does */
10925 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10926 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10927
10928 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10929 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10930
10931 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10932 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
c28fe1ec 10933 if (PL_compiling.cop_hints_hash) {
cbb1fbea 10934 HINTS_REFCNT_LOCK;
c28fe1ec 10935 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
cbb1fbea
NC
10936 HINTS_REFCNT_UNLOCK;
10937 }
bd81e77b
NC
10938 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10939
10940 /* pseudo environmental stuff */
10941 PL_origargc = proto_perl->Iorigargc;
10942 PL_origargv = proto_perl->Iorigargv;
10943
10944 param->stashes = newAV(); /* Setup array of objects to call clone on */
10945
10946 /* Set tainting stuff before PerlIO_debug can possibly get called */
10947 PL_tainting = proto_perl->Itainting;
10948 PL_taint_warn = proto_perl->Itaint_warn;
10949
10950#ifdef PERLIO_LAYERS
10951 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10952 PerlIO_clone(aTHX_ proto_perl, param);
10953#endif
10954
10955 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10956 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10957 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10958 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10959 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10960 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10961
10962 /* switches */
10963 PL_minus_c = proto_perl->Iminus_c;
10964 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10965 PL_localpatches = proto_perl->Ilocalpatches;
10966 PL_splitstr = proto_perl->Isplitstr;
10967 PL_preprocess = proto_perl->Ipreprocess;
10968 PL_minus_n = proto_perl->Iminus_n;
10969 PL_minus_p = proto_perl->Iminus_p;
10970 PL_minus_l = proto_perl->Iminus_l;
10971 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10972 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10973 PL_minus_F = proto_perl->Iminus_F;
10974 PL_doswitches = proto_perl->Idoswitches;
10975 PL_dowarn = proto_perl->Idowarn;
10976 PL_doextract = proto_perl->Idoextract;
10977 PL_sawampersand = proto_perl->Isawampersand;
10978 PL_unsafe = proto_perl->Iunsafe;
10979 PL_inplace = SAVEPV(proto_perl->Iinplace);
10980 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10981 PL_perldb = proto_perl->Iperldb;
10982 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10983 PL_exit_flags = proto_perl->Iexit_flags;
10984
10985 /* magical thingies */
10986 /* XXX time(&PL_basetime) when asked for? */
10987 PL_basetime = proto_perl->Ibasetime;
10988 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
10989
10990 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
10991 PL_statusvalue = proto_perl->Istatusvalue;
10992#ifdef VMS
10993 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
10994#else
10995 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
10996#endif
10997 PL_encoding = sv_dup(proto_perl->Iencoding, param);
10998
10999 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
11000 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
11001 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
11002
84da74a7 11003
f9f4320a 11004 /* RE engine related */
84da74a7
YO
11005 Zero(&PL_reg_state, 1, struct re_save_state);
11006 PL_reginterp_cnt = 0;
11007 PL_regmatch_slab = NULL;
11008
bd81e77b
NC
11009 /* Clone the regex array */
11010 PL_regex_padav = newAV();
11011 {
11012 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 11013 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 11014 IV i;
7f466ec7 11015 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 11016 for(i = 1; i <= len; i++) {
7a5b473e
AL
11017 const SV * const regex = regexen[i];
11018 SV * const sv =
11019 SvREPADTMP(regex)
11020 ? sv_dup_inc(regex, param)
11021 : SvREFCNT_inc(
f8149455 11022 newSViv(PTR2IV(CALLREGDUPE(
7a5b473e
AL
11023 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11024 ;
60790534
DM
11025 if (SvFLAGS(regex) & SVf_BREAK)
11026 SvFLAGS(sv) |= SVf_BREAK; /* unrefcnted PL_curpm */
7a5b473e 11027 av_push(PL_regex_padav, sv);
bd81e77b
NC
11028 }
11029 }
11030 PL_regex_pad = AvARRAY(PL_regex_padav);
11031
11032 /* shortcuts to various I/O objects */
11033 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11034 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11035 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11036 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11037 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11038 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11039
bd81e77b
NC
11040 /* shortcuts to regexp stuff */
11041 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11042
bd81e77b
NC
11043 /* shortcuts to misc objects */
11044 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11045
bd81e77b
NC
11046 /* shortcuts to debugging objects */
11047 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11048 PL_DBline = gv_dup(proto_perl->IDBline, param);
11049 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11050 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11051 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11052 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11053 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11054 PL_lineary = av_dup(proto_perl->Ilineary, param);
11055 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11056
bd81e77b
NC
11057 /* symbol tables */
11058 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11059 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11060 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11061 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11062 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11063
11064 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11065 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11066 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
3c10abe3
AG
11067 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
11068 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
bd81e77b
NC
11069 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11070 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11071 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11072
11073 PL_sub_generation = proto_perl->Isub_generation;
11074
11075 /* funky return mechanisms */
11076 PL_forkprocess = proto_perl->Iforkprocess;
11077
11078 /* subprocess state */
11079 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11080
11081 /* internal state */
11082 PL_maxo = proto_perl->Imaxo;
11083 if (proto_perl->Iop_mask)
11084 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11085 else
bd61b366 11086 PL_op_mask = NULL;
bd81e77b
NC
11087 /* PL_asserting = proto_perl->Iasserting; */
11088
11089 /* current interpreter roots */
11090 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
d3c72c2a 11091 OP_REFCNT_LOCK;
bd81e77b 11092 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
d3c72c2a 11093 OP_REFCNT_UNLOCK;
bd81e77b
NC
11094 PL_main_start = proto_perl->Imain_start;
11095 PL_eval_root = proto_perl->Ieval_root;
11096 PL_eval_start = proto_perl->Ieval_start;
11097
11098 /* runtime control stuff */
11099 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11100 PL_copline = proto_perl->Icopline;
11101
11102 PL_filemode = proto_perl->Ifilemode;
11103 PL_lastfd = proto_perl->Ilastfd;
11104 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11105 PL_Argv = NULL;
bd61b366 11106 PL_Cmd = NULL;
bd81e77b 11107 PL_gensym = proto_perl->Igensym;
bd81e77b
NC
11108 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11109 PL_laststatval = proto_perl->Ilaststatval;
11110 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11111 PL_mess_sv = NULL;
bd81e77b
NC
11112
11113 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11114
11115 /* interpreter atexit processing */
11116 PL_exitlistlen = proto_perl->Iexitlistlen;
11117 if (PL_exitlistlen) {
11118 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11119 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11120 }
bd81e77b
NC
11121 else
11122 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11123
11124 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11125 if (PL_my_cxt_size) {
f16dd614
DM
11126 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11127 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
53d44271
JH
11128#ifdef PERL_GLOBAL_STRUCT_PRIVATE
11129 Newx(PL_my_cxt_keys, PL_my_cxt_size, char *);
11130 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
11131#endif
f16dd614 11132 }
53d44271 11133 else {
f16dd614 11134 PL_my_cxt_list = (void**)NULL;
53d44271
JH
11135#ifdef PERL_GLOBAL_STRUCT_PRIVATE
11136 PL_my_cxt_keys = (void**)NULL;
11137#endif
11138 }
bd81e77b
NC
11139 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11140 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11141 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11142
11143 PL_profiledata = NULL;
11144 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11145 /* PL_rsfp_filters entries have fake IoDIRP() */
11146 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11147
bd81e77b 11148 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11149
bd81e77b 11150 PAD_CLONE_VARS(proto_perl, param);
9660f481 11151
bd81e77b
NC
11152#ifdef HAVE_INTERP_INTERN
11153 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11154#endif
645c22ef 11155
bd81e77b
NC
11156 /* more statics moved here */
11157 PL_generation = proto_perl->Igeneration;
11158 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11159
bd81e77b
NC
11160 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11161 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11162
bd81e77b
NC
11163 PL_uid = proto_perl->Iuid;
11164 PL_euid = proto_perl->Ieuid;
11165 PL_gid = proto_perl->Igid;
11166 PL_egid = proto_perl->Iegid;
11167 PL_nomemok = proto_perl->Inomemok;
11168 PL_an = proto_perl->Ian;
11169 PL_evalseq = proto_perl->Ievalseq;
11170 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11171 PL_origalen = proto_perl->Iorigalen;
11172#ifdef PERL_USES_PL_PIDSTATUS
11173 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11174#endif
11175 PL_osname = SAVEPV(proto_perl->Iosname);
11176 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11177
bd81e77b 11178 PL_runops = proto_perl->Irunops;
6a78b4db 11179
bd81e77b 11180 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11181
bd81e77b
NC
11182#ifdef CSH
11183 PL_cshlen = proto_perl->Icshlen;
11184 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11185#endif
645c22ef 11186
199e78b7
DM
11187 PL_parser = parser_dup(proto_perl->Iparser, param);
11188
bd81e77b 11189 PL_lex_state = proto_perl->Ilex_state;
645c22ef 11190
5db06880
NC
11191#ifdef PERL_MAD
11192 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
5336380d 11193 PL_curforce = proto_perl->Icurforce;
5db06880 11194#else
bd81e77b
NC
11195 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11196 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11197 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11198#endif
c43294b8 11199
db4997f0
NC
11200 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11201 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11202 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11203 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11204 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11205 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11206 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11207 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11208 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
bd81e77b 11209 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
bd81e77b
NC
11210
11211 PL_expect = proto_perl->Iexpect;
11212
bd81e77b 11213 PL_multi_end = proto_perl->Imulti_end;
bd81e77b
NC
11214
11215 PL_error_count = proto_perl->Ierror_count;
11216 PL_subline = proto_perl->Isubline;
11217 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11218
db4997f0
NC
11219 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11220 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11221 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11222 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11223 PL_last_lop_op = proto_perl->Ilast_lop_op;
bd81e77b
NC
11224 PL_in_my = proto_perl->Iin_my;
11225 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11226#ifdef FCRYPT
11227 PL_cryptseen = proto_perl->Icryptseen;
11228#endif
1d7c1841 11229
bd81e77b 11230 PL_hints = proto_perl->Ihints;
1d7c1841 11231
bd81e77b 11232 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11233
bd81e77b
NC
11234#ifdef USE_LOCALE_COLLATE
11235 PL_collation_ix = proto_perl->Icollation_ix;
11236 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11237 PL_collation_standard = proto_perl->Icollation_standard;
11238 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11239 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11240#endif /* USE_LOCALE_COLLATE */
1d7c1841 11241
bd81e77b
NC
11242#ifdef USE_LOCALE_NUMERIC
11243 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11244 PL_numeric_standard = proto_perl->Inumeric_standard;
11245 PL_numeric_local = proto_perl->Inumeric_local;
11246 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11247#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11248
bd81e77b
NC
11249 /* utf8 character classes */
11250 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11251 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11252 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11253 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11254 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11255 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11256 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11257 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11258 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11259 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11260 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11261 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11262 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11263 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11264 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11265 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11266 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11267 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11268 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11269 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11270
bd81e77b
NC
11271 /* Did the locale setup indicate UTF-8? */
11272 PL_utf8locale = proto_perl->Iutf8locale;
11273 /* Unicode features (see perlrun/-C) */
11274 PL_unicode = proto_perl->Iunicode;
1d7c1841 11275
bd81e77b
NC
11276 /* Pre-5.8 signals control */
11277 PL_signals = proto_perl->Isignals;
1d7c1841 11278
bd81e77b
NC
11279 /* times() ticks per second */
11280 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11281
bd81e77b
NC
11282 /* Recursion stopper for PerlIO_find_layer */
11283 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11284
bd81e77b
NC
11285 /* sort() routine */
11286 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11287
bd81e77b
NC
11288 /* Not really needed/useful since the reenrant_retint is "volatile",
11289 * but do it for consistency's sake. */
11290 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11291
bd81e77b
NC
11292 /* Hooks to shared SVs and locks. */
11293 PL_sharehook = proto_perl->Isharehook;
11294 PL_lockhook = proto_perl->Ilockhook;
11295 PL_unlockhook = proto_perl->Iunlockhook;
11296 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11297
bd81e77b
NC
11298 PL_runops_std = proto_perl->Irunops_std;
11299 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11300
bd81e77b
NC
11301#ifdef THREADS_HAVE_PIDS
11302 PL_ppid = proto_perl->Ippid;
11303#endif
1d7c1841 11304
bd81e77b 11305 /* swatch cache */
5c284bb0 11306 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11307 PL_last_swash_klen = 0;
11308 PL_last_swash_key[0]= '\0';
11309 PL_last_swash_tmps = (U8*)NULL;
11310 PL_last_swash_slen = 0;
1d7c1841 11311
bd81e77b
NC
11312 PL_glob_index = proto_perl->Iglob_index;
11313 PL_srand_called = proto_perl->Isrand_called;
11b79775 11314 PL_uudmap[(U32) 'M'] = 0; /* reinits on demand */
bd61b366 11315 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11316
bd81e77b
NC
11317 if (proto_perl->Ipsig_pend) {
11318 Newxz(PL_psig_pend, SIG_SIZE, int);
11319 }
11320 else {
11321 PL_psig_pend = (int*)NULL;
11322 }
05ec9bb3 11323
bd81e77b
NC
11324 if (proto_perl->Ipsig_ptr) {
11325 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11326 Newxz(PL_psig_name, SIG_SIZE, SV*);
11327 for (i = 1; i < SIG_SIZE; i++) {
11328 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11329 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11330 }
11331 }
11332 else {
11333 PL_psig_ptr = (SV**)NULL;
11334 PL_psig_name = (SV**)NULL;
11335 }
05ec9bb3 11336
bd81e77b 11337 /* thrdvar.h stuff */
1d7c1841 11338
bd81e77b
NC
11339 if (flags & CLONEf_COPY_STACKS) {
11340 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11341 PL_tmps_ix = proto_perl->Ttmps_ix;
11342 PL_tmps_max = proto_perl->Ttmps_max;
11343 PL_tmps_floor = proto_perl->Ttmps_floor;
11344 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11345 i = 0;
11346 while (i <= PL_tmps_ix) {
11347 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11348 ++i;
11349 }
d2d73c3e 11350
bd81e77b
NC
11351 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11352 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11353 Newxz(PL_markstack, i, I32);
11354 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11355 - proto_perl->Tmarkstack);
11356 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11357 - proto_perl->Tmarkstack);
11358 Copy(proto_perl->Tmarkstack, PL_markstack,
11359 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11360
bd81e77b
NC
11361 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11362 * NOTE: unlike the others! */
11363 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11364 PL_scopestack_max = proto_perl->Tscopestack_max;
11365 Newxz(PL_scopestack, PL_scopestack_max, I32);
11366 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11367
bd81e77b
NC
11368 /* NOTE: si_dup() looks at PL_markstack */
11369 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11370
bd81e77b
NC
11371 /* PL_curstack = PL_curstackinfo->si_stack; */
11372 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11373 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11374
bd81e77b
NC
11375 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11376 PL_stack_base = AvARRAY(PL_curstack);
11377 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11378 - proto_perl->Tstack_base);
11379 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11380
bd81e77b
NC
11381 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11382 * NOTE: unlike the others! */
11383 PL_savestack_ix = proto_perl->Tsavestack_ix;
11384 PL_savestack_max = proto_perl->Tsavestack_max;
11385 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11386 PL_savestack = ss_dup(proto_perl, param);
11387 }
11388 else {
11389 init_stacks();
11390 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11391
11392 /* although we're not duplicating the tmps stack, we should still
11393 * add entries for any SVs on the tmps stack that got cloned by a
11394 * non-refcount means (eg a temp in @_); otherwise they will be
11395 * orphaned
11396 */
11397 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11398 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11399 proto_perl->Ttmps_stack[i]);
11400 if (nsv && !SvREFCNT(nsv)) {
11401 EXTEND_MORTAL(1);
b37c2d43 11402 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11403 }
11404 }
bd81e77b 11405 }
1d7c1841 11406
bd81e77b
NC
11407 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11408 PL_top_env = &PL_start_env;
1d7c1841 11409
bd81e77b 11410 PL_op = proto_perl->Top;
4a4c6fe3 11411
a0714e2c 11412 PL_Sv = NULL;
bd81e77b
NC
11413 PL_Xpv = (XPV*)NULL;
11414 PL_na = proto_perl->Tna;
1fcf4c12 11415
bd81e77b
NC
11416 PL_statbuf = proto_perl->Tstatbuf;
11417 PL_statcache = proto_perl->Tstatcache;
11418 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11419 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11420#ifdef HAS_TIMES
11421 PL_timesbuf = proto_perl->Ttimesbuf;
11422#endif
1d7c1841 11423
bd81e77b
NC
11424 PL_tainted = proto_perl->Ttainted;
11425 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11426 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11427 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11428 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11429 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11430 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11431 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11432 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11433 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11434
bd81e77b
NC
11435 PL_restartop = proto_perl->Trestartop;
11436 PL_in_eval = proto_perl->Tin_eval;
11437 PL_delaymagic = proto_perl->Tdelaymagic;
11438 PL_dirty = proto_perl->Tdirty;
11439 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11440
bd81e77b 11441 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11442 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11443 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11444 PL_lastgotoprobe = NULL;
bd81e77b 11445 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11446
bd81e77b
NC
11447 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11448 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11449 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11450 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11451 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11452 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11453
bd81e77b 11454 /* regex stuff */
1d7c1841 11455
bd81e77b
NC
11456 PL_screamfirst = NULL;
11457 PL_screamnext = NULL;
11458 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11459 PL_lastscream = NULL;
1d7c1841 11460
bd81e77b 11461 PL_watchaddr = NULL;
bd61b366 11462 PL_watchok = NULL;
1d7c1841 11463
bd81e77b 11464 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11465 PL_colorset = 0; /* reinits PL_colors[] */
11466 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11467
84da74a7 11468
1d7c1841 11469
bd81e77b
NC
11470 /* Pluggable optimizer */
11471 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11472
bd81e77b 11473 PL_stashcache = newHV();
1d7c1841 11474
bd81e77b
NC
11475 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11476 ptr_table_free(PL_ptr_table);
11477 PL_ptr_table = NULL;
11478 }
1d7c1841 11479
bd81e77b
NC
11480 /* Call the ->CLONE method, if it exists, for each of the stashes
11481 identified by sv_dup() above.
11482 */
11483 while(av_len(param->stashes) != -1) {
11484 HV* const stash = (HV*) av_shift(param->stashes);
11485 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11486 if (cloner && GvCV(cloner)) {
11487 dSP;
11488 ENTER;
11489 SAVETMPS;
11490 PUSHMARK(SP);
11491 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11492 PUTBACK;
11493 call_sv((SV*)GvCV(cloner), G_DISCARD);
11494 FREETMPS;
11495 LEAVE;
11496 }
1d7c1841 11497 }
1d7c1841 11498
bd81e77b 11499 SvREFCNT_dec(param->stashes);
1d7c1841 11500
bd81e77b
NC
11501 /* orphaned? eg threads->new inside BEGIN or use */
11502 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11503 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11504 SAVEFREESV(PL_compcv);
11505 }
dd2155a4 11506
bd81e77b
NC
11507 return my_perl;
11508}
1d7c1841 11509
bd81e77b 11510#endif /* USE_ITHREADS */
1d7c1841 11511
bd81e77b
NC
11512/*
11513=head1 Unicode Support
1d7c1841 11514
bd81e77b 11515=for apidoc sv_recode_to_utf8
1d7c1841 11516
bd81e77b
NC
11517The encoding is assumed to be an Encode object, on entry the PV
11518of the sv is assumed to be octets in that encoding, and the sv
11519will be converted into Unicode (and UTF-8).
1d7c1841 11520
bd81e77b
NC
11521If the sv already is UTF-8 (or if it is not POK), or if the encoding
11522is not a reference, nothing is done to the sv. If the encoding is not
11523an C<Encode::XS> Encoding object, bad things will happen.
11524(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11525
bd81e77b 11526The PV of the sv is returned.
1d7c1841 11527
bd81e77b 11528=cut */
1d7c1841 11529
bd81e77b
NC
11530char *
11531Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11532{
11533 dVAR;
11534 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11535 SV *uni;
11536 STRLEN len;
11537 const char *s;
11538 dSP;
11539 ENTER;
11540 SAVETMPS;
11541 save_re_context();
11542 PUSHMARK(sp);
11543 EXTEND(SP, 3);
11544 XPUSHs(encoding);
11545 XPUSHs(sv);
11546/*
11547 NI-S 2002/07/09
11548 Passing sv_yes is wrong - it needs to be or'ed set of constants
11549 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11550 remove converted chars from source.
1d7c1841 11551
bd81e77b 11552 Both will default the value - let them.
1d7c1841 11553
bd81e77b
NC
11554 XPUSHs(&PL_sv_yes);
11555*/
11556 PUTBACK;
11557 call_method("decode", G_SCALAR);
11558 SPAGAIN;
11559 uni = POPs;
11560 PUTBACK;
11561 s = SvPV_const(uni, len);
11562 if (s != SvPVX_const(sv)) {
11563 SvGROW(sv, len + 1);
11564 Move(s, SvPVX(sv), len + 1, char);
11565 SvCUR_set(sv, len);
11566 }
11567 FREETMPS;
11568 LEAVE;
11569 SvUTF8_on(sv);
11570 return SvPVX(sv);
389edf32 11571 }
bd81e77b
NC
11572 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11573}
1d7c1841 11574
bd81e77b
NC
11575/*
11576=for apidoc sv_cat_decode
1d7c1841 11577
bd81e77b
NC
11578The encoding is assumed to be an Encode object, the PV of the ssv is
11579assumed to be octets in that encoding and decoding the input starts
11580from the position which (PV + *offset) pointed to. The dsv will be
11581concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11582when the string tstr appears in decoding output or the input ends on
11583the PV of the ssv. The value which the offset points will be modified
11584to the last input position on the ssv.
1d7c1841 11585
bd81e77b 11586Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11587
bd81e77b
NC
11588=cut */
11589
11590bool
11591Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11592 SV *ssv, int *offset, char *tstr, int tlen)
11593{
11594 dVAR;
11595 bool ret = FALSE;
11596 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11597 SV *offsv;
11598 dSP;
11599 ENTER;
11600 SAVETMPS;
11601 save_re_context();
11602 PUSHMARK(sp);
11603 EXTEND(SP, 6);
11604 XPUSHs(encoding);
11605 XPUSHs(dsv);
11606 XPUSHs(ssv);
11607 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11608 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11609 PUTBACK;
11610 call_method("cat_decode", G_SCALAR);
11611 SPAGAIN;
11612 ret = SvTRUE(TOPs);
11613 *offset = SvIV(offsv);
11614 PUTBACK;
11615 FREETMPS;
11616 LEAVE;
389edf32 11617 }
bd81e77b
NC
11618 else
11619 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11620 return ret;
1d7c1841 11621
bd81e77b 11622}
1d7c1841 11623
bd81e77b
NC
11624/* ---------------------------------------------------------------------
11625 *
11626 * support functions for report_uninit()
11627 */
1d7c1841 11628
bd81e77b
NC
11629/* the maxiumum size of array or hash where we will scan looking
11630 * for the undefined element that triggered the warning */
1d7c1841 11631
bd81e77b 11632#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11633
bd81e77b
NC
11634/* Look for an entry in the hash whose value has the same SV as val;
11635 * If so, return a mortal copy of the key. */
1d7c1841 11636
bd81e77b
NC
11637STATIC SV*
11638S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11639{
11640 dVAR;
11641 register HE **array;
11642 I32 i;
6c3182a5 11643
bd81e77b
NC
11644 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11645 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11646 return NULL;
6c3182a5 11647
bd81e77b 11648 array = HvARRAY(hv);
6c3182a5 11649
bd81e77b
NC
11650 for (i=HvMAX(hv); i>0; i--) {
11651 register HE *entry;
11652 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11653 if (HeVAL(entry) != val)
11654 continue;
11655 if ( HeVAL(entry) == &PL_sv_undef ||
11656 HeVAL(entry) == &PL_sv_placeholder)
11657 continue;
11658 if (!HeKEY(entry))
a0714e2c 11659 return NULL;
bd81e77b
NC
11660 if (HeKLEN(entry) == HEf_SVKEY)
11661 return sv_mortalcopy(HeKEY_sv(entry));
11662 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11663 }
11664 }
a0714e2c 11665 return NULL;
bd81e77b 11666}
6c3182a5 11667
bd81e77b
NC
11668/* Look for an entry in the array whose value has the same SV as val;
11669 * If so, return the index, otherwise return -1. */
6c3182a5 11670
bd81e77b
NC
11671STATIC I32
11672S_find_array_subscript(pTHX_ AV *av, SV* val)
11673{
97aff369 11674 dVAR;
bd81e77b
NC
11675 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11676 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11677 return -1;
57c6e6d2 11678
4a021917
AL
11679 if (val != &PL_sv_undef) {
11680 SV ** const svp = AvARRAY(av);
11681 I32 i;
11682
11683 for (i=AvFILLp(av); i>=0; i--)
11684 if (svp[i] == val)
11685 return i;
bd81e77b
NC
11686 }
11687 return -1;
11688}
15a5279a 11689
bd81e77b
NC
11690/* S_varname(): return the name of a variable, optionally with a subscript.
11691 * If gv is non-zero, use the name of that global, along with gvtype (one
11692 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11693 * targ. Depending on the value of the subscript_type flag, return:
11694 */
bce260cd 11695
bd81e77b
NC
11696#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11697#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11698#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11699#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11700
bd81e77b
NC
11701STATIC SV*
11702S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11703 SV* keyname, I32 aindex, int subscript_type)
11704{
1d7c1841 11705
bd81e77b
NC
11706 SV * const name = sv_newmortal();
11707 if (gv) {
11708 char buffer[2];
11709 buffer[0] = gvtype;
11710 buffer[1] = 0;
1d7c1841 11711
bd81e77b 11712 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11713
bd81e77b 11714 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11715
bd81e77b
NC
11716 if ((unsigned int)SvPVX(name)[1] <= 26) {
11717 buffer[0] = '^';
11718 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11719
bd81e77b
NC
11720 /* Swap the 1 unprintable control character for the 2 byte pretty
11721 version - ie substr($name, 1, 1) = $buffer; */
11722 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11723 }
bd81e77b
NC
11724 }
11725 else {
11726 U32 unused;
11727 CV * const cv = find_runcv(&unused);
11728 SV *sv;
11729 AV *av;
1d7c1841 11730
bd81e77b 11731 if (!cv || !CvPADLIST(cv))
a0714e2c 11732 return NULL;
bd81e77b
NC
11733 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11734 sv = *av_fetch(av, targ, FALSE);
f8503592 11735 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
bd81e77b 11736 }
1d7c1841 11737
bd81e77b 11738 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11739 SV * const sv = newSV(0);
bd81e77b
NC
11740 *SvPVX(name) = '$';
11741 Perl_sv_catpvf(aTHX_ name, "{%s}",
11742 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11743 SvREFCNT_dec(sv);
11744 }
11745 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11746 *SvPVX(name) = '$';
11747 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11748 }
11749 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11750 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11751
bd81e77b
NC
11752 return name;
11753}
1d7c1841 11754
1d7c1841 11755
bd81e77b
NC
11756/*
11757=for apidoc find_uninit_var
1d7c1841 11758
bd81e77b
NC
11759Find the name of the undefined variable (if any) that caused the operator o
11760to issue a "Use of uninitialized value" warning.
11761If match is true, only return a name if it's value matches uninit_sv.
11762So roughly speaking, if a unary operator (such as OP_COS) generates a
11763warning, then following the direct child of the op may yield an
11764OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11765other hand, with OP_ADD there are two branches to follow, so we only print
11766the variable name if we get an exact match.
1d7c1841 11767
bd81e77b 11768The name is returned as a mortal SV.
1d7c1841 11769
bd81e77b
NC
11770Assumes that PL_op is the op that originally triggered the error, and that
11771PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11772
bd81e77b
NC
11773=cut
11774*/
1d7c1841 11775
bd81e77b
NC
11776STATIC SV *
11777S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11778{
11779 dVAR;
11780 SV *sv;
11781 AV *av;
11782 GV *gv;
11783 OP *o, *o2, *kid;
1d7c1841 11784
bd81e77b
NC
11785 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11786 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11787 return NULL;
1d7c1841 11788
bd81e77b 11789 switch (obase->op_type) {
1d7c1841 11790
bd81e77b
NC
11791 case OP_RV2AV:
11792 case OP_RV2HV:
11793 case OP_PADAV:
11794 case OP_PADHV:
11795 {
11796 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11797 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11798 I32 index = 0;
a0714e2c 11799 SV *keysv = NULL;
bd81e77b 11800 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11801
bd81e77b
NC
11802 if (pad) { /* @lex, %lex */
11803 sv = PAD_SVl(obase->op_targ);
a0714e2c 11804 gv = NULL;
bd81e77b
NC
11805 }
11806 else {
11807 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11808 /* @global, %global */
11809 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11810 if (!gv)
11811 break;
11812 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11813 }
11814 else /* @{expr}, %{expr} */
11815 return find_uninit_var(cUNOPx(obase)->op_first,
11816 uninit_sv, match);
11817 }
1d7c1841 11818
bd81e77b
NC
11819 /* attempt to find a match within the aggregate */
11820 if (hash) {
d4c19fe8 11821 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11822 if (keysv)
11823 subscript_type = FUV_SUBSCRIPT_HASH;
11824 }
11825 else {
e15d5972 11826 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11827 if (index >= 0)
11828 subscript_type = FUV_SUBSCRIPT_ARRAY;
11829 }
1d7c1841 11830
bd81e77b
NC
11831 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11832 break;
1d7c1841 11833
bd81e77b
NC
11834 return varname(gv, hash ? '%' : '@', obase->op_targ,
11835 keysv, index, subscript_type);
11836 }
1d7c1841 11837
bd81e77b
NC
11838 case OP_PADSV:
11839 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11840 break;
a0714e2c
SS
11841 return varname(NULL, '$', obase->op_targ,
11842 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11843
bd81e77b
NC
11844 case OP_GVSV:
11845 gv = cGVOPx_gv(obase);
11846 if (!gv || (match && GvSV(gv) != uninit_sv))
11847 break;
a0714e2c 11848 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11849
bd81e77b
NC
11850 case OP_AELEMFAST:
11851 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11852 if (match) {
11853 SV **svp;
11854 av = (AV*)PAD_SV(obase->op_targ);
11855 if (!av || SvRMAGICAL(av))
11856 break;
11857 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11858 if (!svp || *svp != uninit_sv)
11859 break;
11860 }
a0714e2c
SS
11861 return varname(NULL, '$', obase->op_targ,
11862 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11863 }
11864 else {
11865 gv = cGVOPx_gv(obase);
11866 if (!gv)
11867 break;
11868 if (match) {
11869 SV **svp;
11870 av = GvAV(gv);
11871 if (!av || SvRMAGICAL(av))
11872 break;
11873 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11874 if (!svp || *svp != uninit_sv)
11875 break;
11876 }
11877 return varname(gv, '$', 0,
a0714e2c 11878 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11879 }
11880 break;
1d7c1841 11881
bd81e77b
NC
11882 case OP_EXISTS:
11883 o = cUNOPx(obase)->op_first;
11884 if (!o || o->op_type != OP_NULL ||
11885 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11886 break;
11887 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11888
bd81e77b
NC
11889 case OP_AELEM:
11890 case OP_HELEM:
11891 if (PL_op == obase)
11892 /* $a[uninit_expr] or $h{uninit_expr} */
11893 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11894
a0714e2c 11895 gv = NULL;
bd81e77b
NC
11896 o = cBINOPx(obase)->op_first;
11897 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11898
bd81e77b 11899 /* get the av or hv, and optionally the gv */
a0714e2c 11900 sv = NULL;
bd81e77b
NC
11901 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11902 sv = PAD_SV(o->op_targ);
11903 }
11904 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11905 && cUNOPo->op_first->op_type == OP_GV)
11906 {
11907 gv = cGVOPx_gv(cUNOPo->op_first);
11908 if (!gv)
11909 break;
11910 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11911 }
11912 if (!sv)
11913 break;
11914
11915 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11916 /* index is constant */
11917 if (match) {
11918 if (SvMAGICAL(sv))
11919 break;
11920 if (obase->op_type == OP_HELEM) {
11921 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11922 if (!he || HeVAL(he) != uninit_sv)
11923 break;
11924 }
11925 else {
00b6aa41 11926 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11927 if (!svp || *svp != uninit_sv)
11928 break;
11929 }
11930 }
11931 if (obase->op_type == OP_HELEM)
11932 return varname(gv, '%', o->op_targ,
11933 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11934 else
a0714e2c 11935 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11936 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11937 }
11938 else {
11939 /* index is an expression;
11940 * attempt to find a match within the aggregate */
11941 if (obase->op_type == OP_HELEM) {
d4c19fe8 11942 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11943 if (keysv)
11944 return varname(gv, '%', o->op_targ,
11945 keysv, 0, FUV_SUBSCRIPT_HASH);
11946 }
11947 else {
d4c19fe8 11948 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11949 if (index >= 0)
11950 return varname(gv, '@', o->op_targ,
a0714e2c 11951 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11952 }
11953 if (match)
11954 break;
11955 return varname(gv,
11956 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
11957 ? '@' : '%',
a0714e2c 11958 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 11959 }
bd81e77b 11960 break;
dc507217 11961
bd81e77b
NC
11962 case OP_AASSIGN:
11963 /* only examine RHS */
11964 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 11965
bd81e77b
NC
11966 case OP_OPEN:
11967 o = cUNOPx(obase)->op_first;
11968 if (o->op_type == OP_PUSHMARK)
11969 o = o->op_sibling;
1d7c1841 11970
bd81e77b
NC
11971 if (!o->op_sibling) {
11972 /* one-arg version of open is highly magical */
a0ae6670 11973
bd81e77b
NC
11974 if (o->op_type == OP_GV) { /* open FOO; */
11975 gv = cGVOPx_gv(o);
11976 if (match && GvSV(gv) != uninit_sv)
11977 break;
11978 return varname(gv, '$', 0,
a0714e2c 11979 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
11980 }
11981 /* other possibilities not handled are:
11982 * open $x; or open my $x; should return '${*$x}'
11983 * open expr; should return '$'.expr ideally
11984 */
11985 break;
11986 }
11987 goto do_op;
ccfc67b7 11988
bd81e77b
NC
11989 /* ops where $_ may be an implicit arg */
11990 case OP_TRANS:
11991 case OP_SUBST:
11992 case OP_MATCH:
11993 if ( !(obase->op_flags & OPf_STACKED)) {
11994 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
11995 ? PAD_SVl(obase->op_targ)
11996 : DEFSV))
11997 {
11998 sv = sv_newmortal();
11999 sv_setpvn(sv, "$_", 2);
12000 return sv;
12001 }
12002 }
12003 goto do_op;
9f4817db 12004
bd81e77b
NC
12005 case OP_PRTF:
12006 case OP_PRINT:
12007 /* skip filehandle as it can't produce 'undef' warning */
12008 o = cUNOPx(obase)->op_first;
12009 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12010 o = o->op_sibling->op_sibling;
12011 goto do_op2;
9f4817db 12012
9f4817db 12013
bd81e77b
NC
12014 case OP_RV2SV:
12015 case OP_CUSTOM:
12016 case OP_ENTERSUB:
12017 match = 1; /* XS or custom code could trigger random warnings */
12018 goto do_op;
9f4817db 12019
bd81e77b
NC
12020 case OP_SCHOMP:
12021 case OP_CHOMP:
12022 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12023 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12024 /*FALLTHROUGH*/
5d170f3a 12025
bd81e77b
NC
12026 default:
12027 do_op:
12028 if (!(obase->op_flags & OPf_KIDS))
12029 break;
12030 o = cUNOPx(obase)->op_first;
12031
12032 do_op2:
12033 if (!o)
12034 break;
f9893866 12035
bd81e77b
NC
12036 /* if all except one arg are constant, or have no side-effects,
12037 * or are optimized away, then it's unambiguous */
5f66b61c 12038 o2 = NULL;
bd81e77b 12039 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
12040 if (kid) {
12041 const OPCODE type = kid->op_type;
12042 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
12043 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12044 || (type == OP_PUSHMARK)
bd81e77b 12045 )
bd81e77b 12046 continue;
e15d5972 12047 }
bd81e77b 12048 if (o2) { /* more than one found */
5f66b61c 12049 o2 = NULL;
bd81e77b
NC
12050 break;
12051 }
12052 o2 = kid;
12053 }
12054 if (o2)
12055 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12056
bd81e77b
NC
12057 /* scan all args */
12058 while (o) {
12059 sv = find_uninit_var(o, uninit_sv, 1);
12060 if (sv)
12061 return sv;
12062 o = o->op_sibling;
d0063567 12063 }
bd81e77b 12064 break;
f9893866 12065 }
a0714e2c 12066 return NULL;
9f4817db
JH
12067}
12068
220e2d4e 12069
bd81e77b
NC
12070/*
12071=for apidoc report_uninit
68795e93 12072
bd81e77b 12073Print appropriate "Use of uninitialized variable" warning
220e2d4e 12074
bd81e77b
NC
12075=cut
12076*/
220e2d4e 12077
bd81e77b
NC
12078void
12079Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12080{
97aff369 12081 dVAR;
bd81e77b 12082 if (PL_op) {
a0714e2c 12083 SV* varname = NULL;
bd81e77b
NC
12084 if (uninit_sv) {
12085 varname = find_uninit_var(PL_op, uninit_sv,0);
12086 if (varname)
12087 sv_insert(varname, 0, 0, " ", 1);
12088 }
12089 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12090 varname ? SvPV_nolen_const(varname) : "",
12091 " in ", OP_DESC(PL_op));
220e2d4e 12092 }
a73e8557 12093 else
bd81e77b
NC
12094 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12095 "", "", "");
220e2d4e 12096}
f9893866 12097
241d1a3b
NC
12098/*
12099 * Local variables:
12100 * c-indentation-style: bsd
12101 * c-basic-offset: 4
12102 * indent-tabs-mode: t
12103 * End:
12104 *
37442d52
RGS
12105 * ex: set ts=8 sts=4 sw=4 noet:
12106 */