This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
More doc nits on Switch, and bump version to 2.11 for upcoming CPAN release.
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b94e2f88 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 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;
823a54a3
AL
446 if (SvROK(ref)) {
447 SV * const target = SvRV(ref);
448 if (SvOBJECT(target)) {
449 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
450 if (SvWEAKREF(ref)) {
451 sv_del_backref(target, ref);
452 SvWEAKREF_off(ref);
453 SvRV_set(ref, NULL);
454 } else {
455 SvROK_off(ref);
456 SvRV_set(ref, NULL);
457 SvREFCNT_dec(target);
458 }
645c22ef
DM
459 }
460 }
461
462 /* XXX Might want to check arrays, etc. */
463}
464
465/* called by sv_clean_objs() for each live SV */
466
467#ifndef DISABLE_DESTRUCTOR_KLUDGE
468static void
acfe0abc 469do_clean_named_objs(pTHX_ SV *sv)
645c22ef 470{
97aff369 471 dVAR;
f7877b28 472 if (SvTYPE(sv) == SVt_PVGV && isGV_with_GP(sv) && GvGP(sv)) {
c69033f2
NC
473 if ((
474#ifdef PERL_DONT_CREATE_GVSV
475 GvSV(sv) &&
476#endif
477 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
478 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
479 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
480 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
481 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
482 {
483 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 484 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
485 SvREFCNT_dec(sv);
486 }
487 }
488}
489#endif
490
491/*
492=for apidoc sv_clean_objs
493
494Attempt to destroy all objects not yet freed
495
496=cut
497*/
498
4561caa4 499void
864dbfa3 500Perl_sv_clean_objs(pTHX)
4561caa4 501{
97aff369 502 dVAR;
3280af22 503 PL_in_clean_objs = TRUE;
055972dc 504 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 505#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 506 /* some barnacles may yet remain, clinging to typeglobs */
055972dc 507 visit(do_clean_named_objs, SVt_PVGV, SVTYPEMASK);
4561caa4 508#endif
3280af22 509 PL_in_clean_objs = FALSE;
4561caa4
CS
510}
511
645c22ef
DM
512/* called by sv_clean_all() for each live SV */
513
514static void
acfe0abc 515do_clean_all(pTHX_ SV *sv)
645c22ef 516{
97aff369 517 dVAR;
645c22ef
DM
518 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
519 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b 520 if (PL_comppad == (AV*)sv) {
7d49f689 521 PL_comppad = NULL;
4608196e 522 PL_curpad = NULL;
0e705b3b 523 }
645c22ef
DM
524 SvREFCNT_dec(sv);
525}
526
527/*
528=for apidoc sv_clean_all
529
530Decrement the refcnt of each remaining SV, possibly triggering a
531cleanup. This function may have to be called multiple times to free
ff276b08 532SVs which are in complex self-referential hierarchies.
645c22ef
DM
533
534=cut
535*/
536
5226ed68 537I32
864dbfa3 538Perl_sv_clean_all(pTHX)
8990e307 539{
97aff369 540 dVAR;
5226ed68 541 I32 cleaned;
3280af22 542 PL_in_clean_all = TRUE;
055972dc 543 cleaned = visit(do_clean_all, 0,0);
3280af22 544 PL_in_clean_all = FALSE;
5226ed68 545 return cleaned;
8990e307 546}
463ee0b2 547
5e258f8c
JC
548/*
549 ARENASETS: a meta-arena implementation which separates arena-info
550 into struct arena_set, which contains an array of struct
551 arena_descs, each holding info for a single arena. By separating
552 the meta-info from the arena, we recover the 1st slot, formerly
553 borrowed for list management. The arena_set is about the size of an
554 arena, avoiding the needless malloc overhead of a naive linked-list
555
556 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
557 memory in the last arena-set (1/2 on average). In trade, we get
558 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284
JC
559 smaller types). The recovery of the wasted space allows use of
560 small arenas for large, rare body types,
5e258f8c 561*/
5e258f8c 562struct arena_desc {
398c677b
NC
563 char *arena; /* the raw storage, allocated aligned */
564 size_t size; /* its size ~4k typ */
565 int unit_type; /* useful for arena audits */
5e258f8c
JC
566 /* info for sv-heads (eventually)
567 int count, flags;
568 */
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;
582 int set_size; /* ie ARENAS_PER_SET */
583 int curr; /* index of next available arena-desc */
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;
93e68bfb 601 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
JC
615 {
616 struct arena_set *next, *aroot = (struct arena_set*) PL_body_arenas;
617
618 for (; aroot; aroot = next) {
96a5add6 619 const int max = aroot->curr;
5e258f8c
JC
620 for (i=0; i<max; i++) {
621 assert(aroot->set[i].arena);
622 Safefree(aroot->set[i].arena);
623 }
624 next = aroot->next;
625 Safefree(aroot);
626 }
627 }
dc8220bf 628 PL_body_arenas = 0;
fdda85ca 629
232d1c15 630 for (i=0; i<PERL_ARENA_ROOTS_SIZE; i++)
93e68bfb 631 PL_body_roots[i] = 0;
93e68bfb 632
43c5f42d 633 Safefree(PL_nice_chunk);
bd61b366 634 PL_nice_chunk = NULL;
3280af22
NIS
635 PL_nice_chunk_size = 0;
636 PL_sv_arenaroot = 0;
637 PL_sv_root = 0;
4633a7c4
LW
638}
639
bd81e77b
NC
640/*
641 Here are mid-level routines that manage the allocation of bodies out
642 of the various arenas. There are 5 kinds of arenas:
29489e7c 643
bd81e77b
NC
644 1. SV-head arenas, which are discussed and handled above
645 2. regular body arenas
646 3. arenas for reduced-size bodies
647 4. Hash-Entry arenas
648 5. pte arenas (thread related)
29489e7c 649
bd81e77b
NC
650 Arena types 2 & 3 are chained by body-type off an array of
651 arena-root pointers, which is indexed by svtype. Some of the
652 larger/less used body types are malloced singly, since a large
653 unused block of them is wasteful. Also, several svtypes dont have
654 bodies; the data fits into the sv-head itself. The arena-root
655 pointer thus has a few unused root-pointers (which may be hijacked
656 later for arena types 4,5)
29489e7c 657
bd81e77b
NC
658 3 differs from 2 as an optimization; some body types have several
659 unused fields in the front of the structure (which are kept in-place
660 for consistency). These bodies can be allocated in smaller chunks,
661 because the leading fields arent accessed. Pointers to such bodies
662 are decremented to point at the unused 'ghost' memory, knowing that
663 the pointers are used with offsets to the real memory.
29489e7c 664
bd81e77b
NC
665 HE, HEK arenas are managed separately, with separate code, but may
666 be merge-able later..
667
668 PTE arenas are not sv-bodies, but they share these mid-level
669 mechanics, so are considered here. The new mid-level mechanics rely
670 on the sv_type of the body being allocated, so we just reserve one
671 of the unused body-slots for PTEs, then use it in those (2) PTE
672 contexts below (line ~10k)
673*/
674
bd26d9a3 675/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
676 TBD: export properly for hv.c: S_more_he().
677*/
678void*
679Perl_get_arena(pTHX_ int arena_size)
680{
7a89be66 681 dVAR;
5e258f8c 682 struct arena_desc* adesc;
476a1e16 683 struct arena_set *newroot, **aroot = (struct arena_set**) &PL_body_arenas;
5e258f8c
JC
684 int curr;
685
476a1e16
JC
686 /* shouldnt need this
687 if (!arena_size) arena_size = PERL_ARENA_SIZE;
688 */
5e258f8c
JC
689
690 /* may need new arena-set to hold new arena */
476a1e16 691 if (!*aroot || (*aroot)->curr >= (*aroot)->set_size) {
5e258f8c
JC
692 Newxz(newroot, 1, struct arena_set);
693 newroot->set_size = ARENAS_PER_SET;
476a1e16
JC
694 newroot->next = *aroot;
695 *aroot = newroot;
ca0270c4 696 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)*aroot));
5e258f8c
JC
697 }
698
699 /* ok, now have arena-set with at least 1 empty/available arena-desc */
476a1e16
JC
700 curr = (*aroot)->curr++;
701 adesc = &((*aroot)->set[curr]);
5e258f8c
JC
702 assert(!adesc->arena);
703
5e258f8c
JC
704 Newxz(adesc->arena, arena_size, char);
705 adesc->size = arena_size;
d2a0f284
JC
706 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
707 curr, adesc->arena, arena_size));
5e258f8c
JC
708
709 return adesc->arena;
5e258f8c
JC
710}
711
53c1dcc0 712
bd81e77b 713/* return a thing to the free list */
29489e7c 714
bd81e77b
NC
715#define del_body(thing, root) \
716 STMT_START { \
00b6aa41 717 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
718 LOCK_SV_MUTEX; \
719 *thing_copy = *root; \
720 *root = (void*)thing_copy; \
721 UNLOCK_SV_MUTEX; \
722 } STMT_END
29489e7c 723
bd81e77b 724/*
d2a0f284
JC
725
726=head1 SV-Body Allocation
727
728Allocation of SV-bodies is similar to SV-heads, differing as follows;
729the allocation mechanism is used for many body types, so is somewhat
730more complicated, it uses arena-sets, and has no need for still-live
731SV detection.
732
733At the outermost level, (new|del)_X*V macros return bodies of the
734appropriate type. These macros call either (new|del)_body_type or
735(new|del)_body_allocated macro pairs, depending on specifics of the
736type. Most body types use the former pair, the latter pair is used to
737allocate body types with "ghost fields".
738
739"ghost fields" are fields that are unused in certain types, and
740consequently dont need to actually exist. They are declared because
741they're part of a "base type", which allows use of functions as
742methods. The simplest examples are AVs and HVs, 2 aggregate types
743which don't use the fields which support SCALAR semantics.
744
745For these types, the arenas are carved up into *_allocated size
746chunks, we thus avoid wasted memory for those unaccessed members.
747When bodies are allocated, we adjust the pointer back in memory by the
748size of the bit not allocated, so it's as if we allocated the full
749structure. (But things will all go boom if you write to the part that
750is "not there", because you'll be overwriting the last members of the
751preceding structure in memory.)
752
753We calculate the correction using the STRUCT_OFFSET macro. For
754example, if xpv_allocated is the same structure as XPV then the two
755OFFSETs sum to zero, and the pointer is unchanged. If the allocated
756structure is smaller (no initial NV actually allocated) then the net
757effect is to subtract the size of the NV from the pointer, to return a
758new pointer as if an initial NV were actually allocated.
759
760This is the same trick as was used for NV and IV bodies. Ironically it
761doesn't need to be used for NV bodies any more, because NV is now at
762the start of the structure. IV bodies don't need it either, because
763they are no longer allocated.
764
765In turn, the new_body_* allocators call S_new_body(), which invokes
766new_body_inline macro, which takes a lock, and takes a body off the
767linked list at PL_body_roots[sv_type], calling S_more_bodies() if
768necessary to refresh an empty list. Then the lock is released, and
769the body is returned.
770
771S_more_bodies calls get_arena(), and carves it up into an array of N
772bodies, which it strings into a linked list. It looks up arena-size
773and body-size from the body_details table described below, thus
774supporting the multiple body-types.
775
776If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
777the (new|del)_X*V macros are mapped directly to malloc/free.
778
779*/
780
781/*
782
783For each sv-type, struct body_details bodies_by_type[] carries
784parameters which control these aspects of SV handling:
785
786Arena_size determines whether arenas are used for this body type, and if
787so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
788zero, forcing individual mallocs and frees.
789
790Body_size determines how big a body is, and therefore how many fit into
791each arena. Offset carries the body-pointer adjustment needed for
792*_allocated body types, and is used in *_allocated macros.
793
794But its main purpose is to parameterize info needed in
795Perl_sv_upgrade(). The info here dramatically simplifies the function
796vs the implementation in 5.8.7, making it table-driven. All fields
797are used for this, except for arena_size.
798
799For the sv-types that have no bodies, arenas are not used, so those
800PL_body_roots[sv_type] are unused, and can be overloaded. In
801something of a special case, SVt_NULL is borrowed for HE arenas;
802PL_body_roots[SVt_NULL] is filled by S_more_he, but the
803bodies_by_type[SVt_NULL] slot is not used, as the table is not
804available in hv.c,
805
806PTEs also use arenas, but are never seen in Perl_sv_upgrade.
807Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
808they can just use the same allocation semantics. At first, PTEs were
809also overloaded to a non-body sv-type, but this yielded hard-to-find
810malloc bugs, so was simplified by claiming a new slot. This choice
811has no consequence at this time.
812
29489e7c
DM
813*/
814
bd81e77b 815struct body_details {
0fb58b32 816 U8 body_size; /* Size to allocate */
10666ae3 817 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 818 U8 offset;
10666ae3
NC
819 unsigned int type : 4; /* We have space for a sanity check. */
820 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
821 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
822 unsigned int arena : 1; /* Allocated from an arena */
823 size_t arena_size; /* Size of arena to allocate */
bd81e77b 824};
29489e7c 825
bd81e77b
NC
826#define HADNV FALSE
827#define NONV TRUE
29489e7c 828
d2a0f284 829
bd81e77b
NC
830#ifdef PURIFY
831/* With -DPURFIY we allocate everything directly, and don't use arenas.
832 This seems a rather elegant way to simplify some of the code below. */
833#define HASARENA FALSE
834#else
835#define HASARENA TRUE
836#endif
837#define NOARENA FALSE
29489e7c 838
d2a0f284
JC
839/* Size the arenas to exactly fit a given number of bodies. A count
840 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
841 simplifying the default. If count > 0, the arena is sized to fit
842 only that many bodies, allowing arenas to be used for large, rare
843 bodies (XPVFM, XPVIO) without undue waste. The arena size is
844 limited by PERL_ARENA_SIZE, so we can safely oversize the
845 declarations.
846 */
95db5f15
MB
847#define FIT_ARENA0(body_size) \
848 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
849#define FIT_ARENAn(count,body_size) \
850 ( count * body_size <= PERL_ARENA_SIZE) \
851 ? count * body_size \
852 : FIT_ARENA0 (body_size)
853#define FIT_ARENA(count,body_size) \
854 count \
855 ? FIT_ARENAn (count, body_size) \
856 : FIT_ARENA0 (body_size)
d2a0f284 857
bd81e77b 858/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 859
bd81e77b
NC
860typedef struct {
861 STRLEN xpv_cur;
862 STRLEN xpv_len;
863} xpv_allocated;
29489e7c 864
bd81e77b 865to make its members accessible via a pointer to (say)
29489e7c 866
bd81e77b
NC
867struct xpv {
868 NV xnv_nv;
869 STRLEN xpv_cur;
870 STRLEN xpv_len;
871};
29489e7c 872
bd81e77b 873*/
29489e7c 874
bd81e77b
NC
875#define relative_STRUCT_OFFSET(longer, shorter, member) \
876 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 877
bd81e77b
NC
878/* Calculate the length to copy. Specifically work out the length less any
879 final padding the compiler needed to add. See the comment in sv_upgrade
880 for why copying the padding proved to be a bug. */
29489e7c 881
bd81e77b
NC
882#define copy_length(type, last_member) \
883 STRUCT_OFFSET(type, last_member) \
884 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 885
bd81e77b 886static const struct body_details bodies_by_type[] = {
10666ae3
NC
887 { sizeof(HE), 0, 0, SVt_NULL,
888 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284
JC
889
890 /* IVs are in the head, so the allocation size is 0.
891 However, the slot is overloaded for PTEs. */
892 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
893 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 894 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
895 NOARENA /* IVS don't need an arena */,
896 /* But PTEs need to know the size of their arena */
897 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
898 },
899
bd81e77b 900 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 901 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
902 FIT_ARENA(0, sizeof(NV)) },
903
904 /* RVs are in the head now. */
10666ae3 905 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 906
bd81e77b 907 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
908 { sizeof(xpv_allocated),
909 copy_length(XPV, xpv_len)
910 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
911 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 912 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 913
bd81e77b 914 /* 12 */
d2a0f284
JC
915 { sizeof(xpviv_allocated),
916 copy_length(XPVIV, xiv_u)
917 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
918 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 919 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 920
bd81e77b 921 /* 20 */
10666ae3 922 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
923 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
924
bd81e77b 925 /* 28 */
10666ae3 926 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
927 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
928
bd81e77b 929 /* 36 */
10666ae3 930 { sizeof(XPVBM), sizeof(XPVBM), 0, SVt_PVBM, TRUE, HADNV,
d2a0f284
JC
931 HASARENA, FIT_ARENA(0, sizeof(XPVBM)) },
932
bd81e77b 933 /* 48 */
10666ae3 934 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
935 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
936
bd81e77b 937 /* 64 */
10666ae3 938 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
939 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
940
941 { sizeof(xpvav_allocated),
942 copy_length(XPVAV, xmg_stash)
943 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
944 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 945 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
946
947 { sizeof(xpvhv_allocated),
948 copy_length(XPVHV, xmg_stash)
949 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
950 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 951 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 952
c84c4652 953 /* 56 */
4115f141 954 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 955 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 956 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 957
4115f141 958 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 959 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 960 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
961
962 /* XPVIO is 84 bytes, fits 48x */
10666ae3 963 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 964 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 965};
29489e7c 966
d2a0f284
JC
967#define new_body_type(sv_type) \
968 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 969
bd81e77b
NC
970#define del_body_type(p, sv_type) \
971 del_body(p, &PL_body_roots[sv_type])
29489e7c 972
29489e7c 973
bd81e77b 974#define new_body_allocated(sv_type) \
d2a0f284 975 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 976 - bodies_by_type[sv_type].offset)
29489e7c 977
bd81e77b
NC
978#define del_body_allocated(p, sv_type) \
979 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 980
29489e7c 981
bd81e77b
NC
982#define my_safemalloc(s) (void*)safemalloc(s)
983#define my_safecalloc(s) (void*)safecalloc(s, 1)
984#define my_safefree(p) safefree((char*)p)
29489e7c 985
bd81e77b 986#ifdef PURIFY
29489e7c 987
bd81e77b
NC
988#define new_XNV() my_safemalloc(sizeof(XPVNV))
989#define del_XNV(p) my_safefree(p)
29489e7c 990
bd81e77b
NC
991#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
992#define del_XPVNV(p) my_safefree(p)
29489e7c 993
bd81e77b
NC
994#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
995#define del_XPVAV(p) my_safefree(p)
29489e7c 996
bd81e77b
NC
997#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
998#define del_XPVHV(p) my_safefree(p)
29489e7c 999
bd81e77b
NC
1000#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1001#define del_XPVMG(p) my_safefree(p)
29489e7c 1002
bd81e77b
NC
1003#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1004#define del_XPVGV(p) my_safefree(p)
29489e7c 1005
bd81e77b 1006#else /* !PURIFY */
29489e7c 1007
bd81e77b
NC
1008#define new_XNV() new_body_type(SVt_NV)
1009#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1010
bd81e77b
NC
1011#define new_XPVNV() new_body_type(SVt_PVNV)
1012#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1013
bd81e77b
NC
1014#define new_XPVAV() new_body_allocated(SVt_PVAV)
1015#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1016
bd81e77b
NC
1017#define new_XPVHV() new_body_allocated(SVt_PVHV)
1018#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1019
bd81e77b
NC
1020#define new_XPVMG() new_body_type(SVt_PVMG)
1021#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1022
bd81e77b
NC
1023#define new_XPVGV() new_body_type(SVt_PVGV)
1024#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1025
bd81e77b 1026#endif /* PURIFY */
93e68bfb 1027
bd81e77b 1028/* no arena for you! */
93e68bfb 1029
bd81e77b 1030#define new_NOARENA(details) \
d2a0f284 1031 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1032#define new_NOARENAZ(details) \
d2a0f284
JC
1033 my_safecalloc((details)->body_size + (details)->offset)
1034
0b2d3faa 1035#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
10666ae3
NC
1036static bool done_sanity_check;
1037#endif
1038
d2a0f284
JC
1039STATIC void *
1040S_more_bodies (pTHX_ svtype sv_type)
1041{
1042 dVAR;
1043 void ** const root = &PL_body_roots[sv_type];
96a5add6 1044 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1045 const size_t body_size = bdp->body_size;
1046 char *start;
1047 const char *end;
1048
1049 assert(bdp->arena_size);
10666ae3 1050
0b2d3faa
JH
1051#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1052 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1053 * variables like done_sanity_check. */
10666ae3 1054 if (!done_sanity_check) {
ea471437 1055 unsigned int i = SVt_LAST;
10666ae3
NC
1056
1057 done_sanity_check = TRUE;
1058
1059 while (i--)
1060 assert (bodies_by_type[i].type == i);
1061 }
1062#endif
1063
d2a0f284
JC
1064 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size);
1065
1066 end = start + bdp->arena_size - body_size;
1067
d2a0f284
JC
1068 /* computed count doesnt reflect the 1st slot reservation */
1069 DEBUG_m(PerlIO_printf(Perl_debug_log,
1070 "arena %p end %p arena-size %d type %d size %d ct %d\n",
0e84aef4
JH
1071 start, end,
1072 (int)bdp->arena_size, sv_type, (int)body_size,
1073 (int)bdp->arena_size / (int)body_size));
d2a0f284
JC
1074
1075 *root = (void *)start;
1076
1077 while (start < end) {
1078 char * const next = start + body_size;
1079 *(void**) start = (void *)next;
1080 start = next;
1081 }
1082 *(void **)start = 0;
1083
1084 return *root;
1085}
1086
1087/* grab a new thing from the free list, allocating more if necessary.
1088 The inline version is used for speed in hot routines, and the
1089 function using it serves the rest (unless PURIFY).
1090*/
1091#define new_body_inline(xpv, sv_type) \
1092 STMT_START { \
1093 void ** const r3wt = &PL_body_roots[sv_type]; \
1094 LOCK_SV_MUTEX; \
11b79775
DD
1095 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1096 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284
JC
1097 *(r3wt) = *(void**)(xpv); \
1098 UNLOCK_SV_MUTEX; \
1099 } STMT_END
1100
1101#ifndef PURIFY
1102
1103STATIC void *
1104S_new_body(pTHX_ svtype sv_type)
1105{
1106 dVAR;
1107 void *xpv;
1108 new_body_inline(xpv, sv_type);
1109 return xpv;
1110}
1111
1112#endif
93e68bfb 1113
bd81e77b
NC
1114/*
1115=for apidoc sv_upgrade
93e68bfb 1116
bd81e77b
NC
1117Upgrade an SV to a more complex form. Generally adds a new body type to the
1118SV, then copies across as much information as possible from the old body.
1119You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1120
bd81e77b 1121=cut
93e68bfb 1122*/
93e68bfb 1123
bd81e77b 1124void
42d0e0b7 1125Perl_sv_upgrade(pTHX_ register SV *sv, svtype new_type)
cac9b346 1126{
97aff369 1127 dVAR;
bd81e77b
NC
1128 void* old_body;
1129 void* new_body;
42d0e0b7 1130 const svtype old_type = SvTYPE(sv);
d2a0f284 1131 const struct body_details *new_type_details;
bd81e77b
NC
1132 const struct body_details *const old_type_details
1133 = bodies_by_type + old_type;
cac9b346 1134
bd81e77b
NC
1135 if (new_type != SVt_PV && SvIsCOW(sv)) {
1136 sv_force_normal_flags(sv, 0);
1137 }
cac9b346 1138
bd81e77b
NC
1139 if (old_type == new_type)
1140 return;
cac9b346 1141
bd81e77b
NC
1142 if (old_type > new_type)
1143 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1144 (int)old_type, (int)new_type);
cac9b346 1145
cac9b346 1146
bd81e77b 1147 old_body = SvANY(sv);
de042e1d 1148
bd81e77b
NC
1149 /* Copying structures onto other structures that have been neatly zeroed
1150 has a subtle gotcha. Consider XPVMG
cac9b346 1151
bd81e77b
NC
1152 +------+------+------+------+------+-------+-------+
1153 | NV | CUR | LEN | IV | MAGIC | STASH |
1154 +------+------+------+------+------+-------+-------+
1155 0 4 8 12 16 20 24 28
645c22ef 1156
bd81e77b
NC
1157 where NVs are aligned to 8 bytes, so that sizeof that structure is
1158 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1159
bd81e77b
NC
1160 +------+------+------+------+------+-------+-------+------+
1161 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1162 +------+------+------+------+------+-------+-------+------+
1163 0 4 8 12 16 20 24 28 32
08742458 1164
bd81e77b 1165 so what happens if you allocate memory for this structure:
30f9da9e 1166
bd81e77b
NC
1167 +------+------+------+------+------+-------+-------+------+------+...
1168 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1169 +------+------+------+------+------+-------+-------+------+------+...
1170 0 4 8 12 16 20 24 28 32 36
bfc44f79 1171
bd81e77b
NC
1172 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1173 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1174 started out as zero once, but it's quite possible that it isn't. So now,
1175 rather than a nicely zeroed GP, you have it pointing somewhere random.
1176 Bugs ensue.
bfc44f79 1177
bd81e77b
NC
1178 (In fact, GP ends up pointing at a previous GP structure, because the
1179 principle cause of the padding in XPVMG getting garbage is a copy of
1180 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 1181
bd81e77b
NC
1182 So we are careful and work out the size of used parts of all the
1183 structures. */
bfc44f79 1184
bd81e77b
NC
1185 switch (old_type) {
1186 case SVt_NULL:
1187 break;
1188 case SVt_IV:
1189 if (new_type < SVt_PVIV) {
1190 new_type = (new_type == SVt_NV)
1191 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1192 }
1193 break;
1194 case SVt_NV:
1195 if (new_type < SVt_PVNV) {
1196 new_type = SVt_PVNV;
bd81e77b
NC
1197 }
1198 break;
1199 case SVt_RV:
1200 break;
1201 case SVt_PV:
1202 assert(new_type > SVt_PV);
1203 assert(SVt_IV < SVt_PV);
1204 assert(SVt_NV < SVt_PV);
1205 break;
1206 case SVt_PVIV:
1207 break;
1208 case SVt_PVNV:
1209 break;
1210 case SVt_PVMG:
1211 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1212 there's no way that it can be safely upgraded, because perl.c
1213 expects to Safefree(SvANY(PL_mess_sv)) */
1214 assert(sv != PL_mess_sv);
1215 /* This flag bit is used to mean other things in other scalar types.
1216 Given that it only has meaning inside the pad, it shouldn't be set
1217 on anything that can get upgraded. */
00b1698f 1218 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1219 break;
1220 default:
1221 if (old_type_details->cant_upgrade)
c81225bc
NC
1222 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1223 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1224 }
2fa1109b 1225 new_type_details = bodies_by_type + new_type;
645c22ef 1226
bd81e77b
NC
1227 SvFLAGS(sv) &= ~SVTYPEMASK;
1228 SvFLAGS(sv) |= new_type;
932e9ff9 1229
ab4416c0
NC
1230 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1231 the return statements above will have triggered. */
1232 assert (new_type != SVt_NULL);
bd81e77b 1233 switch (new_type) {
bd81e77b
NC
1234 case SVt_IV:
1235 assert(old_type == SVt_NULL);
1236 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1237 SvIV_set(sv, 0);
1238 return;
1239 case SVt_NV:
1240 assert(old_type == SVt_NULL);
1241 SvANY(sv) = new_XNV();
1242 SvNV_set(sv, 0);
1243 return;
1244 case SVt_RV:
1245 assert(old_type == SVt_NULL);
1246 SvANY(sv) = &sv->sv_u.svu_rv;
1247 SvRV_set(sv, 0);
1248 return;
1249 case SVt_PVHV:
bd81e77b 1250 case SVt_PVAV:
d2a0f284 1251 assert(new_type_details->body_size);
c1ae03ae
NC
1252
1253#ifndef PURIFY
1254 assert(new_type_details->arena);
d2a0f284 1255 assert(new_type_details->arena_size);
c1ae03ae 1256 /* This points to the start of the allocated area. */
d2a0f284
JC
1257 new_body_inline(new_body, new_type);
1258 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1259 new_body = ((char *)new_body) - new_type_details->offset;
1260#else
1261 /* We always allocated the full length item with PURIFY. To do this
1262 we fake things so that arena is false for all 16 types.. */
1263 new_body = new_NOARENAZ(new_type_details);
1264#endif
1265 SvANY(sv) = new_body;
1266 if (new_type == SVt_PVAV) {
1267 AvMAX(sv) = -1;
1268 AvFILLp(sv) = -1;
1269 AvREAL_only(sv);
1270 }
aeb18a1e 1271
bd81e77b
NC
1272 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1273 The target created by newSVrv also is, and it can have magic.
1274 However, it never has SvPVX set.
1275 */
1276 if (old_type >= SVt_RV) {
1277 assert(SvPVX_const(sv) == 0);
1278 }
aeb18a1e 1279
bd81e77b 1280 if (old_type >= SVt_PVMG) {
e736a858 1281 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1282 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1283 } else {
1284 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1285 }
1286 break;
93e68bfb 1287
93e68bfb 1288
bd81e77b
NC
1289 case SVt_PVIV:
1290 /* XXX Is this still needed? Was it ever needed? Surely as there is
1291 no route from NV to PVIV, NOK can never be true */
1292 assert(!SvNOKp(sv));
1293 assert(!SvNOK(sv));
1294 case SVt_PVIO:
1295 case SVt_PVFM:
1296 case SVt_PVBM:
1297 case SVt_PVGV:
1298 case SVt_PVCV:
1299 case SVt_PVLV:
1300 case SVt_PVMG:
1301 case SVt_PVNV:
1302 case SVt_PV:
93e68bfb 1303
d2a0f284 1304 assert(new_type_details->body_size);
bd81e77b
NC
1305 /* We always allocated the full length item with PURIFY. To do this
1306 we fake things so that arena is false for all 16 types.. */
1307 if(new_type_details->arena) {
1308 /* This points to the start of the allocated area. */
d2a0f284
JC
1309 new_body_inline(new_body, new_type);
1310 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1311 new_body = ((char *)new_body) - new_type_details->offset;
1312 } else {
1313 new_body = new_NOARENAZ(new_type_details);
1314 }
1315 SvANY(sv) = new_body;
5e2fc214 1316
bd81e77b 1317 if (old_type_details->copy) {
f9ba3d20
NC
1318 /* There is now the potential for an upgrade from something without
1319 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1320 int offset = old_type_details->offset;
1321 int length = old_type_details->copy;
1322
1323 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1324 const int difference
f9ba3d20
NC
1325 = new_type_details->offset - old_type_details->offset;
1326 offset += difference;
1327 length -= difference;
1328 }
1329 assert (length >= 0);
1330
1331 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1332 char);
bd81e77b
NC
1333 }
1334
1335#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1336 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1337 * correct 0.0 for us. Otherwise, if the old body didn't have an
1338 * NV slot, but the new one does, then we need to initialise the
1339 * freshly created NV slot with whatever the correct bit pattern is
1340 * for 0.0 */
1341 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1342 SvNV_set(sv, 0);
82048762 1343#endif
5e2fc214 1344
bd81e77b 1345 if (new_type == SVt_PVIO)
f2524eef 1346 IoPAGE_LEN(sv) = 60;
bd81e77b 1347 if (old_type < SVt_RV)
6136c704 1348 SvPV_set(sv, NULL);
bd81e77b
NC
1349 break;
1350 default:
afd78fd5
JH
1351 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1352 (unsigned long)new_type);
bd81e77b 1353 }
73171d91 1354
d2a0f284
JC
1355 if (old_type_details->arena) {
1356 /* If there was an old body, then we need to free it.
1357 Note that there is an assumption that all bodies of types that
1358 can be upgraded came from arenas. Only the more complex non-
1359 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1360#ifdef PURIFY
1361 my_safefree(old_body);
1362#else
1363 del_body((void*)((char*)old_body + old_type_details->offset),
1364 &PL_body_roots[old_type]);
1365#endif
1366 }
1367}
73171d91 1368
bd81e77b
NC
1369/*
1370=for apidoc sv_backoff
73171d91 1371
bd81e77b
NC
1372Remove any string offset. You should normally use the C<SvOOK_off> macro
1373wrapper instead.
73171d91 1374
bd81e77b 1375=cut
73171d91
NC
1376*/
1377
bd81e77b
NC
1378int
1379Perl_sv_backoff(pTHX_ register SV *sv)
1380{
96a5add6 1381 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1382 assert(SvOOK(sv));
1383 assert(SvTYPE(sv) != SVt_PVHV);
1384 assert(SvTYPE(sv) != SVt_PVAV);
1385 if (SvIVX(sv)) {
1386 const char * const s = SvPVX_const(sv);
1387 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1388 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1389 SvIV_set(sv, 0);
1390 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1391 }
1392 SvFLAGS(sv) &= ~SVf_OOK;
1393 return 0;
1394}
73171d91 1395
bd81e77b
NC
1396/*
1397=for apidoc sv_grow
73171d91 1398
bd81e77b
NC
1399Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1400upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1401Use the C<SvGROW> wrapper instead.
93e68bfb 1402
bd81e77b
NC
1403=cut
1404*/
93e68bfb 1405
bd81e77b
NC
1406char *
1407Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1408{
1409 register char *s;
93e68bfb 1410
5db06880
NC
1411 if (PL_madskills && newlen >= 0x100000) {
1412 PerlIO_printf(Perl_debug_log,
1413 "Allocation too large: %"UVxf"\n", (UV)newlen);
1414 }
bd81e77b
NC
1415#ifdef HAS_64K_LIMIT
1416 if (newlen >= 0x10000) {
1417 PerlIO_printf(Perl_debug_log,
1418 "Allocation too large: %"UVxf"\n", (UV)newlen);
1419 my_exit(1);
1420 }
1421#endif /* HAS_64K_LIMIT */
1422 if (SvROK(sv))
1423 sv_unref(sv);
1424 if (SvTYPE(sv) < SVt_PV) {
1425 sv_upgrade(sv, SVt_PV);
1426 s = SvPVX_mutable(sv);
1427 }
1428 else if (SvOOK(sv)) { /* pv is offset? */
1429 sv_backoff(sv);
1430 s = SvPVX_mutable(sv);
1431 if (newlen > SvLEN(sv))
1432 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1433#ifdef HAS_64K_LIMIT
1434 if (newlen >= 0x10000)
1435 newlen = 0xFFFF;
1436#endif
1437 }
1438 else
1439 s = SvPVX_mutable(sv);
aeb18a1e 1440
bd81e77b
NC
1441 if (newlen > SvLEN(sv)) { /* need more room? */
1442 newlen = PERL_STRLEN_ROUNDUP(newlen);
1443 if (SvLEN(sv) && s) {
1444#ifdef MYMALLOC
1445 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1446 if (newlen <= l) {
1447 SvLEN_set(sv, l);
1448 return s;
1449 } else
1450#endif
10edeb5d 1451 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1452 }
1453 else {
10edeb5d 1454 s = (char*)safemalloc(newlen);
bd81e77b
NC
1455 if (SvPVX_const(sv) && SvCUR(sv)) {
1456 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1457 }
1458 }
1459 SvPV_set(sv, s);
1460 SvLEN_set(sv, newlen);
1461 }
1462 return s;
1463}
aeb18a1e 1464
bd81e77b
NC
1465/*
1466=for apidoc sv_setiv
932e9ff9 1467
bd81e77b
NC
1468Copies an integer into the given SV, upgrading first if necessary.
1469Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1470
bd81e77b
NC
1471=cut
1472*/
463ee0b2 1473
bd81e77b
NC
1474void
1475Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1476{
97aff369 1477 dVAR;
bd81e77b
NC
1478 SV_CHECK_THINKFIRST_COW_DROP(sv);
1479 switch (SvTYPE(sv)) {
1480 case SVt_NULL:
1481 sv_upgrade(sv, SVt_IV);
1482 break;
1483 case SVt_NV:
1484 sv_upgrade(sv, SVt_PVNV);
1485 break;
1486 case SVt_RV:
1487 case SVt_PV:
1488 sv_upgrade(sv, SVt_PVIV);
1489 break;
463ee0b2 1490
bd81e77b
NC
1491 case SVt_PVGV:
1492 case SVt_PVAV:
1493 case SVt_PVHV:
1494 case SVt_PVCV:
1495 case SVt_PVFM:
1496 case SVt_PVIO:
1497 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1498 OP_DESC(PL_op));
42d0e0b7 1499 default: NOOP;
bd81e77b
NC
1500 }
1501 (void)SvIOK_only(sv); /* validate number */
1502 SvIV_set(sv, i);
1503 SvTAINT(sv);
1504}
932e9ff9 1505
bd81e77b
NC
1506/*
1507=for apidoc sv_setiv_mg
d33b2eba 1508
bd81e77b 1509Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1510
bd81e77b
NC
1511=cut
1512*/
d33b2eba 1513
bd81e77b
NC
1514void
1515Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1516{
1517 sv_setiv(sv,i);
1518 SvSETMAGIC(sv);
1519}
727879eb 1520
bd81e77b
NC
1521/*
1522=for apidoc sv_setuv
d33b2eba 1523
bd81e77b
NC
1524Copies an unsigned integer into the given SV, upgrading first if necessary.
1525Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1526
bd81e77b
NC
1527=cut
1528*/
d33b2eba 1529
bd81e77b
NC
1530void
1531Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1532{
1533 /* With these two if statements:
1534 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1535
bd81e77b
NC
1536 without
1537 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1538
bd81e77b
NC
1539 If you wish to remove them, please benchmark to see what the effect is
1540 */
1541 if (u <= (UV)IV_MAX) {
1542 sv_setiv(sv, (IV)u);
1543 return;
1544 }
1545 sv_setiv(sv, 0);
1546 SvIsUV_on(sv);
1547 SvUV_set(sv, u);
1548}
d33b2eba 1549
bd81e77b
NC
1550/*
1551=for apidoc sv_setuv_mg
727879eb 1552
bd81e77b 1553Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1554
bd81e77b
NC
1555=cut
1556*/
5e2fc214 1557
bd81e77b
NC
1558void
1559Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1560{
1561 sv_setiv(sv, 0);
1562 SvIsUV_on(sv);
1563 sv_setuv(sv,u);
1564 SvSETMAGIC(sv);
1565}
5e2fc214 1566
954c1994 1567/*
bd81e77b 1568=for apidoc sv_setnv
954c1994 1569
bd81e77b
NC
1570Copies a double into the given SV, upgrading first if necessary.
1571Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1572
1573=cut
1574*/
1575
63f97190 1576void
bd81e77b 1577Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1578{
97aff369 1579 dVAR;
bd81e77b
NC
1580 SV_CHECK_THINKFIRST_COW_DROP(sv);
1581 switch (SvTYPE(sv)) {
79072805 1582 case SVt_NULL:
79072805 1583 case SVt_IV:
bd81e77b 1584 sv_upgrade(sv, SVt_NV);
79072805 1585 break;
ed6116ce 1586 case SVt_RV:
79072805 1587 case SVt_PV:
79072805 1588 case SVt_PVIV:
bd81e77b 1589 sv_upgrade(sv, SVt_PVNV);
79072805 1590 break;
bd4b1eb5 1591
bd4b1eb5 1592 case SVt_PVGV:
bd81e77b
NC
1593 case SVt_PVAV:
1594 case SVt_PVHV:
79072805 1595 case SVt_PVCV:
bd81e77b
NC
1596 case SVt_PVFM:
1597 case SVt_PVIO:
1598 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1599 OP_NAME(PL_op));
42d0e0b7 1600 default: NOOP;
2068cd4d 1601 }
bd81e77b
NC
1602 SvNV_set(sv, num);
1603 (void)SvNOK_only(sv); /* validate number */
1604 SvTAINT(sv);
79072805
LW
1605}
1606
645c22ef 1607/*
bd81e77b 1608=for apidoc sv_setnv_mg
645c22ef 1609
bd81e77b 1610Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1611
1612=cut
1613*/
1614
bd81e77b
NC
1615void
1616Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1617{
bd81e77b
NC
1618 sv_setnv(sv,num);
1619 SvSETMAGIC(sv);
79072805
LW
1620}
1621
bd81e77b
NC
1622/* Print an "isn't numeric" warning, using a cleaned-up,
1623 * printable version of the offending string
1624 */
954c1994 1625
bd81e77b
NC
1626STATIC void
1627S_not_a_number(pTHX_ SV *sv)
79072805 1628{
97aff369 1629 dVAR;
bd81e77b
NC
1630 SV *dsv;
1631 char tmpbuf[64];
1632 const char *pv;
94463019
JH
1633
1634 if (DO_UTF8(sv)) {
396482e1 1635 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1636 pv = sv_uni_display(dsv, sv, 10, 0);
1637 } else {
1638 char *d = tmpbuf;
551405c4 1639 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1640 /* each *s can expand to 4 chars + "...\0",
1641 i.e. need room for 8 chars */
ecdeb87c 1642
00b6aa41
AL
1643 const char *s = SvPVX_const(sv);
1644 const char * const end = s + SvCUR(sv);
1645 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1646 int ch = *s & 0xFF;
1647 if (ch & 128 && !isPRINT_LC(ch)) {
1648 *d++ = 'M';
1649 *d++ = '-';
1650 ch &= 127;
1651 }
1652 if (ch == '\n') {
1653 *d++ = '\\';
1654 *d++ = 'n';
1655 }
1656 else if (ch == '\r') {
1657 *d++ = '\\';
1658 *d++ = 'r';
1659 }
1660 else if (ch == '\f') {
1661 *d++ = '\\';
1662 *d++ = 'f';
1663 }
1664 else if (ch == '\\') {
1665 *d++ = '\\';
1666 *d++ = '\\';
1667 }
1668 else if (ch == '\0') {
1669 *d++ = '\\';
1670 *d++ = '0';
1671 }
1672 else if (isPRINT_LC(ch))
1673 *d++ = ch;
1674 else {
1675 *d++ = '^';
1676 *d++ = toCTRL(ch);
1677 }
1678 }
1679 if (s < end) {
1680 *d++ = '.';
1681 *d++ = '.';
1682 *d++ = '.';
1683 }
1684 *d = '\0';
1685 pv = tmpbuf;
a0d0e21e 1686 }
a0d0e21e 1687
533c011a 1688 if (PL_op)
9014280d 1689 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1690 "Argument \"%s\" isn't numeric in %s", pv,
1691 OP_DESC(PL_op));
a0d0e21e 1692 else
9014280d 1693 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1694 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1695}
1696
c2988b20
NC
1697/*
1698=for apidoc looks_like_number
1699
645c22ef
DM
1700Test if the content of an SV looks like a number (or is a number).
1701C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1702non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1703
1704=cut
1705*/
1706
1707I32
1708Perl_looks_like_number(pTHX_ SV *sv)
1709{
a3b680e6 1710 register const char *sbegin;
c2988b20
NC
1711 STRLEN len;
1712
1713 if (SvPOK(sv)) {
3f7c398e 1714 sbegin = SvPVX_const(sv);
c2988b20
NC
1715 len = SvCUR(sv);
1716 }
1717 else if (SvPOKp(sv))
83003860 1718 sbegin = SvPV_const(sv, len);
c2988b20 1719 else
e0ab1c0e 1720 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1721 return grok_number(sbegin, len, NULL);
1722}
25da4f38 1723
19f6321d
NC
1724STATIC bool
1725S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1726{
1727 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1728 SV *const buffer = sv_newmortal();
1729
1730 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1731 is on. */
1732 SvFAKE_off(gv);
1733 gv_efullname3(buffer, gv, "*");
1734 SvFLAGS(gv) |= wasfake;
1735
675c862f
AL
1736 /* We know that all GVs stringify to something that is not-a-number,
1737 so no need to test that. */
1738 if (ckWARN(WARN_NUMERIC))
1739 not_a_number(buffer);
1740 /* We just want something true to return, so that S_sv_2iuv_common
1741 can tail call us and return true. */
19f6321d 1742 return TRUE;
675c862f
AL
1743}
1744
1745STATIC char *
19f6321d 1746S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1747{
1748 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1749 SV *const buffer = sv_newmortal();
1750
1751 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1752 is on. */
1753 SvFAKE_off(gv);
1754 gv_efullname3(buffer, gv, "*");
1755 SvFLAGS(gv) |= wasfake;
1756
1757 assert(SvPOK(buffer));
a6d61a6c
NC
1758 if (len) {
1759 *len = SvCUR(buffer);
1760 }
675c862f 1761 return SvPVX(buffer);
180488f8
NC
1762}
1763
25da4f38
IZ
1764/* Actually, ISO C leaves conversion of UV to IV undefined, but
1765 until proven guilty, assume that things are not that bad... */
1766
645c22ef
DM
1767/*
1768 NV_PRESERVES_UV:
1769
1770 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1771 an IV (an assumption perl has been based on to date) it becomes necessary
1772 to remove the assumption that the NV always carries enough precision to
1773 recreate the IV whenever needed, and that the NV is the canonical form.
1774 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1775 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1776 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1777 1) to distinguish between IV/UV/NV slots that have cached a valid
1778 conversion where precision was lost and IV/UV/NV slots that have a
1779 valid conversion which has lost no precision
645c22ef 1780 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1781 would lose precision, the precise conversion (or differently
1782 imprecise conversion) is also performed and cached, to prevent
1783 requests for different numeric formats on the same SV causing
1784 lossy conversion chains. (lossless conversion chains are perfectly
1785 acceptable (still))
1786
1787
1788 flags are used:
1789 SvIOKp is true if the IV slot contains a valid value
1790 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1791 SvNOKp is true if the NV slot contains a valid value
1792 SvNOK is true only if the NV value is accurate
1793
1794 so
645c22ef 1795 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1796 IV(or UV) would lose accuracy over a direct conversion from PV to
1797 IV(or UV). If it would, cache both conversions, return NV, but mark
1798 SV as IOK NOKp (ie not NOK).
1799
645c22ef 1800 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1801 NV would lose accuracy over a direct conversion from PV to NV. If it
1802 would, cache both conversions, flag similarly.
1803
1804 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1805 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1806 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1807 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1808 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1809
645c22ef
DM
1810 The benefit of this is that operations such as pp_add know that if
1811 SvIOK is true for both left and right operands, then integer addition
1812 can be used instead of floating point (for cases where the result won't
1813 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1814 loss of precision compared with integer addition.
1815
1816 * making IV and NV equal status should make maths accurate on 64 bit
1817 platforms
1818 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1819 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1820 looking for SvIOK and checking for overflow will not outweigh the
1821 fp to integer speedup)
1822 * will slow down integer operations (callers of SvIV) on "inaccurate"
1823 values, as the change from SvIOK to SvIOKp will cause a call into
1824 sv_2iv each time rather than a macro access direct to the IV slot
1825 * should speed up number->string conversion on integers as IV is
645c22ef 1826 favoured when IV and NV are equally accurate
28e5dec8
JH
1827
1828 ####################################################################
645c22ef
DM
1829 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1830 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1831 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1832 ####################################################################
1833
645c22ef 1834 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1835 performance ratio.
1836*/
1837
1838#ifndef NV_PRESERVES_UV
645c22ef
DM
1839# define IS_NUMBER_UNDERFLOW_IV 1
1840# define IS_NUMBER_UNDERFLOW_UV 2
1841# define IS_NUMBER_IV_AND_UV 2
1842# define IS_NUMBER_OVERFLOW_IV 4
1843# define IS_NUMBER_OVERFLOW_UV 5
1844
1845/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1846
1847/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1848STATIC int
645c22ef 1849S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1850{
97aff369 1851 dVAR;
b57a0404 1852 PERL_UNUSED_ARG(numtype); /* Used only under DEBUGGING? */
3f7c398e 1853 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
1854 if (SvNVX(sv) < (NV)IV_MIN) {
1855 (void)SvIOKp_on(sv);
1856 (void)SvNOK_on(sv);
45977657 1857 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1858 return IS_NUMBER_UNDERFLOW_IV;
1859 }
1860 if (SvNVX(sv) > (NV)UV_MAX) {
1861 (void)SvIOKp_on(sv);
1862 (void)SvNOK_on(sv);
1863 SvIsUV_on(sv);
607fa7f2 1864 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1865 return IS_NUMBER_OVERFLOW_UV;
1866 }
c2988b20
NC
1867 (void)SvIOKp_on(sv);
1868 (void)SvNOK_on(sv);
1869 /* Can't use strtol etc to convert this string. (See truth table in
1870 sv_2iv */
1871 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1872 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1873 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1874 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1875 } else {
1876 /* Integer is imprecise. NOK, IOKp */
1877 }
1878 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1879 }
1880 SvIsUV_on(sv);
607fa7f2 1881 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1882 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1883 if (SvUVX(sv) == UV_MAX) {
1884 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1885 possibly be preserved by NV. Hence, it must be overflow.
1886 NOK, IOKp */
1887 return IS_NUMBER_OVERFLOW_UV;
1888 }
1889 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1890 } else {
1891 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1892 }
c2988b20 1893 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1894}
645c22ef
DM
1895#endif /* !NV_PRESERVES_UV*/
1896
af359546
NC
1897STATIC bool
1898S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1899 dVAR;
af359546 1900 if (SvNOKp(sv)) {
28e5dec8
JH
1901 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1902 * without also getting a cached IV/UV from it at the same time
1903 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1904 * IV or UV at same time to avoid this. */
1905 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1906
1907 if (SvTYPE(sv) == SVt_NV)
1908 sv_upgrade(sv, SVt_PVNV);
1909
28e5dec8
JH
1910 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1911 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1912 certainly cast into the IV range at IV_MAX, whereas the correct
1913 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1914 cases go to UV */
cab190d4
JD
1915#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1916 if (Perl_isnan(SvNVX(sv))) {
1917 SvUV_set(sv, 0);
1918 SvIsUV_on(sv);
fdbe6d7c 1919 return FALSE;
cab190d4 1920 }
cab190d4 1921#endif
28e5dec8 1922 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1923 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1924 if (SvNVX(sv) == (NV) SvIVX(sv)
1925#ifndef NV_PRESERVES_UV
1926 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1927 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1928 /* Don't flag it as "accurately an integer" if the number
1929 came from a (by definition imprecise) NV operation, and
1930 we're outside the range of NV integer precision */
1931#endif
1932 ) {
1933 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1934 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1935 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1936 PTR2UV(sv),
1937 SvNVX(sv),
1938 SvIVX(sv)));
1939
1940 } else {
1941 /* IV not precise. No need to convert from PV, as NV
1942 conversion would already have cached IV if it detected
1943 that PV->IV would be better than PV->NV->IV
1944 flags already correct - don't set public IOK. */
1945 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1946 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1947 PTR2UV(sv),
1948 SvNVX(sv),
1949 SvIVX(sv)));
1950 }
1951 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1952 but the cast (NV)IV_MIN rounds to a the value less (more
1953 negative) than IV_MIN which happens to be equal to SvNVX ??
1954 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1955 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1956 (NV)UVX == NVX are both true, but the values differ. :-(
1957 Hopefully for 2s complement IV_MIN is something like
1958 0x8000000000000000 which will be exact. NWC */
d460ef45 1959 }
25da4f38 1960 else {
607fa7f2 1961 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1962 if (
1963 (SvNVX(sv) == (NV) SvUVX(sv))
1964#ifndef NV_PRESERVES_UV
1965 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1966 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1967 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1968 /* Don't flag it as "accurately an integer" if the number
1969 came from a (by definition imprecise) NV operation, and
1970 we're outside the range of NV integer precision */
1971#endif
1972 )
1973 SvIOK_on(sv);
25da4f38 1974 SvIsUV_on(sv);
1c846c1f 1975 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1976 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1977 PTR2UV(sv),
57def98f
JH
1978 SvUVX(sv),
1979 SvUVX(sv)));
25da4f38 1980 }
748a9306
LW
1981 }
1982 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1983 UV value;
504618e9 1984 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1985 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1986 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1987 the same as the direct translation of the initial string
1988 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1989 be careful to ensure that the value with the .456 is around if the
1990 NV value is requested in the future).
1c846c1f 1991
af359546 1992 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1993 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1994 cache the NV if we are sure it's not needed.
25da4f38 1995 */
16b7a9a4 1996
c2988b20
NC
1997 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1998 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1999 == IS_NUMBER_IN_UV) {
5e045b90 2000 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
2001 if (SvTYPE(sv) < SVt_PVIV)
2002 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 2003 (void)SvIOK_on(sv);
c2988b20
NC
2004 } else if (SvTYPE(sv) < SVt_PVNV)
2005 sv_upgrade(sv, SVt_PVNV);
28e5dec8 2006
f2524eef 2007 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2008 we aren't going to call atof() below. If NVs don't preserve UVs
2009 then the value returned may have more precision than atof() will
2010 return, even though value isn't perfectly accurate. */
2011 if ((numtype & (IS_NUMBER_IN_UV
2012#ifdef NV_PRESERVES_UV
2013 | IS_NUMBER_NOT_INT
2014#endif
2015 )) == IS_NUMBER_IN_UV) {
2016 /* This won't turn off the public IOK flag if it was set above */
2017 (void)SvIOKp_on(sv);
2018
2019 if (!(numtype & IS_NUMBER_NEG)) {
2020 /* positive */;
2021 if (value <= (UV)IV_MAX) {
45977657 2022 SvIV_set(sv, (IV)value);
c2988b20 2023 } else {
af359546 2024 /* it didn't overflow, and it was positive. */
607fa7f2 2025 SvUV_set(sv, value);
c2988b20
NC
2026 SvIsUV_on(sv);
2027 }
2028 } else {
2029 /* 2s complement assumption */
2030 if (value <= (UV)IV_MIN) {
45977657 2031 SvIV_set(sv, -(IV)value);
c2988b20
NC
2032 } else {
2033 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2034 I'm assuming it will be rare. */
c2988b20
NC
2035 if (SvTYPE(sv) < SVt_PVNV)
2036 sv_upgrade(sv, SVt_PVNV);
2037 SvNOK_on(sv);
2038 SvIOK_off(sv);
2039 SvIOKp_on(sv);
9d6ce603 2040 SvNV_set(sv, -(NV)value);
45977657 2041 SvIV_set(sv, IV_MIN);
c2988b20
NC
2042 }
2043 }
2044 }
2045 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2046 will be in the previous block to set the IV slot, and the next
2047 block to set the NV slot. So no else here. */
2048
2049 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2050 != IS_NUMBER_IN_UV) {
2051 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2052 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2053
c2988b20
NC
2054 if (! numtype && ckWARN(WARN_NUMERIC))
2055 not_a_number(sv);
28e5dec8 2056
65202027 2057#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2058 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2059 PTR2UV(sv), SvNVX(sv)));
65202027 2060#else
1779d84d 2061 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2062 PTR2UV(sv), SvNVX(sv)));
65202027 2063#endif
28e5dec8 2064
28e5dec8 2065#ifdef NV_PRESERVES_UV
af359546
NC
2066 (void)SvIOKp_on(sv);
2067 (void)SvNOK_on(sv);
2068 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2069 SvIV_set(sv, I_V(SvNVX(sv)));
2070 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2071 SvIOK_on(sv);
2072 } else {
6f207bd3 2073 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2074 }
2075 /* UV will not work better than IV */
2076 } else {
2077 if (SvNVX(sv) > (NV)UV_MAX) {
2078 SvIsUV_on(sv);
2079 /* Integer is inaccurate. NOK, IOKp, is UV */
2080 SvUV_set(sv, UV_MAX);
af359546
NC
2081 } else {
2082 SvUV_set(sv, U_V(SvNVX(sv)));
2083 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2084 NV preservse UV so can do correct comparison. */
2085 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2086 SvIOK_on(sv);
af359546 2087 } else {
6f207bd3 2088 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2089 }
2090 }
4b0c9573 2091 SvIsUV_on(sv);
af359546 2092 }
28e5dec8 2093#else /* NV_PRESERVES_UV */
c2988b20
NC
2094 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2095 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2096 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2097 grok_number above. The NV slot has just been set using
2098 Atof. */
560b0c46 2099 SvNOK_on(sv);
c2988b20
NC
2100 assert (SvIOKp(sv));
2101 } else {
2102 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2103 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2104 /* Small enough to preserve all bits. */
2105 (void)SvIOKp_on(sv);
2106 SvNOK_on(sv);
45977657 2107 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2108 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2109 SvIOK_on(sv);
2110 /* Assumption: first non-preserved integer is < IV_MAX,
2111 this NV is in the preserved range, therefore: */
2112 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2113 < (UV)IV_MAX)) {
32fdb065 2114 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
2115 }
2116 } else {
2117 /* IN_UV NOT_INT
2118 0 0 already failed to read UV.
2119 0 1 already failed to read UV.
2120 1 0 you won't get here in this case. IV/UV
2121 slot set, public IOK, Atof() unneeded.
2122 1 1 already read UV.
2123 so there's no point in sv_2iuv_non_preserve() attempting
2124 to use atol, strtol, strtoul etc. */
40a17c4c 2125 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2126 }
2127 }
28e5dec8 2128#endif /* NV_PRESERVES_UV */
25da4f38 2129 }
af359546
NC
2130 }
2131 else {
675c862f 2132 if (isGV_with_GP(sv))
a0933d07 2133 return glob_2number((GV *)sv);
180488f8 2134
af359546
NC
2135 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2136 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2137 report_uninit(sv);
2138 }
25da4f38
IZ
2139 if (SvTYPE(sv) < SVt_IV)
2140 /* Typically the caller expects that sv_any is not NULL now. */
2141 sv_upgrade(sv, SVt_IV);
af359546
NC
2142 /* Return 0 from the caller. */
2143 return TRUE;
2144 }
2145 return FALSE;
2146}
2147
2148/*
2149=for apidoc sv_2iv_flags
2150
2151Return the integer value of an SV, doing any necessary string
2152conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2153Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2154
2155=cut
2156*/
2157
2158IV
2159Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2160{
97aff369 2161 dVAR;
af359546 2162 if (!sv)
a0d0e21e 2163 return 0;
af359546
NC
2164 if (SvGMAGICAL(sv)) {
2165 if (flags & SV_GMAGIC)
2166 mg_get(sv);
2167 if (SvIOKp(sv))
2168 return SvIVX(sv);
2169 if (SvNOKp(sv)) {
2170 return I_V(SvNVX(sv));
2171 }
71c558c3
NC
2172 if (SvPOKp(sv) && SvLEN(sv)) {
2173 UV value;
2174 const int numtype
2175 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2176
2177 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2178 == IS_NUMBER_IN_UV) {
2179 /* It's definitely an integer */
2180 if (numtype & IS_NUMBER_NEG) {
2181 if (value < (UV)IV_MIN)
2182 return -(IV)value;
2183 } else {
2184 if (value < (UV)IV_MAX)
2185 return (IV)value;
2186 }
2187 }
2188 if (!numtype) {
2189 if (ckWARN(WARN_NUMERIC))
2190 not_a_number(sv);
2191 }
2192 return I_V(Atof(SvPVX_const(sv)));
2193 }
1c7ff15e
NC
2194 if (SvROK(sv)) {
2195 goto return_rok;
af359546 2196 }
1c7ff15e
NC
2197 assert(SvTYPE(sv) >= SVt_PVMG);
2198 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2199 } else if (SvTHINKFIRST(sv)) {
af359546 2200 if (SvROK(sv)) {
1c7ff15e 2201 return_rok:
af359546
NC
2202 if (SvAMAGIC(sv)) {
2203 SV * const tmpstr=AMG_CALLun(sv,numer);
2204 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2205 return SvIV(tmpstr);
2206 }
2207 }
2208 return PTR2IV(SvRV(sv));
2209 }
2210 if (SvIsCOW(sv)) {
2211 sv_force_normal_flags(sv, 0);
2212 }
2213 if (SvREADONLY(sv) && !SvOK(sv)) {
2214 if (ckWARN(WARN_UNINITIALIZED))
2215 report_uninit(sv);
2216 return 0;
2217 }
2218 }
2219 if (!SvIOKp(sv)) {
2220 if (S_sv_2iuv_common(aTHX_ sv))
2221 return 0;
79072805 2222 }
1d7c1841
GS
2223 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2224 PTR2UV(sv),SvIVX(sv)));
25da4f38 2225 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2226}
2227
645c22ef 2228/*
891f9566 2229=for apidoc sv_2uv_flags
645c22ef
DM
2230
2231Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2232conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2233Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2234
2235=cut
2236*/
2237
ff68c719 2238UV
891f9566 2239Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2240{
97aff369 2241 dVAR;
ff68c719 2242 if (!sv)
2243 return 0;
2244 if (SvGMAGICAL(sv)) {
891f9566
YST
2245 if (flags & SV_GMAGIC)
2246 mg_get(sv);
ff68c719 2247 if (SvIOKp(sv))
2248 return SvUVX(sv);
2249 if (SvNOKp(sv))
2250 return U_V(SvNVX(sv));
71c558c3
NC
2251 if (SvPOKp(sv) && SvLEN(sv)) {
2252 UV value;
2253 const int numtype
2254 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2255
2256 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2257 == IS_NUMBER_IN_UV) {
2258 /* It's definitely an integer */
2259 if (!(numtype & IS_NUMBER_NEG))
2260 return value;
2261 }
2262 if (!numtype) {
2263 if (ckWARN(WARN_NUMERIC))
2264 not_a_number(sv);
2265 }
2266 return U_V(Atof(SvPVX_const(sv)));
2267 }
1c7ff15e
NC
2268 if (SvROK(sv)) {
2269 goto return_rok;
3fe9a6f1 2270 }
1c7ff15e
NC
2271 assert(SvTYPE(sv) >= SVt_PVMG);
2272 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2273 } else if (SvTHINKFIRST(sv)) {
ff68c719 2274 if (SvROK(sv)) {
1c7ff15e 2275 return_rok:
deb46114
NC
2276 if (SvAMAGIC(sv)) {
2277 SV *const tmpstr = AMG_CALLun(sv,numer);
2278 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2279 return SvUV(tmpstr);
2280 }
2281 }
2282 return PTR2UV(SvRV(sv));
ff68c719 2283 }
765f542d
NC
2284 if (SvIsCOW(sv)) {
2285 sv_force_normal_flags(sv, 0);
8a818333 2286 }
0336b60e 2287 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2288 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2289 report_uninit(sv);
ff68c719 2290 return 0;
2291 }
2292 }
af359546
NC
2293 if (!SvIOKp(sv)) {
2294 if (S_sv_2iuv_common(aTHX_ sv))
2295 return 0;
ff68c719 2296 }
25da4f38 2297
1d7c1841
GS
2298 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2299 PTR2UV(sv),SvUVX(sv)));
25da4f38 2300 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2301}
2302
645c22ef
DM
2303/*
2304=for apidoc sv_2nv
2305
2306Return the num value of an SV, doing any necessary string or integer
2307conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2308macros.
2309
2310=cut
2311*/
2312
65202027 2313NV
864dbfa3 2314Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2315{
97aff369 2316 dVAR;
79072805
LW
2317 if (!sv)
2318 return 0.0;
8990e307 2319 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2320 mg_get(sv);
2321 if (SvNOKp(sv))
2322 return SvNVX(sv);
0aa395f8 2323 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2324 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2325 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2326 not_a_number(sv);
3f7c398e 2327 return Atof(SvPVX_const(sv));
a0d0e21e 2328 }
25da4f38 2329 if (SvIOKp(sv)) {
1c846c1f 2330 if (SvIsUV(sv))
65202027 2331 return (NV)SvUVX(sv);
25da4f38 2332 else
65202027 2333 return (NV)SvIVX(sv);
47a72cb8
NC
2334 }
2335 if (SvROK(sv)) {
2336 goto return_rok;
2337 }
2338 assert(SvTYPE(sv) >= SVt_PVMG);
2339 /* This falls through to the report_uninit near the end of the
2340 function. */
2341 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2342 if (SvROK(sv)) {
47a72cb8 2343 return_rok:
deb46114
NC
2344 if (SvAMAGIC(sv)) {
2345 SV *const tmpstr = AMG_CALLun(sv,numer);
2346 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2347 return SvNV(tmpstr);
2348 }
2349 }
2350 return PTR2NV(SvRV(sv));
a0d0e21e 2351 }
765f542d
NC
2352 if (SvIsCOW(sv)) {
2353 sv_force_normal_flags(sv, 0);
8a818333 2354 }
0336b60e 2355 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2356 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2357 report_uninit(sv);
ed6116ce
LW
2358 return 0.0;
2359 }
79072805
LW
2360 }
2361 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2362 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2363 sv_upgrade(sv, SVt_NV);
906f284f 2364#ifdef USE_LONG_DOUBLE
097ee67d 2365 DEBUG_c({
f93f4e46 2366 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2367 PerlIO_printf(Perl_debug_log,
2368 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2369 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2370 RESTORE_NUMERIC_LOCAL();
2371 });
65202027 2372#else
572bbb43 2373 DEBUG_c({
f93f4e46 2374 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2375 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2376 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2377 RESTORE_NUMERIC_LOCAL();
2378 });
572bbb43 2379#endif
79072805
LW
2380 }
2381 else if (SvTYPE(sv) < SVt_PVNV)
2382 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2383 if (SvNOKp(sv)) {
2384 return SvNVX(sv);
61604483 2385 }
59d8ce62 2386 if (SvIOKp(sv)) {
9d6ce603 2387 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2388#ifdef NV_PRESERVES_UV
2389 SvNOK_on(sv);
2390#else
2391 /* Only set the public NV OK flag if this NV preserves the IV */
2392 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2393 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2394 : (SvIVX(sv) == I_V(SvNVX(sv))))
2395 SvNOK_on(sv);
2396 else
2397 SvNOKp_on(sv);
2398#endif
93a17b20 2399 }
748a9306 2400 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2401 UV value;
3f7c398e 2402 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2403 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2404 not_a_number(sv);
28e5dec8 2405#ifdef NV_PRESERVES_UV
c2988b20
NC
2406 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2407 == IS_NUMBER_IN_UV) {
5e045b90 2408 /* It's definitely an integer */
9d6ce603 2409 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2410 } else
3f7c398e 2411 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2412 SvNOK_on(sv);
2413#else
3f7c398e 2414 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2415 /* Only set the public NV OK flag if this NV preserves the value in
2416 the PV at least as well as an IV/UV would.
2417 Not sure how to do this 100% reliably. */
2418 /* if that shift count is out of range then Configure's test is
2419 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2420 UV_BITS */
2421 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2422 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2423 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2424 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2425 /* Can't use strtol etc to convert this string, so don't try.
2426 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2427 SvNOK_on(sv);
2428 } else {
2429 /* value has been set. It may not be precise. */
2430 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2431 /* 2s complement assumption for (UV)IV_MIN */
2432 SvNOK_on(sv); /* Integer is too negative. */
2433 } else {
2434 SvNOKp_on(sv);
2435 SvIOKp_on(sv);
6fa402ec 2436
c2988b20 2437 if (numtype & IS_NUMBER_NEG) {
45977657 2438 SvIV_set(sv, -(IV)value);
c2988b20 2439 } else if (value <= (UV)IV_MAX) {
45977657 2440 SvIV_set(sv, (IV)value);
c2988b20 2441 } else {
607fa7f2 2442 SvUV_set(sv, value);
c2988b20
NC
2443 SvIsUV_on(sv);
2444 }
2445
2446 if (numtype & IS_NUMBER_NOT_INT) {
2447 /* I believe that even if the original PV had decimals,
2448 they are lost beyond the limit of the FP precision.
2449 However, neither is canonical, so both only get p
2450 flags. NWC, 2000/11/25 */
2451 /* Both already have p flags, so do nothing */
2452 } else {
66a1b24b 2453 const NV nv = SvNVX(sv);
c2988b20
NC
2454 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2455 if (SvIVX(sv) == I_V(nv)) {
2456 SvNOK_on(sv);
c2988b20 2457 } else {
c2988b20
NC
2458 /* It had no "." so it must be integer. */
2459 }
00b6aa41 2460 SvIOK_on(sv);
c2988b20
NC
2461 } else {
2462 /* between IV_MAX and NV(UV_MAX).
2463 Could be slightly > UV_MAX */
6fa402ec 2464
c2988b20
NC
2465 if (numtype & IS_NUMBER_NOT_INT) {
2466 /* UV and NV both imprecise. */
2467 } else {
66a1b24b 2468 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2469
2470 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2471 SvNOK_on(sv);
c2988b20 2472 }
00b6aa41 2473 SvIOK_on(sv);
c2988b20
NC
2474 }
2475 }
2476 }
2477 }
2478 }
28e5dec8 2479#endif /* NV_PRESERVES_UV */
93a17b20 2480 }
79072805 2481 else {
f7877b28 2482 if (isGV_with_GP(sv)) {
19f6321d 2483 glob_2number((GV *)sv);
180488f8
NC
2484 return 0.0;
2485 }
2486
041457d9 2487 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2488 report_uninit(sv);
7e25a7e9
NC
2489 assert (SvTYPE(sv) >= SVt_NV);
2490 /* Typically the caller expects that sv_any is not NULL now. */
2491 /* XXX Ilya implies that this is a bug in callers that assume this
2492 and ideally should be fixed. */
a0d0e21e 2493 return 0.0;
79072805 2494 }
572bbb43 2495#if defined(USE_LONG_DOUBLE)
097ee67d 2496 DEBUG_c({
f93f4e46 2497 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2498 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2499 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2500 RESTORE_NUMERIC_LOCAL();
2501 });
65202027 2502#else
572bbb43 2503 DEBUG_c({
f93f4e46 2504 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2505 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2506 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2507 RESTORE_NUMERIC_LOCAL();
2508 });
572bbb43 2509#endif
463ee0b2 2510 return SvNVX(sv);
79072805
LW
2511}
2512
645c22ef
DM
2513/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2514 * UV as a string towards the end of buf, and return pointers to start and
2515 * end of it.
2516 *
2517 * We assume that buf is at least TYPE_CHARS(UV) long.
2518 */
2519
864dbfa3 2520static char *
aec46f14 2521S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2522{
25da4f38 2523 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2524 char * const ebuf = ptr;
25da4f38 2525 int sign;
25da4f38
IZ
2526
2527 if (is_uv)
2528 sign = 0;
2529 else if (iv >= 0) {
2530 uv = iv;
2531 sign = 0;
2532 } else {
2533 uv = -iv;
2534 sign = 1;
2535 }
2536 do {
eb160463 2537 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2538 } while (uv /= 10);
2539 if (sign)
2540 *--ptr = '-';
2541 *peob = ebuf;
2542 return ptr;
2543}
2544
645c22ef
DM
2545/*
2546=for apidoc sv_2pv_flags
2547
ff276b08 2548Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2549If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2550if necessary.
2551Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2552usually end up here too.
2553
2554=cut
2555*/
2556
8d6d96c1
HS
2557char *
2558Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2559{
97aff369 2560 dVAR;
79072805 2561 register char *s;
79072805 2562
463ee0b2 2563 if (!sv) {
cdb061a3
NC
2564 if (lp)
2565 *lp = 0;
73d840c0 2566 return (char *)"";
463ee0b2 2567 }
8990e307 2568 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2569 if (flags & SV_GMAGIC)
2570 mg_get(sv);
463ee0b2 2571 if (SvPOKp(sv)) {
cdb061a3
NC
2572 if (lp)
2573 *lp = SvCUR(sv);
10516c54
NC
2574 if (flags & SV_MUTABLE_RETURN)
2575 return SvPVX_mutable(sv);
4d84ee25
NC
2576 if (flags & SV_CONST_RETURN)
2577 return (char *)SvPVX_const(sv);
463ee0b2
LW
2578 return SvPVX(sv);
2579 }
75dfc8ec
NC
2580 if (SvIOKp(sv) || SvNOKp(sv)) {
2581 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2582 STRLEN len;
2583
2584 if (SvIOKp(sv)) {
e80fed9d 2585 len = SvIsUV(sv)
d9fad198
JH
2586 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2587 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2588 } else {
e8ada2d0
NC
2589 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2590 len = strlen(tbuf);
75dfc8ec 2591 }
b5b886f0
NC
2592 assert(!SvROK(sv));
2593 {
75dfc8ec
NC
2594 dVAR;
2595
2596#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2597 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2598 tbuf[0] = '0';
2599 tbuf[1] = 0;
75dfc8ec
NC
2600 len = 1;
2601 }
2602#endif
2603 SvUPGRADE(sv, SVt_PV);
2604 if (lp)
2605 *lp = len;
2606 s = SvGROW_mutable(sv, len + 1);
2607 SvCUR_set(sv, len);
2608 SvPOKp_on(sv);
10edeb5d 2609 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2610 }
463ee0b2 2611 }
1c7ff15e
NC
2612 if (SvROK(sv)) {
2613 goto return_rok;
2614 }
2615 assert(SvTYPE(sv) >= SVt_PVMG);
2616 /* This falls through to the report_uninit near the end of the
2617 function. */
2618 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2619 if (SvROK(sv)) {
1c7ff15e 2620 return_rok:
deb46114
NC
2621 if (SvAMAGIC(sv)) {
2622 SV *const tmpstr = AMG_CALLun(sv,string);
2623 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2624 /* Unwrap this: */
2625 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2626 */
2627
2628 char *pv;
2629 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2630 if (flags & SV_CONST_RETURN) {
2631 pv = (char *) SvPVX_const(tmpstr);
2632 } else {
2633 pv = (flags & SV_MUTABLE_RETURN)
2634 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2635 }
2636 if (lp)
2637 *lp = SvCUR(tmpstr);
50adf7d2 2638 } else {
deb46114 2639 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2640 }
deb46114
NC
2641 if (SvUTF8(tmpstr))
2642 SvUTF8_on(sv);
2643 else
2644 SvUTF8_off(sv);
2645 return pv;
50adf7d2 2646 }
deb46114
NC
2647 }
2648 {
fafee734
NC
2649 STRLEN len;
2650 char *retval;
2651 char *buffer;
f9277f47 2652 MAGIC *mg;
d8eae41e
NC
2653 const SV *const referent = (SV*)SvRV(sv);
2654
2655 if (!referent) {
fafee734
NC
2656 len = 7;
2657 retval = buffer = savepvn("NULLREF", len);
042dae7a
NC
2658 } else if (SvTYPE(referent) == SVt_PVMG
2659 && ((SvFLAGS(referent) &
2660 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2661 == (SVs_OBJECT|SVs_SMG))
de8c5301
YO
2662 && (mg = mg_find(referent, PERL_MAGIC_qr)))
2663 {
2664 char *str = NULL;
2665 I32 haseval = 0;
60df1e07 2666 U32 flags = 0;
de8c5301
YO
2667 (str) = CALLREG_AS_STR(mg,lp,&flags,&haseval);
2668 if (flags & 1)
2669 SvUTF8_on(sv);
2670 else
2671 SvUTF8_off(sv);
2672 PL_reginterp_cnt += haseval;
2673 return str;
d8eae41e
NC
2674 } else {
2675 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2676 const STRLEN typelen = strlen(typestr);
2677 UV addr = PTR2UV(referent);
2678 const char *stashname = NULL;
2679 STRLEN stashnamelen = 0; /* hush, gcc */
2680 const char *buffer_end;
d8eae41e 2681
d8eae41e 2682 if (SvOBJECT(referent)) {
fafee734
NC
2683 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2684
2685 if (name) {
2686 stashname = HEK_KEY(name);
2687 stashnamelen = HEK_LEN(name);
2688
2689 if (HEK_UTF8(name)) {
2690 SvUTF8_on(sv);
2691 } else {
2692 SvUTF8_off(sv);
2693 }
2694 } else {
2695 stashname = "__ANON__";
2696 stashnamelen = 8;
2697 }
2698 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2699 + 2 * sizeof(UV) + 2 /* )\0 */;
2700 } else {
2701 len = typelen + 3 /* (0x */
2702 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2703 }
fafee734
NC
2704
2705 Newx(buffer, len, char);
2706 buffer_end = retval = buffer + len;
2707
2708 /* Working backwards */
2709 *--retval = '\0';
2710 *--retval = ')';
2711 do {
2712 *--retval = PL_hexdigit[addr & 15];
2713 } while (addr >>= 4);
2714 *--retval = 'x';
2715 *--retval = '0';
2716 *--retval = '(';
2717
2718 retval -= typelen;
2719 memcpy(retval, typestr, typelen);
2720
2721 if (stashname) {
2722 *--retval = '=';
2723 retval -= stashnamelen;
2724 memcpy(retval, stashname, stashnamelen);
2725 }
2726 /* retval may not neccesarily have reached the start of the
2727 buffer here. */
2728 assert (retval >= buffer);
2729
2730 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2731 }
042dae7a 2732 if (lp)
fafee734
NC
2733 *lp = len;
2734 SAVEFREEPV(buffer);
2735 return retval;
463ee0b2 2736 }
79072805 2737 }
0336b60e 2738 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2739 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2740 report_uninit(sv);
cdb061a3
NC
2741 if (lp)
2742 *lp = 0;
73d840c0 2743 return (char *)"";
79072805 2744 }
79072805 2745 }
28e5dec8
JH
2746 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2747 /* I'm assuming that if both IV and NV are equally valid then
2748 converting the IV is going to be more efficient */
e1ec3a88
AL
2749 const U32 isIOK = SvIOK(sv);
2750 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2751 char buf[TYPE_CHARS(UV)];
2752 char *ebuf, *ptr;
2753
2754 if (SvTYPE(sv) < SVt_PVIV)
2755 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2756 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2757 /* inlined from sv_setpvn */
2758 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2759 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2760 SvCUR_set(sv, ebuf - ptr);
2761 s = SvEND(sv);
2762 *s = '\0';
2763 if (isIOK)
2764 SvIOK_on(sv);
2765 else
2766 SvIOKp_on(sv);
2767 if (isUIOK)
2768 SvIsUV_on(sv);
2769 }
2770 else if (SvNOKp(sv)) {
c81271c3 2771 const int olderrno = errno;
79072805
LW
2772 if (SvTYPE(sv) < SVt_PVNV)
2773 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2774 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2775 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2776 /* some Xenix systems wipe out errno here */
79072805 2777#ifdef apollo
463ee0b2 2778 if (SvNVX(sv) == 0.0)
d1307786 2779 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2780 else
2781#endif /*apollo*/
bbce6d69 2782 {
2d4389e4 2783 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2784 }
79072805 2785 errno = olderrno;
a0d0e21e
LW
2786#ifdef FIXNEGATIVEZERO
2787 if (*s == '-' && s[1] == '0' && !s[2])
d1307786 2788 my_strlcpy(s, "0", SvLEN(s));
a0d0e21e 2789#endif
79072805
LW
2790 while (*s) s++;
2791#ifdef hcx
2792 if (s[-1] == '.')
46fc3d4c 2793 *--s = '\0';
79072805
LW
2794#endif
2795 }
79072805 2796 else {
675c862f 2797 if (isGV_with_GP(sv))
19f6321d 2798 return glob_2pv((GV *)sv, lp);
180488f8 2799
041457d9 2800 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2801 report_uninit(sv);
cdb061a3 2802 if (lp)
00b6aa41 2803 *lp = 0;
25da4f38
IZ
2804 if (SvTYPE(sv) < SVt_PV)
2805 /* Typically the caller expects that sv_any is not NULL now. */
2806 sv_upgrade(sv, SVt_PV);
73d840c0 2807 return (char *)"";
79072805 2808 }
cdb061a3 2809 {
823a54a3 2810 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2811 if (lp)
2812 *lp = len;
2813 SvCUR_set(sv, len);
2814 }
79072805 2815 SvPOK_on(sv);
1d7c1841 2816 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2817 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2818 if (flags & SV_CONST_RETURN)
2819 return (char *)SvPVX_const(sv);
10516c54
NC
2820 if (flags & SV_MUTABLE_RETURN)
2821 return SvPVX_mutable(sv);
463ee0b2
LW
2822 return SvPVX(sv);
2823}
2824
645c22ef 2825/*
6050d10e
JP
2826=for apidoc sv_copypv
2827
2828Copies a stringified representation of the source SV into the
2829destination SV. Automatically performs any necessary mg_get and
54f0641b 2830coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2831UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2832sv_2pv[_flags] but operates directly on an SV instead of just the
2833string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2834would lose the UTF-8'ness of the PV.
2835
2836=cut
2837*/
2838
2839void
2840Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2841{
446eaa42 2842 STRLEN len;
53c1dcc0 2843 const char * const s = SvPV_const(ssv,len);
cb50f42d 2844 sv_setpvn(dsv,s,len);
446eaa42 2845 if (SvUTF8(ssv))
cb50f42d 2846 SvUTF8_on(dsv);
446eaa42 2847 else
cb50f42d 2848 SvUTF8_off(dsv);
6050d10e
JP
2849}
2850
2851/*
645c22ef
DM
2852=for apidoc sv_2pvbyte
2853
2854Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2855to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2856side-effect.
2857
2858Usually accessed via the C<SvPVbyte> macro.
2859
2860=cut
2861*/
2862
7340a771
GS
2863char *
2864Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2865{
0875d2fe 2866 sv_utf8_downgrade(sv,0);
97972285 2867 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2868}
2869
645c22ef 2870/*
035cbb0e
RGS
2871=for apidoc sv_2pvutf8
2872
2873Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2874to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2875
2876Usually accessed via the C<SvPVutf8> macro.
2877
2878=cut
2879*/
645c22ef 2880
7340a771
GS
2881char *
2882Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2883{
035cbb0e
RGS
2884 sv_utf8_upgrade(sv);
2885 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2886}
1c846c1f 2887
7ee2227d 2888
645c22ef
DM
2889/*
2890=for apidoc sv_2bool
2891
2892This function is only called on magical items, and is only used by
8cf8f3d1 2893sv_true() or its macro equivalent.
645c22ef
DM
2894
2895=cut
2896*/
2897
463ee0b2 2898bool
864dbfa3 2899Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2900{
97aff369 2901 dVAR;
5b295bef 2902 SvGETMAGIC(sv);
463ee0b2 2903
a0d0e21e
LW
2904 if (!SvOK(sv))
2905 return 0;
2906 if (SvROK(sv)) {
fabdb6c0
AL
2907 if (SvAMAGIC(sv)) {
2908 SV * const tmpsv = AMG_CALLun(sv,bool_);
2909 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2910 return (bool)SvTRUE(tmpsv);
2911 }
2912 return SvRV(sv) != 0;
a0d0e21e 2913 }
463ee0b2 2914 if (SvPOKp(sv)) {
53c1dcc0
AL
2915 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2916 if (Xpvtmp &&
339049b0 2917 (*sv->sv_u.svu_pv > '0' ||
11343788 2918 Xpvtmp->xpv_cur > 1 ||
339049b0 2919 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2920 return 1;
2921 else
2922 return 0;
2923 }
2924 else {
2925 if (SvIOKp(sv))
2926 return SvIVX(sv) != 0;
2927 else {
2928 if (SvNOKp(sv))
2929 return SvNVX(sv) != 0.0;
180488f8 2930 else {
f7877b28 2931 if (isGV_with_GP(sv))
180488f8
NC
2932 return TRUE;
2933 else
2934 return FALSE;
2935 }
463ee0b2
LW
2936 }
2937 }
79072805
LW
2938}
2939
c461cf8f
JH
2940/*
2941=for apidoc sv_utf8_upgrade
2942
78ea37eb 2943Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2944Forces the SV to string form if it is not already.
4411f3b6
NIS
2945Always sets the SvUTF8 flag to avoid future validity checks even
2946if all the bytes have hibit clear.
c461cf8f 2947
13a6c0e0
JH
2948This is not as a general purpose byte encoding to Unicode interface:
2949use the Encode extension for that.
2950
8d6d96c1
HS
2951=for apidoc sv_utf8_upgrade_flags
2952
78ea37eb 2953Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2954Forces the SV to string form if it is not already.
8d6d96c1
HS
2955Always sets the SvUTF8 flag to avoid future validity checks even
2956if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2957will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2958C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2959
13a6c0e0
JH
2960This is not as a general purpose byte encoding to Unicode interface:
2961use the Encode extension for that.
2962
8d6d96c1
HS
2963=cut
2964*/
2965
2966STRLEN
2967Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2968{
97aff369 2969 dVAR;
808c356f
RGS
2970 if (sv == &PL_sv_undef)
2971 return 0;
e0e62c2a
NIS
2972 if (!SvPOK(sv)) {
2973 STRLEN len = 0;
d52b7888
NC
2974 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2975 (void) sv_2pv_flags(sv,&len, flags);
2976 if (SvUTF8(sv))
2977 return len;
2978 } else {
2979 (void) SvPV_force(sv,len);
2980 }
e0e62c2a 2981 }
4411f3b6 2982
f5cee72b 2983 if (SvUTF8(sv)) {
5fec3b1d 2984 return SvCUR(sv);
f5cee72b 2985 }
5fec3b1d 2986
765f542d
NC
2987 if (SvIsCOW(sv)) {
2988 sv_force_normal_flags(sv, 0);
db42d148
NIS
2989 }
2990
88632417 2991 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2992 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2993 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2994 /* This function could be much more efficient if we
2995 * had a FLAG in SVs to signal if there are any hibit
2996 * chars in the PV. Given that there isn't such a flag
2997 * make the loop as fast as possible. */
00b6aa41 2998 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 2999 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3000 const U8 *t = s;
c4e7c712
NC
3001
3002 while (t < e) {
53c1dcc0 3003 const U8 ch = *t++;
00b6aa41
AL
3004 /* Check for hi bit */
3005 if (!NATIVE_IS_INVARIANT(ch)) {
3006 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3007 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3008
3009 SvPV_free(sv); /* No longer using what was there before. */
3010 SvPV_set(sv, (char*)recoded);
3011 SvCUR_set(sv, len - 1);
3012 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3013 break;
00b6aa41 3014 }
c4e7c712
NC
3015 }
3016 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3017 SvUTF8_on(sv);
560a288e 3018 }
4411f3b6 3019 return SvCUR(sv);
560a288e
GS
3020}
3021
c461cf8f
JH
3022/*
3023=for apidoc sv_utf8_downgrade
3024
78ea37eb
TS
3025Attempts to convert the PV of an SV from characters to bytes.
3026If the PV contains a character beyond byte, this conversion will fail;
3027in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3028true, croaks.
3029
13a6c0e0
JH
3030This is not as a general purpose Unicode to byte encoding interface:
3031use the Encode extension for that.
3032
c461cf8f
JH
3033=cut
3034*/
3035
560a288e
GS
3036bool
3037Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3038{
97aff369 3039 dVAR;
78ea37eb 3040 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3041 if (SvCUR(sv)) {
03cfe0ae 3042 U8 *s;
652088fc 3043 STRLEN len;
fa301091 3044
765f542d
NC
3045 if (SvIsCOW(sv)) {
3046 sv_force_normal_flags(sv, 0);
3047 }
03cfe0ae
NIS
3048 s = (U8 *) SvPV(sv, len);
3049 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3050 if (fail_ok)
3051 return FALSE;
3052 else {
3053 if (PL_op)
3054 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3055 OP_DESC(PL_op));
fa301091
JH
3056 else
3057 Perl_croak(aTHX_ "Wide character");
3058 }
4b3603a4 3059 }
b162af07 3060 SvCUR_set(sv, len);
67e989fb 3061 }
560a288e 3062 }
ffebcc3e 3063 SvUTF8_off(sv);
560a288e
GS
3064 return TRUE;
3065}
3066
c461cf8f
JH
3067/*
3068=for apidoc sv_utf8_encode
3069
78ea37eb
TS
3070Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3071flag off so that it looks like octets again.
c461cf8f
JH
3072
3073=cut
3074*/
3075
560a288e
GS
3076void
3077Perl_sv_utf8_encode(pTHX_ register SV *sv)
3078{
4c94c214
NC
3079 if (SvIsCOW(sv)) {
3080 sv_force_normal_flags(sv, 0);
3081 }
3082 if (SvREADONLY(sv)) {
3083 Perl_croak(aTHX_ PL_no_modify);
3084 }
a5f5288a 3085 (void) sv_utf8_upgrade(sv);
560a288e
GS
3086 SvUTF8_off(sv);
3087}
3088
4411f3b6
NIS
3089/*
3090=for apidoc sv_utf8_decode
3091
78ea37eb
TS
3092If the PV of the SV is an octet sequence in UTF-8
3093and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3094so that it looks like a character. If the PV contains only single-byte
3095characters, the C<SvUTF8> flag stays being off.
3096Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3097
3098=cut
3099*/
3100
560a288e
GS
3101bool
3102Perl_sv_utf8_decode(pTHX_ register SV *sv)
3103{
78ea37eb 3104 if (SvPOKp(sv)) {
93524f2b
NC
3105 const U8 *c;
3106 const U8 *e;
9cbac4c7 3107
645c22ef
DM
3108 /* The octets may have got themselves encoded - get them back as
3109 * bytes
3110 */
3111 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3112 return FALSE;
3113
3114 /* it is actually just a matter of turning the utf8 flag on, but
3115 * we want to make sure everything inside is valid utf8 first.
3116 */
93524f2b 3117 c = (const U8 *) SvPVX_const(sv);
63cd0674 3118 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3119 return FALSE;
93524f2b 3120 e = (const U8 *) SvEND(sv);
511c2ff0 3121 while (c < e) {
b64e5050 3122 const U8 ch = *c++;
c4d5f83a 3123 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3124 SvUTF8_on(sv);
3125 break;
3126 }
560a288e 3127 }
560a288e
GS
3128 }
3129 return TRUE;
3130}
3131
954c1994
GS
3132/*
3133=for apidoc sv_setsv
3134
645c22ef
DM
3135Copies the contents of the source SV C<ssv> into the destination SV
3136C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3137function if the source SV needs to be reused. Does not handle 'set' magic.
3138Loosely speaking, it performs a copy-by-value, obliterating any previous
3139content of the destination.
3140
3141You probably want to use one of the assortment of wrappers, such as
3142C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3143C<SvSetMagicSV_nosteal>.
3144
8d6d96c1
HS
3145=for apidoc sv_setsv_flags
3146
645c22ef
DM
3147Copies the contents of the source SV C<ssv> into the destination SV
3148C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3149function if the source SV needs to be reused. Does not handle 'set' magic.
3150Loosely speaking, it performs a copy-by-value, obliterating any previous
3151content of the destination.
3152If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3153C<ssv> if appropriate, else not. If the C<flags> parameter has the
3154C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3155and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3156
3157You probably want to use one of the assortment of wrappers, such as
3158C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3159C<SvSetMagicSV_nosteal>.
3160
3161This is the primary function for copying scalars, and most other
3162copy-ish functions and macros use this underneath.
8d6d96c1
HS
3163
3164=cut
3165*/
3166
5d0301b7 3167static void
2eb42952 3168S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3169{
3170 if (dtype != SVt_PVGV) {
3171 const char * const name = GvNAME(sstr);
3172 const STRLEN len = GvNAMELEN(sstr);
3173 /* don't upgrade SVt_PVLV: it can hold a glob */
f7877b28
NC
3174 if (dtype != SVt_PVLV) {
3175 if (dtype >= SVt_PV) {
3176 SvPV_free(dstr);
3177 SvPV_set(dstr, 0);
3178 SvLEN_set(dstr, 0);
3179 SvCUR_set(dstr, 0);
3180 }
5d0301b7 3181 sv_upgrade(dstr, SVt_PVGV);
dedf8e73
NC
3182 (void)SvOK_off(dstr);
3183 SvSCREAM_on(dstr);
f7877b28 3184 }
5d0301b7
NC
3185 GvSTASH(dstr) = GvSTASH(sstr);
3186 if (GvSTASH(dstr))
3187 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3188 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3189 SvFAKE_on(dstr); /* can coerce to non-glob */
3190 }
3191
3192#ifdef GV_UNIQUE_CHECK
3193 if (GvUNIQUE((GV*)dstr)) {
3194 Perl_croak(aTHX_ PL_no_modify);
3195 }
3196#endif
3197
f7877b28
NC
3198 gp_free((GV*)dstr);
3199 SvSCREAM_off(dstr);
5d0301b7 3200 (void)SvOK_off(dstr);
f7877b28 3201 SvSCREAM_on(dstr);
dedf8e73 3202 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3203 GvGP(dstr) = gp_ref(GvGP(sstr));
3204 if (SvTAINTED(sstr))
3205 SvTAINT(dstr);
3206 if (GvIMPORTED(dstr) != GVf_IMPORTED
3207 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3208 {
3209 GvIMPORTED_on(dstr);
3210 }
3211 GvMULTI_on(dstr);
3212 return;
3213}
3214
b8473700 3215static void
2eb42952 3216S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3217 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3218 SV *dref = NULL;
3219 const int intro = GvINTRO(dstr);
2440974c 3220 SV **location;
3386d083 3221 U8 import_flag = 0;
27242d61
NC
3222 const U32 stype = SvTYPE(sref);
3223
b8473700
NC
3224
3225#ifdef GV_UNIQUE_CHECK
3226 if (GvUNIQUE((GV*)dstr)) {
3227 Perl_croak(aTHX_ PL_no_modify);
3228 }
3229#endif
3230
3231 if (intro) {
3232 GvINTRO_off(dstr); /* one-shot flag */
3233 GvLINE(dstr) = CopLINE(PL_curcop);
3234 GvEGV(dstr) = (GV*)dstr;
3235 }
3236 GvMULTI_on(dstr);
27242d61 3237 switch (stype) {
b8473700 3238 case SVt_PVCV:
27242d61
NC
3239 location = (SV **) &GvCV(dstr);
3240 import_flag = GVf_IMPORTED_CV;
3241 goto common;
3242 case SVt_PVHV:
3243 location = (SV **) &GvHV(dstr);
3244 import_flag = GVf_IMPORTED_HV;
3245 goto common;
3246 case SVt_PVAV:
3247 location = (SV **) &GvAV(dstr);
3248 import_flag = GVf_IMPORTED_AV;
3249 goto common;
3250 case SVt_PVIO:
3251 location = (SV **) &GvIOp(dstr);
3252 goto common;
3253 case SVt_PVFM:
3254 location = (SV **) &GvFORM(dstr);
3255 default:
3256 location = &GvSV(dstr);
3257 import_flag = GVf_IMPORTED_SV;
3258 common:
b8473700 3259 if (intro) {
27242d61
NC
3260 if (stype == SVt_PVCV) {
3261 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3262 SvREFCNT_dec(GvCV(dstr));
3263 GvCV(dstr) = NULL;
3264 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3265 PL_sub_generation++;
3266 }
b8473700 3267 }
27242d61 3268 SAVEGENERICSV(*location);
b8473700
NC
3269 }
3270 else
27242d61
NC
3271 dref = *location;
3272 if (stype == SVt_PVCV && *location != sref) {
3273 CV* const cv = (CV*)*location;
b8473700
NC
3274 if (cv) {
3275 if (!GvCVGEN((GV*)dstr) &&
3276 (CvROOT(cv) || CvXSUB(cv)))
3277 {
3278 /* Redefining a sub - warning is mandatory if
3279 it was a const and its value changed. */
3280 if (CvCONST(cv) && CvCONST((CV*)sref)
3281 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3282 NOOP;
b8473700
NC
3283 /* They are 2 constant subroutines generated from
3284 the same constant. This probably means that
3285 they are really the "same" proxy subroutine
3286 instantiated in 2 places. Most likely this is
3287 when a constant is exported twice. Don't warn.
3288 */
3289 }
3290 else if (ckWARN(WARN_REDEFINE)
3291 || (CvCONST(cv)
3292 && (!CvCONST((CV*)sref)
3293 || sv_cmp(cv_const_sv(cv),
3294 cv_const_sv((CV*)sref))))) {
3295 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3296 (const char *)
3297 (CvCONST(cv)
3298 ? "Constant subroutine %s::%s redefined"
3299 : "Subroutine %s::%s redefined"),
b8473700
NC
3300 HvNAME_get(GvSTASH((GV*)dstr)),
3301 GvENAME((GV*)dstr));
3302 }
3303 }
3304 if (!intro)
cbf82dd0
NC
3305 cv_ckproto_len(cv, (GV*)dstr,
3306 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3307 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3308 }
b8473700
NC
3309 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3310 GvASSUMECV_on(dstr);
3311 PL_sub_generation++;
3312 }
2440974c 3313 *location = sref;
3386d083
NC
3314 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3315 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3316 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3317 }
3318 break;
3319 }
b37c2d43 3320 SvREFCNT_dec(dref);
b8473700
NC
3321 if (SvTAINTED(sstr))
3322 SvTAINT(dstr);
3323 return;
3324}
3325
8d6d96c1
HS
3326void
3327Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3328{
97aff369 3329 dVAR;
8990e307
LW
3330 register U32 sflags;
3331 register int dtype;
42d0e0b7 3332 register svtype stype;
463ee0b2 3333
79072805
LW
3334 if (sstr == dstr)
3335 return;
29f4f0ab
NC
3336
3337 if (SvIS_FREED(dstr)) {
3338 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3339 " to a freed scalar %p", sstr, dstr);
3340 }
765f542d 3341 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3342 if (!sstr)
3280af22 3343 sstr = &PL_sv_undef;
29f4f0ab
NC
3344 if (SvIS_FREED(sstr)) {
3345 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p", sstr,
3346 dstr);
3347 }
8990e307
LW
3348 stype = SvTYPE(sstr);
3349 dtype = SvTYPE(dstr);
79072805 3350
a0d0e21e 3351 SvAMAGIC_off(dstr);
7a5fa8a2 3352 if ( SvVOK(dstr) )
ece467f9
JP
3353 {
3354 /* need to nuke the magic */
3355 mg_free(dstr);
3356 SvRMAGICAL_off(dstr);
3357 }
9e7bc3e8 3358
463ee0b2 3359 /* There's a lot of redundancy below but we're going for speed here */
79072805 3360
8990e307 3361 switch (stype) {
79072805 3362 case SVt_NULL:
aece5585 3363 undef_sstr:
20408e3c
GS
3364 if (dtype != SVt_PVGV) {
3365 (void)SvOK_off(dstr);
3366 return;
3367 }
3368 break;
463ee0b2 3369 case SVt_IV:
aece5585
GA
3370 if (SvIOK(sstr)) {
3371 switch (dtype) {
3372 case SVt_NULL:
8990e307 3373 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3374 break;
3375 case SVt_NV:
aece5585
GA
3376 case SVt_RV:
3377 case SVt_PV:
a0d0e21e 3378 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
3379 break;
3380 }
3381 (void)SvIOK_only(dstr);
45977657 3382 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3383 if (SvIsUV(sstr))
3384 SvIsUV_on(dstr);
37c25af0
NC
3385 /* SvTAINTED can only be true if the SV has taint magic, which in
3386 turn means that the SV type is PVMG (or greater). This is the
3387 case statement for SVt_IV, so this cannot be true (whatever gcov
3388 may say). */
3389 assert(!SvTAINTED(sstr));
aece5585 3390 return;
8990e307 3391 }
aece5585
GA
3392 goto undef_sstr;
3393
463ee0b2 3394 case SVt_NV:
aece5585
GA
3395 if (SvNOK(sstr)) {
3396 switch (dtype) {
3397 case SVt_NULL:
3398 case SVt_IV:
8990e307 3399 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3400 break;
3401 case SVt_RV:
3402 case SVt_PV:
3403 case SVt_PVIV:
a0d0e21e 3404 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
3405 break;
3406 }
9d6ce603 3407 SvNV_set(dstr, SvNVX(sstr));
aece5585 3408 (void)SvNOK_only(dstr);
37c25af0
NC
3409 /* SvTAINTED can only be true if the SV has taint magic, which in
3410 turn means that the SV type is PVMG (or greater). This is the
3411 case statement for SVt_NV, so this cannot be true (whatever gcov
3412 may say). */
3413 assert(!SvTAINTED(sstr));
aece5585 3414 return;
8990e307 3415 }
aece5585
GA
3416 goto undef_sstr;
3417
ed6116ce 3418 case SVt_RV:
8990e307 3419 if (dtype < SVt_RV)
ed6116ce 3420 sv_upgrade(dstr, SVt_RV);
ed6116ce 3421 break;
fc36a67e 3422 case SVt_PVFM:
f8c7b90f 3423#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3424 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3425 if (dtype < SVt_PVIV)
3426 sv_upgrade(dstr, SVt_PVIV);
3427 break;
3428 }
3429 /* Fall through */
3430#endif
3431 case SVt_PV:
8990e307 3432 if (dtype < SVt_PV)
463ee0b2 3433 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3434 break;
3435 case SVt_PVIV:
8990e307 3436 if (dtype < SVt_PVIV)
463ee0b2 3437 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3438 break;
3439 case SVt_PVNV:
8990e307 3440 if (dtype < SVt_PVNV)
463ee0b2 3441 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3442 break;
489f7bfe 3443 default:
a3b680e6
AL
3444 {
3445 const char * const type = sv_reftype(sstr,0);
533c011a 3446 if (PL_op)
a3b680e6 3447 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3448 else
a3b680e6
AL
3449 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3450 }
4633a7c4
LW
3451 break;
3452
79072805 3453 case SVt_PVGV:
8990e307 3454 if (dtype <= SVt_PVGV) {
d4c19fe8 3455 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3456 return;
79072805 3457 }
5f66b61c 3458 /*FALLTHROUGH*/
79072805 3459
489f7bfe
NC
3460 case SVt_PVMG:
3461 case SVt_PVLV:
3462 case SVt_PVBM:
8d6d96c1 3463 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3464 mg_get(sstr);
1d9c78c6 3465 if (SvTYPE(sstr) != stype) {
973f89ab 3466 stype = SvTYPE(sstr);
b8c701c1 3467 if (stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3468 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3469 return;
3470 }
973f89ab
CS
3471 }
3472 }
ded42b9f 3473 if (stype == SVt_PVLV)
862a34c6 3474 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3475 else
42d0e0b7 3476 SvUPGRADE(dstr, (svtype)stype);
79072805
LW
3477 }
3478
ff920335
NC
3479 /* dstr may have been upgraded. */
3480 dtype = SvTYPE(dstr);
8990e307
LW
3481 sflags = SvFLAGS(sstr);
3482
3483 if (sflags & SVf_ROK) {
acaa9288
NC
3484 if (dtype == SVt_PVGV &&
3485 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
3486 sstr = SvRV(sstr);
3487 if (sstr == dstr) {
3488 if (GvIMPORTED(dstr) != GVf_IMPORTED
3489 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3490 {
3491 GvIMPORTED_on(dstr);
3492 }
3493 GvMULTI_on(dstr);
3494 return;
3495 }
d4c19fe8 3496 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3497 return;
3498 }
3499
8990e307 3500 if (dtype >= SVt_PV) {
b8c701c1 3501 if (dtype == SVt_PVGV) {
d4c19fe8 3502 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3503 return;
3504 }
3f7c398e 3505 if (SvPVX_const(dstr)) {
8bd4d4c5 3506 SvPV_free(dstr);
b162af07
SP
3507 SvLEN_set(dstr, 0);
3508 SvCUR_set(dstr, 0);
a0d0e21e 3509 }
8990e307 3510 }
a0d0e21e 3511 (void)SvOK_off(dstr);
b162af07 3512 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
96d4b0ee 3513 SvFLAGS(dstr) |= sflags & SVf_ROK;
dfd48732
NC
3514 assert(!(sflags & SVp_NOK));
3515 assert(!(sflags & SVp_IOK));
3516 assert(!(sflags & SVf_NOK));
3517 assert(!(sflags & SVf_IOK));
ed6116ce 3518 }
c0c44674
NC
3519 else if (dtype == SVt_PVGV) {
3520 if (!(sflags & SVf_OK)) {
3521 if (ckWARN(WARN_MISC))
3522 Perl_warner(aTHX_ packWARN(WARN_MISC),
3523 "Undefined value assigned to typeglob");
3524 }
3525 else {
3526 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3527 if (dstr != (SV*)gv) {
3528 if (GvGP(dstr))
3529 gp_free((GV*)dstr);
3530 GvGP(dstr) = gp_ref(GvGP(gv));
3531 }
3532 }
3533 }
8990e307 3534 else if (sflags & SVp_POK) {
765f542d 3535 bool isSwipe = 0;
79072805
LW
3536
3537 /*
3538 * Check to see if we can just swipe the string. If so, it's a
3539 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3540 * It might even be a win on short strings if SvPVX_const(dstr)
3541 * has to be allocated and SvPVX_const(sstr) has to be freed.
34482cd6
NC
3542 * Likewise if we can set up COW rather than doing an actual copy, we
3543 * drop to the else clause, as the swipe code and the COW setup code
3544 * have much in common.
79072805
LW
3545 */
3546
120fac95
NC
3547 /* Whichever path we take through the next code, we want this true,
3548 and doing it now facilitates the COW check. */
3549 (void)SvPOK_only(dstr);
3550
765f542d 3551 if (
34482cd6
NC
3552 /* If we're already COW then this clause is not true, and if COW
3553 is allowed then we drop down to the else and make dest COW
3554 with us. If caller hasn't said that we're allowed to COW
3555 shared hash keys then we don't do the COW setup, even if the
3556 source scalar is a shared hash key scalar. */
3557 (((flags & SV_COW_SHARED_HASH_KEYS)
3558 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
3559 : 1 /* If making a COW copy is forbidden then the behaviour we
3560 desire is as if the source SV isn't actually already
3561 COW, even if it is. So we act as if the source flags
3562 are not COW, rather than actually testing them. */
3563 )
f8c7b90f 3564#ifndef PERL_OLD_COPY_ON_WRITE
34482cd6
NC
3565 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
3566 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
3567 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
3568 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
3569 but in turn, it's somewhat dead code, never expected to go
3570 live, but more kept as a placeholder on how to do it better
3571 in a newer implementation. */
3572 /* If we are COW and dstr is a suitable target then we drop down
3573 into the else and make dest a COW of us. */
b8f9541a
NC
3574 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3575#endif
3576 )
765f542d 3577 &&
765f542d
NC
3578 !(isSwipe =
3579 (sflags & SVs_TEMP) && /* slated for free anyway? */
3580 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3581 (!(flags & SV_NOSTEAL)) &&
3582 /* and we're allowed to steal temps */
765f542d
NC
3583 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3584 SvLEN(sstr) && /* and really is a string */
645c22ef 3585 /* and won't be needed again, potentially */
765f542d 3586 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3587#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3588 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3589 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3590 && SvTYPE(sstr) >= SVt_PVIV)
3591#endif
3592 ) {
3593 /* Failed the swipe test, and it's not a shared hash key either.
3594 Have to copy the string. */
3595 STRLEN len = SvCUR(sstr);
3596 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3597 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3598 SvCUR_set(dstr, len);
3599 *SvEND(dstr) = '\0';
765f542d 3600 } else {
f8c7b90f 3601 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3602 be true in here. */
765f542d
NC
3603 /* Either it's a shared hash key, or it's suitable for
3604 copy-on-write or we can swipe the string. */
46187eeb 3605 if (DEBUG_C_TEST) {
ed252734 3606 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3607 sv_dump(sstr);
3608 sv_dump(dstr);
46187eeb 3609 }
f8c7b90f 3610#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3611 if (!isSwipe) {
3612 /* I believe I should acquire a global SV mutex if
3613 it's a COW sv (not a shared hash key) to stop
3614 it going un copy-on-write.
3615 If the source SV has gone un copy on write between up there
3616 and down here, then (assert() that) it is of the correct
3617 form to make it copy on write again */
3618 if ((sflags & (SVf_FAKE | SVf_READONLY))
3619 != (SVf_FAKE | SVf_READONLY)) {
3620 SvREADONLY_on(sstr);
3621 SvFAKE_on(sstr);
3622 /* Make the source SV into a loop of 1.
3623 (about to become 2) */
a29f6d03 3624 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3625 }
3626 }
3627#endif
3628 /* Initial code is common. */
94010e71
NC
3629 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3630 SvPV_free(dstr);
79072805 3631 }
765f542d 3632
765f542d
NC
3633 if (!isSwipe) {
3634 /* making another shared SV. */
3635 STRLEN cur = SvCUR(sstr);
3636 STRLEN len = SvLEN(sstr);
f8c7b90f 3637#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3638 if (len) {
b8f9541a 3639 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3640 /* SvIsCOW_normal */
3641 /* splice us in between source and next-after-source. */
a29f6d03
NC
3642 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3643 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3644 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3645 } else
3646#endif
3647 {
765f542d 3648 /* SvIsCOW_shared_hash */
46187eeb
NC
3649 DEBUG_C(PerlIO_printf(Perl_debug_log,
3650 "Copy on write: Sharing hash\n"));
b8f9541a 3651
bdd68bc3 3652 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3653 SvPV_set(dstr,
d1db91c6 3654 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3655 }
87a1ef3d
SP
3656 SvLEN_set(dstr, len);
3657 SvCUR_set(dstr, cur);
765f542d
NC
3658 SvREADONLY_on(dstr);
3659 SvFAKE_on(dstr);
3660 /* Relesase a global SV mutex. */
3661 }
3662 else
765f542d 3663 { /* Passes the swipe test. */
78d1e721 3664 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3665 SvLEN_set(dstr, SvLEN(sstr));
3666 SvCUR_set(dstr, SvCUR(sstr));
3667
3668 SvTEMP_off(dstr);
3669 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3670 SvPV_set(sstr, NULL);
765f542d
NC
3671 SvLEN_set(sstr, 0);
3672 SvCUR_set(sstr, 0);
3673 SvTEMP_off(sstr);
3674 }
3675 }
8990e307 3676 if (sflags & SVp_NOK) {
9d6ce603 3677 SvNV_set(dstr, SvNVX(sstr));
79072805 3678 }
8990e307 3679 if (sflags & SVp_IOK) {
23525414
NC
3680 SvRELEASE_IVX(dstr);
3681 SvIV_set(dstr, SvIVX(sstr));
3682 /* Must do this otherwise some other overloaded use of 0x80000000
3683 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3684 if (sflags & SVf_IVisUV)
25da4f38 3685 SvIsUV_on(dstr);
79072805 3686 }
96d4b0ee 3687 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4f2da183 3688 {
b0a11fe1 3689 const MAGIC * const smg = SvVSTRING_mg(sstr);
4f2da183
NC
3690 if (smg) {
3691 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3692 smg->mg_ptr, smg->mg_len);
3693 SvRMAGICAL_on(dstr);
3694 }
7a5fa8a2 3695 }
79072805 3696 }
5d581361 3697 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3698 (void)SvOK_off(dstr);
96d4b0ee 3699 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
5d581361
NC
3700 if (sflags & SVp_IOK) {
3701 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3702 SvIV_set(dstr, SvIVX(sstr));
3703 }
3332b3c1 3704 if (sflags & SVp_NOK) {
9d6ce603 3705 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3706 }
3707 }
79072805 3708 else {
f7877b28 3709 if (isGV_with_GP(sstr)) {
180488f8
NC
3710 /* This stringification rule for globs is spread in 3 places.
3711 This feels bad. FIXME. */
3712 const U32 wasfake = sflags & SVf_FAKE;
3713
3714 /* FAKE globs can get coerced, so need to turn this off
3715 temporarily if it is on. */
3716 SvFAKE_off(sstr);
3717 gv_efullname3(dstr, (GV *)sstr, "*");
3718 SvFLAGS(sstr) |= wasfake;
3719 }
20408e3c
GS
3720 else
3721 (void)SvOK_off(dstr);
a0d0e21e 3722 }
27c9684d
AP
3723 if (SvTAINTED(sstr))
3724 SvTAINT(dstr);
79072805
LW
3725}
3726
954c1994
GS
3727/*
3728=for apidoc sv_setsv_mg
3729
3730Like C<sv_setsv>, but also handles 'set' magic.
3731
3732=cut
3733*/
3734
79072805 3735void
864dbfa3 3736Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3737{
3738 sv_setsv(dstr,sstr);
3739 SvSETMAGIC(dstr);
3740}
3741
f8c7b90f 3742#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3743SV *
3744Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3745{
3746 STRLEN cur = SvCUR(sstr);
3747 STRLEN len = SvLEN(sstr);
3748 register char *new_pv;
3749
3750 if (DEBUG_C_TEST) {
3751 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3752 sstr, dstr);
3753 sv_dump(sstr);
3754 if (dstr)
3755 sv_dump(dstr);
3756 }
3757
3758 if (dstr) {
3759 if (SvTHINKFIRST(dstr))
3760 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3761 else if (SvPVX_const(dstr))
3762 Safefree(SvPVX_const(dstr));
ed252734
NC
3763 }
3764 else
3765 new_SV(dstr);
862a34c6 3766 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3767
3768 assert (SvPOK(sstr));
3769 assert (SvPOKp(sstr));
3770 assert (!SvIOK(sstr));
3771 assert (!SvIOKp(sstr));
3772 assert (!SvNOK(sstr));
3773 assert (!SvNOKp(sstr));
3774
3775 if (SvIsCOW(sstr)) {
3776
3777 if (SvLEN(sstr) == 0) {
3778 /* source is a COW shared hash key. */
ed252734
NC
3779 DEBUG_C(PerlIO_printf(Perl_debug_log,
3780 "Fast copy on write: Sharing hash\n"));
d1db91c6 3781 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3782 goto common_exit;
3783 }
3784 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3785 } else {
3786 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3787 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3788 SvREADONLY_on(sstr);
3789 SvFAKE_on(sstr);
3790 DEBUG_C(PerlIO_printf(Perl_debug_log,
3791 "Fast copy on write: Converting sstr to COW\n"));
3792 SV_COW_NEXT_SV_SET(dstr, sstr);
3793 }
3794 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3795 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3796
3797 common_exit:
3798 SvPV_set(dstr, new_pv);
3799 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3800 if (SvUTF8(sstr))
3801 SvUTF8_on(dstr);
87a1ef3d
SP
3802 SvLEN_set(dstr, len);
3803 SvCUR_set(dstr, cur);
ed252734
NC
3804 if (DEBUG_C_TEST) {
3805 sv_dump(dstr);
3806 }
3807 return dstr;
3808}
3809#endif
3810
954c1994
GS
3811/*
3812=for apidoc sv_setpvn
3813
3814Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3815bytes to be copied. If the C<ptr> argument is NULL the SV will become
3816undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3817
3818=cut
3819*/
3820
ef50df4b 3821void
864dbfa3 3822Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3823{
97aff369 3824 dVAR;
c6f8c383 3825 register char *dptr;
22c522df 3826
765f542d 3827 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3828 if (!ptr) {
a0d0e21e 3829 (void)SvOK_off(sv);
463ee0b2
LW
3830 return;
3831 }
22c522df
JH
3832 else {
3833 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3834 const IV iv = len;
9c5ffd7c
JH
3835 if (iv < 0)
3836 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3837 }
862a34c6 3838 SvUPGRADE(sv, SVt_PV);
c6f8c383 3839
5902b6a9 3840 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3841 Move(ptr,dptr,len,char);
3842 dptr[len] = '\0';
79072805 3843 SvCUR_set(sv, len);
1aa99e6b 3844 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3845 SvTAINT(sv);
79072805
LW
3846}
3847
954c1994
GS
3848/*
3849=for apidoc sv_setpvn_mg
3850
3851Like C<sv_setpvn>, but also handles 'set' magic.
3852
3853=cut
3854*/
3855
79072805 3856void
864dbfa3 3857Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3858{
3859 sv_setpvn(sv,ptr,len);
3860 SvSETMAGIC(sv);
3861}
3862
954c1994
GS
3863/*
3864=for apidoc sv_setpv
3865
3866Copies a string into an SV. The string must be null-terminated. Does not
3867handle 'set' magic. See C<sv_setpv_mg>.
3868
3869=cut
3870*/
3871
ef50df4b 3872void
864dbfa3 3873Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3874{
97aff369 3875 dVAR;
79072805
LW
3876 register STRLEN len;
3877
765f542d 3878 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3879 if (!ptr) {
a0d0e21e 3880 (void)SvOK_off(sv);
463ee0b2
LW
3881 return;
3882 }
79072805 3883 len = strlen(ptr);
862a34c6 3884 SvUPGRADE(sv, SVt_PV);
c6f8c383 3885
79072805 3886 SvGROW(sv, len + 1);
463ee0b2 3887 Move(ptr,SvPVX(sv),len+1,char);
79072805 3888 SvCUR_set(sv, len);
1aa99e6b 3889 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3890 SvTAINT(sv);
3891}
3892
954c1994
GS
3893/*
3894=for apidoc sv_setpv_mg
3895
3896Like C<sv_setpv>, but also handles 'set' magic.
3897
3898=cut
3899*/
3900
463ee0b2 3901void
864dbfa3 3902Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3903{
3904 sv_setpv(sv,ptr);
3905 SvSETMAGIC(sv);
3906}
3907
954c1994 3908/*
47518d95 3909=for apidoc sv_usepvn_flags
954c1994 3910
794a0d33
JH
3911Tells an SV to use C<ptr> to find its string value. Normally the
3912string is stored inside the SV but sv_usepvn allows the SV to use an
3913outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
3914by C<malloc>. The string length, C<len>, must be supplied. By default
3915this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
3916so that pointer should not be freed or used by the programmer after
3917giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
3918that pointer (e.g. ptr + 1) be used.
3919
3920If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3921SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 3922will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 3923C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3924
3925=cut
3926*/
3927
ef50df4b 3928void
47518d95 3929Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3930{
97aff369 3931 dVAR;
1936d2a7 3932 STRLEN allocate;
765f542d 3933 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3934 SvUPGRADE(sv, SVt_PV);
463ee0b2 3935 if (!ptr) {
a0d0e21e 3936 (void)SvOK_off(sv);
47518d95
NC
3937 if (flags & SV_SMAGIC)
3938 SvSETMAGIC(sv);
463ee0b2
LW
3939 return;
3940 }
3f7c398e 3941 if (SvPVX_const(sv))
8bd4d4c5 3942 SvPV_free(sv);
1936d2a7 3943
0b7042f9 3944#ifdef DEBUGGING
2e90b4cd
NC
3945 if (flags & SV_HAS_TRAILING_NUL)
3946 assert(ptr[len] == '\0');
0b7042f9 3947#endif
2e90b4cd 3948
c1c21316 3949 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3950 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
cbf82dd0
NC
3951 if (flags & SV_HAS_TRAILING_NUL) {
3952 /* It's long enough - do nothing.
3953 Specfically Perl_newCONSTSUB is relying on this. */
3954 } else {
69d25b4f 3955#ifdef DEBUGGING
69d25b4f 3956 /* Force a move to shake out bugs in callers. */
10edeb5d 3957 char *new_ptr = (char*)safemalloc(allocate);
69d25b4f
NC
3958 Copy(ptr, new_ptr, len, char);
3959 PoisonFree(ptr,len,char);
3960 Safefree(ptr);
3961 ptr = new_ptr;
69d25b4f 3962#else
10edeb5d 3963 ptr = (char*) saferealloc (ptr, allocate);
69d25b4f 3964#endif
cbf82dd0 3965 }
f880fe2f 3966 SvPV_set(sv, ptr);
463ee0b2 3967 SvCUR_set(sv, len);
1936d2a7 3968 SvLEN_set(sv, allocate);
c1c21316
NC
3969 if (!(flags & SV_HAS_TRAILING_NUL)) {
3970 *SvEND(sv) = '\0';
3971 }
1aa99e6b 3972 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3973 SvTAINT(sv);
47518d95
NC
3974 if (flags & SV_SMAGIC)
3975 SvSETMAGIC(sv);
ef50df4b
GS
3976}
3977
f8c7b90f 3978#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3979/* Need to do this *after* making the SV normal, as we need the buffer
3980 pointer to remain valid until after we've copied it. If we let go too early,
3981 another thread could invalidate it by unsharing last of the same hash key
3982 (which it can do by means other than releasing copy-on-write Svs)
3983 or by changing the other copy-on-write SVs in the loop. */
3984STATIC void
bdd68bc3 3985S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3986{
3987 if (len) { /* this SV was SvIsCOW_normal(sv) */
3988 /* we need to find the SV pointing to us. */
cf5629ad 3989 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3990
765f542d
NC
3991 if (current == sv) {
3992 /* The SV we point to points back to us (there were only two of us
3993 in the loop.)
3994 Hence other SV is no longer copy on write either. */
3995 SvFAKE_off(after);
3996 SvREADONLY_off(after);
3997 } else {
3998 /* We need to follow the pointers around the loop. */
3999 SV *next;
4000 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4001 assert (next);
4002 current = next;
4003 /* don't loop forever if the structure is bust, and we have
4004 a pointer into a closed loop. */
4005 assert (current != after);
3f7c398e 4006 assert (SvPVX_const(current) == pvx);
765f542d
NC
4007 }
4008 /* Make the SV before us point to the SV after us. */
a29f6d03 4009 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
4010 }
4011 } else {
bdd68bc3 4012 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
4013 }
4014}
4015
4016int
4017Perl_sv_release_IVX(pTHX_ register SV *sv)
4018{
4019 if (SvIsCOW(sv))
4020 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
4021 SvOOK_off(sv);
4022 return 0;
765f542d
NC
4023}
4024#endif
645c22ef
DM
4025/*
4026=for apidoc sv_force_normal_flags
4027
4028Undo various types of fakery on an SV: if the PV is a shared string, make
4029a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4030an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4031we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4032then a copy-on-write scalar drops its PV buffer (if any) and becomes
4033SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4034set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4035C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4036with flags set to 0.
645c22ef
DM
4037
4038=cut
4039*/
4040
6fc92669 4041void
840a7b70 4042Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4043{
97aff369 4044 dVAR;
f8c7b90f 4045#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4046 if (SvREADONLY(sv)) {
4047 /* At this point I believe I should acquire a global SV mutex. */
4048 if (SvFAKE(sv)) {
b64e5050 4049 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4050 const STRLEN len = SvLEN(sv);
4051 const STRLEN cur = SvCUR(sv);
a28509cc 4052 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4053 if (DEBUG_C_TEST) {
4054 PerlIO_printf(Perl_debug_log,
4055 "Copy on write: Force normal %ld\n",
4056 (long) flags);
e419cbc5 4057 sv_dump(sv);
46187eeb 4058 }
765f542d
NC
4059 SvFAKE_off(sv);
4060 SvREADONLY_off(sv);
9f653bb5 4061 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4062 SvPV_set(sv, NULL);
87a1ef3d 4063 SvLEN_set(sv, 0);
765f542d
NC
4064 if (flags & SV_COW_DROP_PV) {
4065 /* OK, so we don't need to copy our buffer. */
4066 SvPOK_off(sv);
4067 } else {
4068 SvGROW(sv, cur + 1);
4069 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4070 SvCUR_set(sv, cur);
765f542d
NC
4071 *SvEND(sv) = '\0';
4072 }
bdd68bc3 4073 sv_release_COW(sv, pvx, len, next);
46187eeb 4074 if (DEBUG_C_TEST) {
e419cbc5 4075 sv_dump(sv);
46187eeb 4076 }
765f542d 4077 }
923e4eb5 4078 else if (IN_PERL_RUNTIME)
765f542d
NC
4079 Perl_croak(aTHX_ PL_no_modify);
4080 /* At this point I believe that I can drop the global SV mutex. */
4081 }
4082#else
2213622d 4083 if (SvREADONLY(sv)) {
1c846c1f 4084 if (SvFAKE(sv)) {
b64e5050 4085 const char * const pvx = SvPVX_const(sv);
66a1b24b 4086 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4087 SvFAKE_off(sv);
4088 SvREADONLY_off(sv);
bd61b366 4089 SvPV_set(sv, NULL);
66a1b24b 4090 SvLEN_set(sv, 0);
1c846c1f 4091 SvGROW(sv, len + 1);
706aa1c9 4092 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4093 *SvEND(sv) = '\0';
bdd68bc3 4094 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4095 }
923e4eb5 4096 else if (IN_PERL_RUNTIME)
cea2e8a9 4097 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4098 }
765f542d 4099#endif
2213622d 4100 if (SvROK(sv))
840a7b70 4101 sv_unref_flags(sv, flags);
6fc92669
GS
4102 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4103 sv_unglob(sv);
0f15f207 4104}
1c846c1f 4105
645c22ef 4106/*
954c1994
GS
4107=for apidoc sv_chop
4108
1c846c1f 4109Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4110SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4111the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4112string. Uses the "OOK hack".
3f7c398e 4113Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4114refer to the same chunk of data.
954c1994
GS
4115
4116=cut
4117*/
4118
79072805 4119void
f54cb97a 4120Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4121{
4122 register STRLEN delta;
a0d0e21e 4123 if (!ptr || !SvPOKp(sv))
79072805 4124 return;
3f7c398e 4125 delta = ptr - SvPVX_const(sv);
2213622d 4126 SV_CHECK_THINKFIRST(sv);
79072805
LW
4127 if (SvTYPE(sv) < SVt_PVIV)
4128 sv_upgrade(sv,SVt_PVIV);
4129
4130 if (!SvOOK(sv)) {
50483b2c 4131 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4132 const char *pvx = SvPVX_const(sv);
a28509cc 4133 const STRLEN len = SvCUR(sv);
50483b2c 4134 SvGROW(sv, len + 1);
706aa1c9 4135 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4136 *SvEND(sv) = '\0';
4137 }
45977657 4138 SvIV_set(sv, 0);
a4bfb290
AB
4139 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4140 and we do that anyway inside the SvNIOK_off
4141 */
7a5fa8a2 4142 SvFLAGS(sv) |= SVf_OOK;
79072805 4143 }
a4bfb290 4144 SvNIOK_off(sv);
b162af07
SP
4145 SvLEN_set(sv, SvLEN(sv) - delta);
4146 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4147 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4148 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4149}
4150
954c1994
GS
4151/*
4152=for apidoc sv_catpvn
4153
4154Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4155C<len> indicates number of bytes to copy. If the SV has the UTF-8
4156status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4157Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4158
8d6d96c1
HS
4159=for apidoc sv_catpvn_flags
4160
4161Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4162C<len> indicates number of bytes to copy. If the SV has the UTF-8
4163status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4164If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4165appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4166in terms of this function.
4167
4168=cut
4169*/
4170
4171void
4172Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4173{
97aff369 4174 dVAR;
8d6d96c1 4175 STRLEN dlen;
fabdb6c0 4176 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4177
8d6d96c1
HS
4178 SvGROW(dsv, dlen + slen + 1);
4179 if (sstr == dstr)
3f7c398e 4180 sstr = SvPVX_const(dsv);
8d6d96c1 4181 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4182 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4183 *SvEND(dsv) = '\0';
4184 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4185 SvTAINT(dsv);
bddd5118
NC
4186 if (flags & SV_SMAGIC)
4187 SvSETMAGIC(dsv);
79072805
LW
4188}
4189
954c1994 4190/*
954c1994
GS
4191=for apidoc sv_catsv
4192
13e8c8e3
JH
4193Concatenates the string from SV C<ssv> onto the end of the string in
4194SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4195not 'set' magic. See C<sv_catsv_mg>.
954c1994 4196
8d6d96c1
HS
4197=for apidoc sv_catsv_flags
4198
4199Concatenates the string from SV C<ssv> onto the end of the string in
4200SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4201bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4202and C<sv_catsv_nomg> are implemented in terms of this function.
4203
4204=cut */
4205
ef50df4b 4206void
8d6d96c1 4207Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4208{
97aff369 4209 dVAR;
bddd5118 4210 if (ssv) {
00b6aa41
AL
4211 STRLEN slen;
4212 const char *spv = SvPV_const(ssv, slen);
4213 if (spv) {
bddd5118
NC
4214 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4215 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4216 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4217 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4218 dsv->sv_flags doesn't have that bit set.
4fd84b44 4219 Andy Dougherty 12 Oct 2001
bddd5118
NC
4220 */
4221 const I32 sutf8 = DO_UTF8(ssv);
4222 I32 dutf8;
13e8c8e3 4223
bddd5118
NC
4224 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4225 mg_get(dsv);
4226 dutf8 = DO_UTF8(dsv);
8d6d96c1 4227
bddd5118
NC
4228 if (dutf8 != sutf8) {
4229 if (dutf8) {
4230 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4231 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4232
bddd5118
NC
4233 sv_utf8_upgrade(csv);
4234 spv = SvPV_const(csv, slen);
4235 }
4236 else
4237 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4238 }
bddd5118 4239 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4240 }
560a288e 4241 }
bddd5118
NC
4242 if (flags & SV_SMAGIC)
4243 SvSETMAGIC(dsv);
79072805
LW
4244}
4245
954c1994 4246/*
954c1994
GS
4247=for apidoc sv_catpv
4248
4249Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4250If the SV has the UTF-8 status set, then the bytes appended should be
4251valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4252
d5ce4a7c 4253=cut */
954c1994 4254
ef50df4b 4255void
0c981600 4256Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4257{
97aff369 4258 dVAR;
79072805 4259 register STRLEN len;
463ee0b2 4260 STRLEN tlen;
748a9306 4261 char *junk;
79072805 4262
0c981600 4263 if (!ptr)
79072805 4264 return;
748a9306 4265 junk = SvPV_force(sv, tlen);
0c981600 4266 len = strlen(ptr);
463ee0b2 4267 SvGROW(sv, tlen + len + 1);
0c981600 4268 if (ptr == junk)
3f7c398e 4269 ptr = SvPVX_const(sv);
0c981600 4270 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4271 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4272 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4273 SvTAINT(sv);
79072805
LW
4274}
4275
954c1994
GS
4276/*
4277=for apidoc sv_catpv_mg
4278
4279Like C<sv_catpv>, but also handles 'set' magic.
4280
4281=cut
4282*/
4283
ef50df4b 4284void
0c981600 4285Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4286{
0c981600 4287 sv_catpv(sv,ptr);
ef50df4b
GS
4288 SvSETMAGIC(sv);
4289}
4290
645c22ef
DM
4291/*
4292=for apidoc newSV
4293
561b68a9
SH
4294Creates a new SV. A non-zero C<len> parameter indicates the number of
4295bytes of preallocated string space the SV should have. An extra byte for a
4296trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4297space is allocated.) The reference count for the new SV is set to 1.
4298
4299In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4300parameter, I<x>, a debug aid which allowed callers to identify themselves.
4301This aid has been superseded by a new build option, PERL_MEM_LOG (see
4302L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4303modules supporting older perls.
645c22ef
DM
4304
4305=cut
4306*/
4307
79072805 4308SV *
864dbfa3 4309Perl_newSV(pTHX_ STRLEN len)
79072805 4310{
97aff369 4311 dVAR;
79072805 4312 register SV *sv;
1c846c1f 4313
4561caa4 4314 new_SV(sv);
79072805
LW
4315 if (len) {
4316 sv_upgrade(sv, SVt_PV);
4317 SvGROW(sv, len + 1);
4318 }
4319 return sv;
4320}
954c1994 4321/*
92110913 4322=for apidoc sv_magicext
954c1994 4323
68795e93 4324Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4325supplied vtable and returns a pointer to the magic added.
92110913 4326
2d8d5d5a
SH
4327Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4328In particular, you can add magic to SvREADONLY SVs, and add more than
4329one instance of the same 'how'.
645c22ef 4330
2d8d5d5a
SH
4331If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4332stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4333special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4334to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4335
2d8d5d5a 4336(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4337
4338=cut
4339*/
92110913 4340MAGIC *
92e67595 4341Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4342 const char* name, I32 namlen)
79072805 4343{
97aff369 4344 dVAR;
79072805 4345 MAGIC* mg;
68795e93 4346
92110913 4347 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4348 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4349 }
a02a5408 4350 Newxz(mg, 1, MAGIC);
79072805 4351 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4352 SvMAGIC_set(sv, mg);
75f9d97a 4353
05f95b08
SB
4354 /* Sometimes a magic contains a reference loop, where the sv and
4355 object refer to each other. To prevent a reference loop that
4356 would prevent such objects being freed, we look for such loops
4357 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4358
4359 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4360 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4361
4362 */
14befaf4
DM
4363 if (!obj || obj == sv ||
4364 how == PERL_MAGIC_arylen ||
4365 how == PERL_MAGIC_qr ||
8d2f4536 4366 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4367 (SvTYPE(obj) == SVt_PVGV &&
4368 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4369 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4370 GvFORM(obj) == (CV*)sv)))
75f9d97a 4371 {
8990e307 4372 mg->mg_obj = obj;
75f9d97a 4373 }
85e6fe83 4374 else {
b37c2d43 4375 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4376 mg->mg_flags |= MGf_REFCOUNTED;
4377 }
b5ccf5f2
YST
4378
4379 /* Normal self-ties simply pass a null object, and instead of
4380 using mg_obj directly, use the SvTIED_obj macro to produce a
4381 new RV as needed. For glob "self-ties", we are tieing the PVIO
4382 with an RV obj pointing to the glob containing the PVIO. In
4383 this case, to avoid a reference loop, we need to weaken the
4384 reference.
4385 */
4386
4387 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4388 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4389 {
4390 sv_rvweaken(obj);
4391 }
4392
79072805 4393 mg->mg_type = how;
565764a8 4394 mg->mg_len = namlen;
9cbac4c7 4395 if (name) {
92110913 4396 if (namlen > 0)
1edc1566 4397 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4398 else if (namlen == HEf_SVKEY)
b37c2d43 4399 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4400 else
92110913 4401 mg->mg_ptr = (char *) name;
9cbac4c7 4402 }
92110913 4403 mg->mg_virtual = vtable;
68795e93 4404
92110913
NIS
4405 mg_magical(sv);
4406 if (SvGMAGICAL(sv))
4407 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4408 return mg;
4409}
4410
4411/*
4412=for apidoc sv_magic
1c846c1f 4413
92110913
NIS
4414Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4415then adds a new magic item of type C<how> to the head of the magic list.
4416
2d8d5d5a
SH
4417See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4418handling of the C<name> and C<namlen> arguments.
4419
4509d3fb
SB
4420You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4421to add more than one instance of the same 'how'.
4422
92110913
NIS
4423=cut
4424*/
4425
4426void
4427Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4428{
97aff369 4429 dVAR;
92e67595 4430 MGVTBL *vtable;
92110913 4431 MAGIC* mg;
92110913 4432
f8c7b90f 4433#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4434 if (SvIsCOW(sv))
4435 sv_force_normal_flags(sv, 0);
4436#endif
92110913 4437 if (SvREADONLY(sv)) {
d8084ca5
DM
4438 if (
4439 /* its okay to attach magic to shared strings; the subsequent
4440 * upgrade to PVMG will unshare the string */
4441 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4442
4443 && IN_PERL_RUNTIME
92110913
NIS
4444 && how != PERL_MAGIC_regex_global
4445 && how != PERL_MAGIC_bm
4446 && how != PERL_MAGIC_fm
4447 && how != PERL_MAGIC_sv
e6469971 4448 && how != PERL_MAGIC_backref
92110913
NIS
4449 )
4450 {
4451 Perl_croak(aTHX_ PL_no_modify);
4452 }
4453 }
4454 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4455 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4456 /* sv_magic() refuses to add a magic of the same 'how' as an
4457 existing one
92110913 4458 */
2a509ed3 4459 if (how == PERL_MAGIC_taint) {
92110913 4460 mg->mg_len |= 1;
2a509ed3
NC
4461 /* Any scalar which already had taint magic on which someone
4462 (erroneously?) did SvIOK_on() or similar will now be
4463 incorrectly sporting public "OK" flags. */
4464 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4465 }
92110913
NIS
4466 return;
4467 }
4468 }
68795e93 4469
79072805 4470 switch (how) {
14befaf4 4471 case PERL_MAGIC_sv:
92110913 4472 vtable = &PL_vtbl_sv;
79072805 4473 break;
14befaf4 4474 case PERL_MAGIC_overload:
92110913 4475 vtable = &PL_vtbl_amagic;
a0d0e21e 4476 break;
14befaf4 4477 case PERL_MAGIC_overload_elem:
92110913 4478 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4479 break;
14befaf4 4480 case PERL_MAGIC_overload_table:
92110913 4481 vtable = &PL_vtbl_ovrld;
a0d0e21e 4482 break;
14befaf4 4483 case PERL_MAGIC_bm:
92110913 4484 vtable = &PL_vtbl_bm;
79072805 4485 break;
14befaf4 4486 case PERL_MAGIC_regdata:
92110913 4487 vtable = &PL_vtbl_regdata;
6cef1e77 4488 break;
31e9c0d8
RGS
4489 case PERL_MAGIC_regdata_names:
4490 vtable = &PL_vtbl_regdata_names;
4491 break;
14befaf4 4492 case PERL_MAGIC_regdatum:
92110913 4493 vtable = &PL_vtbl_regdatum;
6cef1e77 4494 break;
14befaf4 4495 case PERL_MAGIC_env:
92110913 4496 vtable = &PL_vtbl_env;
79072805 4497 break;
14befaf4 4498 case PERL_MAGIC_fm:
92110913 4499 vtable = &PL_vtbl_fm;
55497cff 4500 break;
14befaf4 4501 case PERL_MAGIC_envelem:
92110913 4502 vtable = &PL_vtbl_envelem;
79072805 4503 break;
14befaf4 4504 case PERL_MAGIC_regex_global:
92110913 4505 vtable = &PL_vtbl_mglob;
93a17b20 4506 break;
14befaf4 4507 case PERL_MAGIC_isa:
92110913 4508 vtable = &PL_vtbl_isa;
463ee0b2 4509 break;
14befaf4 4510 case PERL_MAGIC_isaelem:
92110913 4511 vtable = &PL_vtbl_isaelem;
463ee0b2 4512 break;
14befaf4 4513 case PERL_MAGIC_nkeys:
92110913 4514 vtable = &PL_vtbl_nkeys;
16660edb 4515 break;
14befaf4 4516 case PERL_MAGIC_dbfile:
aec46f14 4517 vtable = NULL;
93a17b20 4518 break;
14befaf4 4519 case PERL_MAGIC_dbline:
92110913 4520 vtable = &PL_vtbl_dbline;
79072805 4521 break;
36477c24 4522#ifdef USE_LOCALE_COLLATE
14befaf4 4523 case PERL_MAGIC_collxfrm:
92110913 4524 vtable = &PL_vtbl_collxfrm;
bbce6d69 4525 break;
36477c24 4526#endif /* USE_LOCALE_COLLATE */
14befaf4 4527 case PERL_MAGIC_tied:
92110913 4528 vtable = &PL_vtbl_pack;
463ee0b2 4529 break;
14befaf4
DM
4530 case PERL_MAGIC_tiedelem:
4531 case PERL_MAGIC_tiedscalar:
92110913 4532 vtable = &PL_vtbl_packelem;
463ee0b2 4533 break;
14befaf4 4534 case PERL_MAGIC_qr:
92110913 4535 vtable = &PL_vtbl_regexp;
c277df42 4536 break;
b3ca2e83
NC
4537 case PERL_MAGIC_hints:
4538 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4539 case PERL_MAGIC_sig:
92110913 4540 vtable = &PL_vtbl_sig;
79072805 4541 break;
14befaf4 4542 case PERL_MAGIC_sigelem:
92110913 4543 vtable = &PL_vtbl_sigelem;
79072805 4544 break;
14befaf4 4545 case PERL_MAGIC_taint:
92110913 4546 vtable = &PL_vtbl_taint;
463ee0b2 4547 break;
14befaf4 4548 case PERL_MAGIC_uvar:
92110913 4549 vtable = &PL_vtbl_uvar;
79072805 4550 break;
14befaf4 4551 case PERL_MAGIC_vec:
92110913 4552 vtable = &PL_vtbl_vec;
79072805 4553 break;
a3874608 4554 case PERL_MAGIC_arylen_p:
bfcb3514 4555 case PERL_MAGIC_rhash:
8d2f4536 4556 case PERL_MAGIC_symtab:
ece467f9 4557 case PERL_MAGIC_vstring:
aec46f14 4558 vtable = NULL;
ece467f9 4559 break;
7e8c5dac
HS
4560 case PERL_MAGIC_utf8:
4561 vtable = &PL_vtbl_utf8;
4562 break;
14befaf4 4563 case PERL_MAGIC_substr:
92110913 4564 vtable = &PL_vtbl_substr;
79072805 4565 break;
14befaf4 4566 case PERL_MAGIC_defelem:
92110913 4567 vtable = &PL_vtbl_defelem;
5f05dabc 4568 break;
14befaf4 4569 case PERL_MAGIC_arylen:
92110913 4570 vtable = &PL_vtbl_arylen;
79072805 4571 break;
14befaf4 4572 case PERL_MAGIC_pos:
92110913 4573 vtable = &PL_vtbl_pos;
a0d0e21e 4574 break;
14befaf4 4575 case PERL_MAGIC_backref:
92110913 4576 vtable = &PL_vtbl_backref;
810b8aa5 4577 break;
b3ca2e83
NC
4578 case PERL_MAGIC_hintselem:
4579 vtable = &PL_vtbl_hintselem;
4580 break;
14befaf4
DM
4581 case PERL_MAGIC_ext:
4582 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4583 /* Useful for attaching extension internal data to perl vars. */
4584 /* Note that multiple extensions may clash if magical scalars */
4585 /* etc holding private data from one are passed to another. */
aec46f14 4586 vtable = NULL;
a0d0e21e 4587 break;
79072805 4588 default:
14befaf4 4589 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4590 }
68795e93 4591
92110913 4592 /* Rest of work is done else where */
aec46f14 4593 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4594
92110913
NIS
4595 switch (how) {
4596 case PERL_MAGIC_taint:
4597 mg->mg_len = 1;
4598 break;
4599 case PERL_MAGIC_ext:
4600 case PERL_MAGIC_dbfile:
4601 SvRMAGICAL_on(sv);
4602 break;
4603 }
463ee0b2
LW
4604}
4605
c461cf8f
JH
4606/*
4607=for apidoc sv_unmagic
4608
645c22ef 4609Removes all magic of type C<type> from an SV.
c461cf8f
JH
4610
4611=cut
4612*/
4613
463ee0b2 4614int
864dbfa3 4615Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4616{
4617 MAGIC* mg;
4618 MAGIC** mgp;
91bba347 4619 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4620 return 0;
064cf529 4621 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4622 for (mg = *mgp; mg; mg = *mgp) {
4623 if (mg->mg_type == type) {
e1ec3a88 4624 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4625 *mgp = mg->mg_moremagic;
1d7c1841 4626 if (vtbl && vtbl->svt_free)
fc0dc3b3 4627 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4628 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4629 if (mg->mg_len > 0)
1edc1566 4630 Safefree(mg->mg_ptr);
565764a8 4631 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4632 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4633 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4634 Safefree(mg->mg_ptr);
9cbac4c7 4635 }
a0d0e21e
LW
4636 if (mg->mg_flags & MGf_REFCOUNTED)
4637 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4638 Safefree(mg);
4639 }
4640 else
4641 mgp = &mg->mg_moremagic;
79072805 4642 }
91bba347 4643 if (!SvMAGIC(sv)) {
463ee0b2 4644 SvMAGICAL_off(sv);
c268c2a6 4645 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4646 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4647 }
4648
4649 return 0;
79072805
LW
4650}
4651
c461cf8f
JH
4652/*
4653=for apidoc sv_rvweaken
4654
645c22ef
DM
4655Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4656referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4657push a back-reference to this RV onto the array of backreferences
1e73acc8
AS
4658associated with that magic. If the RV is magical, set magic will be
4659called after the RV is cleared.
c461cf8f
JH
4660
4661=cut
4662*/
4663
810b8aa5 4664SV *
864dbfa3 4665Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4666{
4667 SV *tsv;
4668 if (!SvOK(sv)) /* let undefs pass */
4669 return sv;
4670 if (!SvROK(sv))
cea2e8a9 4671 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4672 else if (SvWEAKREF(sv)) {
810b8aa5 4673 if (ckWARN(WARN_MISC))
9014280d 4674 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4675 return sv;
4676 }
4677 tsv = SvRV(sv);
e15faf7d 4678 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4679 SvWEAKREF_on(sv);
1c846c1f 4680 SvREFCNT_dec(tsv);
810b8aa5
GS
4681 return sv;
4682}
4683
645c22ef
DM
4684/* Give tsv backref magic if it hasn't already got it, then push a
4685 * back-reference to sv onto the array associated with the backref magic.
4686 */
4687
e15faf7d
NC
4688void
4689Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4690{
97aff369 4691 dVAR;
810b8aa5 4692 AV *av;
86f55936
NC
4693
4694 if (SvTYPE(tsv) == SVt_PVHV) {
4695 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4696
4697 av = *avp;
4698 if (!av) {
4699 /* There is no AV in the offical place - try a fixup. */
4700 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4701
4702 if (mg) {
4703 /* Aha. They've got it stowed in magic. Bring it back. */
4704 av = (AV*)mg->mg_obj;
4705 /* Stop mg_free decreasing the refernce count. */
4706 mg->mg_obj = NULL;
4707 /* Stop mg_free even calling the destructor, given that
4708 there's no AV to free up. */
4709 mg->mg_virtual = 0;
4710 sv_unmagic(tsv, PERL_MAGIC_backref);
4711 } else {
4712 av = newAV();
4713 AvREAL_off(av);
b37c2d43 4714 SvREFCNT_inc_simple_void(av);
86f55936
NC
4715 }
4716 *avp = av;
4717 }
4718 } else {
4719 const MAGIC *const mg
4720 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4721 if (mg)
4722 av = (AV*)mg->mg_obj;
4723 else {
4724 av = newAV();
4725 AvREAL_off(av);
4726 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4727 /* av now has a refcnt of 2, which avoids it getting freed
4728 * before us during global cleanup. The extra ref is removed
4729 * by magic_killbackrefs() when tsv is being freed */
4730 }
810b8aa5 4731 }
d91d49e8 4732 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4733 av_extend(av, AvFILLp(av)+1);
4734 }
4735 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4736}
4737
645c22ef
DM
4738/* delete a back-reference to ourselves from the backref magic associated
4739 * with the SV we point to.
4740 */
4741
1c846c1f 4742STATIC void
e15faf7d 4743S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4744{
97aff369 4745 dVAR;
86f55936 4746 AV *av = NULL;
810b8aa5
GS
4747 SV **svp;
4748 I32 i;
86f55936
NC
4749
4750 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4751 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4752 /* We mustn't attempt to "fix up" the hash here by moving the
4753 backreference array back to the hv_aux structure, as that is stored
4754 in the main HvARRAY(), and hfreentries assumes that no-one
4755 reallocates HvARRAY() while it is running. */
86f55936
NC
4756 }
4757 if (!av) {
4758 const MAGIC *const mg
4759 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4760 if (mg)
4761 av = (AV *)mg->mg_obj;
4762 }
4763 if (!av) {
e15faf7d
NC
4764 if (PL_in_clean_all)
4765 return;
cea2e8a9 4766 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4767 }
4768
4769 if (SvIS_FREED(av))
4770 return;
4771
810b8aa5 4772 svp = AvARRAY(av);
6a76db8b
NC
4773 /* We shouldn't be in here more than once, but for paranoia reasons lets
4774 not assume this. */
4775 for (i = AvFILLp(av); i >= 0; i--) {
4776 if (svp[i] == sv) {
4777 const SSize_t fill = AvFILLp(av);
4778 if (i != fill) {
4779 /* We weren't the last entry.
4780 An unordered list has this property that you can take the
4781 last element off the end to fill the hole, and it's still
4782 an unordered list :-)
4783 */
4784 svp[i] = svp[fill];
4785 }
a0714e2c 4786 svp[fill] = NULL;
6a76db8b
NC
4787 AvFILLp(av) = fill - 1;
4788 }
4789 }
810b8aa5
GS
4790}
4791
86f55936
NC
4792int
4793Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4794{
4795 SV **svp = AvARRAY(av);
4796
4797 PERL_UNUSED_ARG(sv);
4798
4799 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4800 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4801 if (svp && !SvIS_FREED(av)) {
4802 SV *const *const last = svp + AvFILLp(av);
4803
4804 while (svp <= last) {
4805 if (*svp) {
4806 SV *const referrer = *svp;
4807 if (SvWEAKREF(referrer)) {
4808 /* XXX Should we check that it hasn't changed? */
4809 SvRV_set(referrer, 0);
4810 SvOK_off(referrer);
4811 SvWEAKREF_off(referrer);
1e73acc8 4812 SvSETMAGIC(referrer);
86f55936
NC
4813 } else if (SvTYPE(referrer) == SVt_PVGV ||
4814 SvTYPE(referrer) == SVt_PVLV) {
4815 /* You lookin' at me? */
4816 assert(GvSTASH(referrer));
4817 assert(GvSTASH(referrer) == (HV*)sv);
4818 GvSTASH(referrer) = 0;
4819 } else {
4820 Perl_croak(aTHX_
4821 "panic: magic_killbackrefs (flags=%"UVxf")",
4822 (UV)SvFLAGS(referrer));
4823 }
4824
a0714e2c 4825 *svp = NULL;
86f55936
NC
4826 }
4827 svp++;
4828 }
4829 }
4830 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4831 return 0;
4832}
4833
954c1994
GS
4834/*
4835=for apidoc sv_insert
4836
4837Inserts a string at the specified offset/length within the SV. Similar to
4838the Perl substr() function.
4839
4840=cut
4841*/
4842
79072805 4843void
e1ec3a88 4844Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4845{
97aff369 4846 dVAR;
79072805
LW
4847 register char *big;
4848 register char *mid;
4849 register char *midend;
4850 register char *bigend;
4851 register I32 i;
6ff81951 4852 STRLEN curlen;
1c846c1f 4853
79072805 4854
8990e307 4855 if (!bigstr)
cea2e8a9 4856 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4857 SvPV_force(bigstr, curlen);
60fa28ff 4858 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4859 if (offset + len > curlen) {
4860 SvGROW(bigstr, offset+len+1);
93524f2b 4861 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4862 SvCUR_set(bigstr, offset+len);
4863 }
79072805 4864
69b47968 4865 SvTAINT(bigstr);
79072805
LW
4866 i = littlelen - len;
4867 if (i > 0) { /* string might grow */
a0d0e21e 4868 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4869 mid = big + offset + len;
4870 midend = bigend = big + SvCUR(bigstr);
4871 bigend += i;
4872 *bigend = '\0';
4873 while (midend > mid) /* shove everything down */
4874 *--bigend = *--midend;
4875 Move(little,big+offset,littlelen,char);
b162af07 4876 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4877 SvSETMAGIC(bigstr);
4878 return;
4879 }
4880 else if (i == 0) {
463ee0b2 4881 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4882 SvSETMAGIC(bigstr);
4883 return;
4884 }
4885
463ee0b2 4886 big = SvPVX(bigstr);
79072805
LW
4887 mid = big + offset;
4888 midend = mid + len;
4889 bigend = big + SvCUR(bigstr);
4890
4891 if (midend > bigend)
cea2e8a9 4892 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4893
4894 if (mid - big > bigend - midend) { /* faster to shorten from end */
4895 if (littlelen) {
4896 Move(little, mid, littlelen,char);
4897 mid += littlelen;
4898 }
4899 i = bigend - midend;
4900 if (i > 0) {
4901 Move(midend, mid, i,char);
4902 mid += i;
4903 }
4904 *mid = '\0';
4905 SvCUR_set(bigstr, mid - big);
4906 }
155aba94 4907 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4908 midend -= littlelen;
4909 mid = midend;
4910 sv_chop(bigstr,midend-i);
4911 big += i;
4912 while (i--)
4913 *--midend = *--big;
4914 if (littlelen)
4915 Move(little, mid, littlelen,char);
4916 }
4917 else if (littlelen) {
4918 midend -= littlelen;
4919 sv_chop(bigstr,midend);
4920 Move(little,midend,littlelen,char);
4921 }
4922 else {
4923 sv_chop(bigstr,midend);
4924 }
4925 SvSETMAGIC(bigstr);
4926}
4927
c461cf8f
JH
4928/*
4929=for apidoc sv_replace
4930
4931Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4932The target SV physically takes over ownership of the body of the source SV
4933and inherits its flags; however, the target keeps any magic it owns,
4934and any magic in the source is discarded.
ff276b08 4935Note that this is a rather specialist SV copying operation; most of the
645c22ef 4936time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4937
4938=cut
4939*/
79072805
LW
4940
4941void
864dbfa3 4942Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4943{
97aff369 4944 dVAR;
a3b680e6 4945 const U32 refcnt = SvREFCNT(sv);
765f542d 4946 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4947 if (SvREFCNT(nsv) != 1) {
7437becc 4948 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4949 UVuf " != 1)", (UV) SvREFCNT(nsv));
4950 }
93a17b20 4951 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4952 if (SvMAGICAL(nsv))
4953 mg_free(nsv);
4954 else
4955 sv_upgrade(nsv, SVt_PVMG);
b162af07 4956 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4957 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4958 SvMAGICAL_off(sv);
b162af07 4959 SvMAGIC_set(sv, NULL);
93a17b20 4960 }
79072805
LW
4961 SvREFCNT(sv) = 0;
4962 sv_clear(sv);
477f5d66 4963 assert(!SvREFCNT(sv));
fd0854ff
DM
4964#ifdef DEBUG_LEAKING_SCALARS
4965 sv->sv_flags = nsv->sv_flags;
4966 sv->sv_any = nsv->sv_any;
4967 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4968 sv->sv_u = nsv->sv_u;
fd0854ff 4969#else
79072805 4970 StructCopy(nsv,sv,SV);
fd0854ff 4971#endif
7b2c381c
NC
4972 /* Currently could join these into one piece of pointer arithmetic, but
4973 it would be unclear. */
4974 if(SvTYPE(sv) == SVt_IV)
4975 SvANY(sv)
339049b0 4976 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4977 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4978 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4979 }
4980
fd0854ff 4981
f8c7b90f 4982#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4983 if (SvIsCOW_normal(nsv)) {
4984 /* We need to follow the pointers around the loop to make the
4985 previous SV point to sv, rather than nsv. */
4986 SV *next;
4987 SV *current = nsv;
4988 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4989 assert(next);
4990 current = next;
3f7c398e 4991 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4992 }
4993 /* Make the SV before us point to the SV after us. */
4994 if (DEBUG_C_TEST) {
4995 PerlIO_printf(Perl_debug_log, "previous is\n");
4996 sv_dump(current);
a29f6d03
NC
4997 PerlIO_printf(Perl_debug_log,
4998 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4999 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5000 }
a29f6d03 5001 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
5002 }
5003#endif
79072805 5004 SvREFCNT(sv) = refcnt;
1edc1566 5005 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 5006 SvREFCNT(nsv) = 0;
463ee0b2 5007 del_SV(nsv);
79072805
LW
5008}
5009
c461cf8f
JH
5010/*
5011=for apidoc sv_clear
5012
645c22ef
DM
5013Clear an SV: call any destructors, free up any memory used by the body,
5014and free the body itself. The SV's head is I<not> freed, although
5015its type is set to all 1's so that it won't inadvertently be assumed
5016to be live during global destruction etc.
5017This function should only be called when REFCNT is zero. Most of the time
5018you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5019instead.
c461cf8f
JH
5020
5021=cut
5022*/
5023
79072805 5024void
864dbfa3 5025Perl_sv_clear(pTHX_ register SV *sv)
79072805 5026{
27da23d5 5027 dVAR;
82bb6deb 5028 const U32 type = SvTYPE(sv);
8edfc514
NC
5029 const struct body_details *const sv_type_details
5030 = bodies_by_type + type;
82bb6deb 5031
79072805
LW
5032 assert(sv);
5033 assert(SvREFCNT(sv) == 0);
5034
d2a0f284
JC
5035 if (type <= SVt_IV) {
5036 /* See the comment in sv.h about the collusion between this early
5037 return and the overloading of the NULL and IV slots in the size
5038 table. */
82bb6deb 5039 return;
d2a0f284 5040 }
82bb6deb 5041
ed6116ce 5042 if (SvOBJECT(sv)) {
3280af22 5043 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5044 dSP;
893645bd 5045 HV* stash;
d460ef45 5046 do {
b464bac0 5047 CV* destructor;
4e8e7886 5048 stash = SvSTASH(sv);
32251b26 5049 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5050 if (destructor) {
1b6737cc 5051 SV* const tmpref = newRV(sv);
5cc433a6 5052 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5053 ENTER;
e788e7d3 5054 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5055 EXTEND(SP, 2);
5056 PUSHMARK(SP);
5cc433a6 5057 PUSHs(tmpref);
4e8e7886 5058 PUTBACK;
44389ee9 5059 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5060
5061
d3acc0f7 5062 POPSTACK;
3095d977 5063 SPAGAIN;
4e8e7886 5064 LEAVE;
5cc433a6
AB
5065 if(SvREFCNT(tmpref) < 2) {
5066 /* tmpref is not kept alive! */
5067 SvREFCNT(sv)--;
b162af07 5068 SvRV_set(tmpref, NULL);
5cc433a6
AB
5069 SvROK_off(tmpref);
5070 }
5071 SvREFCNT_dec(tmpref);
4e8e7886
GS
5072 }
5073 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5074
6f44e0a4
JP
5075
5076 if (SvREFCNT(sv)) {
5077 if (PL_in_clean_objs)
cea2e8a9 5078 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5079 HvNAME_get(stash));
6f44e0a4
JP
5080 /* DESTROY gave object new lease on life */
5081 return;
5082 }
a0d0e21e 5083 }
4e8e7886 5084
a0d0e21e 5085 if (SvOBJECT(sv)) {
4e8e7886 5086 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5087 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5088 if (type != SVt_PVIO)
3280af22 5089 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5090 }
463ee0b2 5091 }
82bb6deb 5092 if (type >= SVt_PVMG) {
885ffcb3
NC
5093 if ((type == SVt_PVMG || type == SVt_PVGV) && SvPAD_OUR(sv)) {
5094 SvREFCNT_dec(OURSTASH(sv));
e736a858 5095 } else if (SvMAGIC(sv))
524189f1 5096 mg_free(sv);
00b1698f 5097 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5098 SvREFCNT_dec(SvSTASH(sv));
5099 }
82bb6deb 5100 switch (type) {
8990e307 5101 case SVt_PVIO:
df0bd2f4
GS
5102 if (IoIFP(sv) &&
5103 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5104 IoIFP(sv) != PerlIO_stdout() &&
5105 IoIFP(sv) != PerlIO_stderr())
93578b34 5106 {
f2b5be74 5107 io_close((IO*)sv, FALSE);
93578b34 5108 }
1d7c1841 5109 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5110 PerlDir_close(IoDIRP(sv));
1d7c1841 5111 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5112 Safefree(IoTOP_NAME(sv));
5113 Safefree(IoFMT_NAME(sv));
5114 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5115 goto freescalar;
79072805 5116 case SVt_PVBM:
a0d0e21e 5117 goto freescalar;
79072805 5118 case SVt_PVCV:
748a9306 5119 case SVt_PVFM:
85e6fe83 5120 cv_undef((CV*)sv);
a0d0e21e 5121 goto freescalar;
79072805 5122 case SVt_PVHV:
86f55936 5123 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5124 hv_undef((HV*)sv);
a0d0e21e 5125 break;
79072805 5126 case SVt_PVAV:
85e6fe83 5127 av_undef((AV*)sv);
a0d0e21e 5128 break;
02270b4e 5129 case SVt_PVLV:
dd28f7bb
DM
5130 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5131 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5132 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5133 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5134 }
5135 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5136 SvREFCNT_dec(LvTARG(sv));
02270b4e 5137 goto freescalar;
a0d0e21e 5138 case SVt_PVGV:
1edc1566 5139 gp_free((GV*)sv);
acda4c6a
NC
5140 if (GvNAME_HEK(sv)) {
5141 unshare_hek(GvNAME_HEK(sv));
5142 }
893645bd
NC
5143 /* If we're in a stash, we don't own a reference to it. However it does
5144 have a back reference to us, which needs to be cleared. */
5145 if (GvSTASH(sv))
5146 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5147 case SVt_PVMG:
79072805
LW
5148 case SVt_PVNV:
5149 case SVt_PVIV:
a0d0e21e 5150 freescalar:
5228ca4e
NC
5151 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5152 if (SvOOK(sv)) {
93524f2b 5153 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5154 /* Don't even bother with turning off the OOK flag. */
5155 }
79072805 5156 case SVt_PV:
a0d0e21e 5157 case SVt_RV:
810b8aa5 5158 if (SvROK(sv)) {
b37c2d43 5159 SV * const target = SvRV(sv);
810b8aa5 5160 if (SvWEAKREF(sv))
e15faf7d 5161 sv_del_backref(target, sv);
810b8aa5 5162 else
e15faf7d 5163 SvREFCNT_dec(target);
810b8aa5 5164 }
f8c7b90f 5165#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5166 else if (SvPVX_const(sv)) {
765f542d
NC
5167 if (SvIsCOW(sv)) {
5168 /* I believe I need to grab the global SV mutex here and
5169 then recheck the COW status. */
46187eeb
NC
5170 if (DEBUG_C_TEST) {
5171 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5172 sv_dump(sv);
46187eeb 5173 }
bdd68bc3
NC
5174 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5175 SV_COW_NEXT_SV(sv));
765f542d
NC
5176 /* And drop it here. */
5177 SvFAKE_off(sv);
5178 } else if (SvLEN(sv)) {
3f7c398e 5179 Safefree(SvPVX_const(sv));
765f542d
NC
5180 }
5181 }
5182#else
3f7c398e 5183 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5184 Safefree(SvPVX_mutable(sv));
3f7c398e 5185 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5186 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5187 SvFAKE_off(sv);
5188 }
765f542d 5189#endif
79072805
LW
5190 break;
5191 case SVt_NV:
79072805
LW
5192 break;
5193 }
5194
893645bd
NC
5195 SvFLAGS(sv) &= SVf_BREAK;
5196 SvFLAGS(sv) |= SVTYPEMASK;
5197
8edfc514 5198 if (sv_type_details->arena) {
b9502f15 5199 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5200 &PL_body_roots[type]);
5201 }
d2a0f284 5202 else if (sv_type_details->body_size) {
8edfc514
NC
5203 my_safefree(SvANY(sv));
5204 }
79072805
LW
5205}
5206
645c22ef
DM
5207/*
5208=for apidoc sv_newref
5209
5210Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5211instead.
5212
5213=cut
5214*/
5215
79072805 5216SV *
864dbfa3 5217Perl_sv_newref(pTHX_ SV *sv)
79072805 5218{
96a5add6 5219 PERL_UNUSED_CONTEXT;
463ee0b2 5220 if (sv)
4db098f4 5221 (SvREFCNT(sv))++;
79072805
LW
5222 return sv;
5223}
5224
c461cf8f
JH
5225/*
5226=for apidoc sv_free
5227
645c22ef
DM
5228Decrement an SV's reference count, and if it drops to zero, call
5229C<sv_clear> to invoke destructors and free up any memory used by
5230the body; finally, deallocate the SV's head itself.
5231Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5232
5233=cut
5234*/
5235
79072805 5236void
864dbfa3 5237Perl_sv_free(pTHX_ SV *sv)
79072805 5238{
27da23d5 5239 dVAR;
79072805
LW
5240 if (!sv)
5241 return;
a0d0e21e
LW
5242 if (SvREFCNT(sv) == 0) {
5243 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5244 /* this SV's refcnt has been artificially decremented to
5245 * trigger cleanup */
a0d0e21e 5246 return;
3280af22 5247 if (PL_in_clean_all) /* All is fair */
1edc1566 5248 return;
d689ffdd
JP
5249 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5250 /* make sure SvREFCNT(sv)==0 happens very seldom */
5251 SvREFCNT(sv) = (~(U32)0)/2;
5252 return;
5253 }
41e4abd8 5254 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5255 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5256 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5257 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5258#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5259 Perl_dump_sv_child(aTHX_ sv);
5260#endif
5261 }
79072805
LW
5262 return;
5263 }
4db098f4 5264 if (--(SvREFCNT(sv)) > 0)
8990e307 5265 return;
8c4d3c90
NC
5266 Perl_sv_free2(aTHX_ sv);
5267}
5268
5269void
5270Perl_sv_free2(pTHX_ SV *sv)
5271{
27da23d5 5272 dVAR;
463ee0b2
LW
5273#ifdef DEBUGGING
5274 if (SvTEMP(sv)) {
0453d815 5275 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5276 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5277 "Attempt to free temp prematurely: SV 0x%"UVxf
5278 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5279 return;
79072805 5280 }
463ee0b2 5281#endif
d689ffdd
JP
5282 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5283 /* make sure SvREFCNT(sv)==0 happens very seldom */
5284 SvREFCNT(sv) = (~(U32)0)/2;
5285 return;
5286 }
79072805 5287 sv_clear(sv);
477f5d66
CS
5288 if (! SvREFCNT(sv))
5289 del_SV(sv);
79072805
LW
5290}
5291
954c1994
GS
5292/*
5293=for apidoc sv_len
5294
645c22ef
DM
5295Returns the length of the string in the SV. Handles magic and type
5296coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5297
5298=cut
5299*/
5300
79072805 5301STRLEN
864dbfa3 5302Perl_sv_len(pTHX_ register SV *sv)
79072805 5303{
463ee0b2 5304 STRLEN len;
79072805
LW
5305
5306 if (!sv)
5307 return 0;
5308
8990e307 5309 if (SvGMAGICAL(sv))
565764a8 5310 len = mg_length(sv);
8990e307 5311 else
4d84ee25 5312 (void)SvPV_const(sv, len);
463ee0b2 5313 return len;
79072805
LW
5314}
5315
c461cf8f
JH
5316/*
5317=for apidoc sv_len_utf8
5318
5319Returns the number of characters in the string in an SV, counting wide
1e54db1a 5320UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5321
5322=cut
5323*/
5324
7e8c5dac
HS
5325/*
5326 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5327 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5328 * (Note that the mg_len is not the length of the mg_ptr field.
5329 * This allows the cache to store the character length of the string without
5330 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5331 *
7e8c5dac
HS
5332 */
5333
a0ed51b3 5334STRLEN
864dbfa3 5335Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5336{
a0ed51b3
LW
5337 if (!sv)
5338 return 0;
5339
a0ed51b3 5340 if (SvGMAGICAL(sv))
b76347f2 5341 return mg_length(sv);
a0ed51b3 5342 else
b76347f2 5343 {
26346457 5344 STRLEN len;
e62f0680 5345 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5346
26346457
NC
5347 if (PL_utf8cache) {
5348 STRLEN ulen;
5349 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5350
5351 if (mg && mg->mg_len != -1) {
5352 ulen = mg->mg_len;
5353 if (PL_utf8cache < 0) {
5354 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5355 if (real != ulen) {
5356 /* Need to turn the assertions off otherwise we may
5357 recurse infinitely while printing error messages.
5358 */
5359 SAVEI8(PL_utf8cache);
5360 PL_utf8cache = 0;
f5992bc4
RB
5361 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
5362 " real %"UVuf" for %"SVf,
95b63a38 5363 (UV) ulen, (UV) real, (void*)sv);
26346457
NC
5364 }
5365 }
5366 }
5367 else {
5368 ulen = Perl_utf8_length(aTHX_ s, s + len);
5369 if (!SvREADONLY(sv)) {
5370 if (!mg) {
5371 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5372 &PL_vtbl_utf8, 0, 0);
5373 }
cb9e20bb 5374 assert(mg);
26346457 5375 mg->mg_len = ulen;
cb9e20bb 5376 }
cb9e20bb 5377 }
26346457 5378 return ulen;
7e8c5dac 5379 }
26346457 5380 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5381 }
5382}
5383
9564a3bd
NC
5384/* Walk forwards to find the byte corresponding to the passed in UTF-8
5385 offset. */
bdf30dd6 5386static STRLEN
721e86b6 5387S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
bdf30dd6
NC
5388 STRLEN uoffset)
5389{
5390 const U8 *s = start;
5391
5392 while (s < send && uoffset--)
5393 s += UTF8SKIP(s);
5394 if (s > send) {
5395 /* This is the existing behaviour. Possibly it should be a croak, as
5396 it's actually a bounds error */
5397 s = send;
5398 }
5399 return s - start;
5400}
5401
9564a3bd
NC
5402/* Given the length of the string in both bytes and UTF-8 characters, decide
5403 whether to walk forwards or backwards to find the byte corresponding to
5404 the passed in UTF-8 offset. */
c336ad0b 5405static STRLEN
721e86b6 5406S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
c336ad0b
NC
5407 STRLEN uoffset, STRLEN uend)
5408{
5409 STRLEN backw = uend - uoffset;
5410 if (uoffset < 2 * backw) {
25a8a4ef 5411 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5412 forward (that's where the 2 * backw comes from).
5413 (The real figure of course depends on the UTF-8 data.) */
721e86b6 5414 return sv_pos_u2b_forwards(start, send, uoffset);
c336ad0b
NC
5415 }
5416
5417 while (backw--) {
5418 send--;
5419 while (UTF8_IS_CONTINUATION(*send))
5420 send--;
5421 }
5422 return send - start;
5423}
5424
9564a3bd
NC
5425/* For the string representation of the given scalar, find the byte
5426 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5427 give another position in the string, *before* the sought offset, which
5428 (which is always true, as 0, 0 is a valid pair of positions), which should
5429 help reduce the amount of linear searching.
5430 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5431 will be used to reduce the amount of linear searching. The cache will be
5432 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5433static STRLEN
5434S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5435 const U8 *const send, STRLEN uoffset,
5436 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5437 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5438 bool found = FALSE;
5439
75c33c12
NC
5440 assert (uoffset >= uoffset0);
5441
c336ad0b 5442 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5443 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5444 if ((*mgp)->mg_ptr) {
5445 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5446 if (cache[0] == uoffset) {
5447 /* An exact match. */
5448 return cache[1];
5449 }
ab455f60
NC
5450 if (cache[2] == uoffset) {
5451 /* An exact match. */
5452 return cache[3];
5453 }
668af93f
NC
5454
5455 if (cache[0] < uoffset) {
d8b2e1f9
NC
5456 /* The cache already knows part of the way. */
5457 if (cache[0] > uoffset0) {
5458 /* The cache knows more than the passed in pair */
5459 uoffset0 = cache[0];
5460 boffset0 = cache[1];
5461 }
5462 if ((*mgp)->mg_len != -1) {
5463 /* And we know the end too. */
5464 boffset = boffset0
721e86b6 5465 + sv_pos_u2b_midway(start + boffset0, send,
d8b2e1f9
NC
5466 uoffset - uoffset0,
5467 (*mgp)->mg_len - uoffset0);
5468 } else {
5469 boffset = boffset0
721e86b6 5470 + sv_pos_u2b_forwards(start + boffset0,
d8b2e1f9
NC
5471 send, uoffset - uoffset0);
5472 }
dd7c5fd3
NC
5473 }
5474 else if (cache[2] < uoffset) {
5475 /* We're between the two cache entries. */
5476 if (cache[2] > uoffset0) {
5477 /* and the cache knows more than the passed in pair */
5478 uoffset0 = cache[2];
5479 boffset0 = cache[3];
5480 }
5481
668af93f 5482 boffset = boffset0
721e86b6 5483 + sv_pos_u2b_midway(start + boffset0,
668af93f
NC
5484 start + cache[1],
5485 uoffset - uoffset0,
5486 cache[0] - uoffset0);
dd7c5fd3
NC
5487 } else {
5488 boffset = boffset0
721e86b6 5489 + sv_pos_u2b_midway(start + boffset0,
dd7c5fd3
NC
5490 start + cache[3],
5491 uoffset - uoffset0,
5492 cache[2] - uoffset0);
d8b2e1f9 5493 }
668af93f 5494 found = TRUE;
d8b2e1f9
NC
5495 }
5496 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5497 /* If we can take advantage of a passed in offset, do so. */
5498 /* In fact, offset0 is either 0, or less than offset, so don't
5499 need to worry about the other possibility. */
5500 boffset = boffset0
721e86b6 5501 + sv_pos_u2b_midway(start + boffset0, send,
75c33c12
NC
5502 uoffset - uoffset0,
5503 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5504 found = TRUE;
5505 }
28ccbf94 5506 }
c336ad0b
NC
5507
5508 if (!found || PL_utf8cache < 0) {
75c33c12 5509 const STRLEN real_boffset
721e86b6 5510 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
75c33c12
NC
5511 send, uoffset - uoffset0);
5512
c336ad0b
NC
5513 if (found && PL_utf8cache < 0) {
5514 if (real_boffset != boffset) {
5515 /* Need to turn the assertions off otherwise we may recurse
5516 infinitely while printing error messages. */
5517 SAVEI8(PL_utf8cache);
5518 PL_utf8cache = 0;
f5992bc4
RB
5519 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
5520 " real %"UVuf" for %"SVf,
95b63a38 5521 (UV) boffset, (UV) real_boffset, (void*)sv);
c336ad0b
NC
5522 }
5523 }
5524 boffset = real_boffset;
28ccbf94 5525 }
0905937d 5526
ab455f60 5527 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5528 return boffset;
5529}
5530
9564a3bd
NC
5531
5532/*
5533=for apidoc sv_pos_u2b
5534
5535Converts the value pointed to by offsetp from a count of UTF-8 chars from
5536the start of the string, to a count of the equivalent number of bytes; if
5537lenp is non-zero, it does the same to lenp, but this time starting from
5538the offset, rather than from the start of the string. Handles magic and
5539type coercion.
5540
5541=cut
5542*/
5543
5544/*
5545 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5546 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5547 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5548 *
5549 */
5550
a0ed51b3 5551void
864dbfa3 5552Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5553{
245d4a47 5554 const U8 *start;
a0ed51b3
LW
5555 STRLEN len;
5556
5557 if (!sv)
5558 return;
5559
245d4a47 5560 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5561 if (len) {
bdf30dd6
NC
5562 STRLEN uoffset = (STRLEN) *offsetp;
5563 const U8 * const send = start + len;
0905937d 5564 MAGIC *mg = NULL;
721e86b6 5565 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
28ccbf94 5566 uoffset, 0, 0);
bdf30dd6
NC
5567
5568 *offsetp = (I32) boffset;
5569
5570 if (lenp) {
28ccbf94 5571 /* Convert the relative offset to absolute. */
721e86b6
AL
5572 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5573 const STRLEN boffset2
5574 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
28ccbf94 5575 uoffset, boffset) - boffset;
bdf30dd6 5576
28ccbf94 5577 *lenp = boffset2;
bdf30dd6 5578 }
7e8c5dac
HS
5579 }
5580 else {
5581 *offsetp = 0;
5582 if (lenp)
5583 *lenp = 0;
a0ed51b3 5584 }
e23c8137 5585
a0ed51b3
LW
5586 return;
5587}
5588
9564a3bd
NC
5589/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5590 byte length pairing. The (byte) length of the total SV is passed in too,
5591 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5592 may not have updated SvCUR, so we can't rely on reading it directly.
5593
5594 The proffered utf8/byte length pairing isn't used if the cache already has
5595 two pairs, and swapping either for the proffered pair would increase the
5596 RMS of the intervals between known byte offsets.
5597
5598 The cache itself consists of 4 STRLEN values
5599 0: larger UTF-8 offset
5600 1: corresponding byte offset
5601 2: smaller UTF-8 offset
5602 3: corresponding byte offset
5603
5604 Unused cache pairs have the value 0, 0.
5605 Keeping the cache "backwards" means that the invariant of
5606 cache[0] >= cache[2] is maintained even with empty slots, which means that
5607 the code that uses it doesn't need to worry if only 1 entry has actually
5608 been set to non-zero. It also makes the "position beyond the end of the
5609 cache" logic much simpler, as the first slot is always the one to start
5610 from.
645c22ef 5611*/
ec07b5e0 5612static void
ab455f60
NC
5613S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5614 STRLEN blen)
ec07b5e0
NC
5615{
5616 STRLEN *cache;
5617 if (SvREADONLY(sv))
5618 return;
5619
5620 if (!*mgp) {
5621 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5622 0);
5623 (*mgp)->mg_len = -1;
5624 }
5625 assert(*mgp);
5626
5627 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5628 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5629 (*mgp)->mg_ptr = (char *) cache;
5630 }
5631 assert(cache);
5632
5633 if (PL_utf8cache < 0) {
ef816a78 5634 const U8 *start = (const U8 *) SvPVX_const(sv);
6448472a 5635 const STRLEN realutf8 = utf8_length(start, start + byte);
ec07b5e0
NC
5636
5637 if (realutf8 != utf8) {
5638 /* Need to turn the assertions off otherwise we may recurse
5639 infinitely while printing error messages. */
5640 SAVEI8(PL_utf8cache);
5641 PL_utf8cache = 0;
f5992bc4
RB
5642 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
5643 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, (void*)sv);
ec07b5e0
NC
5644 }
5645 }
ab455f60
NC
5646
5647 /* Cache is held with the later position first, to simplify the code
5648 that deals with unbounded ends. */
5649
5650 ASSERT_UTF8_CACHE(cache);
5651 if (cache[1] == 0) {
5652 /* Cache is totally empty */
5653 cache[0] = utf8;
5654 cache[1] = byte;
5655 } else if (cache[3] == 0) {
5656 if (byte > cache[1]) {
5657 /* New one is larger, so goes first. */
5658 cache[2] = cache[0];
5659 cache[3] = cache[1];
5660 cache[0] = utf8;
5661 cache[1] = byte;
5662 } else {
5663 cache[2] = utf8;
5664 cache[3] = byte;
5665 }
5666 } else {
5667#define THREEWAY_SQUARE(a,b,c,d) \
5668 ((float)((d) - (c))) * ((float)((d) - (c))) \
5669 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5670 + ((float)((b) - (a))) * ((float)((b) - (a)))
5671
5672 /* Cache has 2 slots in use, and we know three potential pairs.
5673 Keep the two that give the lowest RMS distance. Do the
5674 calcualation in bytes simply because we always know the byte
5675 length. squareroot has the same ordering as the positive value,
5676 so don't bother with the actual square root. */
5677 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5678 if (byte > cache[1]) {
5679 /* New position is after the existing pair of pairs. */
5680 const float keep_earlier
5681 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5682 const float keep_later
5683 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5684
5685 if (keep_later < keep_earlier) {
5686 if (keep_later < existing) {
5687 cache[2] = cache[0];
5688 cache[3] = cache[1];
5689 cache[0] = utf8;
5690 cache[1] = byte;
5691 }
5692 }
5693 else {
5694 if (keep_earlier < existing) {
5695 cache[0] = utf8;
5696 cache[1] = byte;
5697 }
5698 }
5699 }
57d7fbf1
NC
5700 else if (byte > cache[3]) {
5701 /* New position is between the existing pair of pairs. */
5702 const float keep_earlier
5703 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5704 const float keep_later
5705 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5706
5707 if (keep_later < keep_earlier) {
5708 if (keep_later < existing) {
5709 cache[2] = utf8;
5710 cache[3] = byte;
5711 }
5712 }
5713 else {
5714 if (keep_earlier < existing) {
5715 cache[0] = utf8;
5716 cache[1] = byte;
5717 }
5718 }
5719 }
5720 else {
5721 /* New position is before the existing pair of pairs. */
5722 const float keep_earlier
5723 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5724 const float keep_later
5725 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5726
5727 if (keep_later < keep_earlier) {
5728 if (keep_later < existing) {
5729 cache[2] = utf8;
5730 cache[3] = byte;
5731 }
5732 }
5733 else {
5734 if (keep_earlier < existing) {
5735 cache[0] = cache[2];
5736 cache[1] = cache[3];
5737 cache[2] = utf8;
5738 cache[3] = byte;
5739 }
5740 }
5741 }
ab455f60 5742 }
0905937d 5743 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5744}
5745
ec07b5e0 5746/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5747 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5748 backward is half the speed of walking forward. */
ec07b5e0
NC
5749static STRLEN
5750S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5751 STRLEN endu)
5752{
5753 const STRLEN forw = target - s;
5754 STRLEN backw = end - target;
5755
5756 if (forw < 2 * backw) {
6448472a 5757 return utf8_length(s, target);
ec07b5e0
NC
5758 }
5759
5760 while (end > target) {
5761 end--;
5762 while (UTF8_IS_CONTINUATION(*end)) {
5763 end--;
5764 }
5765 endu--;
5766 }
5767 return endu;
5768}
5769
9564a3bd
NC
5770/*
5771=for apidoc sv_pos_b2u
5772
5773Converts the value pointed to by offsetp from a count of bytes from the
5774start of the string, to a count of the equivalent number of UTF-8 chars.
5775Handles magic and type coercion.
5776
5777=cut
5778*/
5779
5780/*
5781 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5782 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5783 * byte offsets.
5784 *
5785 */
a0ed51b3 5786void
7e8c5dac 5787Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5788{
83003860 5789 const U8* s;
ec07b5e0 5790 const STRLEN byte = *offsetp;
7087a21c 5791 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5792 STRLEN blen;
ec07b5e0
NC
5793 MAGIC* mg = NULL;
5794 const U8* send;
a922f900 5795 bool found = FALSE;
a0ed51b3
LW
5796
5797 if (!sv)
5798 return;
5799
ab455f60 5800 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5801
ab455f60 5802 if (blen < byte)
ec07b5e0 5803 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5804
ec07b5e0 5805 send = s + byte;
a67d7df9 5806
ffca234a
NC
5807 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5808 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5809 if (mg->mg_ptr) {
d4c19fe8 5810 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5811 if (cache[1] == byte) {
ec07b5e0
NC
5812 /* An exact match. */
5813 *offsetp = cache[0];
ec07b5e0 5814 return;
7e8c5dac 5815 }
ab455f60
NC
5816 if (cache[3] == byte) {
5817 /* An exact match. */
5818 *offsetp = cache[2];
5819 return;
5820 }
668af93f
NC
5821
5822 if (cache[1] < byte) {
ec07b5e0 5823 /* We already know part of the way. */
b9f984a5
NC
5824 if (mg->mg_len != -1) {
5825 /* Actually, we know the end too. */
5826 len = cache[0]
5827 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5828 s + blen, mg->mg_len - cache[0]);
b9f984a5 5829 } else {
6448472a 5830 len = cache[0] + utf8_length(s + cache[1], send);
b9f984a5 5831 }
7e8c5dac 5832 }
9f985e4c
NC
5833 else if (cache[3] < byte) {
5834 /* We're between the two cached pairs, so we do the calculation
5835 offset by the byte/utf-8 positions for the earlier pair,
5836 then add the utf-8 characters from the string start to
5837 there. */
5838 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5839 s + cache[1], cache[0] - cache[2])
5840 + cache[2];
5841
5842 }
5843 else { /* cache[3] > byte */
5844 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5845 cache[2]);
7e8c5dac 5846
7e8c5dac 5847 }
ec07b5e0 5848 ASSERT_UTF8_CACHE(cache);
a922f900 5849 found = TRUE;
ffca234a 5850 } else if (mg->mg_len != -1) {
ab455f60 5851 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5852 found = TRUE;
7e8c5dac 5853 }
a0ed51b3 5854 }
a922f900 5855 if (!found || PL_utf8cache < 0) {
6448472a 5856 const STRLEN real_len = utf8_length(s, send);
a922f900
NC
5857
5858 if (found && PL_utf8cache < 0) {
5859 if (len != real_len) {
5860 /* Need to turn the assertions off otherwise we may recurse
5861 infinitely while printing error messages. */
5862 SAVEI8(PL_utf8cache);
5863 PL_utf8cache = 0;
f5992bc4
RB
5864 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
5865 " real %"UVuf" for %"SVf,
95b63a38 5866 (UV) len, (UV) real_len, (void*)sv);
a922f900
NC
5867 }
5868 }
5869 len = real_len;
ec07b5e0
NC
5870 }
5871 *offsetp = len;
5872
ab455f60 5873 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5874}
5875
954c1994
GS
5876/*
5877=for apidoc sv_eq
5878
5879Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5880identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5881coerce its args to strings if necessary.
954c1994
GS
5882
5883=cut
5884*/
5885
79072805 5886I32
e01b9e88 5887Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5888{
97aff369 5889 dVAR;
e1ec3a88 5890 const char *pv1;
463ee0b2 5891 STRLEN cur1;
e1ec3a88 5892 const char *pv2;
463ee0b2 5893 STRLEN cur2;
e01b9e88 5894 I32 eq = 0;
bd61b366 5895 char *tpv = NULL;
a0714e2c 5896 SV* svrecode = NULL;
79072805 5897
e01b9e88 5898 if (!sv1) {
79072805
LW
5899 pv1 = "";
5900 cur1 = 0;
5901 }
ced497e2
YST
5902 else {
5903 /* if pv1 and pv2 are the same, second SvPV_const call may
5904 * invalidate pv1, so we may need to make a copy */
5905 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
5906 pv1 = SvPV_const(sv1, cur1);
5907 sv1 = sv_2mortal(newSVpvn(pv1, cur1));
5908 if (SvUTF8(sv2)) SvUTF8_on(sv1);
5909 }
4d84ee25 5910 pv1 = SvPV_const(sv1, cur1);
ced497e2 5911 }
79072805 5912
e01b9e88
SC
5913 if (!sv2){
5914 pv2 = "";
5915 cur2 = 0;
92d29cee 5916 }
e01b9e88 5917 else
4d84ee25 5918 pv2 = SvPV_const(sv2, cur2);
79072805 5919
cf48d248 5920 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5921 /* Differing utf8ness.
5922 * Do not UTF8size the comparands as a side-effect. */
5923 if (PL_encoding) {
5924 if (SvUTF8(sv1)) {
553e1bcc
AT
5925 svrecode = newSVpvn(pv2, cur2);
5926 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5927 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5928 }
5929 else {
553e1bcc
AT
5930 svrecode = newSVpvn(pv1, cur1);
5931 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5932 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5933 }
5934 /* Now both are in UTF-8. */
0a1bd7ac
DM
5935 if (cur1 != cur2) {
5936 SvREFCNT_dec(svrecode);
799ef3cb 5937 return FALSE;
0a1bd7ac 5938 }
799ef3cb
JH
5939 }
5940 else {
5941 bool is_utf8 = TRUE;
5942
5943 if (SvUTF8(sv1)) {
5944 /* sv1 is the UTF-8 one,
5945 * if is equal it must be downgrade-able */
9d4ba2ae 5946 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5947 &cur1, &is_utf8);
5948 if (pv != pv1)
553e1bcc 5949 pv1 = tpv = pv;
799ef3cb
JH
5950 }
5951 else {
5952 /* sv2 is the UTF-8 one,
5953 * if is equal it must be downgrade-able */
9d4ba2ae 5954 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5955 &cur2, &is_utf8);
5956 if (pv != pv2)
553e1bcc 5957 pv2 = tpv = pv;
799ef3cb
JH
5958 }
5959 if (is_utf8) {
5960 /* Downgrade not possible - cannot be eq */
bf694877 5961 assert (tpv == 0);
799ef3cb
JH
5962 return FALSE;
5963 }
5964 }
cf48d248
JH
5965 }
5966
5967 if (cur1 == cur2)
765f542d 5968 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5969
b37c2d43 5970 SvREFCNT_dec(svrecode);
553e1bcc
AT
5971 if (tpv)
5972 Safefree(tpv);
cf48d248 5973
e01b9e88 5974 return eq;
79072805
LW
5975}
5976
954c1994
GS
5977/*
5978=for apidoc sv_cmp
5979
5980Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5981string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5982C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5983coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5984
5985=cut
5986*/
5987
79072805 5988I32
e01b9e88 5989Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5990{
97aff369 5991 dVAR;
560a288e 5992 STRLEN cur1, cur2;
e1ec3a88 5993 const char *pv1, *pv2;
bd61b366 5994 char *tpv = NULL;
cf48d248 5995 I32 cmp;
a0714e2c 5996 SV *svrecode = NULL;
560a288e 5997
e01b9e88
SC
5998 if (!sv1) {
5999 pv1 = "";
560a288e
GS
6000 cur1 = 0;
6001 }
e01b9e88 6002 else
4d84ee25 6003 pv1 = SvPV_const(sv1, cur1);
560a288e 6004
553e1bcc 6005 if (!sv2) {
e01b9e88 6006 pv2 = "";
560a288e
GS
6007 cur2 = 0;
6008 }
e01b9e88 6009 else
4d84ee25 6010 pv2 = SvPV_const(sv2, cur2);
79072805 6011
cf48d248 6012 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6013 /* Differing utf8ness.
6014 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6015 if (SvUTF8(sv1)) {
799ef3cb 6016 if (PL_encoding) {
553e1bcc
AT
6017 svrecode = newSVpvn(pv2, cur2);
6018 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6019 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6020 }
6021 else {
e1ec3a88 6022 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6023 }
cf48d248
JH
6024 }
6025 else {
799ef3cb 6026 if (PL_encoding) {
553e1bcc
AT
6027 svrecode = newSVpvn(pv1, cur1);
6028 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6029 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6030 }
6031 else {
e1ec3a88 6032 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6033 }
cf48d248
JH
6034 }
6035 }
6036
e01b9e88 6037 if (!cur1) {
cf48d248 6038 cmp = cur2 ? -1 : 0;
e01b9e88 6039 } else if (!cur2) {
cf48d248
JH
6040 cmp = 1;
6041 } else {
e1ec3a88 6042 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6043
6044 if (retval) {
cf48d248 6045 cmp = retval < 0 ? -1 : 1;
e01b9e88 6046 } else if (cur1 == cur2) {
cf48d248
JH
6047 cmp = 0;
6048 } else {
6049 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6050 }
cf48d248 6051 }
16660edb 6052
b37c2d43 6053 SvREFCNT_dec(svrecode);
553e1bcc
AT
6054 if (tpv)
6055 Safefree(tpv);
cf48d248
JH
6056
6057 return cmp;
bbce6d69 6058}
16660edb 6059
c461cf8f
JH
6060/*
6061=for apidoc sv_cmp_locale
6062
645c22ef
DM
6063Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6064'use bytes' aware, handles get magic, and will coerce its args to strings
6065if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6066
6067=cut
6068*/
6069
bbce6d69 6070I32
864dbfa3 6071Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6072{
97aff369 6073 dVAR;
36477c24 6074#ifdef USE_LOCALE_COLLATE
16660edb 6075
bbce6d69 6076 char *pv1, *pv2;
6077 STRLEN len1, len2;
6078 I32 retval;
16660edb 6079
3280af22 6080 if (PL_collation_standard)
bbce6d69 6081 goto raw_compare;
16660edb 6082
bbce6d69 6083 len1 = 0;
8ac85365 6084 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6085 len2 = 0;
8ac85365 6086 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6087
bbce6d69 6088 if (!pv1 || !len1) {
6089 if (pv2 && len2)
6090 return -1;
6091 else
6092 goto raw_compare;
6093 }
6094 else {
6095 if (!pv2 || !len2)
6096 return 1;
6097 }
16660edb 6098
bbce6d69 6099 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6100
bbce6d69 6101 if (retval)
16660edb 6102 return retval < 0 ? -1 : 1;
6103
bbce6d69 6104 /*
6105 * When the result of collation is equality, that doesn't mean
6106 * that there are no differences -- some locales exclude some
6107 * characters from consideration. So to avoid false equalities,
6108 * we use the raw string as a tiebreaker.
6109 */
16660edb 6110
bbce6d69 6111 raw_compare:
5f66b61c 6112 /*FALLTHROUGH*/
16660edb 6113
36477c24 6114#endif /* USE_LOCALE_COLLATE */
16660edb 6115
bbce6d69 6116 return sv_cmp(sv1, sv2);
6117}
79072805 6118
645c22ef 6119
36477c24 6120#ifdef USE_LOCALE_COLLATE
645c22ef 6121
7a4c00b4 6122/*
645c22ef
DM
6123=for apidoc sv_collxfrm
6124
6125Add Collate Transform magic to an SV if it doesn't already have it.
6126
6127Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6128scalar data of the variable, but transformed to such a format that a normal
6129memory comparison can be used to compare the data according to the locale
6130settings.
6131
6132=cut
6133*/
6134
bbce6d69 6135char *
864dbfa3 6136Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6137{
97aff369 6138 dVAR;
7a4c00b4 6139 MAGIC *mg;
16660edb 6140
14befaf4 6141 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6142 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6143 const char *s;
6144 char *xf;
bbce6d69 6145 STRLEN len, xlen;
6146
7a4c00b4 6147 if (mg)
6148 Safefree(mg->mg_ptr);
93524f2b 6149 s = SvPV_const(sv, len);
bbce6d69 6150 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6151 if (SvREADONLY(sv)) {
6152 SAVEFREEPV(xf);
6153 *nxp = xlen;
3280af22 6154 return xf + sizeof(PL_collation_ix);
ff0cee69 6155 }
7a4c00b4 6156 if (! mg) {
d83f0a82
NC
6157#ifdef PERL_OLD_COPY_ON_WRITE
6158 if (SvIsCOW(sv))
6159 sv_force_normal_flags(sv, 0);
6160#endif
6161 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6162 0, 0);
7a4c00b4 6163 assert(mg);
bbce6d69 6164 }
7a4c00b4 6165 mg->mg_ptr = xf;
565764a8 6166 mg->mg_len = xlen;
7a4c00b4 6167 }
6168 else {
ff0cee69 6169 if (mg) {
6170 mg->mg_ptr = NULL;
565764a8 6171 mg->mg_len = -1;
ff0cee69 6172 }
bbce6d69 6173 }
6174 }
7a4c00b4 6175 if (mg && mg->mg_ptr) {
565764a8 6176 *nxp = mg->mg_len;
3280af22 6177 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6178 }
6179 else {
6180 *nxp = 0;
6181 return NULL;
16660edb 6182 }
79072805
LW
6183}
6184
36477c24 6185#endif /* USE_LOCALE_COLLATE */
bbce6d69 6186
c461cf8f
JH
6187/*
6188=for apidoc sv_gets
6189
6190Get a line from the filehandle and store it into the SV, optionally
6191appending to the currently-stored string.
6192
6193=cut
6194*/
6195
79072805 6196char *
864dbfa3 6197Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6198{
97aff369 6199 dVAR;
e1ec3a88 6200 const char *rsptr;
c07a80fd 6201 STRLEN rslen;
6202 register STDCHAR rslast;
6203 register STDCHAR *bp;
6204 register I32 cnt;
9c5ffd7c 6205 I32 i = 0;
8bfdd7d9 6206 I32 rspara = 0;
c07a80fd 6207
bc44a8a2
NC
6208 if (SvTHINKFIRST(sv))
6209 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6210 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6211 from <>.
6212 However, perlbench says it's slower, because the existing swipe code
6213 is faster than copy on write.
6214 Swings and roundabouts. */
862a34c6 6215 SvUPGRADE(sv, SVt_PV);
99491443 6216
ff68c719 6217 SvSCREAM_off(sv);
efd8b2ba
AE
6218
6219 if (append) {
6220 if (PerlIO_isutf8(fp)) {
6221 if (!SvUTF8(sv)) {
6222 sv_utf8_upgrade_nomg(sv);
6223 sv_pos_u2b(sv,&append,0);
6224 }
6225 } else if (SvUTF8(sv)) {
561b68a9 6226 SV * const tsv = newSV(0);
efd8b2ba
AE
6227 sv_gets(tsv, fp, 0);
6228 sv_utf8_upgrade_nomg(tsv);
6229 SvCUR_set(sv,append);
6230 sv_catsv(sv,tsv);
6231 sv_free(tsv);
6232 goto return_string_or_null;
6233 }
6234 }
6235
6236 SvPOK_only(sv);
6237 if (PerlIO_isutf8(fp))
6238 SvUTF8_on(sv);
c07a80fd 6239
923e4eb5 6240 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6241 /* we always read code in line mode */
6242 rsptr = "\n";
6243 rslen = 1;
6244 }
6245 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6246 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6247 of amount we are going to read -- may result in mallocing
6248 more memory than we really need if the layers below reduce
6249 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6250 */
e311fd51 6251 Stat_t st;
e468d35b 6252 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6253 const Off_t offset = PerlIO_tell(fp);
58f1856e 6254 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6255 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6256 }
6257 }
c07a80fd 6258 rsptr = NULL;
6259 rslen = 0;
6260 }
3280af22 6261 else if (RsRECORD(PL_rs)) {
e311fd51 6262 I32 bytesread;
5b2b9c68 6263 char *buffer;
acbd132f 6264 U32 recsize;
5b2b9c68
HM
6265
6266 /* Grab the size of the record we're getting */
acbd132f 6267 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6268 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6269 /* Go yank in */
6270#ifdef VMS
6271 /* VMS wants read instead of fread, because fread doesn't respect */
6272 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6273 /* doing, but we've got no other real choice - except avoid stdio
6274 as implementation - perhaps write a :vms layer ?
6275 */
5b2b9c68
HM
6276 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6277#else
6278 bytesread = PerlIO_read(fp, buffer, recsize);
6279#endif
27e6ca2d
AE
6280 if (bytesread < 0)
6281 bytesread = 0;
e311fd51 6282 SvCUR_set(sv, bytesread += append);
e670df4e 6283 buffer[bytesread] = '\0';
efd8b2ba 6284 goto return_string_or_null;
5b2b9c68 6285 }
3280af22 6286 else if (RsPARA(PL_rs)) {
c07a80fd 6287 rsptr = "\n\n";
6288 rslen = 2;
8bfdd7d9 6289 rspara = 1;
c07a80fd 6290 }
7d59b7e4
NIS
6291 else {
6292 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6293 if (PerlIO_isutf8(fp)) {
6294 rsptr = SvPVutf8(PL_rs, rslen);
6295 }
6296 else {
6297 if (SvUTF8(PL_rs)) {
6298 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6299 Perl_croak(aTHX_ "Wide character in $/");
6300 }
6301 }
93524f2b 6302 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6303 }
6304 }
6305
c07a80fd 6306 rslast = rslen ? rsptr[rslen - 1] : '\0';
6307
8bfdd7d9 6308 if (rspara) { /* have to do this both before and after */
79072805 6309 do { /* to make sure file boundaries work right */
760ac839 6310 if (PerlIO_eof(fp))
a0d0e21e 6311 return 0;
760ac839 6312 i = PerlIO_getc(fp);
79072805 6313 if (i != '\n') {
a0d0e21e
LW
6314 if (i == -1)
6315 return 0;
760ac839 6316 PerlIO_ungetc(fp,i);
79072805
LW
6317 break;
6318 }
6319 } while (i != EOF);
6320 }
c07a80fd 6321
760ac839
LW
6322 /* See if we know enough about I/O mechanism to cheat it ! */
6323
6324 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6325 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6326 enough here - and may even be a macro allowing compile
6327 time optimization.
6328 */
6329
6330 if (PerlIO_fast_gets(fp)) {
6331
6332 /*
6333 * We're going to steal some values from the stdio struct
6334 * and put EVERYTHING in the innermost loop into registers.
6335 */
6336 register STDCHAR *ptr;
6337 STRLEN bpx;
6338 I32 shortbuffered;
6339
16660edb 6340#if defined(VMS) && defined(PERLIO_IS_STDIO)
6341 /* An ungetc()d char is handled separately from the regular
6342 * buffer, so we getc() it back out and stuff it in the buffer.
6343 */
6344 i = PerlIO_getc(fp);
6345 if (i == EOF) return 0;
6346 *(--((*fp)->_ptr)) = (unsigned char) i;
6347 (*fp)->_cnt++;
6348#endif
c07a80fd 6349
c2960299 6350 /* Here is some breathtakingly efficient cheating */
c07a80fd 6351
a20bf0c3 6352 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6353 /* make sure we have the room */
7a5fa8a2 6354 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6355 /* Not room for all of it
7a5fa8a2 6356 if we are looking for a separator and room for some
e468d35b
NIS
6357 */
6358 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6359 /* just process what we have room for */
79072805
LW
6360 shortbuffered = cnt - SvLEN(sv) + append + 1;
6361 cnt -= shortbuffered;
6362 }
6363 else {
6364 shortbuffered = 0;
bbce6d69 6365 /* remember that cnt can be negative */
eb160463 6366 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6367 }
6368 }
7a5fa8a2 6369 else
79072805 6370 shortbuffered = 0;
3f7c398e 6371 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6372 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6373 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6374 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6375 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6376 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6377 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6378 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6379 for (;;) {
6380 screamer:
93a17b20 6381 if (cnt > 0) {
c07a80fd 6382 if (rslen) {
760ac839
LW
6383 while (cnt > 0) { /* this | eat */
6384 cnt--;
c07a80fd 6385 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6386 goto thats_all_folks; /* screams | sed :-) */
6387 }
6388 }
6389 else {
1c846c1f
NIS
6390 Copy(ptr, bp, cnt, char); /* this | eat */
6391 bp += cnt; /* screams | dust */
c07a80fd 6392 ptr += cnt; /* louder | sed :-) */
a5f75d66 6393 cnt = 0;
93a17b20 6394 }
79072805
LW
6395 }
6396
748a9306 6397 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6398 cnt = shortbuffered;
6399 shortbuffered = 0;
3f7c398e 6400 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6401 SvCUR_set(sv, bpx);
6402 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6403 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6404 continue;
6405 }
6406
16660edb 6407 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6408 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6409 PTR2UV(ptr),(long)cnt));
cc00df79 6410 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6411#if 0
16660edb 6412 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6413 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6414 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6415 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6416#endif
1c846c1f 6417 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6418 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6419 another abstraction. */
760ac839 6420 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6421#if 0
16660edb 6422 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6423 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6424 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6425 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6426#endif
a20bf0c3
JH
6427 cnt = PerlIO_get_cnt(fp);
6428 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6429 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6430 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6431
748a9306
LW
6432 if (i == EOF) /* all done for ever? */
6433 goto thats_really_all_folks;
6434
3f7c398e 6435 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6436 SvCUR_set(sv, bpx);
6437 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6438 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6439
eb160463 6440 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6441
c07a80fd 6442 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6443 goto thats_all_folks;
79072805
LW
6444 }
6445
6446thats_all_folks:
3f7c398e 6447 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6448 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6449 goto screamer; /* go back to the fray */
79072805
LW
6450thats_really_all_folks:
6451 if (shortbuffered)
6452 cnt += shortbuffered;
16660edb 6453 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6454 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6455 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6456 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6457 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6458 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6459 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6460 *bp = '\0';
3f7c398e 6461 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6462 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6463 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6464 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6465 }
6466 else
79072805 6467 {
6edd2cd5 6468 /*The big, slow, and stupid way. */
27da23d5 6469#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6470 STDCHAR *buf = NULL;
a02a5408 6471 Newx(buf, 8192, STDCHAR);
6edd2cd5 6472 assert(buf);
4d2c4e07 6473#else
6edd2cd5 6474 STDCHAR buf[8192];
4d2c4e07 6475#endif
79072805 6476
760ac839 6477screamer2:
c07a80fd 6478 if (rslen) {
00b6aa41 6479 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6480 bp = buf;
eb160463 6481 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6482 ; /* keep reading */
6483 cnt = bp - buf;
c07a80fd 6484 }
6485 else {
760ac839 6486 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6487 /* Accomodate broken VAXC compiler, which applies U8 cast to
6488 * both args of ?: operator, causing EOF to change into 255
6489 */
37be0adf 6490 if (cnt > 0)
cbe9e203
JH
6491 i = (U8)buf[cnt - 1];
6492 else
37be0adf 6493 i = EOF;
c07a80fd 6494 }
79072805 6495
cbe9e203
JH
6496 if (cnt < 0)
6497 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6498 if (append)
6499 sv_catpvn(sv, (char *) buf, cnt);
6500 else
6501 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6502
6503 if (i != EOF && /* joy */
6504 (!rslen ||
6505 SvCUR(sv) < rslen ||
3f7c398e 6506 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6507 {
6508 append = -1;
63e4d877
CS
6509 /*
6510 * If we're reading from a TTY and we get a short read,
6511 * indicating that the user hit his EOF character, we need
6512 * to notice it now, because if we try to read from the TTY
6513 * again, the EOF condition will disappear.
6514 *
6515 * The comparison of cnt to sizeof(buf) is an optimization
6516 * that prevents unnecessary calls to feof().
6517 *
6518 * - jik 9/25/96
6519 */
bb7a0f54 6520 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 6521 goto screamer2;
79072805 6522 }
6edd2cd5 6523
27da23d5 6524#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6525 Safefree(buf);
6526#endif
79072805
LW
6527 }
6528
8bfdd7d9 6529 if (rspara) { /* have to do this both before and after */
c07a80fd 6530 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6531 i = PerlIO_getc(fp);
79072805 6532 if (i != '\n') {
760ac839 6533 PerlIO_ungetc(fp,i);
79072805
LW
6534 break;
6535 }
6536 }
6537 }
c07a80fd 6538
efd8b2ba 6539return_string_or_null:
bd61b366 6540 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6541}
6542
954c1994
GS
6543/*
6544=for apidoc sv_inc
6545
645c22ef
DM
6546Auto-increment of the value in the SV, doing string to numeric conversion
6547if necessary. Handles 'get' magic.
954c1994
GS
6548
6549=cut
6550*/
6551
79072805 6552void
864dbfa3 6553Perl_sv_inc(pTHX_ register SV *sv)
79072805 6554{
97aff369 6555 dVAR;
79072805 6556 register char *d;
463ee0b2 6557 int flags;
79072805
LW
6558
6559 if (!sv)
6560 return;
5b295bef 6561 SvGETMAGIC(sv);
ed6116ce 6562 if (SvTHINKFIRST(sv)) {
765f542d
NC
6563 if (SvIsCOW(sv))
6564 sv_force_normal_flags(sv, 0);
0f15f207 6565 if (SvREADONLY(sv)) {
923e4eb5 6566 if (IN_PERL_RUNTIME)
cea2e8a9 6567 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6568 }
a0d0e21e 6569 if (SvROK(sv)) {
b5be31e9 6570 IV i;
9e7bc3e8
JD
6571 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6572 return;
56431972 6573 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6574 sv_unref(sv);
6575 sv_setiv(sv, i);
a0d0e21e 6576 }
ed6116ce 6577 }
8990e307 6578 flags = SvFLAGS(sv);
28e5dec8
JH
6579 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6580 /* It's (privately or publicly) a float, but not tested as an
6581 integer, so test it to see. */
d460ef45 6582 (void) SvIV(sv);
28e5dec8
JH
6583 flags = SvFLAGS(sv);
6584 }
6585 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6586 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6587#ifdef PERL_PRESERVE_IVUV
28e5dec8 6588 oops_its_int:
59d8ce62 6589#endif
25da4f38
IZ
6590 if (SvIsUV(sv)) {
6591 if (SvUVX(sv) == UV_MAX)
a1e868e7 6592 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6593 else
6594 (void)SvIOK_only_UV(sv);
607fa7f2 6595 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6596 } else {
6597 if (SvIVX(sv) == IV_MAX)
28e5dec8 6598 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6599 else {
6600 (void)SvIOK_only(sv);
45977657 6601 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6602 }
55497cff 6603 }
79072805
LW
6604 return;
6605 }
28e5dec8
JH
6606 if (flags & SVp_NOK) {
6607 (void)SvNOK_only(sv);
9d6ce603 6608 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6609 return;
6610 }
6611
3f7c398e 6612 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6613 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6614 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6615 (void)SvIOK_only(sv);
45977657 6616 SvIV_set(sv, 1);
79072805
LW
6617 return;
6618 }
463ee0b2 6619 d = SvPVX(sv);
79072805
LW
6620 while (isALPHA(*d)) d++;
6621 while (isDIGIT(*d)) d++;
6622 if (*d) {
28e5dec8 6623#ifdef PERL_PRESERVE_IVUV
d1be9408 6624 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6625 warnings. Probably ought to make the sv_iv_please() that does
6626 the conversion if possible, and silently. */
504618e9 6627 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6628 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6629 /* Need to try really hard to see if it's an integer.
6630 9.22337203685478e+18 is an integer.
6631 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6632 so $a="9.22337203685478e+18"; $a+0; $a++
6633 needs to be the same as $a="9.22337203685478e+18"; $a++
6634 or we go insane. */
d460ef45 6635
28e5dec8
JH
6636 (void) sv_2iv(sv);
6637 if (SvIOK(sv))
6638 goto oops_its_int;
6639
6640 /* sv_2iv *should* have made this an NV */
6641 if (flags & SVp_NOK) {
6642 (void)SvNOK_only(sv);
9d6ce603 6643 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6644 return;
6645 }
6646 /* I don't think we can get here. Maybe I should assert this
6647 And if we do get here I suspect that sv_setnv will croak. NWC
6648 Fall through. */
6649#if defined(USE_LONG_DOUBLE)
6650 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 6651 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6652#else
1779d84d 6653 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 6654 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6655#endif
6656 }
6657#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6658 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6659 return;
6660 }
6661 d--;
3f7c398e 6662 while (d >= SvPVX_const(sv)) {
79072805
LW
6663 if (isDIGIT(*d)) {
6664 if (++*d <= '9')
6665 return;
6666 *(d--) = '0';
6667 }
6668 else {
9d116dd7
JH
6669#ifdef EBCDIC
6670 /* MKS: The original code here died if letters weren't consecutive.
6671 * at least it didn't have to worry about non-C locales. The
6672 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6673 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6674 * [A-Za-z] are accepted by isALPHA in the C locale.
6675 */
6676 if (*d != 'z' && *d != 'Z') {
6677 do { ++*d; } while (!isALPHA(*d));
6678 return;
6679 }
6680 *(d--) -= 'z' - 'a';
6681#else
79072805
LW
6682 ++*d;
6683 if (isALPHA(*d))
6684 return;
6685 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6686#endif
79072805
LW
6687 }
6688 }
6689 /* oh,oh, the number grew */
6690 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6691 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6692 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6693 *d = d[-1];
6694 if (isDIGIT(d[1]))
6695 *d = '1';
6696 else
6697 *d = d[1];
6698}
6699
954c1994
GS
6700/*
6701=for apidoc sv_dec
6702
645c22ef
DM
6703Auto-decrement of the value in the SV, doing string to numeric conversion
6704if necessary. Handles 'get' magic.
954c1994
GS
6705
6706=cut
6707*/
6708
79072805 6709void
864dbfa3 6710Perl_sv_dec(pTHX_ register SV *sv)
79072805 6711{
97aff369 6712 dVAR;
463ee0b2
LW
6713 int flags;
6714
79072805
LW
6715 if (!sv)
6716 return;
5b295bef 6717 SvGETMAGIC(sv);
ed6116ce 6718 if (SvTHINKFIRST(sv)) {
765f542d
NC
6719 if (SvIsCOW(sv))
6720 sv_force_normal_flags(sv, 0);
0f15f207 6721 if (SvREADONLY(sv)) {
923e4eb5 6722 if (IN_PERL_RUNTIME)
cea2e8a9 6723 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6724 }
a0d0e21e 6725 if (SvROK(sv)) {
b5be31e9 6726 IV i;
9e7bc3e8
JD
6727 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6728 return;
56431972 6729 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6730 sv_unref(sv);
6731 sv_setiv(sv, i);
a0d0e21e 6732 }
ed6116ce 6733 }
28e5dec8
JH
6734 /* Unlike sv_inc we don't have to worry about string-never-numbers
6735 and keeping them magic. But we mustn't warn on punting */
8990e307 6736 flags = SvFLAGS(sv);
28e5dec8
JH
6737 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6738 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6739#ifdef PERL_PRESERVE_IVUV
28e5dec8 6740 oops_its_int:
59d8ce62 6741#endif
25da4f38
IZ
6742 if (SvIsUV(sv)) {
6743 if (SvUVX(sv) == 0) {
6744 (void)SvIOK_only(sv);
45977657 6745 SvIV_set(sv, -1);
25da4f38
IZ
6746 }
6747 else {
6748 (void)SvIOK_only_UV(sv);
f4eee32f 6749 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6750 }
25da4f38
IZ
6751 } else {
6752 if (SvIVX(sv) == IV_MIN)
65202027 6753 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6754 else {
6755 (void)SvIOK_only(sv);
45977657 6756 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6757 }
55497cff 6758 }
6759 return;
6760 }
28e5dec8 6761 if (flags & SVp_NOK) {
9d6ce603 6762 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6763 (void)SvNOK_only(sv);
6764 return;
6765 }
8990e307 6766 if (!(flags & SVp_POK)) {
ef088171
NC
6767 if ((flags & SVTYPEMASK) < SVt_PVIV)
6768 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6769 SvIV_set(sv, -1);
6770 (void)SvIOK_only(sv);
79072805
LW
6771 return;
6772 }
28e5dec8
JH
6773#ifdef PERL_PRESERVE_IVUV
6774 {
504618e9 6775 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6776 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6777 /* Need to try really hard to see if it's an integer.
6778 9.22337203685478e+18 is an integer.
6779 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6780 so $a="9.22337203685478e+18"; $a+0; $a--
6781 needs to be the same as $a="9.22337203685478e+18"; $a--
6782 or we go insane. */
d460ef45 6783
28e5dec8
JH
6784 (void) sv_2iv(sv);
6785 if (SvIOK(sv))
6786 goto oops_its_int;
6787
6788 /* sv_2iv *should* have made this an NV */
6789 if (flags & SVp_NOK) {
6790 (void)SvNOK_only(sv);
9d6ce603 6791 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6792 return;
6793 }
6794 /* I don't think we can get here. Maybe I should assert this
6795 And if we do get here I suspect that sv_setnv will croak. NWC
6796 Fall through. */
6797#if defined(USE_LONG_DOUBLE)
6798 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 6799 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6800#else
1779d84d 6801 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 6802 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6803#endif
6804 }
6805 }
6806#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6807 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6808}
6809
954c1994
GS
6810/*
6811=for apidoc sv_mortalcopy
6812
645c22ef 6813Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6814The new SV is marked as mortal. It will be destroyed "soon", either by an
6815explicit call to FREETMPS, or by an implicit call at places such as
6816statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6817
6818=cut
6819*/
6820
79072805
LW
6821/* Make a string that will exist for the duration of the expression
6822 * evaluation. Actually, it may have to last longer than that, but
6823 * hopefully we won't free it until it has been assigned to a
6824 * permanent location. */
6825
6826SV *
864dbfa3 6827Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6828{
97aff369 6829 dVAR;
463ee0b2 6830 register SV *sv;
b881518d 6831
4561caa4 6832 new_SV(sv);
79072805 6833 sv_setsv(sv,oldstr);
677b06e3
GS
6834 EXTEND_MORTAL(1);
6835 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6836 SvTEMP_on(sv);
6837 return sv;
6838}
6839
954c1994
GS
6840/*
6841=for apidoc sv_newmortal
6842
645c22ef 6843Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6844set to 1. It will be destroyed "soon", either by an explicit call to
6845FREETMPS, or by an implicit call at places such as statement boundaries.
6846See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6847
6848=cut
6849*/
6850
8990e307 6851SV *
864dbfa3 6852Perl_sv_newmortal(pTHX)
8990e307 6853{
97aff369 6854 dVAR;
8990e307
LW
6855 register SV *sv;
6856
4561caa4 6857 new_SV(sv);
8990e307 6858 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6859 EXTEND_MORTAL(1);
6860 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6861 return sv;
6862}
6863
954c1994
GS
6864/*
6865=for apidoc sv_2mortal
6866
d4236ebc
DM
6867Marks an existing SV as mortal. The SV will be destroyed "soon", either
6868by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6869statement boundaries. SvTEMP() is turned on which means that the SV's
6870string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6871and C<sv_mortalcopy>.
954c1994
GS
6872
6873=cut
6874*/
6875
79072805 6876SV *
864dbfa3 6877Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6878{
27da23d5 6879 dVAR;
79072805 6880 if (!sv)
7a5b473e 6881 return NULL;
d689ffdd 6882 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6883 return sv;
677b06e3
GS
6884 EXTEND_MORTAL(1);
6885 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6886 SvTEMP_on(sv);
79072805
LW
6887 return sv;
6888}
6889
954c1994
GS
6890/*
6891=for apidoc newSVpv
6892
6893Creates a new SV and copies a string into it. The reference count for the
6894SV is set to 1. If C<len> is zero, Perl will compute the length using
6895strlen(). For efficiency, consider using C<newSVpvn> instead.
6896
6897=cut
6898*/
6899
79072805 6900SV *
864dbfa3 6901Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6902{
97aff369 6903 dVAR;
463ee0b2 6904 register SV *sv;
79072805 6905
4561caa4 6906 new_SV(sv);
ddfa59c7 6907 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
79072805
LW
6908 return sv;
6909}
6910
954c1994
GS
6911/*
6912=for apidoc newSVpvn
6913
6914Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6915SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6916string. You are responsible for ensuring that the source string is at least
9e09f5f2 6917C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6918
6919=cut
6920*/
6921
9da1e3b5 6922SV *
864dbfa3 6923Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6924{
97aff369 6925 dVAR;
9da1e3b5
MUN
6926 register SV *sv;
6927
6928 new_SV(sv);
9da1e3b5
MUN
6929 sv_setpvn(sv,s,len);
6930 return sv;
6931}
6932
bd08039b
NC
6933
6934/*
926f8064 6935=for apidoc newSVhek
bd08039b
NC
6936
6937Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6938point to the shared string table where possible. Returns a new (undefined)
6939SV if the hek is NULL.
bd08039b
NC
6940
6941=cut
6942*/
6943
6944SV *
c1b02ed8 6945Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6946{
97aff369 6947 dVAR;
5aaec2b4
NC
6948 if (!hek) {
6949 SV *sv;
6950
6951 new_SV(sv);
6952 return sv;
6953 }
6954
bd08039b
NC
6955 if (HEK_LEN(hek) == HEf_SVKEY) {
6956 return newSVsv(*(SV**)HEK_KEY(hek));
6957 } else {
6958 const int flags = HEK_FLAGS(hek);
6959 if (flags & HVhek_WASUTF8) {
6960 /* Trouble :-)
6961 Andreas would like keys he put in as utf8 to come back as utf8
6962 */
6963 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6964 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6965 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6966
6967 SvUTF8_on (sv);
6968 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6969 return sv;
45e34800 6970 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
bd08039b
NC
6971 /* We don't have a pointer to the hv, so we have to replicate the
6972 flag into every HEK. This hv is using custom a hasing
6973 algorithm. Hence we can't return a shared string scalar, as
6974 that would contain the (wrong) hash value, and might get passed
45e34800
NC
6975 into an hv routine with a regular hash.
6976 Similarly, a hash that isn't using shared hash keys has to have
6977 the flag in every key so that we know not to try to call
6978 share_hek_kek on it. */
bd08039b 6979
b64e5050 6980 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6981 if (HEK_UTF8(hek))
6982 SvUTF8_on (sv);
6983 return sv;
6984 }
6985 /* This will be overwhelminly the most common case. */
409dfe77
NC
6986 {
6987 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6988 more efficient than sharepvn(). */
6989 SV *sv;
6990
6991 new_SV(sv);
6992 sv_upgrade(sv, SVt_PV);
6993 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
6994 SvCUR_set(sv, HEK_LEN(hek));
6995 SvLEN_set(sv, 0);
6996 SvREADONLY_on(sv);
6997 SvFAKE_on(sv);
6998 SvPOK_on(sv);
6999 if (HEK_UTF8(hek))
7000 SvUTF8_on(sv);
7001 return sv;
7002 }
bd08039b
NC
7003 }
7004}
7005
1c846c1f
NIS
7006/*
7007=for apidoc newSVpvn_share
7008
3f7c398e 7009Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7010table. If the string does not already exist in the table, it is created
7011first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7012slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7013otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7014is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7015hash lookup will avoid string compare.
1c846c1f
NIS
7016
7017=cut
7018*/
7019
7020SV *
c3654f1a 7021Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7022{
97aff369 7023 dVAR;
1c846c1f 7024 register SV *sv;
c3654f1a 7025 bool is_utf8 = FALSE;
a51caccf
NC
7026 const char *const orig_src = src;
7027
c3654f1a 7028 if (len < 0) {
77caf834 7029 STRLEN tmplen = -len;
c3654f1a 7030 is_utf8 = TRUE;
75a54232 7031 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7032 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7033 len = tmplen;
7034 }
1c846c1f 7035 if (!hash)
5afd6d42 7036 PERL_HASH(hash, src, len);
1c846c1f 7037 new_SV(sv);
bdd68bc3 7038 sv_upgrade(sv, SVt_PV);
f880fe2f 7039 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7040 SvCUR_set(sv, len);
b162af07 7041 SvLEN_set(sv, 0);
1c846c1f
NIS
7042 SvREADONLY_on(sv);
7043 SvFAKE_on(sv);
7044 SvPOK_on(sv);
c3654f1a
IH
7045 if (is_utf8)
7046 SvUTF8_on(sv);
a51caccf
NC
7047 if (src != orig_src)
7048 Safefree(src);
1c846c1f
NIS
7049 return sv;
7050}
7051
645c22ef 7052
cea2e8a9 7053#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7054
7055/* pTHX_ magic can't cope with varargs, so this is a no-context
7056 * version of the main function, (which may itself be aliased to us).
7057 * Don't access this version directly.
7058 */
7059
46fc3d4c 7060SV *
cea2e8a9 7061Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7062{
cea2e8a9 7063 dTHX;
46fc3d4c 7064 register SV *sv;
7065 va_list args;
46fc3d4c 7066 va_start(args, pat);
c5be433b 7067 sv = vnewSVpvf(pat, &args);
46fc3d4c 7068 va_end(args);
7069 return sv;
7070}
cea2e8a9 7071#endif
46fc3d4c 7072
954c1994
GS
7073/*
7074=for apidoc newSVpvf
7075
645c22ef 7076Creates a new SV and initializes it with the string formatted like
954c1994
GS
7077C<sprintf>.
7078
7079=cut
7080*/
7081
cea2e8a9
GS
7082SV *
7083Perl_newSVpvf(pTHX_ const char* pat, ...)
7084{
7085 register SV *sv;
7086 va_list args;
cea2e8a9 7087 va_start(args, pat);
c5be433b 7088 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7089 va_end(args);
7090 return sv;
7091}
46fc3d4c 7092
645c22ef
DM
7093/* backend for newSVpvf() and newSVpvf_nocontext() */
7094
79072805 7095SV *
c5be433b
GS
7096Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7097{
97aff369 7098 dVAR;
c5be433b
GS
7099 register SV *sv;
7100 new_SV(sv);
4608196e 7101 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7102 return sv;
7103}
7104
954c1994
GS
7105/*
7106=for apidoc newSVnv
7107
7108Creates a new SV and copies a floating point value into it.
7109The reference count for the SV is set to 1.
7110
7111=cut
7112*/
7113
c5be433b 7114SV *
65202027 7115Perl_newSVnv(pTHX_ NV n)
79072805 7116{
97aff369 7117 dVAR;
463ee0b2 7118 register SV *sv;
79072805 7119
4561caa4 7120 new_SV(sv);
79072805
LW
7121 sv_setnv(sv,n);
7122 return sv;
7123}
7124
954c1994
GS
7125/*
7126=for apidoc newSViv
7127
7128Creates a new SV and copies an integer into it. The reference count for the
7129SV is set to 1.
7130
7131=cut
7132*/
7133
79072805 7134SV *
864dbfa3 7135Perl_newSViv(pTHX_ IV i)
79072805 7136{
97aff369 7137 dVAR;
463ee0b2 7138 register SV *sv;
79072805 7139
4561caa4 7140 new_SV(sv);
79072805
LW
7141 sv_setiv(sv,i);
7142 return sv;
7143}
7144
954c1994 7145/*
1a3327fb
JH
7146=for apidoc newSVuv
7147
7148Creates a new SV and copies an unsigned integer into it.
7149The reference count for the SV is set to 1.
7150
7151=cut
7152*/
7153
7154SV *
7155Perl_newSVuv(pTHX_ UV u)
7156{
97aff369 7157 dVAR;
1a3327fb
JH
7158 register SV *sv;
7159
7160 new_SV(sv);
7161 sv_setuv(sv,u);
7162 return sv;
7163}
7164
7165/*
954c1994
GS
7166=for apidoc newRV_noinc
7167
7168Creates an RV wrapper for an SV. The reference count for the original
7169SV is B<not> incremented.
7170
7171=cut
7172*/
7173
2304df62 7174SV *
864dbfa3 7175Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7176{
97aff369 7177 dVAR;
2304df62
AD
7178 register SV *sv;
7179
4561caa4 7180 new_SV(sv);
2304df62 7181 sv_upgrade(sv, SVt_RV);
76e3520e 7182 SvTEMP_off(tmpRef);
b162af07 7183 SvRV_set(sv, tmpRef);
2304df62 7184 SvROK_on(sv);
2304df62
AD
7185 return sv;
7186}
7187
ff276b08 7188/* newRV_inc is the official function name to use now.
645c22ef
DM
7189 * newRV_inc is in fact #defined to newRV in sv.h
7190 */
7191
5f05dabc 7192SV *
7f466ec7 7193Perl_newRV(pTHX_ SV *sv)
5f05dabc 7194{
97aff369 7195 dVAR;
7f466ec7 7196 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7197}
5f05dabc 7198
954c1994
GS
7199/*
7200=for apidoc newSVsv
7201
7202Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7203(Uses C<sv_setsv>).
954c1994
GS
7204
7205=cut
7206*/
7207
79072805 7208SV *
864dbfa3 7209Perl_newSVsv(pTHX_ register SV *old)
79072805 7210{
97aff369 7211 dVAR;
463ee0b2 7212 register SV *sv;
79072805
LW
7213
7214 if (!old)
7a5b473e 7215 return NULL;
8990e307 7216 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7217 if (ckWARN_d(WARN_INTERNAL))
9014280d 7218 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7219 return NULL;
79072805 7220 }
4561caa4 7221 new_SV(sv);
e90aabeb
NC
7222 /* SV_GMAGIC is the default for sv_setv()
7223 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7224 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7225 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7226 return sv;
79072805
LW
7227}
7228
645c22ef
DM
7229/*
7230=for apidoc sv_reset
7231
7232Underlying implementation for the C<reset> Perl function.
7233Note that the perl-level function is vaguely deprecated.
7234
7235=cut
7236*/
7237
79072805 7238void
e1ec3a88 7239Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7240{
27da23d5 7241 dVAR;
4802d5d7 7242 char todo[PERL_UCHAR_MAX+1];
79072805 7243
49d8d3a1
MB
7244 if (!stash)
7245 return;
7246
79072805 7247 if (!*s) { /* reset ?? searches */
aec46f14 7248 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7249 if (mg) {
7250 PMOP *pm = (PMOP *) mg->mg_obj;
7251 while (pm) {
7252 pm->op_pmdynflags &= ~PMdf_USED;
7253 pm = pm->op_pmnext;
7254 }
79072805
LW
7255 }
7256 return;
7257 }
7258
7259 /* reset variables */
7260
7261 if (!HvARRAY(stash))
7262 return;
463ee0b2
LW
7263
7264 Zero(todo, 256, char);
79072805 7265 while (*s) {
b464bac0
AL
7266 I32 max;
7267 I32 i = (unsigned char)*s;
79072805
LW
7268 if (s[1] == '-') {
7269 s += 2;
7270 }
4802d5d7 7271 max = (unsigned char)*s++;
79072805 7272 for ( ; i <= max; i++) {
463ee0b2
LW
7273 todo[i] = 1;
7274 }
a0d0e21e 7275 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7276 HE *entry;
79072805 7277 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7278 entry;
7279 entry = HeNEXT(entry))
7280 {
b464bac0
AL
7281 register GV *gv;
7282 register SV *sv;
7283
1edc1566 7284 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7285 continue;
1edc1566 7286 gv = (GV*)HeVAL(entry);
79072805 7287 sv = GvSV(gv);
e203899d
NC
7288 if (sv) {
7289 if (SvTHINKFIRST(sv)) {
7290 if (!SvREADONLY(sv) && SvROK(sv))
7291 sv_unref(sv);
7292 /* XXX Is this continue a bug? Why should THINKFIRST
7293 exempt us from resetting arrays and hashes? */
7294 continue;
7295 }
7296 SvOK_off(sv);
7297 if (SvTYPE(sv) >= SVt_PV) {
7298 SvCUR_set(sv, 0);
bd61b366 7299 if (SvPVX_const(sv) != NULL)
e203899d
NC
7300 *SvPVX(sv) = '\0';
7301 SvTAINT(sv);
7302 }
79072805
LW
7303 }
7304 if (GvAV(gv)) {
7305 av_clear(GvAV(gv));
7306 }
bfcb3514 7307 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7308#if defined(VMS)
7309 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7310#else /* ! VMS */
463ee0b2 7311 hv_clear(GvHV(gv));
b0269e46
AB
7312# if defined(USE_ENVIRON_ARRAY)
7313 if (gv == PL_envgv)
7314 my_clearenv();
7315# endif /* USE_ENVIRON_ARRAY */
7316#endif /* VMS */
79072805
LW
7317 }
7318 }
7319 }
7320 }
7321}
7322
645c22ef
DM
7323/*
7324=for apidoc sv_2io
7325
7326Using various gambits, try to get an IO from an SV: the IO slot if its a
7327GV; or the recursive result if we're an RV; or the IO slot of the symbol
7328named after the PV if we're a string.
7329
7330=cut
7331*/
7332
46fc3d4c 7333IO*
864dbfa3 7334Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7335{
7336 IO* io;
7337 GV* gv;
7338
7339 switch (SvTYPE(sv)) {
7340 case SVt_PVIO:
7341 io = (IO*)sv;
7342 break;
7343 case SVt_PVGV:
7344 gv = (GV*)sv;
7345 io = GvIO(gv);
7346 if (!io)
cea2e8a9 7347 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7348 break;
7349 default:
7350 if (!SvOK(sv))
cea2e8a9 7351 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7352 if (SvROK(sv))
7353 return sv_2io(SvRV(sv));
f776e3cd 7354 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7355 if (gv)
7356 io = GvIO(gv);
7357 else
7358 io = 0;
7359 if (!io)
95b63a38 7360 Perl_croak(aTHX_ "Bad filehandle: %"SVf, (void*)sv);
46fc3d4c 7361 break;
7362 }
7363 return io;
7364}
7365
645c22ef
DM
7366/*
7367=for apidoc sv_2cv
7368
7369Using various gambits, try to get a CV from an SV; in addition, try if
7370possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7371The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7372
7373=cut
7374*/
7375
79072805 7376CV *
864dbfa3 7377Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7378{
27da23d5 7379 dVAR;
a0714e2c 7380 GV *gv = NULL;
601f1833 7381 CV *cv = NULL;
79072805 7382
85dec29a
NC
7383 if (!sv) {
7384 *st = NULL;
7385 *gvp = NULL;
7386 return NULL;
7387 }
79072805 7388 switch (SvTYPE(sv)) {
79072805
LW
7389 case SVt_PVCV:
7390 *st = CvSTASH(sv);
a0714e2c 7391 *gvp = NULL;
79072805
LW
7392 return (CV*)sv;
7393 case SVt_PVHV:
7394 case SVt_PVAV:
ef58ba18 7395 *st = NULL;
a0714e2c 7396 *gvp = NULL;
601f1833 7397 return NULL;
8990e307
LW
7398 case SVt_PVGV:
7399 gv = (GV*)sv;
a0d0e21e 7400 *gvp = gv;
8990e307
LW
7401 *st = GvESTASH(gv);
7402 goto fix_gv;
7403
79072805 7404 default:
5b295bef 7405 SvGETMAGIC(sv);
a0d0e21e 7406 if (SvROK(sv)) {
823a54a3 7407 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7408 tryAMAGICunDEREF(to_cv);
7409
62f274bf
GS
7410 sv = SvRV(sv);
7411 if (SvTYPE(sv) == SVt_PVCV) {
7412 cv = (CV*)sv;
a0714e2c 7413 *gvp = NULL;
62f274bf
GS
7414 *st = CvSTASH(cv);
7415 return cv;
7416 }
7417 else if(isGV(sv))
7418 gv = (GV*)sv;
7419 else
cea2e8a9 7420 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7421 }
62f274bf 7422 else if (isGV(sv))
79072805
LW
7423 gv = (GV*)sv;
7424 else
7a5fd60d 7425 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7426 *gvp = gv;
ef58ba18
NC
7427 if (!gv) {
7428 *st = NULL;
601f1833 7429 return NULL;
ef58ba18 7430 }
e26df76a
NC
7431 /* Some flags to gv_fetchsv mean don't really create the GV */
7432 if (SvTYPE(gv) != SVt_PVGV) {
7433 *st = NULL;
7434 return NULL;
7435 }
79072805 7436 *st = GvESTASH(gv);
8990e307 7437 fix_gv:
8ebc5c01 7438 if (lref && !GvCVu(gv)) {
4633a7c4 7439 SV *tmpsv;
748a9306 7440 ENTER;
561b68a9 7441 tmpsv = newSV(0);
bd61b366 7442 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7443 /* XXX this is probably not what they think they're getting.
7444 * It has the same effect as "sub name;", i.e. just a forward
7445 * declaration! */
774d564b 7446 newSUB(start_subparse(FALSE, 0),
4633a7c4 7447 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7448 NULL, NULL);
748a9306 7449 LEAVE;
8ebc5c01 7450 if (!GvCVu(gv))
35c1215d 7451 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
95b63a38 7452 (void*)sv);
8990e307 7453 }
8ebc5c01 7454 return GvCVu(gv);
79072805
LW
7455 }
7456}
7457
c461cf8f
JH
7458/*
7459=for apidoc sv_true
7460
7461Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7462Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7463instead use an in-line version.
c461cf8f
JH
7464
7465=cut
7466*/
7467
79072805 7468I32
864dbfa3 7469Perl_sv_true(pTHX_ register SV *sv)
79072805 7470{
8990e307
LW
7471 if (!sv)
7472 return 0;
79072805 7473 if (SvPOK(sv)) {
823a54a3
AL
7474 register const XPV* const tXpv = (XPV*)SvANY(sv);
7475 if (tXpv &&
c2f1de04 7476 (tXpv->xpv_cur > 1 ||
339049b0 7477 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7478 return 1;
7479 else
7480 return 0;
7481 }
7482 else {
7483 if (SvIOK(sv))
463ee0b2 7484 return SvIVX(sv) != 0;
79072805
LW
7485 else {
7486 if (SvNOK(sv))
463ee0b2 7487 return SvNVX(sv) != 0.0;
79072805 7488 else
463ee0b2 7489 return sv_2bool(sv);
79072805
LW
7490 }
7491 }
7492}
79072805 7493
645c22ef 7494/*
c461cf8f
JH
7495=for apidoc sv_pvn_force
7496
7497Get a sensible string out of the SV somehow.
645c22ef
DM
7498A private implementation of the C<SvPV_force> macro for compilers which
7499can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7500
8d6d96c1
HS
7501=for apidoc sv_pvn_force_flags
7502
7503Get a sensible string out of the SV somehow.
7504If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7505appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7506implemented in terms of this function.
645c22ef
DM
7507You normally want to use the various wrapper macros instead: see
7508C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7509
7510=cut
7511*/
7512
7513char *
7514Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7515{
97aff369 7516 dVAR;
6fc92669 7517 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7518 sv_force_normal_flags(sv, 0);
1c846c1f 7519
a0d0e21e 7520 if (SvPOK(sv)) {
13c5b33c
NC
7521 if (lp)
7522 *lp = SvCUR(sv);
a0d0e21e
LW
7523 }
7524 else {
a3b680e6 7525 char *s;
13c5b33c
NC
7526 STRLEN len;
7527
4d84ee25 7528 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7529 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7530 if (PL_op)
7531 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7532 ref, OP_NAME(PL_op));
4d84ee25 7533 else
b64e5050 7534 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7535 }
b64e5050 7536 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7537 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7538 OP_NAME(PL_op));
b64e5050 7539 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7540 if (lp)
7541 *lp = len;
7542
3f7c398e 7543 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7544 if (SvROK(sv))
7545 sv_unref(sv);
862a34c6 7546 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7547 SvGROW(sv, len + 1);
706aa1c9 7548 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7549 SvCUR_set(sv, len);
7550 *SvEND(sv) = '\0';
7551 }
7552 if (!SvPOK(sv)) {
7553 SvPOK_on(sv); /* validate pointer */
7554 SvTAINT(sv);
1d7c1841 7555 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7556 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7557 }
7558 }
4d84ee25 7559 return SvPVX_mutable(sv);
a0d0e21e
LW
7560}
7561
645c22ef 7562/*
645c22ef
DM
7563=for apidoc sv_pvbyten_force
7564
0feed65a 7565The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7566
7567=cut
7568*/
7569
7340a771
GS
7570char *
7571Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7572{
46ec2f14 7573 sv_pvn_force(sv,lp);
ffebcc3e 7574 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7575 *lp = SvCUR(sv);
7576 return SvPVX(sv);
7340a771
GS
7577}
7578
645c22ef 7579/*
c461cf8f
JH
7580=for apidoc sv_pvutf8n_force
7581
0feed65a 7582The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7583
7584=cut
7585*/
7586
7340a771
GS
7587char *
7588Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7589{
46ec2f14 7590 sv_pvn_force(sv,lp);
560a288e 7591 sv_utf8_upgrade(sv);
46ec2f14
TS
7592 *lp = SvCUR(sv);
7593 return SvPVX(sv);
7340a771
GS
7594}
7595
c461cf8f
JH
7596/*
7597=for apidoc sv_reftype
7598
7599Returns a string describing what the SV is a reference to.
7600
7601=cut
7602*/
7603
2b388283 7604const char *
bfed75c6 7605Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7606{
07409e01
NC
7607 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7608 inside return suggests a const propagation bug in g++. */
c86bf373 7609 if (ob && SvOBJECT(sv)) {
1b6737cc 7610 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7611 return name ? name : (char *) "__ANON__";
c86bf373 7612 }
a0d0e21e
LW
7613 else {
7614 switch (SvTYPE(sv)) {
7615 case SVt_NULL:
7616 case SVt_IV:
7617 case SVt_NV:
7618 case SVt_RV:
7619 case SVt_PV:
7620 case SVt_PVIV:
7621 case SVt_PVNV:
7622 case SVt_PVMG:
7623 case SVt_PVBM:
1cb0ed9b 7624 if (SvVOK(sv))
439cb1c4 7625 return "VSTRING";
a0d0e21e
LW
7626 if (SvROK(sv))
7627 return "REF";
7628 else
7629 return "SCALAR";
1cb0ed9b 7630
07409e01 7631 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7632 /* tied lvalues should appear to be
7633 * scalars for backwards compatitbility */
7634 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7635 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7636 case SVt_PVAV: return "ARRAY";
7637 case SVt_PVHV: return "HASH";
7638 case SVt_PVCV: return "CODE";
7639 case SVt_PVGV: return "GLOB";
1d2dff63 7640 case SVt_PVFM: return "FORMAT";
27f9d8f3 7641 case SVt_PVIO: return "IO";
a0d0e21e
LW
7642 default: return "UNKNOWN";
7643 }
7644 }
7645}
7646
954c1994
GS
7647/*
7648=for apidoc sv_isobject
7649
7650Returns a boolean indicating whether the SV is an RV pointing to a blessed
7651object. If the SV is not an RV, or if the object is not blessed, then this
7652will return false.
7653
7654=cut
7655*/
7656
463ee0b2 7657int
864dbfa3 7658Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7659{
68dc0745 7660 if (!sv)
7661 return 0;
5b295bef 7662 SvGETMAGIC(sv);
85e6fe83
LW
7663 if (!SvROK(sv))
7664 return 0;
7665 sv = (SV*)SvRV(sv);
7666 if (!SvOBJECT(sv))
7667 return 0;
7668 return 1;
7669}
7670
954c1994
GS
7671/*
7672=for apidoc sv_isa
7673
7674Returns a boolean indicating whether the SV is blessed into the specified
7675class. This does not check for subtypes; use C<sv_derived_from> to verify
7676an inheritance relationship.
7677
7678=cut
7679*/
7680
85e6fe83 7681int
864dbfa3 7682Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7683{
bfcb3514 7684 const char *hvname;
68dc0745 7685 if (!sv)
7686 return 0;
5b295bef 7687 SvGETMAGIC(sv);
ed6116ce 7688 if (!SvROK(sv))
463ee0b2 7689 return 0;
ed6116ce
LW
7690 sv = (SV*)SvRV(sv);
7691 if (!SvOBJECT(sv))
463ee0b2 7692 return 0;
bfcb3514
NC
7693 hvname = HvNAME_get(SvSTASH(sv));
7694 if (!hvname)
e27ad1f2 7695 return 0;
463ee0b2 7696
bfcb3514 7697 return strEQ(hvname, name);
463ee0b2
LW
7698}
7699
954c1994
GS
7700/*
7701=for apidoc newSVrv
7702
7703Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7704it will be upgraded to one. If C<classname> is non-null then the new SV will
7705be blessed in the specified package. The new SV is returned and its
7706reference count is 1.
7707
7708=cut
7709*/
7710
463ee0b2 7711SV*
864dbfa3 7712Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7713{
97aff369 7714 dVAR;
463ee0b2
LW
7715 SV *sv;
7716
4561caa4 7717 new_SV(sv);
51cf62d8 7718
765f542d 7719 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7720 SvAMAGIC_off(rv);
51cf62d8 7721
0199fce9 7722 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7723 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7724 SvREFCNT(rv) = 0;
7725 sv_clear(rv);
7726 SvFLAGS(rv) = 0;
7727 SvREFCNT(rv) = refcnt;
0199fce9 7728
dc5494d2
NC
7729 sv_upgrade(rv, SVt_RV);
7730 } else if (SvROK(rv)) {
7731 SvREFCNT_dec(SvRV(rv));
7732 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7733 sv_upgrade(rv, SVt_RV);
7734 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7735 SvPV_free(rv);
0199fce9
JD
7736 SvCUR_set(rv, 0);
7737 SvLEN_set(rv, 0);
7738 }
51cf62d8 7739
0c34ef67 7740 SvOK_off(rv);
b162af07 7741 SvRV_set(rv, sv);
ed6116ce 7742 SvROK_on(rv);
463ee0b2 7743
a0d0e21e 7744 if (classname) {
1b6737cc 7745 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7746 (void)sv_bless(rv, stash);
7747 }
7748 return sv;
7749}
7750
954c1994
GS
7751/*
7752=for apidoc sv_setref_pv
7753
7754Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7755argument will be upgraded to an RV. That RV will be modified to point to
7756the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7757into the SV. The C<classname> argument indicates the package for the
bd61b366 7758blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7759will have a reference count of 1, and the RV will be returned.
954c1994
GS
7760
7761Do not use with other Perl types such as HV, AV, SV, CV, because those
7762objects will become corrupted by the pointer copy process.
7763
7764Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7765
7766=cut
7767*/
7768
a0d0e21e 7769SV*
864dbfa3 7770Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7771{
97aff369 7772 dVAR;
189b2af5 7773 if (!pv) {
3280af22 7774 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7775 SvSETMAGIC(rv);
7776 }
a0d0e21e 7777 else
56431972 7778 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7779 return rv;
7780}
7781
954c1994
GS
7782/*
7783=for apidoc sv_setref_iv
7784
7785Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7786argument will be upgraded to an RV. That RV will be modified to point to
7787the new SV. The C<classname> argument indicates the package for the
bd61b366 7788blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7789will have a reference count of 1, and the RV will be returned.
954c1994
GS
7790
7791=cut
7792*/
7793
a0d0e21e 7794SV*
864dbfa3 7795Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7796{
7797 sv_setiv(newSVrv(rv,classname), iv);
7798 return rv;
7799}
7800
954c1994 7801/*
e1c57cef
JH
7802=for apidoc sv_setref_uv
7803
7804Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7805argument will be upgraded to an RV. That RV will be modified to point to
7806the new SV. The C<classname> argument indicates the package for the
bd61b366 7807blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7808will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7809
7810=cut
7811*/
7812
7813SV*
7814Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7815{
7816 sv_setuv(newSVrv(rv,classname), uv);
7817 return rv;
7818}
7819
7820/*
954c1994
GS
7821=for apidoc sv_setref_nv
7822
7823Copies a double into a new SV, optionally blessing the SV. The C<rv>
7824argument will be upgraded to an RV. That RV will be modified to point to
7825the new SV. The C<classname> argument indicates the package for the
bd61b366 7826blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7827will have a reference count of 1, and the RV will be returned.
954c1994
GS
7828
7829=cut
7830*/
7831
a0d0e21e 7832SV*
65202027 7833Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7834{
7835 sv_setnv(newSVrv(rv,classname), nv);
7836 return rv;
7837}
463ee0b2 7838
954c1994
GS
7839/*
7840=for apidoc sv_setref_pvn
7841
7842Copies a string into a new SV, optionally blessing the SV. The length of the
7843string must be specified with C<n>. The C<rv> argument will be upgraded to
7844an RV. That RV will be modified to point to the new SV. The C<classname>
7845argument indicates the package for the blessing. Set C<classname> to
bd61b366 7846C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7847of 1, and the RV will be returned.
954c1994
GS
7848
7849Note that C<sv_setref_pv> copies the pointer while this copies the string.
7850
7851=cut
7852*/
7853
a0d0e21e 7854SV*
1b6737cc 7855Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7856{
7857 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7858 return rv;
7859}
7860
954c1994
GS
7861/*
7862=for apidoc sv_bless
7863
7864Blesses an SV into a specified package. The SV must be an RV. The package
7865must be designated by its stash (see C<gv_stashpv()>). The reference count
7866of the SV is unaffected.
7867
7868=cut
7869*/
7870
a0d0e21e 7871SV*
864dbfa3 7872Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7873{
97aff369 7874 dVAR;
76e3520e 7875 SV *tmpRef;
a0d0e21e 7876 if (!SvROK(sv))
cea2e8a9 7877 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7878 tmpRef = SvRV(sv);
7879 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7880 if (SvREADONLY(tmpRef))
cea2e8a9 7881 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7882 if (SvOBJECT(tmpRef)) {
7883 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7884 --PL_sv_objcount;
76e3520e 7885 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7886 }
a0d0e21e 7887 }
76e3520e
GS
7888 SvOBJECT_on(tmpRef);
7889 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7890 ++PL_sv_objcount;
862a34c6 7891 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7892 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7893
2e3febc6
CS
7894 if (Gv_AMG(stash))
7895 SvAMAGIC_on(sv);
7896 else
7897 SvAMAGIC_off(sv);
a0d0e21e 7898
1edbfb88
AB
7899 if(SvSMAGICAL(tmpRef))
7900 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7901 mg_set(tmpRef);
7902
7903
ecdeb87c 7904
a0d0e21e
LW
7905 return sv;
7906}
7907
645c22ef 7908/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7909 */
7910
76e3520e 7911STATIC void
cea2e8a9 7912S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7913{
97aff369 7914 dVAR;
850fabdf 7915 void *xpvmg;
b37c2d43 7916 SV * const temp = sv_newmortal();
850fabdf 7917
a0d0e21e
LW
7918 assert(SvTYPE(sv) == SVt_PVGV);
7919 SvFAKE_off(sv);
180488f8
NC
7920 gv_efullname3(temp, (GV *) sv, "*");
7921
f7877b28 7922 if (GvGP(sv)) {
1edc1566 7923 gp_free((GV*)sv);
f7877b28 7924 }
e826b3c7 7925 if (GvSTASH(sv)) {
e15faf7d 7926 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7927 GvSTASH(sv) = NULL;
e826b3c7 7928 }
a5f75d66 7929 GvMULTI_off(sv);
acda4c6a
NC
7930 if (GvNAME_HEK(sv)) {
7931 unshare_hek(GvNAME_HEK(sv));
7932 }
dedf8e73 7933 SvSCREAM_off(sv);
850fabdf
GS
7934
7935 /* need to keep SvANY(sv) in the right arena */
7936 xpvmg = new_XPVMG();
7937 StructCopy(SvANY(sv), xpvmg, XPVMG);
7938 del_XPVGV(SvANY(sv));
7939 SvANY(sv) = xpvmg;
7940
a0d0e21e
LW
7941 SvFLAGS(sv) &= ~SVTYPEMASK;
7942 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7943
7944 /* Intentionally not calling any local SET magic, as this isn't so much a
7945 set operation as merely an internal storage change. */
7946 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7947}
7948
954c1994 7949/*
840a7b70 7950=for apidoc sv_unref_flags
954c1994
GS
7951
7952Unsets the RV status of the SV, and decrements the reference count of
7953whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7954as a reversal of C<newSVrv>. The C<cflags> argument can contain
7955C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7956(otherwise the decrementing is conditional on the reference count being
7957different from one or the reference being a readonly SV).
7889fe52 7958See C<SvROK_off>.
954c1994
GS
7959
7960=cut
7961*/
7962
ed6116ce 7963void
e15faf7d 7964Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7965{
b64e5050 7966 SV* const target = SvRV(ref);
810b8aa5 7967
e15faf7d
NC
7968 if (SvWEAKREF(ref)) {
7969 sv_del_backref(target, ref);
7970 SvWEAKREF_off(ref);
7971 SvRV_set(ref, NULL);
810b8aa5
GS
7972 return;
7973 }
e15faf7d
NC
7974 SvRV_set(ref, NULL);
7975 SvROK_off(ref);
7976 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7977 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7978 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7979 SvREFCNT_dec(target);
840a7b70 7980 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7981 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7982}
8990e307 7983
840a7b70 7984/*
645c22ef
DM
7985=for apidoc sv_untaint
7986
7987Untaint an SV. Use C<SvTAINTED_off> instead.
7988=cut
7989*/
7990
bbce6d69 7991void
864dbfa3 7992Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7993{
13f57bf8 7994 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 7995 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 7996 if (mg)
565764a8 7997 mg->mg_len &= ~1;
36477c24 7998 }
bbce6d69 7999}
8000
645c22ef
DM
8001/*
8002=for apidoc sv_tainted
8003
8004Test an SV for taintedness. Use C<SvTAINTED> instead.
8005=cut
8006*/
8007
bbce6d69 8008bool
864dbfa3 8009Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8010{
13f57bf8 8011 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8012 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8013 if (mg && (mg->mg_len & 1) )
36477c24 8014 return TRUE;
8015 }
8016 return FALSE;
bbce6d69 8017}
8018
09540bc3
JH
8019/*
8020=for apidoc sv_setpviv
8021
8022Copies an integer into the given SV, also updating its string value.
8023Does not handle 'set' magic. See C<sv_setpviv_mg>.
8024
8025=cut
8026*/
8027
8028void
8029Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8030{
8031 char buf[TYPE_CHARS(UV)];
8032 char *ebuf;
b64e5050 8033 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8034
8035 sv_setpvn(sv, ptr, ebuf - ptr);
8036}
8037
8038/*
8039=for apidoc sv_setpviv_mg
8040
8041Like C<sv_setpviv>, but also handles 'set' magic.
8042
8043=cut
8044*/
8045
8046void
8047Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8048{
df7eb254 8049 sv_setpviv(sv, iv);
09540bc3
JH
8050 SvSETMAGIC(sv);
8051}
8052
cea2e8a9 8053#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8054
8055/* pTHX_ magic can't cope with varargs, so this is a no-context
8056 * version of the main function, (which may itself be aliased to us).
8057 * Don't access this version directly.
8058 */
8059
cea2e8a9
GS
8060void
8061Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8062{
8063 dTHX;
8064 va_list args;
8065 va_start(args, pat);
c5be433b 8066 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8067 va_end(args);
8068}
8069
645c22ef
DM
8070/* pTHX_ magic can't cope with varargs, so this is a no-context
8071 * version of the main function, (which may itself be aliased to us).
8072 * Don't access this version directly.
8073 */
cea2e8a9
GS
8074
8075void
8076Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8077{
8078 dTHX;
8079 va_list args;
8080 va_start(args, pat);
c5be433b 8081 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8082 va_end(args);
cea2e8a9
GS
8083}
8084#endif
8085
954c1994
GS
8086/*
8087=for apidoc sv_setpvf
8088
bffc3d17
SH
8089Works like C<sv_catpvf> but copies the text into the SV instead of
8090appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8091
8092=cut
8093*/
8094
46fc3d4c 8095void
864dbfa3 8096Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8097{
8098 va_list args;
46fc3d4c 8099 va_start(args, pat);
c5be433b 8100 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8101 va_end(args);
8102}
8103
bffc3d17
SH
8104/*
8105=for apidoc sv_vsetpvf
8106
8107Works like C<sv_vcatpvf> but copies the text into the SV instead of
8108appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8109
8110Usually used via its frontend C<sv_setpvf>.
8111
8112=cut
8113*/
645c22ef 8114
c5be433b
GS
8115void
8116Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8117{
4608196e 8118 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8119}
ef50df4b 8120
954c1994
GS
8121/*
8122=for apidoc sv_setpvf_mg
8123
8124Like C<sv_setpvf>, but also handles 'set' magic.
8125
8126=cut
8127*/
8128
ef50df4b 8129void
864dbfa3 8130Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8131{
8132 va_list args;
ef50df4b 8133 va_start(args, pat);
c5be433b 8134 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8135 va_end(args);
c5be433b
GS
8136}
8137
bffc3d17
SH
8138/*
8139=for apidoc sv_vsetpvf_mg
8140
8141Like C<sv_vsetpvf>, but also handles 'set' magic.
8142
8143Usually used via its frontend C<sv_setpvf_mg>.
8144
8145=cut
8146*/
645c22ef 8147
c5be433b
GS
8148void
8149Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8150{
4608196e 8151 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8152 SvSETMAGIC(sv);
8153}
8154
cea2e8a9 8155#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8156
8157/* pTHX_ magic can't cope with varargs, so this is a no-context
8158 * version of the main function, (which may itself be aliased to us).
8159 * Don't access this version directly.
8160 */
8161
cea2e8a9
GS
8162void
8163Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8164{
8165 dTHX;
8166 va_list args;
8167 va_start(args, pat);
c5be433b 8168 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8169 va_end(args);
8170}
8171
645c22ef
DM
8172/* pTHX_ magic can't cope with varargs, so this is a no-context
8173 * version of the main function, (which may itself be aliased to us).
8174 * Don't access this version directly.
8175 */
8176
cea2e8a9
GS
8177void
8178Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8179{
8180 dTHX;
8181 va_list args;
8182 va_start(args, pat);
c5be433b 8183 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8184 va_end(args);
cea2e8a9
GS
8185}
8186#endif
8187
954c1994
GS
8188/*
8189=for apidoc sv_catpvf
8190
d5ce4a7c
GA
8191Processes its arguments like C<sprintf> and appends the formatted
8192output to an SV. If the appended data contains "wide" characters
8193(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8194and characters >255 formatted with %c), the original SV might get
bffc3d17 8195upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8196C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8197valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8198
d5ce4a7c 8199=cut */
954c1994 8200
46fc3d4c 8201void
864dbfa3 8202Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8203{
8204 va_list args;
46fc3d4c 8205 va_start(args, pat);
c5be433b 8206 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8207 va_end(args);
8208}
8209
bffc3d17
SH
8210/*
8211=for apidoc sv_vcatpvf
8212
8213Processes its arguments like C<vsprintf> and appends the formatted output
8214to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8215
8216Usually used via its frontend C<sv_catpvf>.
8217
8218=cut
8219*/
645c22ef 8220
ef50df4b 8221void
c5be433b
GS
8222Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8223{
4608196e 8224 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8225}
8226
954c1994
GS
8227/*
8228=for apidoc sv_catpvf_mg
8229
8230Like C<sv_catpvf>, but also handles 'set' magic.
8231
8232=cut
8233*/
8234
c5be433b 8235void
864dbfa3 8236Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8237{
8238 va_list args;
ef50df4b 8239 va_start(args, pat);
c5be433b 8240 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8241 va_end(args);
c5be433b
GS
8242}
8243
bffc3d17
SH
8244/*
8245=for apidoc sv_vcatpvf_mg
8246
8247Like C<sv_vcatpvf>, but also handles 'set' magic.
8248
8249Usually used via its frontend C<sv_catpvf_mg>.
8250
8251=cut
8252*/
645c22ef 8253
c5be433b
GS
8254void
8255Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8256{
4608196e 8257 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8258 SvSETMAGIC(sv);
8259}
8260
954c1994
GS
8261/*
8262=for apidoc sv_vsetpvfn
8263
bffc3d17 8264Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8265appending it.
8266
bffc3d17 8267Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8268
954c1994
GS
8269=cut
8270*/
8271
46fc3d4c 8272void
7d5ea4e7 8273Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8274{
8275 sv_setpvn(sv, "", 0);
7d5ea4e7 8276 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8277}
8278
2d00ba3b 8279STATIC I32
9dd79c3f 8280S_expect_number(pTHX_ char** pattern)
211dfcf1 8281{
97aff369 8282 dVAR;
211dfcf1
HS
8283 I32 var = 0;
8284 switch (**pattern) {
8285 case '1': case '2': case '3':
8286 case '4': case '5': case '6':
8287 case '7': case '8': case '9':
2fba7546
GA
8288 var = *(*pattern)++ - '0';
8289 while (isDIGIT(**pattern)) {
5f66b61c 8290 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8291 if (tmp < var)
8292 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8293 var = tmp;
8294 }
211dfcf1
HS
8295 }
8296 return var;
8297}
211dfcf1 8298
c445ea15
AL
8299STATIC char *
8300S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8301{
a3b680e6 8302 const int neg = nv < 0;
4151a5fe 8303 UV uv;
4151a5fe
IZ
8304
8305 if (neg)
8306 nv = -nv;
8307 if (nv < UV_MAX) {
b464bac0 8308 char *p = endbuf;
4151a5fe 8309 nv += 0.5;
028f8eaa 8310 uv = (UV)nv;
4151a5fe
IZ
8311 if (uv & 1 && uv == nv)
8312 uv--; /* Round to even */
8313 do {
a3b680e6 8314 const unsigned dig = uv % 10;
4151a5fe
IZ
8315 *--p = '0' + dig;
8316 } while (uv /= 10);
8317 if (neg)
8318 *--p = '-';
8319 *len = endbuf - p;
8320 return p;
8321 }
bd61b366 8322 return NULL;
4151a5fe
IZ
8323}
8324
8325
954c1994
GS
8326/*
8327=for apidoc sv_vcatpvfn
8328
8329Processes its arguments like C<vsprintf> and appends the formatted output
8330to an SV. Uses an array of SVs if the C style variable argument list is
8331missing (NULL). When running with taint checks enabled, indicates via
8332C<maybe_tainted> if results are untrustworthy (often due to the use of
8333locales).
8334
bffc3d17 8335Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8336
954c1994
GS
8337=cut
8338*/
8339
8896765a
RB
8340
8341#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8342 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8343 vec_utf8 = DO_UTF8(vecsv);
8344
1ef29b0e
RGS
8345/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8346
46fc3d4c 8347void
7d5ea4e7 8348Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8349{
97aff369 8350 dVAR;
46fc3d4c 8351 char *p;
8352 char *q;
a3b680e6 8353 const char *patend;
fc36a67e 8354 STRLEN origlen;
46fc3d4c 8355 I32 svix = 0;
27da23d5 8356 static const char nullstr[] = "(null)";
a0714e2c 8357 SV *argsv = NULL;
b464bac0
AL
8358 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8359 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8360 SV *nsv = NULL;
4151a5fe
IZ
8361 /* Times 4: a decimal digit takes more than 3 binary digits.
8362 * NV_DIG: mantissa takes than many decimal digits.
8363 * Plus 32: Playing safe. */
8364 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8365 /* large enough for "%#.#f" --chip */
8366 /* what about long double NVs? --jhi */
db79b45b 8367
53c1dcc0
AL
8368 PERL_UNUSED_ARG(maybe_tainted);
8369
46fc3d4c 8370 /* no matter what, this is a string now */
fc36a67e 8371 (void)SvPV_force(sv, origlen);
46fc3d4c 8372
8896765a 8373 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8374 if (patlen == 0)
8375 return;
0dbb1585 8376 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8377 if (args) {
8378 const char * const s = va_arg(*args, char*);
8379 sv_catpv(sv, s ? s : nullstr);
8380 }
8381 else if (svix < svmax) {
8382 sv_catsv(sv, *svargs);
2d03de9c
AL
8383 }
8384 return;
0dbb1585 8385 }
8896765a
RB
8386 if (args && patlen == 3 && pat[0] == '%' &&
8387 pat[1] == '-' && pat[2] == 'p') {
8388 argsv = va_arg(*args, SV*);
8389 sv_catsv(sv, argsv);
8896765a 8390 return;
46fc3d4c 8391 }
8392
1d917b39 8393#ifndef USE_LONG_DOUBLE
4151a5fe 8394 /* special-case "%.<number>[gf]" */
7af36d83 8395 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8396 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8397 unsigned digits = 0;
8398 const char *pp;
8399
8400 pp = pat + 2;
8401 while (*pp >= '0' && *pp <= '9')
8402 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8403 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8404 NV nv;
8405
7af36d83 8406 if (svix < svmax)
4151a5fe
IZ
8407 nv = SvNV(*svargs);
8408 else
8409 return;
8410 if (*pp == 'g') {
2873255c
NC
8411 /* Add check for digits != 0 because it seems that some
8412 gconverts are buggy in this case, and we don't yet have
8413 a Configure test for this. */
8414 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8415 /* 0, point, slack */
2e59c212 8416 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8417 sv_catpv(sv, ebuf);
8418 if (*ebuf) /* May return an empty string for digits==0 */
8419 return;
8420 }
8421 } else if (!digits) {
8422 STRLEN l;
8423
8424 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8425 sv_catpvn(sv, p, l);
8426 return;
8427 }
8428 }
8429 }
8430 }
1d917b39 8431#endif /* !USE_LONG_DOUBLE */
4151a5fe 8432
2cf2cfc6 8433 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8434 has_utf8 = TRUE;
2cf2cfc6 8435
46fc3d4c 8436 patend = (char*)pat + patlen;
8437 for (p = (char*)pat; p < patend; p = q) {
8438 bool alt = FALSE;
8439 bool left = FALSE;
b22c7a20 8440 bool vectorize = FALSE;
211dfcf1 8441 bool vectorarg = FALSE;
2cf2cfc6 8442 bool vec_utf8 = FALSE;
46fc3d4c 8443 char fill = ' ';
8444 char plus = 0;
8445 char intsize = 0;
8446 STRLEN width = 0;
fc36a67e 8447 STRLEN zeros = 0;
46fc3d4c 8448 bool has_precis = FALSE;
8449 STRLEN precis = 0;
c445ea15 8450 const I32 osvix = svix;
2cf2cfc6 8451 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8452#ifdef HAS_LDBL_SPRINTF_BUG
8453 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8454 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8455 bool fix_ldbl_sprintf_bug = FALSE;
8456#endif
205f51d8 8457
46fc3d4c 8458 char esignbuf[4];
89ebb4a3 8459 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8460 STRLEN esignlen = 0;
8461
bd61b366 8462 const char *eptr = NULL;
fc36a67e 8463 STRLEN elen = 0;
a0714e2c 8464 SV *vecsv = NULL;
4608196e 8465 const U8 *vecstr = NULL;
b22c7a20 8466 STRLEN veclen = 0;
934abaf1 8467 char c = 0;
46fc3d4c 8468 int i;
9c5ffd7c 8469 unsigned base = 0;
8c8eb53c
RB
8470 IV iv = 0;
8471 UV uv = 0;
9e5b023a
JH
8472 /* we need a long double target in case HAS_LONG_DOUBLE but
8473 not USE_LONG_DOUBLE
8474 */
35fff930 8475#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8476 long double nv;
8477#else
65202027 8478 NV nv;
9e5b023a 8479#endif
46fc3d4c 8480 STRLEN have;
8481 STRLEN need;
8482 STRLEN gap;
7af36d83 8483 const char *dotstr = ".";
b22c7a20 8484 STRLEN dotstrlen = 1;
211dfcf1 8485 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8486 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8487 I32 epix = 0; /* explicit precision index */
8488 I32 evix = 0; /* explicit vector index */
eb3fce90 8489 bool asterisk = FALSE;
46fc3d4c 8490
211dfcf1 8491 /* echo everything up to the next format specification */
46fc3d4c 8492 for (q = p; q < patend && *q != '%'; ++q) ;
8493 if (q > p) {
db79b45b
JH
8494 if (has_utf8 && !pat_utf8)
8495 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8496 else
8497 sv_catpvn(sv, p, q - p);
46fc3d4c 8498 p = q;
8499 }
8500 if (q++ >= patend)
8501 break;
8502
211dfcf1
HS
8503/*
8504 We allow format specification elements in this order:
8505 \d+\$ explicit format parameter index
8506 [-+ 0#]+ flags
a472f209 8507 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8508 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8509 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8510 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8511 [hlqLV] size
8896765a
RB
8512 [%bcdefginopsuxDFOUX] format (mandatory)
8513*/
8514
8515 if (args) {
8516/*
8517 As of perl5.9.3, printf format checking is on by default.
8518 Internally, perl uses %p formats to provide an escape to
8519 some extended formatting. This block deals with those
8520 extensions: if it does not match, (char*)q is reset and
8521 the normal format processing code is used.
8522
8523 Currently defined extensions are:
8524 %p include pointer address (standard)
8525 %-p (SVf) include an SV (previously %_)
8526 %-<num>p include an SV with precision <num>
8527 %1p (VDf) include a v-string (as %vd)
8528 %<num>p reserved for future extensions
8529
8530 Robin Barker 2005-07-14
211dfcf1 8531*/
8896765a
RB
8532 char* r = q;
8533 bool sv = FALSE;
8534 STRLEN n = 0;
8535 if (*q == '-')
8536 sv = *q++;
c445ea15 8537 n = expect_number(&q);
8896765a
RB
8538 if (*q++ == 'p') {
8539 if (sv) { /* SVf */
8540 if (n) {
8541 precis = n;
8542 has_precis = TRUE;
8543 }
8544 argsv = va_arg(*args, SV*);
8545 eptr = SvPVx_const(argsv, elen);
8546 if (DO_UTF8(argsv))
8547 is_utf8 = TRUE;
8548 goto string;
8549 }
8550#if vdNUMBER
8551 else if (n == vdNUMBER) { /* VDf */
8552 vectorize = TRUE;
8553 VECTORIZE_ARGS
8554 goto format_vd;
8555 }
8556#endif
8557 else if (n) {
8558 if (ckWARN_d(WARN_INTERNAL))
8559 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8560 "internal %%<num>p might conflict with future printf extensions");
8561 }
8562 }
8563 q = r;
8564 }
8565
c445ea15 8566 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8567 if (*q == '$') {
8568 ++q;
8569 efix = width;
8570 } else {
8571 goto gotwidth;
8572 }
8573 }
8574
fc36a67e 8575 /* FLAGS */
8576
46fc3d4c 8577 while (*q) {
8578 switch (*q) {
8579 case ' ':
8580 case '+':
9911cee9
TS
8581 if (plus == '+' && *q == ' ') /* '+' over ' ' */
8582 q++;
8583 else
8584 plus = *q++;
46fc3d4c 8585 continue;
8586
8587 case '-':
8588 left = TRUE;
8589 q++;
8590 continue;
8591
8592 case '0':
8593 fill = *q++;
8594 continue;
8595
8596 case '#':
8597 alt = TRUE;
8598 q++;
8599 continue;
8600
fc36a67e 8601 default:
8602 break;
8603 }
8604 break;
8605 }
46fc3d4c 8606
211dfcf1 8607 tryasterisk:
eb3fce90 8608 if (*q == '*') {
211dfcf1 8609 q++;
c445ea15 8610 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8611 if (*q++ != '$')
8612 goto unknown;
eb3fce90 8613 asterisk = TRUE;
211dfcf1
HS
8614 }
8615 if (*q == 'v') {
eb3fce90 8616 q++;
211dfcf1
HS
8617 if (vectorize)
8618 goto unknown;
9cbac4c7 8619 if ((vectorarg = asterisk)) {
211dfcf1
HS
8620 evix = ewix;
8621 ewix = 0;
8622 asterisk = FALSE;
8623 }
8624 vectorize = TRUE;
8625 goto tryasterisk;
eb3fce90
JH
8626 }
8627
211dfcf1 8628 if (!asterisk)
858a90f9 8629 {
7a5fa8a2 8630 if( *q == '0' )
f3583277 8631 fill = *q++;
c445ea15 8632 width = expect_number(&q);
858a90f9 8633 }
211dfcf1
HS
8634
8635 if (vectorize) {
8636 if (vectorarg) {
8637 if (args)
8638 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8639 else if (evix) {
8640 vecsv = (evix > 0 && evix <= svmax)
8641 ? svargs[evix-1] : &PL_sv_undef;
8642 } else {
8643 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8644 }
245d4a47 8645 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8646 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8647 bad with tied or overloaded values that return UTF8. */
211dfcf1 8648 if (DO_UTF8(vecsv))
2cf2cfc6 8649 is_utf8 = TRUE;
640283f5
NC
8650 else if (has_utf8) {
8651 vecsv = sv_mortalcopy(vecsv);
8652 sv_utf8_upgrade(vecsv);
8653 dotstr = SvPV_const(vecsv, dotstrlen);
8654 is_utf8 = TRUE;
8655 }
211dfcf1
HS
8656 }
8657 if (args) {
8896765a 8658 VECTORIZE_ARGS
eb3fce90 8659 }
7ad96abb 8660 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8661 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8662 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8663 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8664
8665 /* if this is a version object, we need to convert
8666 * back into v-string notation and then let the
8667 * vectorize happen normally
d7aa5382 8668 */
96b8f7ce
JP
8669 if (sv_derived_from(vecsv, "version")) {
8670 char *version = savesvpv(vecsv);
34ba6322
SP
8671 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8672 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8673 "vector argument not supported with alpha versions");
8674 goto unknown;
8675 }
96b8f7ce
JP
8676 vecsv = sv_newmortal();
8677 /* scan_vstring is expected to be called during
8678 * tokenization, so we need to fake up the end
8679 * of the buffer for it
8680 */
8681 PL_bufend = version + veclen;
8682 scan_vstring(version, vecsv);
8683 vecstr = (U8*)SvPV_const(vecsv, veclen);
8684 vec_utf8 = DO_UTF8(vecsv);
8685 Safefree(version);
d7aa5382 8686 }
211dfcf1
HS
8687 }
8688 else {
8689 vecstr = (U8*)"";
8690 veclen = 0;
8691 }
eb3fce90 8692 }
fc36a67e 8693
eb3fce90 8694 if (asterisk) {
fc36a67e 8695 if (args)
8696 i = va_arg(*args, int);
8697 else
eb3fce90
JH
8698 i = (ewix ? ewix <= svmax : svix < svmax) ?
8699 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8700 left |= (i < 0);
8701 width = (i < 0) ? -i : i;
fc36a67e 8702 }
211dfcf1 8703 gotwidth:
fc36a67e 8704
8705 /* PRECISION */
46fc3d4c 8706
fc36a67e 8707 if (*q == '.') {
8708 q++;
8709 if (*q == '*') {
211dfcf1 8710 q++;
c445ea15 8711 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8712 goto unknown;
8713 /* XXX: todo, support specified precision parameter */
8714 if (epix)
211dfcf1 8715 goto unknown;
46fc3d4c 8716 if (args)
8717 i = va_arg(*args, int);
8718 else
eb3fce90
JH
8719 i = (ewix ? ewix <= svmax : svix < svmax)
8720 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9911cee9
TS
8721 precis = i;
8722 has_precis = !(i < 0);
fc36a67e 8723 }
8724 else {
8725 precis = 0;
8726 while (isDIGIT(*q))
8727 precis = precis * 10 + (*q++ - '0');
9911cee9 8728 has_precis = TRUE;
fc36a67e 8729 }
fc36a67e 8730 }
46fc3d4c 8731
fc36a67e 8732 /* SIZE */
46fc3d4c 8733
fc36a67e 8734 switch (*q) {
c623ac67
GS
8735#ifdef WIN32
8736 case 'I': /* Ix, I32x, and I64x */
8737# ifdef WIN64
8738 if (q[1] == '6' && q[2] == '4') {
8739 q += 3;
8740 intsize = 'q';
8741 break;
8742 }
8743# endif
8744 if (q[1] == '3' && q[2] == '2') {
8745 q += 3;
8746 break;
8747 }
8748# ifdef WIN64
8749 intsize = 'q';
8750# endif
8751 q++;
8752 break;
8753#endif
9e5b023a 8754#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8755 case 'L': /* Ld */
5f66b61c 8756 /*FALLTHROUGH*/
e5c81feb 8757#ifdef HAS_QUAD
6f9bb7fd 8758 case 'q': /* qd */
9e5b023a 8759#endif
6f9bb7fd
GS
8760 intsize = 'q';
8761 q++;
8762 break;
8763#endif
fc36a67e 8764 case 'l':
9e5b023a 8765#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8766 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8767 intsize = 'q';
8768 q += 2;
46fc3d4c 8769 break;
cf2093f6 8770 }
fc36a67e 8771#endif
5f66b61c 8772 /*FALLTHROUGH*/
fc36a67e 8773 case 'h':
5f66b61c 8774 /*FALLTHROUGH*/
fc36a67e 8775 case 'V':
8776 intsize = *q++;
46fc3d4c 8777 break;
8778 }
8779
fc36a67e 8780 /* CONVERSION */
8781
211dfcf1
HS
8782 if (*q == '%') {
8783 eptr = q++;
8784 elen = 1;
26372e71
GA
8785 if (vectorize) {
8786 c = '%';
8787 goto unknown;
8788 }
211dfcf1
HS
8789 goto string;
8790 }
8791
26372e71 8792 if (!vectorize && !args) {
86c51f8b
NC
8793 if (efix) {
8794 const I32 i = efix-1;
8795 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8796 } else {
8797 argsv = (svix >= 0 && svix < svmax)
8798 ? svargs[svix++] : &PL_sv_undef;
8799 }
863811b2 8800 }
211dfcf1 8801
46fc3d4c 8802 switch (c = *q++) {
8803
8804 /* STRINGS */
8805
46fc3d4c 8806 case 'c':
26372e71
GA
8807 if (vectorize)
8808 goto unknown;
8809 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8810 if ((uv > 255 ||
8811 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8812 && !IN_BYTES) {
dfe13c55 8813 eptr = (char*)utf8buf;
9041c2e3 8814 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8815 is_utf8 = TRUE;
7e2040f0
GS
8816 }
8817 else {
8818 c = (char)uv;
8819 eptr = &c;
8820 elen = 1;
a0ed51b3 8821 }
46fc3d4c 8822 goto string;
8823
46fc3d4c 8824 case 's':
26372e71
GA
8825 if (vectorize)
8826 goto unknown;
8827 if (args) {
fc36a67e 8828 eptr = va_arg(*args, char*);
c635e13b 8829 if (eptr)
1d7c1841
GS
8830#ifdef MACOS_TRADITIONAL
8831 /* On MacOS, %#s format is used for Pascal strings */
8832 if (alt)
8833 elen = *eptr++;
8834 else
8835#endif
c635e13b 8836 elen = strlen(eptr);
8837 else {
27da23d5 8838 eptr = (char *)nullstr;
c635e13b 8839 elen = sizeof nullstr - 1;
8840 }
46fc3d4c 8841 }
211dfcf1 8842 else {
4d84ee25 8843 eptr = SvPVx_const(argsv, elen);
7e2040f0 8844 if (DO_UTF8(argsv)) {
59b61096 8845 I32 old_precis = precis;
a0ed51b3
LW
8846 if (has_precis && precis < elen) {
8847 I32 p = precis;
7e2040f0 8848 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8849 precis = p;
8850 }
8851 if (width) { /* fudge width (can't fudge elen) */
59b61096
AV
8852 if (has_precis && precis < elen)
8853 width += precis - old_precis;
8854 else
8855 width += elen - sv_len_utf8(argsv);
a0ed51b3 8856 }
2cf2cfc6 8857 is_utf8 = TRUE;
a0ed51b3
LW
8858 }
8859 }
fc36a67e 8860
46fc3d4c 8861 string:
8862 if (has_precis && elen > precis)
8863 elen = precis;
8864 break;
8865
8866 /* INTEGERS */
8867
fc36a67e 8868 case 'p':
be75b157 8869 if (alt || vectorize)
c2e66d9e 8870 goto unknown;
211dfcf1 8871 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8872 base = 16;
8873 goto integer;
8874
46fc3d4c 8875 case 'D':
29fe7a80 8876#ifdef IV_IS_QUAD
22f3ae8c 8877 intsize = 'q';
29fe7a80 8878#else
46fc3d4c 8879 intsize = 'l';
29fe7a80 8880#endif
5f66b61c 8881 /*FALLTHROUGH*/
46fc3d4c 8882 case 'd':
8883 case 'i':
8896765a
RB
8884#if vdNUMBER
8885 format_vd:
8886#endif
b22c7a20 8887 if (vectorize) {
ba210ebe 8888 STRLEN ulen;
211dfcf1
HS
8889 if (!veclen)
8890 continue;
2cf2cfc6
A
8891 if (vec_utf8)
8892 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8893 UTF8_ALLOW_ANYUV);
b22c7a20 8894 else {
e83d50c9 8895 uv = *vecstr;
b22c7a20
GS
8896 ulen = 1;
8897 }
8898 vecstr += ulen;
8899 veclen -= ulen;
e83d50c9
JP
8900 if (plus)
8901 esignbuf[esignlen++] = plus;
b22c7a20
GS
8902 }
8903 else if (args) {
46fc3d4c 8904 switch (intsize) {
8905 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8906 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8907 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8908 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8909#ifdef HAS_QUAD
8910 case 'q': iv = va_arg(*args, Quad_t); break;
8911#endif
46fc3d4c 8912 }
8913 }
8914 else {
b10c0dba 8915 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8916 switch (intsize) {
b10c0dba
MHM
8917 case 'h': iv = (short)tiv; break;
8918 case 'l': iv = (long)tiv; break;
8919 case 'V':
8920 default: iv = tiv; break;
cf2093f6 8921#ifdef HAS_QUAD
b10c0dba 8922 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8923#endif
46fc3d4c 8924 }
8925 }
e83d50c9
JP
8926 if ( !vectorize ) /* we already set uv above */
8927 {
8928 if (iv >= 0) {
8929 uv = iv;
8930 if (plus)
8931 esignbuf[esignlen++] = plus;
8932 }
8933 else {
8934 uv = -iv;
8935 esignbuf[esignlen++] = '-';
8936 }
46fc3d4c 8937 }
8938 base = 10;
8939 goto integer;
8940
fc36a67e 8941 case 'U':
29fe7a80 8942#ifdef IV_IS_QUAD
22f3ae8c 8943 intsize = 'q';
29fe7a80 8944#else
fc36a67e 8945 intsize = 'l';
29fe7a80 8946#endif
5f66b61c 8947 /*FALLTHROUGH*/
fc36a67e 8948 case 'u':
8949 base = 10;
8950 goto uns_integer;
8951
7ff06cc7 8952 case 'B':
4f19785b
WSI
8953 case 'b':
8954 base = 2;
8955 goto uns_integer;
8956
46fc3d4c 8957 case 'O':
29fe7a80 8958#ifdef IV_IS_QUAD
22f3ae8c 8959 intsize = 'q';
29fe7a80 8960#else
46fc3d4c 8961 intsize = 'l';
29fe7a80 8962#endif
5f66b61c 8963 /*FALLTHROUGH*/
46fc3d4c 8964 case 'o':
8965 base = 8;
8966 goto uns_integer;
8967
8968 case 'X':
46fc3d4c 8969 case 'x':
8970 base = 16;
46fc3d4c 8971
8972 uns_integer:
b22c7a20 8973 if (vectorize) {
ba210ebe 8974 STRLEN ulen;
b22c7a20 8975 vector:
211dfcf1
HS
8976 if (!veclen)
8977 continue;
2cf2cfc6
A
8978 if (vec_utf8)
8979 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8980 UTF8_ALLOW_ANYUV);
b22c7a20 8981 else {
a05b299f 8982 uv = *vecstr;
b22c7a20
GS
8983 ulen = 1;
8984 }
8985 vecstr += ulen;
8986 veclen -= ulen;
8987 }
8988 else if (args) {
46fc3d4c 8989 switch (intsize) {
8990 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8991 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8992 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8993 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8994#ifdef HAS_QUAD
9e3321a5 8995 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8996#endif
46fc3d4c 8997 }
8998 }
8999 else {
b10c0dba 9000 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 9001 switch (intsize) {
b10c0dba
MHM
9002 case 'h': uv = (unsigned short)tuv; break;
9003 case 'l': uv = (unsigned long)tuv; break;
9004 case 'V':
9005 default: uv = tuv; break;
cf2093f6 9006#ifdef HAS_QUAD
b10c0dba 9007 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 9008#endif
46fc3d4c 9009 }
9010 }
9011
9012 integer:
4d84ee25
NC
9013 {
9014 char *ptr = ebuf + sizeof ebuf;
1387f30c
DD
9015 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9016 zeros = 0;
9017
4d84ee25
NC
9018 switch (base) {
9019 unsigned dig;
9020 case 16:
14eb61ab 9021 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
4d84ee25
NC
9022 do {
9023 dig = uv & 15;
9024 *--ptr = p[dig];
9025 } while (uv >>= 4);
1387f30c 9026 if (tempalt) {
4d84ee25
NC
9027 esignbuf[esignlen++] = '0';
9028 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9029 }
9030 break;
9031 case 8:
9032 do {
9033 dig = uv & 7;
9034 *--ptr = '0' + dig;
9035 } while (uv >>= 3);
9036 if (alt && *ptr != '0')
9037 *--ptr = '0';
9038 break;
9039 case 2:
9040 do {
9041 dig = uv & 1;
9042 *--ptr = '0' + dig;
9043 } while (uv >>= 1);
1387f30c 9044 if (tempalt) {
4d84ee25 9045 esignbuf[esignlen++] = '0';
7ff06cc7 9046 esignbuf[esignlen++] = c;
4d84ee25
NC
9047 }
9048 break;
9049 default: /* it had better be ten or less */
9050 do {
9051 dig = uv % base;
9052 *--ptr = '0' + dig;
9053 } while (uv /= base);
9054 break;
46fc3d4c 9055 }
4d84ee25
NC
9056 elen = (ebuf + sizeof ebuf) - ptr;
9057 eptr = ptr;
9058 if (has_precis) {
9059 if (precis > elen)
9060 zeros = precis - elen;
e6bb52fd
TS
9061 else if (precis == 0 && elen == 1 && *eptr == '0'
9062 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
4d84ee25 9063 elen = 0;
9911cee9
TS
9064
9065 /* a precision nullifies the 0 flag. */
9066 if (fill == '0')
9067 fill = ' ';
eda88b6d 9068 }
c10ed8b9 9069 }
46fc3d4c 9070 break;
9071
9072 /* FLOATING POINT */
9073
fc36a67e 9074 case 'F':
9075 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9076 /*FALLTHROUGH*/
46fc3d4c 9077 case 'e': case 'E':
fc36a67e 9078 case 'f':
46fc3d4c 9079 case 'g': case 'G':
26372e71
GA
9080 if (vectorize)
9081 goto unknown;
46fc3d4c 9082
9083 /* This is evil, but floating point is even more evil */
9084
9e5b023a
JH
9085 /* for SV-style calling, we can only get NV
9086 for C-style calling, we assume %f is double;
9087 for simplicity we allow any of %Lf, %llf, %qf for long double
9088 */
9089 switch (intsize) {
9090 case 'V':
9091#if defined(USE_LONG_DOUBLE)
9092 intsize = 'q';
9093#endif
9094 break;
8a2e3f14 9095/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9096 case 'l':
5f66b61c 9097 /*FALLTHROUGH*/
9e5b023a
JH
9098 default:
9099#if defined(USE_LONG_DOUBLE)
9100 intsize = args ? 0 : 'q';
9101#endif
9102 break;
9103 case 'q':
9104#if defined(HAS_LONG_DOUBLE)
9105 break;
9106#else
5f66b61c 9107 /*FALLTHROUGH*/
9e5b023a
JH
9108#endif
9109 case 'h':
9e5b023a
JH
9110 goto unknown;
9111 }
9112
9113 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9114 nv = (args) ?
35fff930
JH
9115#if LONG_DOUBLESIZE > DOUBLESIZE
9116 intsize == 'q' ?
205f51d8
AS
9117 va_arg(*args, long double) :
9118 va_arg(*args, double)
35fff930 9119#else
205f51d8 9120 va_arg(*args, double)
35fff930 9121#endif
9e5b023a 9122 : SvNVx(argsv);
fc36a67e 9123
9124 need = 0;
9125 if (c != 'e' && c != 'E') {
9126 i = PERL_INT_MIN;
9e5b023a
JH
9127 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9128 will cast our (long double) to (double) */
73b309ea 9129 (void)Perl_frexp(nv, &i);
fc36a67e 9130 if (i == PERL_INT_MIN)
cea2e8a9 9131 Perl_die(aTHX_ "panic: frexp");
c635e13b 9132 if (i > 0)
fc36a67e 9133 need = BIT_DIGITS(i);
9134 }
9135 need += has_precis ? precis : 6; /* known default */
20f6aaab 9136
fc36a67e 9137 if (need < width)
9138 need = width;
9139
20f6aaab
AS
9140#ifdef HAS_LDBL_SPRINTF_BUG
9141 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9142 with sfio - Allen <allens@cpan.org> */
9143
9144# ifdef DBL_MAX
9145# define MY_DBL_MAX DBL_MAX
9146# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9147# if DOUBLESIZE >= 8
9148# define MY_DBL_MAX 1.7976931348623157E+308L
9149# else
9150# define MY_DBL_MAX 3.40282347E+38L
9151# endif
9152# endif
9153
9154# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9155# define MY_DBL_MAX_BUG 1L
20f6aaab 9156# else
205f51d8 9157# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9158# endif
20f6aaab 9159
205f51d8
AS
9160# ifdef DBL_MIN
9161# define MY_DBL_MIN DBL_MIN
9162# else /* XXX guessing! -Allen */
9163# if DOUBLESIZE >= 8
9164# define MY_DBL_MIN 2.2250738585072014E-308L
9165# else
9166# define MY_DBL_MIN 1.17549435E-38L
9167# endif
9168# endif
20f6aaab 9169
205f51d8
AS
9170 if ((intsize == 'q') && (c == 'f') &&
9171 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9172 (need < DBL_DIG)) {
9173 /* it's going to be short enough that
9174 * long double precision is not needed */
9175
9176 if ((nv <= 0L) && (nv >= -0L))
9177 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9178 else {
9179 /* would use Perl_fp_class as a double-check but not
9180 * functional on IRIX - see perl.h comments */
9181
9182 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9183 /* It's within the range that a double can represent */
9184#if defined(DBL_MAX) && !defined(DBL_MIN)
9185 if ((nv >= ((long double)1/DBL_MAX)) ||
9186 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9187#endif
205f51d8 9188 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9189 }
205f51d8
AS
9190 }
9191 if (fix_ldbl_sprintf_bug == TRUE) {
9192 double temp;
9193
9194 intsize = 0;
9195 temp = (double)nv;
9196 nv = (NV)temp;
9197 }
20f6aaab 9198 }
205f51d8
AS
9199
9200# undef MY_DBL_MAX
9201# undef MY_DBL_MAX_BUG
9202# undef MY_DBL_MIN
9203
20f6aaab
AS
9204#endif /* HAS_LDBL_SPRINTF_BUG */
9205
46fc3d4c 9206 need += 20; /* fudge factor */
80252599
GS
9207 if (PL_efloatsize < need) {
9208 Safefree(PL_efloatbuf);
9209 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9210 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9211 PL_efloatbuf[0] = '\0';
46fc3d4c 9212 }
9213
4151a5fe
IZ
9214 if ( !(width || left || plus || alt) && fill != '0'
9215 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9216 /* See earlier comment about buggy Gconvert when digits,
9217 aka precis is 0 */
9218 if ( c == 'g' && precis) {
2e59c212 9219 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9220 /* May return an empty string for digits==0 */
9221 if (*PL_efloatbuf) {
9222 elen = strlen(PL_efloatbuf);
4151a5fe 9223 goto float_converted;
4150c189 9224 }
4151a5fe
IZ
9225 } else if ( c == 'f' && !precis) {
9226 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9227 break;
9228 }
9229 }
4d84ee25
NC
9230 {
9231 char *ptr = ebuf + sizeof ebuf;
9232 *--ptr = '\0';
9233 *--ptr = c;
9234 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9235#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9236 if (intsize == 'q') {
9237 /* Copy the one or more characters in a long double
9238 * format before the 'base' ([efgEFG]) character to
9239 * the format string. */
9240 static char const prifldbl[] = PERL_PRIfldbl;
9241 char const *p = prifldbl + sizeof(prifldbl) - 3;
9242 while (p >= prifldbl) { *--ptr = *p--; }
9243 }
65202027 9244#endif
4d84ee25
NC
9245 if (has_precis) {
9246 base = precis;
9247 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9248 *--ptr = '.';
9249 }
9250 if (width) {
9251 base = width;
9252 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9253 }
9254 if (fill == '0')
9255 *--ptr = fill;
9256 if (left)
9257 *--ptr = '-';
9258 if (plus)
9259 *--ptr = plus;
9260 if (alt)
9261 *--ptr = '#';
9262 *--ptr = '%';
9263
9264 /* No taint. Otherwise we are in the strange situation
9265 * where printf() taints but print($float) doesn't.
9266 * --jhi */
9e5b023a 9267#if defined(HAS_LONG_DOUBLE)
4150c189 9268 elen = ((intsize == 'q')
d9fad198
JH
9269 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
9270 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
9e5b023a 9271#else
4150c189 9272 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9273#endif
4d84ee25 9274 }
4151a5fe 9275 float_converted:
80252599 9276 eptr = PL_efloatbuf;
46fc3d4c 9277 break;
9278
fc36a67e 9279 /* SPECIAL */
9280
9281 case 'n':
26372e71
GA
9282 if (vectorize)
9283 goto unknown;
fc36a67e 9284 i = SvCUR(sv) - origlen;
26372e71 9285 if (args) {
c635e13b 9286 switch (intsize) {
9287 case 'h': *(va_arg(*args, short*)) = i; break;
9288 default: *(va_arg(*args, int*)) = i; break;
9289 case 'l': *(va_arg(*args, long*)) = i; break;
9290 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9291#ifdef HAS_QUAD
9292 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9293#endif
c635e13b 9294 }
fc36a67e 9295 }
9dd79c3f 9296 else
211dfcf1 9297 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9298 continue; /* not "break" */
9299
9300 /* UNKNOWN */
9301
46fc3d4c 9302 default:
fc36a67e 9303 unknown:
041457d9
DM
9304 if (!args
9305 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9306 && ckWARN(WARN_PRINTF))
9307 {
c4420975 9308 SV * const msg = sv_newmortal();
35c1215d
NC
9309 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9310 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9311 if (c) {
0f4b6630 9312 if (isPRINT(c))
1c846c1f 9313 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9314 "\"%%%c\"", c & 0xFF);
9315 else
9316 Perl_sv_catpvf(aTHX_ msg,
57def98f 9317 "\"%%\\%03"UVof"\"",
0f4b6630 9318 (UV)c & 0xFF);
0f4b6630 9319 } else
396482e1 9320 sv_catpvs(msg, "end of string");
95b63a38 9321 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, (void*)msg); /* yes, this is reentrant */
c635e13b 9322 }
fb73857a 9323
9324 /* output mangled stuff ... */
9325 if (c == '\0')
9326 --q;
46fc3d4c 9327 eptr = p;
9328 elen = q - p;
fb73857a 9329
9330 /* ... right here, because formatting flags should not apply */
9331 SvGROW(sv, SvCUR(sv) + elen + 1);
9332 p = SvEND(sv);
4459522c 9333 Copy(eptr, p, elen, char);
fb73857a 9334 p += elen;
9335 *p = '\0';
3f7c398e 9336 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9337 svix = osvix;
fb73857a 9338 continue; /* not "break" */
46fc3d4c 9339 }
9340
cc61b222
TS
9341 if (is_utf8 != has_utf8) {
9342 if (is_utf8) {
9343 if (SvCUR(sv))
9344 sv_utf8_upgrade(sv);
9345 }
9346 else {
9347 const STRLEN old_elen = elen;
9348 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
9349 sv_utf8_upgrade(nsv);
9350 eptr = SvPVX_const(nsv);
9351 elen = SvCUR(nsv);
9352
9353 if (width) { /* fudge width (can't fudge elen) */
9354 width += elen - old_elen;
9355 }
9356 is_utf8 = TRUE;
9357 }
9358 }
9359
6c94ec8b 9360 have = esignlen + zeros + elen;
ed2b91d2
GA
9361 if (have < zeros)
9362 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9363
46fc3d4c 9364 need = (have > width ? have : width);
9365 gap = need - have;
9366
d2641cbd
PC
9367 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9368 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9369 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9370 p = SvEND(sv);
9371 if (esignlen && fill == '0') {
53c1dcc0 9372 int i;
eb160463 9373 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9374 *p++ = esignbuf[i];
9375 }
9376 if (gap && !left) {
9377 memset(p, fill, gap);
9378 p += gap;
9379 }
9380 if (esignlen && fill != '0') {
53c1dcc0 9381 int i;
eb160463 9382 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9383 *p++ = esignbuf[i];
9384 }
fc36a67e 9385 if (zeros) {
53c1dcc0 9386 int i;
fc36a67e 9387 for (i = zeros; i; i--)
9388 *p++ = '0';
9389 }
46fc3d4c 9390 if (elen) {
4459522c 9391 Copy(eptr, p, elen, char);
46fc3d4c 9392 p += elen;
9393 }
9394 if (gap && left) {
9395 memset(p, ' ', gap);
9396 p += gap;
9397 }
b22c7a20
GS
9398 if (vectorize) {
9399 if (veclen) {
4459522c 9400 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9401 p += dotstrlen;
9402 }
9403 else
9404 vectorize = FALSE; /* done iterating over vecstr */
9405 }
2cf2cfc6
A
9406 if (is_utf8)
9407 has_utf8 = TRUE;
9408 if (has_utf8)
7e2040f0 9409 SvUTF8_on(sv);
46fc3d4c 9410 *p = '\0';
3f7c398e 9411 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9412 if (vectorize) {
9413 esignlen = 0;
9414 goto vector;
9415 }
46fc3d4c 9416 }
9417}
51371543 9418
645c22ef
DM
9419/* =========================================================================
9420
9421=head1 Cloning an interpreter
9422
9423All the macros and functions in this section are for the private use of
9424the main function, perl_clone().
9425
9426The foo_dup() functions make an exact copy of an existing foo thinngy.
9427During the course of a cloning, a hash table is used to map old addresses
9428to new addresses. The table is created and manipulated with the
9429ptr_table_* functions.
9430
9431=cut
9432
9433============================================================================*/
9434
9435
1d7c1841
GS
9436#if defined(USE_ITHREADS)
9437
d4c19fe8 9438/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9439#ifndef GpREFCNT_inc
9440# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9441#endif
9442
9443
a41cc44e 9444/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
3e07292d
NC
9445 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
9446 If this changes, please unmerge ss_dup. */
d2d73c3e 9447#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9448#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9449#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9450#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9451#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9452#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9453#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9454#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9455#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9456#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9457#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9458#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9459#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9460#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9461
d2d73c3e 9462
d2f185dc
AMS
9463/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9464 regcomp.c. AMS 20010712 */
645c22ef 9465
1d7c1841 9466REGEXP *
53c1dcc0 9467Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9468{
f9f4320a 9469 return CALLREGDUPE(r,param);
1d7c1841
GS
9470}
9471
d2d73c3e 9472/* duplicate a file handle */
645c22ef 9473
1d7c1841 9474PerlIO *
a8fc9800 9475Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9476{
9477 PerlIO *ret;
53c1dcc0
AL
9478
9479 PERL_UNUSED_ARG(type);
73d840c0 9480
1d7c1841
GS
9481 if (!fp)
9482 return (PerlIO*)NULL;
9483
9484 /* look for it in the table first */
9485 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9486 if (ret)
9487 return ret;
9488
9489 /* create anew and remember what it is */
ecdeb87c 9490 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9491 ptr_table_store(PL_ptr_table, fp, ret);
9492 return ret;
9493}
9494
645c22ef
DM
9495/* duplicate a directory handle */
9496
1d7c1841
GS
9497DIR *
9498Perl_dirp_dup(pTHX_ DIR *dp)
9499{
96a5add6 9500 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9501 if (!dp)
9502 return (DIR*)NULL;
9503 /* XXX TODO */
9504 return dp;
9505}
9506
ff276b08 9507/* duplicate a typeglob */
645c22ef 9508
1d7c1841 9509GP *
a8fc9800 9510Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9511{
9512 GP *ret;
b37c2d43 9513
1d7c1841
GS
9514 if (!gp)
9515 return (GP*)NULL;
9516 /* look for it in the table first */
9517 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9518 if (ret)
9519 return ret;
9520
9521 /* create anew and remember what it is */
a02a5408 9522 Newxz(ret, 1, GP);
1d7c1841
GS
9523 ptr_table_store(PL_ptr_table, gp, ret);
9524
9525 /* clone */
9526 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9527 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9528 ret->gp_io = io_dup_inc(gp->gp_io, param);
9529 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9530 ret->gp_av = av_dup_inc(gp->gp_av, param);
9531 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9532 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9533 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9534 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841 9535 ret->gp_line = gp->gp_line;
f4890806 9536 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
1d7c1841
GS
9537 return ret;
9538}
9539
645c22ef
DM
9540/* duplicate a chain of magic */
9541
1d7c1841 9542MAGIC *
a8fc9800 9543Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9544{
cb359b41
JH
9545 MAGIC *mgprev = (MAGIC*)NULL;
9546 MAGIC *mgret;
1d7c1841
GS
9547 if (!mg)
9548 return (MAGIC*)NULL;
9549 /* look for it in the table first */
9550 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9551 if (mgret)
9552 return mgret;
9553
9554 for (; mg; mg = mg->mg_moremagic) {
9555 MAGIC *nmg;
a02a5408 9556 Newxz(nmg, 1, MAGIC);
cb359b41 9557 if (mgprev)
1d7c1841 9558 mgprev->mg_moremagic = nmg;
cb359b41
JH
9559 else
9560 mgret = nmg;
1d7c1841
GS
9561 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9562 nmg->mg_private = mg->mg_private;
9563 nmg->mg_type = mg->mg_type;
9564 nmg->mg_flags = mg->mg_flags;
14befaf4 9565 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9566 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9567 }
05bd4103 9568 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9569 /* The backref AV has its reference count deliberately bumped by
9570 1. */
9571 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9572 }
8d2f4536
NC
9573 else if (mg->mg_type == PERL_MAGIC_symtab) {
9574 nmg->mg_obj = mg->mg_obj;
9575 }
1d7c1841
GS
9576 else {
9577 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9578 ? sv_dup_inc(mg->mg_obj, param)
9579 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9580 }
9581 nmg->mg_len = mg->mg_len;
9582 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9583 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9584 if (mg->mg_len > 0) {
1d7c1841 9585 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9586 if (mg->mg_type == PERL_MAGIC_overload_table &&
9587 AMT_AMAGIC((AMT*)mg->mg_ptr))
9588 {
c445ea15 9589 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9590 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9591 I32 i;
9592 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9593 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9594 }
9595 }
9596 }
9597 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9598 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9599 }
68795e93
NIS
9600 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9601 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9602 }
1d7c1841
GS
9603 mgprev = nmg;
9604 }
9605 return mgret;
9606}
9607
645c22ef
DM
9608/* create a new pointer-mapping table */
9609
1d7c1841
GS
9610PTR_TBL_t *
9611Perl_ptr_table_new(pTHX)
9612{
9613 PTR_TBL_t *tbl;
96a5add6
AL
9614 PERL_UNUSED_CONTEXT;
9615
a02a5408 9616 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9617 tbl->tbl_max = 511;
9618 tbl->tbl_items = 0;
a02a5408 9619 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9620 return tbl;
9621}
9622
7119fd33
NC
9623#define PTR_TABLE_HASH(ptr) \
9624 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9625
93e68bfb
JC
9626/*
9627 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9628 following define) and at call to new_body_inline made below in
9629 Perl_ptr_table_store()
9630 */
9631
9632#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9633
645c22ef
DM
9634/* map an existing pointer using a table */
9635
7bf61b54 9636STATIC PTR_TBL_ENT_t *
b0e6ae5b 9637S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9638 PTR_TBL_ENT_t *tblent;
4373e329 9639 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9640 assert(tbl);
9641 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9642 for (; tblent; tblent = tblent->next) {
9643 if (tblent->oldval == sv)
7bf61b54 9644 return tblent;
1d7c1841 9645 }
d4c19fe8 9646 return NULL;
7bf61b54
NC
9647}
9648
9649void *
9650Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9651{
b0e6ae5b 9652 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9653 PERL_UNUSED_CONTEXT;
d4c19fe8 9654 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9655}
9656
645c22ef
DM
9657/* add a new entry to a pointer-mapping table */
9658
1d7c1841 9659void
44f8325f 9660Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9661{
0c9fdfe0 9662 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9663 PERL_UNUSED_CONTEXT;
1d7c1841 9664
7bf61b54
NC
9665 if (tblent) {
9666 tblent->newval = newsv;
9667 } else {
9668 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9669
d2a0f284
JC
9670 new_body_inline(tblent, PTE_SVSLOT);
9671
7bf61b54
NC
9672 tblent->oldval = oldsv;
9673 tblent->newval = newsv;
9674 tblent->next = tbl->tbl_ary[entry];
9675 tbl->tbl_ary[entry] = tblent;
9676 tbl->tbl_items++;
9677 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9678 ptr_table_split(tbl);
1d7c1841 9679 }
1d7c1841
GS
9680}
9681
645c22ef
DM
9682/* double the hash bucket size of an existing ptr table */
9683
1d7c1841
GS
9684void
9685Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9686{
9687 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9688 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9689 UV newsize = oldsize * 2;
9690 UV i;
96a5add6 9691 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9692
9693 Renew(ary, newsize, PTR_TBL_ENT_t*);
9694 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9695 tbl->tbl_max = --newsize;
9696 tbl->tbl_ary = ary;
9697 for (i=0; i < oldsize; i++, ary++) {
9698 PTR_TBL_ENT_t **curentp, **entp, *ent;
9699 if (!*ary)
9700 continue;
9701 curentp = ary + oldsize;
9702 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9703 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9704 *entp = ent->next;
9705 ent->next = *curentp;
9706 *curentp = ent;
9707 continue;
9708 }
9709 else
9710 entp = &ent->next;
9711 }
9712 }
9713}
9714
645c22ef
DM
9715/* remove all the entries from a ptr table */
9716
a0739874
DM
9717void
9718Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9719{
d5cefff9 9720 if (tbl && tbl->tbl_items) {
c445ea15 9721 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9722 UV riter = tbl->tbl_max;
a0739874 9723
d5cefff9
NC
9724 do {
9725 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9726
d5cefff9 9727 while (entry) {
00b6aa41 9728 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9729 entry = entry->next;
9730 del_pte(oentry);
9731 }
9732 } while (riter--);
a0739874 9733
d5cefff9
NC
9734 tbl->tbl_items = 0;
9735 }
a0739874
DM
9736}
9737
645c22ef
DM
9738/* clear and free a ptr table */
9739
a0739874
DM
9740void
9741Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9742{
9743 if (!tbl) {
9744 return;
9745 }
9746 ptr_table_clear(tbl);
9747 Safefree(tbl->tbl_ary);
9748 Safefree(tbl);
9749}
9750
5bd07a3d 9751
83841fad 9752void
eb86f8b3 9753Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9754{
9755 if (SvROK(sstr)) {
b162af07
SP
9756 SvRV_set(dstr, SvWEAKREF(sstr)
9757 ? sv_dup(SvRV(sstr), param)
9758 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9759
83841fad 9760 }
3f7c398e 9761 else if (SvPVX_const(sstr)) {
83841fad
NIS
9762 /* Has something there */
9763 if (SvLEN(sstr)) {
68795e93 9764 /* Normal PV - clone whole allocated space */
3f7c398e 9765 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9766 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9767 /* Not that normal - actually sstr is copy on write.
9768 But we are a true, independant SV, so: */
9769 SvREADONLY_off(dstr);
9770 SvFAKE_off(dstr);
9771 }
68795e93 9772 }
83841fad
NIS
9773 else {
9774 /* Special case - not normally malloced for some reason */
f7877b28
NC
9775 if (isGV_with_GP(sstr)) {
9776 /* Don't need to do anything here. */
9777 }
9778 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9779 /* A "shared" PV - clone it as "shared" PV */
9780 SvPV_set(dstr,
9781 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9782 param)));
83841fad
NIS
9783 }
9784 else {
9785 /* Some other special case - random pointer */
f880fe2f 9786 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9787 }
83841fad
NIS
9788 }
9789 }
9790 else {
4608196e 9791 /* Copy the NULL */
f880fe2f 9792 if (SvTYPE(dstr) == SVt_RV)
b162af07 9793 SvRV_set(dstr, NULL);
f880fe2f 9794 else
6136c704 9795 SvPV_set(dstr, NULL);
83841fad
NIS
9796 }
9797}
9798
662fb8b2
NC
9799/* duplicate an SV of any type (including AV, HV etc) */
9800
1d7c1841 9801SV *
eb86f8b3 9802Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9803{
27da23d5 9804 dVAR;
1d7c1841
GS
9805 SV *dstr;
9806
9807 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9808 return NULL;
1d7c1841
GS
9809 /* look for it in the table first */
9810 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9811 if (dstr)
9812 return dstr;
9813
0405e91e
AB
9814 if(param->flags & CLONEf_JOIN_IN) {
9815 /** We are joining here so we don't want do clone
9816 something that is bad **/
eb86f8b3
AL
9817 if (SvTYPE(sstr) == SVt_PVHV) {
9818 const char * const hvname = HvNAME_get(sstr);
9819 if (hvname)
9820 /** don't clone stashes if they already exist **/
9821 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9822 }
9823 }
9824
1d7c1841
GS
9825 /* create anew and remember what it is */
9826 new_SV(dstr);
fd0854ff
DM
9827
9828#ifdef DEBUG_LEAKING_SCALARS
9829 dstr->sv_debug_optype = sstr->sv_debug_optype;
9830 dstr->sv_debug_line = sstr->sv_debug_line;
9831 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9832 dstr->sv_debug_cloned = 1;
fd0854ff 9833 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9834#endif
9835
1d7c1841
GS
9836 ptr_table_store(PL_ptr_table, sstr, dstr);
9837
9838 /* clone */
9839 SvFLAGS(dstr) = SvFLAGS(sstr);
9840 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9841 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9842
9843#ifdef DEBUGGING
3f7c398e 9844 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9845 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9846 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9847#endif
9848
9660f481
DM
9849 /* don't clone objects whose class has asked us not to */
9850 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9851 SvFLAGS(dstr) &= ~SVTYPEMASK;
9852 SvOBJECT_off(dstr);
9853 return dstr;
9854 }
9855
1d7c1841
GS
9856 switch (SvTYPE(sstr)) {
9857 case SVt_NULL:
9858 SvANY(dstr) = NULL;
9859 break;
9860 case SVt_IV:
339049b0 9861 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9862 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9863 break;
9864 case SVt_NV:
9865 SvANY(dstr) = new_XNV();
9d6ce603 9866 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9867 break;
9868 case SVt_RV:
339049b0 9869 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9870 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9871 break;
662fb8b2
NC
9872 default:
9873 {
9874 /* These are all the types that need complex bodies allocating. */
662fb8b2 9875 void *new_body;
2bcc16b3
NC
9876 const svtype sv_type = SvTYPE(sstr);
9877 const struct body_details *const sv_type_details
9878 = bodies_by_type + sv_type;
662fb8b2 9879
93e68bfb 9880 switch (sv_type) {
662fb8b2 9881 default:
bb263b4e 9882 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9883 break;
9884
662fb8b2
NC
9885 case SVt_PVGV:
9886 if (GvUNIQUE((GV*)sstr)) {
6f207bd3 9887 NOOP; /* Do sharing here, and fall through */
662fb8b2 9888 }
c22188b4
NC
9889 case SVt_PVIO:
9890 case SVt_PVFM:
9891 case SVt_PVHV:
9892 case SVt_PVAV:
93e68bfb 9893 case SVt_PVBM:
662fb8b2 9894 case SVt_PVCV:
662fb8b2 9895 case SVt_PVLV:
662fb8b2 9896 case SVt_PVMG:
662fb8b2 9897 case SVt_PVNV:
662fb8b2 9898 case SVt_PVIV:
662fb8b2 9899 case SVt_PV:
d2a0f284 9900 assert(sv_type_details->body_size);
c22188b4 9901 if (sv_type_details->arena) {
d2a0f284 9902 new_body_inline(new_body, sv_type);
c22188b4 9903 new_body
b9502f15 9904 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9905 } else {
9906 new_body = new_NOARENA(sv_type_details);
9907 }
1d7c1841 9908 }
662fb8b2
NC
9909 assert(new_body);
9910 SvANY(dstr) = new_body;
9911
2bcc16b3 9912#ifndef PURIFY
b9502f15
NC
9913 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
9914 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 9915 sv_type_details->copy, char);
2bcc16b3
NC
9916#else
9917 Copy(((char*)SvANY(sstr)),
9918 ((char*)SvANY(dstr)),
d2a0f284 9919 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 9920#endif
662fb8b2 9921
f7877b28
NC
9922 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
9923 && !isGV_with_GP(dstr))
662fb8b2
NC
9924 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
9925
9926 /* The Copy above means that all the source (unduplicated) pointers
9927 are now in the destination. We can check the flags and the
9928 pointers in either, but it's possible that there's less cache
9929 missing by always going for the destination.
9930 FIXME - instrument and check that assumption */
f32993d6 9931 if (sv_type >= SVt_PVMG) {
885ffcb3
NC
9932 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
9933 OURSTASH_set(dstr, hv_dup_inc(OURSTASH(dstr), param));
e736a858 9934 } else if (SvMAGIC(dstr))
662fb8b2
NC
9935 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
9936 if (SvSTASH(dstr))
9937 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 9938 }
662fb8b2 9939
f32993d6
NC
9940 /* The cast silences a GCC warning about unhandled types. */
9941 switch ((int)sv_type) {
662fb8b2
NC
9942 case SVt_PV:
9943 break;
9944 case SVt_PVIV:
9945 break;
9946 case SVt_PVNV:
9947 break;
9948 case SVt_PVMG:
9949 break;
9950 case SVt_PVBM:
9951 break;
9952 case SVt_PVLV:
9953 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
9954 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
9955 LvTARG(dstr) = dstr;
9956 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
9957 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
9958 else
9959 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
9960 break;
9961 case SVt_PVGV:
acda4c6a
NC
9962 if (GvNAME_HEK(dstr))
9963 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 9964
e15faf7d
NC
9965 /* Don't call sv_add_backref here as it's going to be created
9966 as part of the magic cloning of the symbol table. */
f7877b28
NC
9967 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
9968 if(isGV_with_GP(sstr)) {
9969 /* Danger Will Robinson - GvGP(dstr) isn't initialised
9970 at the point of this comment. */
9971 GvGP(dstr) = gp_dup(GvGP(sstr), param);
9972 (void)GpREFCNT_inc(GvGP(dstr));
9973 } else
9974 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
9975 break;
9976 case SVt_PVIO:
9977 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
9978 if (IoOFP(dstr) == IoIFP(sstr))
9979 IoOFP(dstr) = IoIFP(dstr);
9980 else
9981 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
9982 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
9983 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
9984 /* I have no idea why fake dirp (rsfps)
9985 should be treated differently but otherwise
9986 we end up with leaks -- sky*/
9987 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
9988 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
9989 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
9990 } else {
9991 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
9992 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
9993 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
9994 if (IoDIRP(dstr)) {
9995 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
9996 } else {
6f207bd3 9997 NOOP;
100ce7e1
NC
9998 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
9999 }
662fb8b2
NC
10000 }
10001 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10002 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10003 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10004 break;
10005 case SVt_PVAV:
10006 if (AvARRAY((AV*)sstr)) {
10007 SV **dst_ary, **src_ary;
10008 SSize_t items = AvFILLp((AV*)sstr) + 1;
10009
10010 src_ary = AvARRAY((AV*)sstr);
a02a5408 10011 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2 10012 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
9c6bc640 10013 AvARRAY((AV*)dstr) = dst_ary;
662fb8b2
NC
10014 AvALLOC((AV*)dstr) = dst_ary;
10015 if (AvREAL((AV*)sstr)) {
10016 while (items-- > 0)
10017 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10018 }
10019 else {
10020 while (items-- > 0)
10021 *dst_ary++ = sv_dup(*src_ary++, param);
10022 }
10023 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10024 while (items-- > 0) {
10025 *dst_ary++ = &PL_sv_undef;
10026 }
bfcb3514 10027 }
662fb8b2 10028 else {
9c6bc640 10029 AvARRAY((AV*)dstr) = NULL;
662fb8b2 10030 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10031 }
662fb8b2
NC
10032 break;
10033 case SVt_PVHV:
7e265ef3
AL
10034 if (HvARRAY((HV*)sstr)) {
10035 STRLEN i = 0;
10036 const bool sharekeys = !!HvSHAREKEYS(sstr);
10037 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10038 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10039 char *darray;
10040 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
10041 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10042 char);
10043 HvARRAY(dstr) = (HE**)darray;
10044 while (i <= sxhv->xhv_max) {
10045 const HE * const source = HvARRAY(sstr)[i];
10046 HvARRAY(dstr)[i] = source
10047 ? he_dup(source, sharekeys, param) : 0;
10048 ++i;
10049 }
10050 if (SvOOK(sstr)) {
10051 HEK *hvname;
10052 const struct xpvhv_aux * const saux = HvAUX(sstr);
10053 struct xpvhv_aux * const daux = HvAUX(dstr);
10054 /* This flag isn't copied. */
10055 /* SvOOK_on(hv) attacks the IV flags. */
10056 SvFLAGS(dstr) |= SVf_OOK;
10057
10058 hvname = saux->xhv_name;
10059 daux->xhv_name = hvname ? hek_dup(hvname, param) : hvname;
10060
10061 daux->xhv_riter = saux->xhv_riter;
10062 daux->xhv_eiter = saux->xhv_eiter
10063 ? he_dup(saux->xhv_eiter,
10064 (bool)!!HvSHAREKEYS(sstr), param) : 0;
10065 daux->xhv_backreferences =
10066 saux->xhv_backreferences
86f55936 10067 ? (AV*) SvREFCNT_inc(
7e265ef3 10068 sv_dup((SV*)saux->xhv_backreferences, param))
86f55936 10069 : 0;
7e265ef3
AL
10070 /* Record stashes for possible cloning in Perl_clone(). */
10071 if (hvname)
10072 av_push(param->stashes, dstr);
662fb8b2 10073 }
662fb8b2 10074 }
7e265ef3 10075 else
797c7171 10076 HvARRAY((HV*)dstr) = NULL;
662fb8b2 10077 break;
662fb8b2 10078 case SVt_PVCV:
bb172083
NC
10079 if (!(param->flags & CLONEf_COPY_STACKS)) {
10080 CvDEPTH(dstr) = 0;
10081 }
10082 case SVt_PVFM:
662fb8b2
NC
10083 /* NOTE: not refcounted */
10084 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10085 OP_REFCNT_LOCK;
d04ba589
NC
10086 if (!CvISXSUB(dstr))
10087 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10088 OP_REFCNT_UNLOCK;
cfae286e 10089 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10090 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10091 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10092 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10093 }
10094 /* don't dup if copying back - CvGV isn't refcounted, so the
10095 * duped GV may never be freed. A bit of a hack! DAPM */
10096 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10097 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10098 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10099 CvOUTSIDE(dstr) =
10100 CvWEAKOUTSIDE(sstr)
10101 ? cv_dup( CvOUTSIDE(dstr), param)
10102 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10103 if (!CvISXSUB(dstr))
662fb8b2
NC
10104 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10105 break;
bfcb3514 10106 }
1d7c1841 10107 }
1d7c1841
GS
10108 }
10109
10110 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10111 ++PL_sv_objcount;
10112
10113 return dstr;
d2d73c3e 10114 }
1d7c1841 10115
645c22ef
DM
10116/* duplicate a context */
10117
1d7c1841 10118PERL_CONTEXT *
a8fc9800 10119Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10120{
10121 PERL_CONTEXT *ncxs;
10122
10123 if (!cxs)
10124 return (PERL_CONTEXT*)NULL;
10125
10126 /* look for it in the table first */
10127 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10128 if (ncxs)
10129 return ncxs;
10130
10131 /* create anew and remember what it is */
a02a5408 10132 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10133 ptr_table_store(PL_ptr_table, cxs, ncxs);
10134
10135 while (ix >= 0) {
c445ea15
AL
10136 PERL_CONTEXT * const cx = &cxs[ix];
10137 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10138 ncx->cx_type = cx->cx_type;
10139 if (CxTYPE(cx) == CXt_SUBST) {
10140 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10141 }
10142 else {
10143 ncx->blk_oldsp = cx->blk_oldsp;
10144 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10145 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10146 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10147 ncx->blk_oldpm = cx->blk_oldpm;
10148 ncx->blk_gimme = cx->blk_gimme;
10149 switch (CxTYPE(cx)) {
10150 case CXt_SUB:
10151 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10152 ? cv_dup_inc(cx->blk_sub.cv, param)
10153 : cv_dup(cx->blk_sub.cv,param));
cc8d50a7 10154 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10155 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10156 : NULL);
d2d73c3e 10157 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841 10158 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
cc8d50a7
NC
10159 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10160 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10161 ncx->blk_sub.retop = cx->blk_sub.retop;
d8d97e70
DM
10162 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
10163 cx->blk_sub.oldcomppad);
1d7c1841
GS
10164 break;
10165 case CXt_EVAL:
10166 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10167 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10168 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10169 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10170 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10171 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10172 break;
10173 case CXt_LOOP:
10174 ncx->blk_loop.label = cx->blk_loop.label;
10175 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
022eaa24 10176 ncx->blk_loop.my_op = cx->blk_loop.my_op;
1d7c1841
GS
10177 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10178 ? cx->blk_loop.iterdata
d2d73c3e 10179 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10180 ncx->blk_loop.oldcomppad
10181 = (PAD*)ptr_table_fetch(PL_ptr_table,
10182 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10183 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10184 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10185 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10186 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10187 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10188 break;
10189 case CXt_FORMAT:
d2d73c3e
AB
10190 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10191 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10192 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
cc8d50a7 10193 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10194 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10195 break;
10196 case CXt_BLOCK:
10197 case CXt_NULL:
10198 break;
10199 }
10200 }
10201 --ix;
10202 }
10203 return ncxs;
10204}
10205
645c22ef
DM
10206/* duplicate a stack info structure */
10207
1d7c1841 10208PERL_SI *
a8fc9800 10209Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10210{
10211 PERL_SI *nsi;
10212
10213 if (!si)
10214 return (PERL_SI*)NULL;
10215
10216 /* look for it in the table first */
10217 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10218 if (nsi)
10219 return nsi;
10220
10221 /* create anew and remember what it is */
a02a5408 10222 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10223 ptr_table_store(PL_ptr_table, si, nsi);
10224
d2d73c3e 10225 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10226 nsi->si_cxix = si->si_cxix;
10227 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10228 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10229 nsi->si_type = si->si_type;
d2d73c3e
AB
10230 nsi->si_prev = si_dup(si->si_prev, param);
10231 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10232 nsi->si_markoff = si->si_markoff;
10233
10234 return nsi;
10235}
10236
10237#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10238#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10239#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10240#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10241#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10242#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10243#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10244#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10245#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10246#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10247#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10248#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10249#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10250#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10251
10252/* XXXXX todo */
10253#define pv_dup_inc(p) SAVEPV(p)
10254#define pv_dup(p) SAVEPV(p)
10255#define svp_dup_inc(p,pp) any_dup(p,pp)
10256
645c22ef
DM
10257/* map any object to the new equivent - either something in the
10258 * ptr table, or something in the interpreter structure
10259 */
10260
1d7c1841 10261void *
53c1dcc0 10262Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10263{
10264 void *ret;
10265
10266 if (!v)
10267 return (void*)NULL;
10268
10269 /* look for it in the table first */
10270 ret = ptr_table_fetch(PL_ptr_table, v);
10271 if (ret)
10272 return ret;
10273
10274 /* see if it is part of the interpreter structure */
10275 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10276 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10277 else {
1d7c1841 10278 ret = v;
05ec9bb3 10279 }
1d7c1841
GS
10280
10281 return ret;
10282}
10283
645c22ef
DM
10284/* duplicate the save stack */
10285
1d7c1841 10286ANY *
a8fc9800 10287Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10288{
53c1dcc0
AL
10289 ANY * const ss = proto_perl->Tsavestack;
10290 const I32 max = proto_perl->Tsavestack_max;
10291 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10292 ANY *nss;
10293 SV *sv;
10294 GV *gv;
10295 AV *av;
10296 HV *hv;
10297 void* ptr;
10298 int intval;
10299 long longval;
10300 GP *gp;
10301 IV iv;
b24356f5 10302 I32 i;
c4e33207 10303 char *c = NULL;
1d7c1841 10304 void (*dptr) (void*);
acfe0abc 10305 void (*dxptr) (pTHX_ void*);
1d7c1841 10306
a02a5408 10307 Newxz(nss, max, ANY);
1d7c1841
GS
10308
10309 while (ix > 0) {
b24356f5
NC
10310 const I32 type = POPINT(ss,ix);
10311 TOPINT(nss,ix) = type;
10312 switch (type) {
3e07292d
NC
10313 case SAVEt_HELEM: /* hash element */
10314 sv = (SV*)POPPTR(ss,ix);
10315 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
10316 /* fall through */
1d7c1841 10317 case SAVEt_ITEM: /* normal string */
a41cc44e 10318 case SAVEt_SV: /* scalar reference */
1d7c1841 10319 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10320 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10321 /* fall through */
10322 case SAVEt_FREESV:
10323 case SAVEt_MORTALIZESV:
1d7c1841 10324 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10325 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10326 break;
05ec9bb3
NIS
10327 case SAVEt_SHARED_PVREF: /* char* in shared space */
10328 c = (char*)POPPTR(ss,ix);
10329 TOPPTR(nss,ix) = savesharedpv(c);
10330 ptr = POPPTR(ss,ix);
10331 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10332 break;
1d7c1841
GS
10333 case SAVEt_GENERIC_SVREF: /* generic sv */
10334 case SAVEt_SVREF: /* scalar reference */
10335 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10336 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10337 ptr = POPPTR(ss,ix);
10338 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10339 break;
a41cc44e 10340 case SAVEt_HV: /* hash reference */
1d7c1841 10341 case SAVEt_AV: /* array reference */
11b79775 10342 sv = (SV*) POPPTR(ss,ix);
337d28f5 10343 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10344 /* fall through */
10345 case SAVEt_COMPPAD:
10346 case SAVEt_NSTAB:
667e2948 10347 sv = (SV*) POPPTR(ss,ix);
3e07292d 10348 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10349 break;
10350 case SAVEt_INT: /* int reference */
10351 ptr = POPPTR(ss,ix);
10352 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10353 intval = (int)POPINT(ss,ix);
10354 TOPINT(nss,ix) = intval;
10355 break;
10356 case SAVEt_LONG: /* long reference */
10357 ptr = POPPTR(ss,ix);
10358 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
3e07292d
NC
10359 /* fall through */
10360 case SAVEt_CLEARSV:
1d7c1841
GS
10361 longval = (long)POPLONG(ss,ix);
10362 TOPLONG(nss,ix) = longval;
10363 break;
10364 case SAVEt_I32: /* I32 reference */
10365 case SAVEt_I16: /* I16 reference */
10366 case SAVEt_I8: /* I8 reference */
88effcc9 10367 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10368 ptr = POPPTR(ss,ix);
10369 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
1ccabee8 10370 i = POPINT(ss,ix);
1d7c1841
GS
10371 TOPINT(nss,ix) = i;
10372 break;
10373 case SAVEt_IV: /* IV reference */
10374 ptr = POPPTR(ss,ix);
10375 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10376 iv = POPIV(ss,ix);
10377 TOPIV(nss,ix) = iv;
10378 break;
a41cc44e
NC
10379 case SAVEt_HPTR: /* HV* reference */
10380 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10381 case SAVEt_SPTR: /* SV* reference */
10382 ptr = POPPTR(ss,ix);
10383 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10384 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10385 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10386 break;
10387 case SAVEt_VPTR: /* random* reference */
10388 ptr = POPPTR(ss,ix);
10389 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10390 ptr = POPPTR(ss,ix);
10391 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10392 break;
b03d03b0 10393 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10394 case SAVEt_PPTR: /* char* reference */
10395 ptr = POPPTR(ss,ix);
10396 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10397 c = (char*)POPPTR(ss,ix);
10398 TOPPTR(nss,ix) = pv_dup(c);
10399 break;
1d7c1841
GS
10400 case SAVEt_GP: /* scalar reference */
10401 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10402 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10403 (void)GpREFCNT_inc(gp);
10404 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10405 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841 10406 break;
1d7c1841
GS
10407 case SAVEt_FREEOP:
10408 ptr = POPPTR(ss,ix);
10409 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10410 /* these are assumed to be refcounted properly */
53c1dcc0 10411 OP *o;
1d7c1841
GS
10412 switch (((OP*)ptr)->op_type) {
10413 case OP_LEAVESUB:
10414 case OP_LEAVESUBLV:
10415 case OP_LEAVEEVAL:
10416 case OP_LEAVE:
10417 case OP_SCOPE:
10418 case OP_LEAVEWRITE:
e977893f
GS
10419 TOPPTR(nss,ix) = ptr;
10420 o = (OP*)ptr;
10421 OpREFCNT_inc(o);
1d7c1841
GS
10422 break;
10423 default:
5f66b61c 10424 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10425 break;
10426 }
10427 }
10428 else
5f66b61c 10429 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10430 break;
10431 case SAVEt_FREEPV:
10432 c = (char*)POPPTR(ss,ix);
10433 TOPPTR(nss,ix) = pv_dup_inc(c);
10434 break;
1d7c1841
GS
10435 case SAVEt_DELETE:
10436 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10437 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10438 c = (char*)POPPTR(ss,ix);
10439 TOPPTR(nss,ix) = pv_dup_inc(c);
3e07292d
NC
10440 /* fall through */
10441 case SAVEt_STACK_POS: /* Position on Perl stack */
1d7c1841
GS
10442 i = POPINT(ss,ix);
10443 TOPINT(nss,ix) = i;
10444 break;
10445 case SAVEt_DESTRUCTOR:
10446 ptr = POPPTR(ss,ix);
10447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10448 dptr = POPDPTR(ss,ix);
8141890a
JH
10449 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10450 any_dup(FPTR2DPTR(void *, dptr),
10451 proto_perl));
1d7c1841
GS
10452 break;
10453 case SAVEt_DESTRUCTOR_X:
10454 ptr = POPPTR(ss,ix);
10455 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10456 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10457 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10458 any_dup(FPTR2DPTR(void *, dxptr),
10459 proto_perl));
1d7c1841
GS
10460 break;
10461 case SAVEt_REGCONTEXT:
10462 case SAVEt_ALLOC:
10463 i = POPINT(ss,ix);
10464 TOPINT(nss,ix) = i;
10465 ix -= i;
10466 break;
1d7c1841
GS
10467 case SAVEt_AELEM: /* array element */
10468 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10469 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10470 i = POPINT(ss,ix);
10471 TOPINT(nss,ix) = i;
10472 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10473 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841 10474 break;
1d7c1841
GS
10475 case SAVEt_OP:
10476 ptr = POPPTR(ss,ix);
10477 TOPPTR(nss,ix) = ptr;
10478 break;
10479 case SAVEt_HINTS:
10480 i = POPINT(ss,ix);
10481 TOPINT(nss,ix) = i;
b3ca2e83 10482 ptr = POPPTR(ss,ix);
080ac856 10483 if (ptr) {
7b6dd8c3 10484 HINTS_REFCNT_LOCK;
080ac856 10485 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10486 HINTS_REFCNT_UNLOCK;
10487 }
cbb1fbea 10488 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10489 if (i & HINT_LOCALIZE_HH) {
10490 hv = (HV*)POPPTR(ss,ix);
10491 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10492 }
1d7c1841 10493 break;
c3564e5c
GS
10494 case SAVEt_PADSV:
10495 longval = (long)POPLONG(ss,ix);
10496 TOPLONG(nss,ix) = longval;
10497 ptr = POPPTR(ss,ix);
10498 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10499 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10500 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10501 break;
a1bb4754 10502 case SAVEt_BOOL:
38d8b13e 10503 ptr = POPPTR(ss,ix);
b9609c01 10504 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10505 longval = (long)POPBOOL(ss,ix);
b9609c01 10506 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10507 break;
8bd2680e
MHM
10508 case SAVEt_SET_SVFLAGS:
10509 i = POPINT(ss,ix);
10510 TOPINT(nss,ix) = i;
10511 i = POPINT(ss,ix);
10512 TOPINT(nss,ix) = i;
10513 sv = (SV*)POPPTR(ss,ix);
10514 TOPPTR(nss,ix) = sv_dup(sv, param);
10515 break;
5bfb7d0e
NC
10516 case SAVEt_RE_STATE:
10517 {
10518 const struct re_save_state *const old_state
10519 = (struct re_save_state *)
10520 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10521 struct re_save_state *const new_state
10522 = (struct re_save_state *)
10523 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10524
10525 Copy(old_state, new_state, 1, struct re_save_state);
10526 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10527
10528 new_state->re_state_bostr
10529 = pv_dup(old_state->re_state_bostr);
10530 new_state->re_state_reginput
10531 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10532 new_state->re_state_regeol
10533 = pv_dup(old_state->re_state_regeol);
10534 new_state->re_state_regstartp
11b79775 10535 = (I32*) any_dup(old_state->re_state_regstartp, proto_perl);
5bfb7d0e 10536 new_state->re_state_regendp
11b79775 10537 = (I32*) any_dup(old_state->re_state_regendp, proto_perl);
5bfb7d0e 10538 new_state->re_state_reglastparen
11b79775
DD
10539 = (U32*) any_dup(old_state->re_state_reglastparen,
10540 proto_perl);
5bfb7d0e 10541 new_state->re_state_reglastcloseparen
11b79775 10542 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
5bfb7d0e 10543 proto_perl);
5bfb7d0e
NC
10544 /* XXX This just has to be broken. The old save_re_context
10545 code did SAVEGENERICPV(PL_reg_start_tmp);
10546 PL_reg_start_tmp is char **.
10547 Look above to what the dup code does for
10548 SAVEt_GENERIC_PVREF
10549 It can never have worked.
10550 So this is merely a faithful copy of the exiting bug: */
10551 new_state->re_state_reg_start_tmp
10552 = (char **) pv_dup((char *)
10553 old_state->re_state_reg_start_tmp);
10554 /* I assume that it only ever "worked" because no-one called
10555 (pseudo)fork while the regexp engine had re-entered itself.
10556 */
5bfb7d0e
NC
10557#ifdef PERL_OLD_COPY_ON_WRITE
10558 new_state->re_state_nrs
10559 = sv_dup(old_state->re_state_nrs, param);
10560#endif
10561 new_state->re_state_reg_magic
11b79775
DD
10562 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
10563 proto_perl);
5bfb7d0e 10564 new_state->re_state_reg_oldcurpm
11b79775
DD
10565 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
10566 proto_perl);
5bfb7d0e 10567 new_state->re_state_reg_curpm
11b79775
DD
10568 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
10569 proto_perl);
5bfb7d0e
NC
10570 new_state->re_state_reg_oldsaved
10571 = pv_dup(old_state->re_state_reg_oldsaved);
10572 new_state->re_state_reg_poscache
10573 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
10574 new_state->re_state_reg_starttry
10575 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
10576 break;
10577 }
68da3b2f
NC
10578 case SAVEt_COMPILE_WARNINGS:
10579 ptr = POPPTR(ss,ix);
10580 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10581 break;
1d7c1841 10582 default:
147bc374
NC
10583 Perl_croak(aTHX_
10584 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
1d7c1841
GS
10585 }
10586 }
10587
bd81e77b
NC
10588 return nss;
10589}
10590
10591
10592/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10593 * flag to the result. This is done for each stash before cloning starts,
10594 * so we know which stashes want their objects cloned */
10595
10596static void
10597do_mark_cloneable_stash(pTHX_ SV *sv)
10598{
10599 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10600 if (hvname) {
10601 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10602 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10603 if (cloner && GvCV(cloner)) {
10604 dSP;
10605 UV status;
10606
10607 ENTER;
10608 SAVETMPS;
10609 PUSHMARK(SP);
10610 XPUSHs(sv_2mortal(newSVhek(hvname)));
10611 PUTBACK;
10612 call_sv((SV*)GvCV(cloner), G_SCALAR);
10613 SPAGAIN;
10614 status = POPu;
10615 PUTBACK;
10616 FREETMPS;
10617 LEAVE;
10618 if (status)
10619 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10620 }
10621 }
10622}
10623
10624
10625
10626/*
10627=for apidoc perl_clone
10628
10629Create and return a new interpreter by cloning the current one.
10630
10631perl_clone takes these flags as parameters:
10632
10633CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10634without it we only clone the data and zero the stacks,
10635with it we copy the stacks and the new perl interpreter is
10636ready to run at the exact same point as the previous one.
10637The pseudo-fork code uses COPY_STACKS while the
10638threads->new doesn't.
10639
10640CLONEf_KEEP_PTR_TABLE
10641perl_clone keeps a ptr_table with the pointer of the old
10642variable as a key and the new variable as a value,
10643this allows it to check if something has been cloned and not
10644clone it again but rather just use the value and increase the
10645refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10646the ptr_table using the function
10647C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10648reason to keep it around is if you want to dup some of your own
10649variable who are outside the graph perl scans, example of this
10650code is in threads.xs create
10651
10652CLONEf_CLONE_HOST
10653This is a win32 thing, it is ignored on unix, it tells perls
10654win32host code (which is c++) to clone itself, this is needed on
10655win32 if you want to run two threads at the same time,
10656if you just want to do some stuff in a separate perl interpreter
10657and then throw it away and return to the original one,
10658you don't need to do anything.
10659
10660=cut
10661*/
10662
10663/* XXX the above needs expanding by someone who actually understands it ! */
10664EXTERN_C PerlInterpreter *
10665perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10666
10667PerlInterpreter *
10668perl_clone(PerlInterpreter *proto_perl, UV flags)
10669{
10670 dVAR;
10671#ifdef PERL_IMPLICIT_SYS
10672
10673 /* perlhost.h so we need to call into it
10674 to clone the host, CPerlHost should have a c interface, sky */
10675
10676 if (flags & CLONEf_CLONE_HOST) {
10677 return perl_clone_host(proto_perl,flags);
10678 }
10679 return perl_clone_using(proto_perl, flags,
10680 proto_perl->IMem,
10681 proto_perl->IMemShared,
10682 proto_perl->IMemParse,
10683 proto_perl->IEnv,
10684 proto_perl->IStdIO,
10685 proto_perl->ILIO,
10686 proto_perl->IDir,
10687 proto_perl->ISock,
10688 proto_perl->IProc);
10689}
10690
10691PerlInterpreter *
10692perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10693 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10694 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10695 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10696 struct IPerlDir* ipD, struct IPerlSock* ipS,
10697 struct IPerlProc* ipP)
10698{
10699 /* XXX many of the string copies here can be optimized if they're
10700 * constants; they need to be allocated as common memory and just
10701 * their pointers copied. */
10702
10703 IV i;
10704 CLONE_PARAMS clone_params;
5f66b61c 10705 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10706
5f66b61c 10707 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10708 /* for each stash, determine whether its objects should be cloned */
10709 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10710 PERL_SET_THX(my_perl);
10711
10712# ifdef DEBUGGING
7e337ee0 10713 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10714 PL_op = NULL;
10715 PL_curcop = NULL;
bd81e77b
NC
10716 PL_markstack = 0;
10717 PL_scopestack = 0;
10718 PL_savestack = 0;
10719 PL_savestack_ix = 0;
10720 PL_savestack_max = -1;
10721 PL_sig_pending = 0;
10722 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10723# else /* !DEBUGGING */
10724 Zero(my_perl, 1, PerlInterpreter);
10725# endif /* DEBUGGING */
10726
10727 /* host pointers */
10728 PL_Mem = ipM;
10729 PL_MemShared = ipMS;
10730 PL_MemParse = ipMP;
10731 PL_Env = ipE;
10732 PL_StdIO = ipStd;
10733 PL_LIO = ipLIO;
10734 PL_Dir = ipD;
10735 PL_Sock = ipS;
10736 PL_Proc = ipP;
10737#else /* !PERL_IMPLICIT_SYS */
10738 IV i;
10739 CLONE_PARAMS clone_params;
10740 CLONE_PARAMS* param = &clone_params;
5f66b61c 10741 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10742 /* for each stash, determine whether its objects should be cloned */
10743 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10744 PERL_SET_THX(my_perl);
10745
10746# ifdef DEBUGGING
7e337ee0 10747 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10748 PL_op = NULL;
10749 PL_curcop = NULL;
bd81e77b
NC
10750 PL_markstack = 0;
10751 PL_scopestack = 0;
10752 PL_savestack = 0;
10753 PL_savestack_ix = 0;
10754 PL_savestack_max = -1;
10755 PL_sig_pending = 0;
10756 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10757# else /* !DEBUGGING */
10758 Zero(my_perl, 1, PerlInterpreter);
10759# endif /* DEBUGGING */
10760#endif /* PERL_IMPLICIT_SYS */
10761 param->flags = flags;
10762 param->proto_perl = proto_perl;
10763
7cb608b5
NC
10764 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10765
fdda85ca 10766 PL_body_arenas = NULL;
bd81e77b
NC
10767 Zero(&PL_body_roots, 1, PL_body_roots);
10768
10769 PL_nice_chunk = NULL;
10770 PL_nice_chunk_size = 0;
10771 PL_sv_count = 0;
10772 PL_sv_objcount = 0;
a0714e2c
SS
10773 PL_sv_root = NULL;
10774 PL_sv_arenaroot = NULL;
bd81e77b
NC
10775
10776 PL_debug = proto_perl->Idebug;
10777
10778 PL_hash_seed = proto_perl->Ihash_seed;
10779 PL_rehash_seed = proto_perl->Irehash_seed;
10780
10781#ifdef USE_REENTRANT_API
10782 /* XXX: things like -Dm will segfault here in perlio, but doing
10783 * PERL_SET_CONTEXT(proto_perl);
10784 * breaks too many other things
10785 */
10786 Perl_reentrant_init(aTHX);
10787#endif
10788
10789 /* create SV map for pointer relocation */
10790 PL_ptr_table = ptr_table_new();
10791
10792 /* initialize these special pointers as early as possible */
10793 SvANY(&PL_sv_undef) = NULL;
10794 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10795 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10796 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10797
10798 SvANY(&PL_sv_no) = new_XPVNV();
10799 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10800 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10801 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10802 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
10803 SvCUR_set(&PL_sv_no, 0);
10804 SvLEN_set(&PL_sv_no, 1);
10805 SvIV_set(&PL_sv_no, 0);
10806 SvNV_set(&PL_sv_no, 0);
10807 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10808
10809 SvANY(&PL_sv_yes) = new_XPVNV();
10810 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10811 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10812 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10813 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
10814 SvCUR_set(&PL_sv_yes, 1);
10815 SvLEN_set(&PL_sv_yes, 2);
10816 SvIV_set(&PL_sv_yes, 1);
10817 SvNV_set(&PL_sv_yes, 1);
10818 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10819
10820 /* create (a non-shared!) shared string table */
10821 PL_strtab = newHV();
10822 HvSHAREKEYS_off(PL_strtab);
10823 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10824 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10825
10826 PL_compiling = proto_perl->Icompiling;
10827
10828 /* These two PVs will be free'd special way so must set them same way op.c does */
10829 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10830 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10831
10832 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10833 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10834
10835 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10836 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
c28fe1ec 10837 if (PL_compiling.cop_hints_hash) {
cbb1fbea 10838 HINTS_REFCNT_LOCK;
c28fe1ec 10839 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
cbb1fbea
NC
10840 HINTS_REFCNT_UNLOCK;
10841 }
bd81e77b
NC
10842 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10843
10844 /* pseudo environmental stuff */
10845 PL_origargc = proto_perl->Iorigargc;
10846 PL_origargv = proto_perl->Iorigargv;
10847
10848 param->stashes = newAV(); /* Setup array of objects to call clone on */
10849
10850 /* Set tainting stuff before PerlIO_debug can possibly get called */
10851 PL_tainting = proto_perl->Itainting;
10852 PL_taint_warn = proto_perl->Itaint_warn;
10853
10854#ifdef PERLIO_LAYERS
10855 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10856 PerlIO_clone(aTHX_ proto_perl, param);
10857#endif
10858
10859 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10860 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10861 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10862 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10863 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10864 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10865
10866 /* switches */
10867 PL_minus_c = proto_perl->Iminus_c;
10868 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10869 PL_localpatches = proto_perl->Ilocalpatches;
10870 PL_splitstr = proto_perl->Isplitstr;
10871 PL_preprocess = proto_perl->Ipreprocess;
10872 PL_minus_n = proto_perl->Iminus_n;
10873 PL_minus_p = proto_perl->Iminus_p;
10874 PL_minus_l = proto_perl->Iminus_l;
10875 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10876 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10877 PL_minus_F = proto_perl->Iminus_F;
10878 PL_doswitches = proto_perl->Idoswitches;
10879 PL_dowarn = proto_perl->Idowarn;
10880 PL_doextract = proto_perl->Idoextract;
10881 PL_sawampersand = proto_perl->Isawampersand;
10882 PL_unsafe = proto_perl->Iunsafe;
10883 PL_inplace = SAVEPV(proto_perl->Iinplace);
10884 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10885 PL_perldb = proto_perl->Iperldb;
10886 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10887 PL_exit_flags = proto_perl->Iexit_flags;
10888
10889 /* magical thingies */
10890 /* XXX time(&PL_basetime) when asked for? */
10891 PL_basetime = proto_perl->Ibasetime;
10892 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
10893
10894 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
10895 PL_statusvalue = proto_perl->Istatusvalue;
10896#ifdef VMS
10897 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
10898#else
10899 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
10900#endif
10901 PL_encoding = sv_dup(proto_perl->Iencoding, param);
10902
10903 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
10904 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
10905 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
10906
84da74a7 10907
f9f4320a 10908 /* RE engine related */
84da74a7
YO
10909 Zero(&PL_reg_state, 1, struct re_save_state);
10910 PL_reginterp_cnt = 0;
10911 PL_regmatch_slab = NULL;
10912
bd81e77b
NC
10913 /* Clone the regex array */
10914 PL_regex_padav = newAV();
10915 {
10916 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 10917 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 10918 IV i;
7f466ec7 10919 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 10920 for(i = 1; i <= len; i++) {
7a5b473e
AL
10921 const SV * const regex = regexen[i];
10922 SV * const sv =
10923 SvREPADTMP(regex)
10924 ? sv_dup_inc(regex, param)
10925 : SvREFCNT_inc(
10926 newSViv(PTR2IV(re_dup(
10927 INT2PTR(REGEXP *, SvIVX(regex)), param))))
10928 ;
10929 av_push(PL_regex_padav, sv);
bd81e77b
NC
10930 }
10931 }
10932 PL_regex_pad = AvARRAY(PL_regex_padav);
10933
10934 /* shortcuts to various I/O objects */
10935 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
10936 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
10937 PL_defgv = gv_dup(proto_perl->Idefgv, param);
10938 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
10939 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
10940 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 10941
bd81e77b
NC
10942 /* shortcuts to regexp stuff */
10943 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 10944
bd81e77b
NC
10945 /* shortcuts to misc objects */
10946 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 10947
bd81e77b
NC
10948 /* shortcuts to debugging objects */
10949 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
10950 PL_DBline = gv_dup(proto_perl->IDBline, param);
10951 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
10952 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
10953 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
10954 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
10955 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
10956 PL_lineary = av_dup(proto_perl->Ilineary, param);
10957 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 10958
bd81e77b
NC
10959 /* symbol tables */
10960 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
10961 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
10962 PL_debstash = hv_dup(proto_perl->Idebstash, param);
10963 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
10964 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
10965
10966 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
10967 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
10968 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
3c10abe3
AG
10969 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
10970 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
bd81e77b
NC
10971 PL_endav = av_dup_inc(proto_perl->Iendav, param);
10972 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
10973 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
10974
10975 PL_sub_generation = proto_perl->Isub_generation;
10976
10977 /* funky return mechanisms */
10978 PL_forkprocess = proto_perl->Iforkprocess;
10979
10980 /* subprocess state */
10981 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
10982
10983 /* internal state */
10984 PL_maxo = proto_perl->Imaxo;
10985 if (proto_perl->Iop_mask)
10986 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
10987 else
bd61b366 10988 PL_op_mask = NULL;
bd81e77b
NC
10989 /* PL_asserting = proto_perl->Iasserting; */
10990
10991 /* current interpreter roots */
10992 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
10993 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
10994 PL_main_start = proto_perl->Imain_start;
10995 PL_eval_root = proto_perl->Ieval_root;
10996 PL_eval_start = proto_perl->Ieval_start;
10997
10998 /* runtime control stuff */
10999 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11000 PL_copline = proto_perl->Icopline;
11001
11002 PL_filemode = proto_perl->Ifilemode;
11003 PL_lastfd = proto_perl->Ilastfd;
11004 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11005 PL_Argv = NULL;
bd61b366 11006 PL_Cmd = NULL;
bd81e77b
NC
11007 PL_gensym = proto_perl->Igensym;
11008 PL_preambled = proto_perl->Ipreambled;
11009 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11010 PL_laststatval = proto_perl->Ilaststatval;
11011 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11012 PL_mess_sv = NULL;
bd81e77b
NC
11013
11014 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11015
11016 /* interpreter atexit processing */
11017 PL_exitlistlen = proto_perl->Iexitlistlen;
11018 if (PL_exitlistlen) {
11019 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11020 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11021 }
bd81e77b
NC
11022 else
11023 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11024
11025 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11026 if (PL_my_cxt_size) {
f16dd614
DM
11027 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11028 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11029 }
11030 else
11031 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11032 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11033 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11034 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11035
11036 PL_profiledata = NULL;
11037 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11038 /* PL_rsfp_filters entries have fake IoDIRP() */
11039 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11040
bd81e77b 11041 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11042
bd81e77b 11043 PAD_CLONE_VARS(proto_perl, param);
9660f481 11044
bd81e77b
NC
11045#ifdef HAVE_INTERP_INTERN
11046 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11047#endif
645c22ef 11048
bd81e77b
NC
11049 /* more statics moved here */
11050 PL_generation = proto_perl->Igeneration;
11051 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11052
bd81e77b
NC
11053 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11054 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11055
bd81e77b
NC
11056 PL_uid = proto_perl->Iuid;
11057 PL_euid = proto_perl->Ieuid;
11058 PL_gid = proto_perl->Igid;
11059 PL_egid = proto_perl->Iegid;
11060 PL_nomemok = proto_perl->Inomemok;
11061 PL_an = proto_perl->Ian;
11062 PL_evalseq = proto_perl->Ievalseq;
11063 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11064 PL_origalen = proto_perl->Iorigalen;
11065#ifdef PERL_USES_PL_PIDSTATUS
11066 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11067#endif
11068 PL_osname = SAVEPV(proto_perl->Iosname);
11069 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11070
bd81e77b 11071 PL_runops = proto_perl->Irunops;
6a78b4db 11072
bd81e77b 11073 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11074
bd81e77b
NC
11075#ifdef CSH
11076 PL_cshlen = proto_perl->Icshlen;
11077 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11078#endif
645c22ef 11079
bd81e77b
NC
11080 PL_lex_state = proto_perl->Ilex_state;
11081 PL_lex_defer = proto_perl->Ilex_defer;
11082 PL_lex_expect = proto_perl->Ilex_expect;
11083 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11084 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11085 PL_lex_starts = proto_perl->Ilex_starts;
11086 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11087 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11088 PL_lex_op = proto_perl->Ilex_op;
11089 PL_lex_inpat = proto_perl->Ilex_inpat;
11090 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11091 PL_lex_brackets = proto_perl->Ilex_brackets;
11092 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11093 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11094 PL_lex_casemods = proto_perl->Ilex_casemods;
11095 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11096 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11097
5db06880
NC
11098#ifdef PERL_MAD
11099 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11100 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11101 PL_realtokenstart = proto_perl->Irealtokenstart;
11102 PL_faketokens = proto_perl->Ifaketokens;
11103 PL_thismad = proto_perl->Ithismad;
11104 PL_thistoken = proto_perl->Ithistoken;
11105 PL_thisopen = proto_perl->Ithisopen;
11106 PL_thisstuff = proto_perl->Ithisstuff;
11107 PL_thisclose = proto_perl->Ithisclose;
11108 PL_thiswhite = proto_perl->Ithiswhite;
11109 PL_nextwhite = proto_perl->Inextwhite;
11110 PL_skipwhite = proto_perl->Iskipwhite;
11111 PL_endwhite = proto_perl->Iendwhite;
11112 PL_curforce = proto_perl->Icurforce;
5db06880 11113#else
bd81e77b
NC
11114 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11115 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11116 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11117#endif
c43294b8 11118
bd81e77b
NC
11119 /* XXX This is probably masking the deeper issue of why
11120 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11121 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11122 * (A little debugging with a watchpoint on it may help.)
11123 */
11124 if (SvANY(proto_perl->Ilinestr)) {
11125 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11126 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11127 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11128 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11129 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11130 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11131 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11132 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11133 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11134 }
11135 else {
561b68a9 11136 PL_linestr = newSV(79);
bd81e77b
NC
11137 sv_upgrade(PL_linestr,SVt_PVIV);
11138 sv_setpvn(PL_linestr,"",0);
11139 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11140 }
11141 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11142 PL_pending_ident = proto_perl->Ipending_ident;
11143 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11144
11145 PL_expect = proto_perl->Iexpect;
11146
11147 PL_multi_start = proto_perl->Imulti_start;
11148 PL_multi_end = proto_perl->Imulti_end;
11149 PL_multi_open = proto_perl->Imulti_open;
11150 PL_multi_close = proto_perl->Imulti_close;
11151
11152 PL_error_count = proto_perl->Ierror_count;
11153 PL_subline = proto_perl->Isubline;
11154 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11155
bd81e77b
NC
11156 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11157 if (SvANY(proto_perl->Ilinestr)) {
11158 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11159 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11160 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11161 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11162 PL_last_lop_op = proto_perl->Ilast_lop_op;
11163 }
11164 else {
11165 PL_last_uni = SvPVX(PL_linestr);
11166 PL_last_lop = SvPVX(PL_linestr);
11167 PL_last_lop_op = 0;
11168 }
11169 PL_in_my = proto_perl->Iin_my;
11170 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11171#ifdef FCRYPT
11172 PL_cryptseen = proto_perl->Icryptseen;
11173#endif
1d7c1841 11174
bd81e77b 11175 PL_hints = proto_perl->Ihints;
1d7c1841 11176
bd81e77b 11177 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11178
bd81e77b
NC
11179#ifdef USE_LOCALE_COLLATE
11180 PL_collation_ix = proto_perl->Icollation_ix;
11181 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11182 PL_collation_standard = proto_perl->Icollation_standard;
11183 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11184 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11185#endif /* USE_LOCALE_COLLATE */
1d7c1841 11186
bd81e77b
NC
11187#ifdef USE_LOCALE_NUMERIC
11188 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11189 PL_numeric_standard = proto_perl->Inumeric_standard;
11190 PL_numeric_local = proto_perl->Inumeric_local;
11191 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11192#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11193
bd81e77b
NC
11194 /* utf8 character classes */
11195 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11196 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11197 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11198 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11199 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11200 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11201 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11202 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11203 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11204 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11205 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11206 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11207 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11208 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11209 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11210 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11211 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11212 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11213 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11214 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11215
bd81e77b
NC
11216 /* Did the locale setup indicate UTF-8? */
11217 PL_utf8locale = proto_perl->Iutf8locale;
11218 /* Unicode features (see perlrun/-C) */
11219 PL_unicode = proto_perl->Iunicode;
1d7c1841 11220
bd81e77b
NC
11221 /* Pre-5.8 signals control */
11222 PL_signals = proto_perl->Isignals;
1d7c1841 11223
bd81e77b
NC
11224 /* times() ticks per second */
11225 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11226
bd81e77b
NC
11227 /* Recursion stopper for PerlIO_find_layer */
11228 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11229
bd81e77b
NC
11230 /* sort() routine */
11231 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11232
bd81e77b
NC
11233 /* Not really needed/useful since the reenrant_retint is "volatile",
11234 * but do it for consistency's sake. */
11235 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11236
bd81e77b
NC
11237 /* Hooks to shared SVs and locks. */
11238 PL_sharehook = proto_perl->Isharehook;
11239 PL_lockhook = proto_perl->Ilockhook;
11240 PL_unlockhook = proto_perl->Iunlockhook;
11241 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11242
bd81e77b
NC
11243 PL_runops_std = proto_perl->Irunops_std;
11244 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11245
bd81e77b
NC
11246#ifdef THREADS_HAVE_PIDS
11247 PL_ppid = proto_perl->Ippid;
11248#endif
1d7c1841 11249
bd81e77b 11250 /* swatch cache */
5c284bb0 11251 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11252 PL_last_swash_klen = 0;
11253 PL_last_swash_key[0]= '\0';
11254 PL_last_swash_tmps = (U8*)NULL;
11255 PL_last_swash_slen = 0;
1d7c1841 11256
bd81e77b
NC
11257 PL_glob_index = proto_perl->Iglob_index;
11258 PL_srand_called = proto_perl->Isrand_called;
11b79775 11259 PL_uudmap[(U32) 'M'] = 0; /* reinits on demand */
bd61b366 11260 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11261
bd81e77b
NC
11262 if (proto_perl->Ipsig_pend) {
11263 Newxz(PL_psig_pend, SIG_SIZE, int);
11264 }
11265 else {
11266 PL_psig_pend = (int*)NULL;
11267 }
05ec9bb3 11268
bd81e77b
NC
11269 if (proto_perl->Ipsig_ptr) {
11270 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11271 Newxz(PL_psig_name, SIG_SIZE, SV*);
11272 for (i = 1; i < SIG_SIZE; i++) {
11273 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11274 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11275 }
11276 }
11277 else {
11278 PL_psig_ptr = (SV**)NULL;
11279 PL_psig_name = (SV**)NULL;
11280 }
05ec9bb3 11281
bd81e77b 11282 /* thrdvar.h stuff */
1d7c1841 11283
bd81e77b
NC
11284 if (flags & CLONEf_COPY_STACKS) {
11285 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11286 PL_tmps_ix = proto_perl->Ttmps_ix;
11287 PL_tmps_max = proto_perl->Ttmps_max;
11288 PL_tmps_floor = proto_perl->Ttmps_floor;
11289 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11290 i = 0;
11291 while (i <= PL_tmps_ix) {
11292 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11293 ++i;
11294 }
d2d73c3e 11295
bd81e77b
NC
11296 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11297 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11298 Newxz(PL_markstack, i, I32);
11299 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11300 - proto_perl->Tmarkstack);
11301 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11302 - proto_perl->Tmarkstack);
11303 Copy(proto_perl->Tmarkstack, PL_markstack,
11304 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11305
bd81e77b
NC
11306 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11307 * NOTE: unlike the others! */
11308 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11309 PL_scopestack_max = proto_perl->Tscopestack_max;
11310 Newxz(PL_scopestack, PL_scopestack_max, I32);
11311 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11312
bd81e77b
NC
11313 /* NOTE: si_dup() looks at PL_markstack */
11314 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11315
bd81e77b
NC
11316 /* PL_curstack = PL_curstackinfo->si_stack; */
11317 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11318 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11319
bd81e77b
NC
11320 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11321 PL_stack_base = AvARRAY(PL_curstack);
11322 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11323 - proto_perl->Tstack_base);
11324 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11325
bd81e77b
NC
11326 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11327 * NOTE: unlike the others! */
11328 PL_savestack_ix = proto_perl->Tsavestack_ix;
11329 PL_savestack_max = proto_perl->Tsavestack_max;
11330 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11331 PL_savestack = ss_dup(proto_perl, param);
11332 }
11333 else {
11334 init_stacks();
11335 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11336
11337 /* although we're not duplicating the tmps stack, we should still
11338 * add entries for any SVs on the tmps stack that got cloned by a
11339 * non-refcount means (eg a temp in @_); otherwise they will be
11340 * orphaned
11341 */
11342 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11343 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11344 proto_perl->Ttmps_stack[i]);
11345 if (nsv && !SvREFCNT(nsv)) {
11346 EXTEND_MORTAL(1);
b37c2d43 11347 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11348 }
11349 }
bd81e77b 11350 }
1d7c1841 11351
bd81e77b
NC
11352 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11353 PL_top_env = &PL_start_env;
1d7c1841 11354
bd81e77b 11355 PL_op = proto_perl->Top;
4a4c6fe3 11356
a0714e2c 11357 PL_Sv = NULL;
bd81e77b
NC
11358 PL_Xpv = (XPV*)NULL;
11359 PL_na = proto_perl->Tna;
1fcf4c12 11360
bd81e77b
NC
11361 PL_statbuf = proto_perl->Tstatbuf;
11362 PL_statcache = proto_perl->Tstatcache;
11363 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11364 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11365#ifdef HAS_TIMES
11366 PL_timesbuf = proto_perl->Ttimesbuf;
11367#endif
1d7c1841 11368
bd81e77b
NC
11369 PL_tainted = proto_perl->Ttainted;
11370 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11371 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11372 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11373 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11374 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11375 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11376 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11377 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11378 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11379
bd81e77b
NC
11380 PL_restartop = proto_perl->Trestartop;
11381 PL_in_eval = proto_perl->Tin_eval;
11382 PL_delaymagic = proto_perl->Tdelaymagic;
11383 PL_dirty = proto_perl->Tdirty;
11384 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11385
bd81e77b 11386 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11387 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11388 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11389 PL_lastgotoprobe = NULL;
bd81e77b 11390 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11391
bd81e77b
NC
11392 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11393 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11394 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11395 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11396 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11397 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11398
bd81e77b 11399 /* regex stuff */
1d7c1841 11400
bd81e77b
NC
11401 PL_screamfirst = NULL;
11402 PL_screamnext = NULL;
11403 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11404 PL_lastscream = NULL;
1d7c1841 11405
bd81e77b 11406 PL_watchaddr = NULL;
bd61b366 11407 PL_watchok = NULL;
1d7c1841 11408
bd81e77b 11409 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11410 PL_colorset = 0; /* reinits PL_colors[] */
11411 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11412
84da74a7 11413
1d7c1841 11414
bd81e77b
NC
11415 /* Pluggable optimizer */
11416 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11417
bd81e77b 11418 PL_stashcache = newHV();
1d7c1841 11419
bd81e77b
NC
11420 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11421 ptr_table_free(PL_ptr_table);
11422 PL_ptr_table = NULL;
11423 }
1d7c1841 11424
bd81e77b
NC
11425 /* Call the ->CLONE method, if it exists, for each of the stashes
11426 identified by sv_dup() above.
11427 */
11428 while(av_len(param->stashes) != -1) {
11429 HV* const stash = (HV*) av_shift(param->stashes);
11430 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11431 if (cloner && GvCV(cloner)) {
11432 dSP;
11433 ENTER;
11434 SAVETMPS;
11435 PUSHMARK(SP);
11436 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11437 PUTBACK;
11438 call_sv((SV*)GvCV(cloner), G_DISCARD);
11439 FREETMPS;
11440 LEAVE;
11441 }
1d7c1841 11442 }
1d7c1841 11443
bd81e77b 11444 SvREFCNT_dec(param->stashes);
1d7c1841 11445
bd81e77b
NC
11446 /* orphaned? eg threads->new inside BEGIN or use */
11447 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11448 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11449 SAVEFREESV(PL_compcv);
11450 }
dd2155a4 11451
bd81e77b
NC
11452 return my_perl;
11453}
1d7c1841 11454
bd81e77b 11455#endif /* USE_ITHREADS */
1d7c1841 11456
bd81e77b
NC
11457/*
11458=head1 Unicode Support
1d7c1841 11459
bd81e77b 11460=for apidoc sv_recode_to_utf8
1d7c1841 11461
bd81e77b
NC
11462The encoding is assumed to be an Encode object, on entry the PV
11463of the sv is assumed to be octets in that encoding, and the sv
11464will be converted into Unicode (and UTF-8).
1d7c1841 11465
bd81e77b
NC
11466If the sv already is UTF-8 (or if it is not POK), or if the encoding
11467is not a reference, nothing is done to the sv. If the encoding is not
11468an C<Encode::XS> Encoding object, bad things will happen.
11469(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11470
bd81e77b 11471The PV of the sv is returned.
1d7c1841 11472
bd81e77b 11473=cut */
1d7c1841 11474
bd81e77b
NC
11475char *
11476Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11477{
11478 dVAR;
11479 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11480 SV *uni;
11481 STRLEN len;
11482 const char *s;
11483 dSP;
11484 ENTER;
11485 SAVETMPS;
11486 save_re_context();
11487 PUSHMARK(sp);
11488 EXTEND(SP, 3);
11489 XPUSHs(encoding);
11490 XPUSHs(sv);
11491/*
11492 NI-S 2002/07/09
11493 Passing sv_yes is wrong - it needs to be or'ed set of constants
11494 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11495 remove converted chars from source.
1d7c1841 11496
bd81e77b 11497 Both will default the value - let them.
1d7c1841 11498
bd81e77b
NC
11499 XPUSHs(&PL_sv_yes);
11500*/
11501 PUTBACK;
11502 call_method("decode", G_SCALAR);
11503 SPAGAIN;
11504 uni = POPs;
11505 PUTBACK;
11506 s = SvPV_const(uni, len);
11507 if (s != SvPVX_const(sv)) {
11508 SvGROW(sv, len + 1);
11509 Move(s, SvPVX(sv), len + 1, char);
11510 SvCUR_set(sv, len);
11511 }
11512 FREETMPS;
11513 LEAVE;
11514 SvUTF8_on(sv);
11515 return SvPVX(sv);
389edf32 11516 }
bd81e77b
NC
11517 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11518}
1d7c1841 11519
bd81e77b
NC
11520/*
11521=for apidoc sv_cat_decode
1d7c1841 11522
bd81e77b
NC
11523The encoding is assumed to be an Encode object, the PV of the ssv is
11524assumed to be octets in that encoding and decoding the input starts
11525from the position which (PV + *offset) pointed to. The dsv will be
11526concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11527when the string tstr appears in decoding output or the input ends on
11528the PV of the ssv. The value which the offset points will be modified
11529to the last input position on the ssv.
1d7c1841 11530
bd81e77b 11531Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11532
bd81e77b
NC
11533=cut */
11534
11535bool
11536Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11537 SV *ssv, int *offset, char *tstr, int tlen)
11538{
11539 dVAR;
11540 bool ret = FALSE;
11541 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11542 SV *offsv;
11543 dSP;
11544 ENTER;
11545 SAVETMPS;
11546 save_re_context();
11547 PUSHMARK(sp);
11548 EXTEND(SP, 6);
11549 XPUSHs(encoding);
11550 XPUSHs(dsv);
11551 XPUSHs(ssv);
11552 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11553 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11554 PUTBACK;
11555 call_method("cat_decode", G_SCALAR);
11556 SPAGAIN;
11557 ret = SvTRUE(TOPs);
11558 *offset = SvIV(offsv);
11559 PUTBACK;
11560 FREETMPS;
11561 LEAVE;
389edf32 11562 }
bd81e77b
NC
11563 else
11564 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11565 return ret;
1d7c1841 11566
bd81e77b 11567}
1d7c1841 11568
bd81e77b
NC
11569/* ---------------------------------------------------------------------
11570 *
11571 * support functions for report_uninit()
11572 */
1d7c1841 11573
bd81e77b
NC
11574/* the maxiumum size of array or hash where we will scan looking
11575 * for the undefined element that triggered the warning */
1d7c1841 11576
bd81e77b 11577#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11578
bd81e77b
NC
11579/* Look for an entry in the hash whose value has the same SV as val;
11580 * If so, return a mortal copy of the key. */
1d7c1841 11581
bd81e77b
NC
11582STATIC SV*
11583S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11584{
11585 dVAR;
11586 register HE **array;
11587 I32 i;
6c3182a5 11588
bd81e77b
NC
11589 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11590 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11591 return NULL;
6c3182a5 11592
bd81e77b 11593 array = HvARRAY(hv);
6c3182a5 11594
bd81e77b
NC
11595 for (i=HvMAX(hv); i>0; i--) {
11596 register HE *entry;
11597 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11598 if (HeVAL(entry) != val)
11599 continue;
11600 if ( HeVAL(entry) == &PL_sv_undef ||
11601 HeVAL(entry) == &PL_sv_placeholder)
11602 continue;
11603 if (!HeKEY(entry))
a0714e2c 11604 return NULL;
bd81e77b
NC
11605 if (HeKLEN(entry) == HEf_SVKEY)
11606 return sv_mortalcopy(HeKEY_sv(entry));
11607 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11608 }
11609 }
a0714e2c 11610 return NULL;
bd81e77b 11611}
6c3182a5 11612
bd81e77b
NC
11613/* Look for an entry in the array whose value has the same SV as val;
11614 * If so, return the index, otherwise return -1. */
6c3182a5 11615
bd81e77b
NC
11616STATIC I32
11617S_find_array_subscript(pTHX_ AV *av, SV* val)
11618{
97aff369 11619 dVAR;
bd81e77b
NC
11620 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11621 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11622 return -1;
57c6e6d2 11623
4a021917
AL
11624 if (val != &PL_sv_undef) {
11625 SV ** const svp = AvARRAY(av);
11626 I32 i;
11627
11628 for (i=AvFILLp(av); i>=0; i--)
11629 if (svp[i] == val)
11630 return i;
bd81e77b
NC
11631 }
11632 return -1;
11633}
15a5279a 11634
bd81e77b
NC
11635/* S_varname(): return the name of a variable, optionally with a subscript.
11636 * If gv is non-zero, use the name of that global, along with gvtype (one
11637 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11638 * targ. Depending on the value of the subscript_type flag, return:
11639 */
bce260cd 11640
bd81e77b
NC
11641#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11642#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11643#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11644#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11645
bd81e77b
NC
11646STATIC SV*
11647S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11648 SV* keyname, I32 aindex, int subscript_type)
11649{
1d7c1841 11650
bd81e77b
NC
11651 SV * const name = sv_newmortal();
11652 if (gv) {
11653 char buffer[2];
11654 buffer[0] = gvtype;
11655 buffer[1] = 0;
1d7c1841 11656
bd81e77b 11657 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11658
bd81e77b 11659 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11660
bd81e77b
NC
11661 if ((unsigned int)SvPVX(name)[1] <= 26) {
11662 buffer[0] = '^';
11663 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11664
bd81e77b
NC
11665 /* Swap the 1 unprintable control character for the 2 byte pretty
11666 version - ie substr($name, 1, 1) = $buffer; */
11667 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11668 }
bd81e77b
NC
11669 }
11670 else {
11671 U32 unused;
11672 CV * const cv = find_runcv(&unused);
11673 SV *sv;
11674 AV *av;
1d7c1841 11675
bd81e77b 11676 if (!cv || !CvPADLIST(cv))
a0714e2c 11677 return NULL;
bd81e77b
NC
11678 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11679 sv = *av_fetch(av, targ, FALSE);
11680 /* SvLEN in a pad name is not to be trusted */
11681 sv_setpv(name, SvPV_nolen_const(sv));
11682 }
1d7c1841 11683
bd81e77b 11684 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11685 SV * const sv = newSV(0);
bd81e77b
NC
11686 *SvPVX(name) = '$';
11687 Perl_sv_catpvf(aTHX_ name, "{%s}",
11688 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11689 SvREFCNT_dec(sv);
11690 }
11691 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11692 *SvPVX(name) = '$';
11693 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11694 }
11695 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11696 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11697
bd81e77b
NC
11698 return name;
11699}
1d7c1841 11700
1d7c1841 11701
bd81e77b
NC
11702/*
11703=for apidoc find_uninit_var
1d7c1841 11704
bd81e77b
NC
11705Find the name of the undefined variable (if any) that caused the operator o
11706to issue a "Use of uninitialized value" warning.
11707If match is true, only return a name if it's value matches uninit_sv.
11708So roughly speaking, if a unary operator (such as OP_COS) generates a
11709warning, then following the direct child of the op may yield an
11710OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11711other hand, with OP_ADD there are two branches to follow, so we only print
11712the variable name if we get an exact match.
1d7c1841 11713
bd81e77b 11714The name is returned as a mortal SV.
1d7c1841 11715
bd81e77b
NC
11716Assumes that PL_op is the op that originally triggered the error, and that
11717PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11718
bd81e77b
NC
11719=cut
11720*/
1d7c1841 11721
bd81e77b
NC
11722STATIC SV *
11723S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11724{
11725 dVAR;
11726 SV *sv;
11727 AV *av;
11728 GV *gv;
11729 OP *o, *o2, *kid;
1d7c1841 11730
bd81e77b
NC
11731 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11732 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11733 return NULL;
1d7c1841 11734
bd81e77b 11735 switch (obase->op_type) {
1d7c1841 11736
bd81e77b
NC
11737 case OP_RV2AV:
11738 case OP_RV2HV:
11739 case OP_PADAV:
11740 case OP_PADHV:
11741 {
11742 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11743 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11744 I32 index = 0;
a0714e2c 11745 SV *keysv = NULL;
bd81e77b 11746 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11747
bd81e77b
NC
11748 if (pad) { /* @lex, %lex */
11749 sv = PAD_SVl(obase->op_targ);
a0714e2c 11750 gv = NULL;
bd81e77b
NC
11751 }
11752 else {
11753 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11754 /* @global, %global */
11755 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11756 if (!gv)
11757 break;
11758 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11759 }
11760 else /* @{expr}, %{expr} */
11761 return find_uninit_var(cUNOPx(obase)->op_first,
11762 uninit_sv, match);
11763 }
1d7c1841 11764
bd81e77b
NC
11765 /* attempt to find a match within the aggregate */
11766 if (hash) {
d4c19fe8 11767 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11768 if (keysv)
11769 subscript_type = FUV_SUBSCRIPT_HASH;
11770 }
11771 else {
e15d5972 11772 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11773 if (index >= 0)
11774 subscript_type = FUV_SUBSCRIPT_ARRAY;
11775 }
1d7c1841 11776
bd81e77b
NC
11777 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11778 break;
1d7c1841 11779
bd81e77b
NC
11780 return varname(gv, hash ? '%' : '@', obase->op_targ,
11781 keysv, index, subscript_type);
11782 }
1d7c1841 11783
bd81e77b
NC
11784 case OP_PADSV:
11785 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11786 break;
a0714e2c
SS
11787 return varname(NULL, '$', obase->op_targ,
11788 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11789
bd81e77b
NC
11790 case OP_GVSV:
11791 gv = cGVOPx_gv(obase);
11792 if (!gv || (match && GvSV(gv) != uninit_sv))
11793 break;
a0714e2c 11794 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11795
bd81e77b
NC
11796 case OP_AELEMFAST:
11797 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11798 if (match) {
11799 SV **svp;
11800 av = (AV*)PAD_SV(obase->op_targ);
11801 if (!av || SvRMAGICAL(av))
11802 break;
11803 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11804 if (!svp || *svp != uninit_sv)
11805 break;
11806 }
a0714e2c
SS
11807 return varname(NULL, '$', obase->op_targ,
11808 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11809 }
11810 else {
11811 gv = cGVOPx_gv(obase);
11812 if (!gv)
11813 break;
11814 if (match) {
11815 SV **svp;
11816 av = GvAV(gv);
11817 if (!av || SvRMAGICAL(av))
11818 break;
11819 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11820 if (!svp || *svp != uninit_sv)
11821 break;
11822 }
11823 return varname(gv, '$', 0,
a0714e2c 11824 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11825 }
11826 break;
1d7c1841 11827
bd81e77b
NC
11828 case OP_EXISTS:
11829 o = cUNOPx(obase)->op_first;
11830 if (!o || o->op_type != OP_NULL ||
11831 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11832 break;
11833 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11834
bd81e77b
NC
11835 case OP_AELEM:
11836 case OP_HELEM:
11837 if (PL_op == obase)
11838 /* $a[uninit_expr] or $h{uninit_expr} */
11839 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11840
a0714e2c 11841 gv = NULL;
bd81e77b
NC
11842 o = cBINOPx(obase)->op_first;
11843 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11844
bd81e77b 11845 /* get the av or hv, and optionally the gv */
a0714e2c 11846 sv = NULL;
bd81e77b
NC
11847 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11848 sv = PAD_SV(o->op_targ);
11849 }
11850 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11851 && cUNOPo->op_first->op_type == OP_GV)
11852 {
11853 gv = cGVOPx_gv(cUNOPo->op_first);
11854 if (!gv)
11855 break;
11856 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11857 }
11858 if (!sv)
11859 break;
11860
11861 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11862 /* index is constant */
11863 if (match) {
11864 if (SvMAGICAL(sv))
11865 break;
11866 if (obase->op_type == OP_HELEM) {
11867 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11868 if (!he || HeVAL(he) != uninit_sv)
11869 break;
11870 }
11871 else {
00b6aa41 11872 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11873 if (!svp || *svp != uninit_sv)
11874 break;
11875 }
11876 }
11877 if (obase->op_type == OP_HELEM)
11878 return varname(gv, '%', o->op_targ,
11879 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11880 else
a0714e2c 11881 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11882 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11883 }
11884 else {
11885 /* index is an expression;
11886 * attempt to find a match within the aggregate */
11887 if (obase->op_type == OP_HELEM) {
d4c19fe8 11888 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11889 if (keysv)
11890 return varname(gv, '%', o->op_targ,
11891 keysv, 0, FUV_SUBSCRIPT_HASH);
11892 }
11893 else {
d4c19fe8 11894 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11895 if (index >= 0)
11896 return varname(gv, '@', o->op_targ,
a0714e2c 11897 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11898 }
11899 if (match)
11900 break;
11901 return varname(gv,
11902 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
11903 ? '@' : '%',
a0714e2c 11904 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 11905 }
bd81e77b 11906 break;
dc507217 11907
bd81e77b
NC
11908 case OP_AASSIGN:
11909 /* only examine RHS */
11910 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 11911
bd81e77b
NC
11912 case OP_OPEN:
11913 o = cUNOPx(obase)->op_first;
11914 if (o->op_type == OP_PUSHMARK)
11915 o = o->op_sibling;
1d7c1841 11916
bd81e77b
NC
11917 if (!o->op_sibling) {
11918 /* one-arg version of open is highly magical */
a0ae6670 11919
bd81e77b
NC
11920 if (o->op_type == OP_GV) { /* open FOO; */
11921 gv = cGVOPx_gv(o);
11922 if (match && GvSV(gv) != uninit_sv)
11923 break;
11924 return varname(gv, '$', 0,
a0714e2c 11925 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
11926 }
11927 /* other possibilities not handled are:
11928 * open $x; or open my $x; should return '${*$x}'
11929 * open expr; should return '$'.expr ideally
11930 */
11931 break;
11932 }
11933 goto do_op;
ccfc67b7 11934
bd81e77b
NC
11935 /* ops where $_ may be an implicit arg */
11936 case OP_TRANS:
11937 case OP_SUBST:
11938 case OP_MATCH:
11939 if ( !(obase->op_flags & OPf_STACKED)) {
11940 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
11941 ? PAD_SVl(obase->op_targ)
11942 : DEFSV))
11943 {
11944 sv = sv_newmortal();
11945 sv_setpvn(sv, "$_", 2);
11946 return sv;
11947 }
11948 }
11949 goto do_op;
9f4817db 11950
bd81e77b
NC
11951 case OP_PRTF:
11952 case OP_PRINT:
11953 /* skip filehandle as it can't produce 'undef' warning */
11954 o = cUNOPx(obase)->op_first;
11955 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
11956 o = o->op_sibling->op_sibling;
11957 goto do_op2;
9f4817db 11958
9f4817db 11959
bd81e77b
NC
11960 case OP_RV2SV:
11961 case OP_CUSTOM:
11962 case OP_ENTERSUB:
11963 match = 1; /* XS or custom code could trigger random warnings */
11964 goto do_op;
9f4817db 11965
bd81e77b
NC
11966 case OP_SCHOMP:
11967 case OP_CHOMP:
11968 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 11969 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 11970 /*FALLTHROUGH*/
5d170f3a 11971
bd81e77b
NC
11972 default:
11973 do_op:
11974 if (!(obase->op_flags & OPf_KIDS))
11975 break;
11976 o = cUNOPx(obase)->op_first;
11977
11978 do_op2:
11979 if (!o)
11980 break;
f9893866 11981
bd81e77b
NC
11982 /* if all except one arg are constant, or have no side-effects,
11983 * or are optimized away, then it's unambiguous */
5f66b61c 11984 o2 = NULL;
bd81e77b 11985 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
11986 if (kid) {
11987 const OPCODE type = kid->op_type;
11988 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
11989 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
11990 || (type == OP_PUSHMARK)
bd81e77b 11991 )
bd81e77b 11992 continue;
e15d5972 11993 }
bd81e77b 11994 if (o2) { /* more than one found */
5f66b61c 11995 o2 = NULL;
bd81e77b
NC
11996 break;
11997 }
11998 o2 = kid;
11999 }
12000 if (o2)
12001 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12002
bd81e77b
NC
12003 /* scan all args */
12004 while (o) {
12005 sv = find_uninit_var(o, uninit_sv, 1);
12006 if (sv)
12007 return sv;
12008 o = o->op_sibling;
d0063567 12009 }
bd81e77b 12010 break;
f9893866 12011 }
a0714e2c 12012 return NULL;
9f4817db
JH
12013}
12014
220e2d4e 12015
bd81e77b
NC
12016/*
12017=for apidoc report_uninit
68795e93 12018
bd81e77b 12019Print appropriate "Use of uninitialized variable" warning
220e2d4e 12020
bd81e77b
NC
12021=cut
12022*/
220e2d4e 12023
bd81e77b
NC
12024void
12025Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12026{
97aff369 12027 dVAR;
bd81e77b 12028 if (PL_op) {
a0714e2c 12029 SV* varname = NULL;
bd81e77b
NC
12030 if (uninit_sv) {
12031 varname = find_uninit_var(PL_op, uninit_sv,0);
12032 if (varname)
12033 sv_insert(varname, 0, 0, " ", 1);
12034 }
12035 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12036 varname ? SvPV_nolen_const(varname) : "",
12037 " in ", OP_DESC(PL_op));
220e2d4e 12038 }
a73e8557 12039 else
bd81e77b
NC
12040 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12041 "", "", "");
220e2d4e 12042}
f9893866 12043
241d1a3b
NC
12044/*
12045 * Local variables:
12046 * c-indentation-style: bsd
12047 * c-basic-offset: 4
12048 * indent-tabs-mode: t
12049 * End:
12050 *
37442d52
RGS
12051 * ex: set ts=8 sts=4 sw=4 noet:
12052 */