This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Upgrade to Encode-2.15
[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{
5e258f8c 681 struct arena_desc* adesc;
476a1e16 682 struct arena_set *newroot, **aroot = (struct arena_set**) &PL_body_arenas;
5e258f8c
JC
683 int curr;
684
476a1e16
JC
685 /* shouldnt need this
686 if (!arena_size) arena_size = PERL_ARENA_SIZE;
687 */
5e258f8c
JC
688
689 /* may need new arena-set to hold new arena */
476a1e16 690 if (!*aroot || (*aroot)->curr >= (*aroot)->set_size) {
5e258f8c
JC
691 Newxz(newroot, 1, struct arena_set);
692 newroot->set_size = ARENAS_PER_SET;
476a1e16
JC
693 newroot->next = *aroot;
694 *aroot = newroot;
695 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", *aroot));
5e258f8c
JC
696 }
697
698 /* ok, now have arena-set with at least 1 empty/available arena-desc */
476a1e16
JC
699 curr = (*aroot)->curr++;
700 adesc = &((*aroot)->set[curr]);
5e258f8c
JC
701 assert(!adesc->arena);
702
5e258f8c
JC
703 Newxz(adesc->arena, arena_size, char);
704 adesc->size = arena_size;
d2a0f284
JC
705 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
706 curr, adesc->arena, arena_size));
5e258f8c
JC
707
708 return adesc->arena;
5e258f8c
JC
709}
710
53c1dcc0 711
bd81e77b 712/* return a thing to the free list */
29489e7c 713
bd81e77b
NC
714#define del_body(thing, root) \
715 STMT_START { \
00b6aa41 716 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
717 LOCK_SV_MUTEX; \
718 *thing_copy = *root; \
719 *root = (void*)thing_copy; \
720 UNLOCK_SV_MUTEX; \
721 } STMT_END
29489e7c 722
bd81e77b 723/*
d2a0f284
JC
724
725=head1 SV-Body Allocation
726
727Allocation of SV-bodies is similar to SV-heads, differing as follows;
728the allocation mechanism is used for many body types, so is somewhat
729more complicated, it uses arena-sets, and has no need for still-live
730SV detection.
731
732At the outermost level, (new|del)_X*V macros return bodies of the
733appropriate type. These macros call either (new|del)_body_type or
734(new|del)_body_allocated macro pairs, depending on specifics of the
735type. Most body types use the former pair, the latter pair is used to
736allocate body types with "ghost fields".
737
738"ghost fields" are fields that are unused in certain types, and
739consequently dont need to actually exist. They are declared because
740they're part of a "base type", which allows use of functions as
741methods. The simplest examples are AVs and HVs, 2 aggregate types
742which don't use the fields which support SCALAR semantics.
743
744For these types, the arenas are carved up into *_allocated size
745chunks, we thus avoid wasted memory for those unaccessed members.
746When bodies are allocated, we adjust the pointer back in memory by the
747size of the bit not allocated, so it's as if we allocated the full
748structure. (But things will all go boom if you write to the part that
749is "not there", because you'll be overwriting the last members of the
750preceding structure in memory.)
751
752We calculate the correction using the STRUCT_OFFSET macro. For
753example, if xpv_allocated is the same structure as XPV then the two
754OFFSETs sum to zero, and the pointer is unchanged. If the allocated
755structure is smaller (no initial NV actually allocated) then the net
756effect is to subtract the size of the NV from the pointer, to return a
757new pointer as if an initial NV were actually allocated.
758
759This is the same trick as was used for NV and IV bodies. Ironically it
760doesn't need to be used for NV bodies any more, because NV is now at
761the start of the structure. IV bodies don't need it either, because
762they are no longer allocated.
763
764In turn, the new_body_* allocators call S_new_body(), which invokes
765new_body_inline macro, which takes a lock, and takes a body off the
766linked list at PL_body_roots[sv_type], calling S_more_bodies() if
767necessary to refresh an empty list. Then the lock is released, and
768the body is returned.
769
770S_more_bodies calls get_arena(), and carves it up into an array of N
771bodies, which it strings into a linked list. It looks up arena-size
772and body-size from the body_details table described below, thus
773supporting the multiple body-types.
774
775If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
776the (new|del)_X*V macros are mapped directly to malloc/free.
777
778*/
779
780/*
781
782For each sv-type, struct body_details bodies_by_type[] carries
783parameters which control these aspects of SV handling:
784
785Arena_size determines whether arenas are used for this body type, and if
786so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
787zero, forcing individual mallocs and frees.
788
789Body_size determines how big a body is, and therefore how many fit into
790each arena. Offset carries the body-pointer adjustment needed for
791*_allocated body types, and is used in *_allocated macros.
792
793But its main purpose is to parameterize info needed in
794Perl_sv_upgrade(). The info here dramatically simplifies the function
795vs the implementation in 5.8.7, making it table-driven. All fields
796are used for this, except for arena_size.
797
798For the sv-types that have no bodies, arenas are not used, so those
799PL_body_roots[sv_type] are unused, and can be overloaded. In
800something of a special case, SVt_NULL is borrowed for HE arenas;
801PL_body_roots[SVt_NULL] is filled by S_more_he, but the
802bodies_by_type[SVt_NULL] slot is not used, as the table is not
803available in hv.c,
804
805PTEs also use arenas, but are never seen in Perl_sv_upgrade.
806Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
807they can just use the same allocation semantics. At first, PTEs were
808also overloaded to a non-body sv-type, but this yielded hard-to-find
809malloc bugs, so was simplified by claiming a new slot. This choice
810has no consequence at this time.
811
29489e7c
DM
812*/
813
bd81e77b 814struct body_details {
0fb58b32 815 U8 body_size; /* Size to allocate */
10666ae3 816 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 817 U8 offset;
10666ae3
NC
818 unsigned int type : 4; /* We have space for a sanity check. */
819 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
820 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
821 unsigned int arena : 1; /* Allocated from an arena */
822 size_t arena_size; /* Size of arena to allocate */
bd81e77b 823};
29489e7c 824
bd81e77b
NC
825#define HADNV FALSE
826#define NONV TRUE
29489e7c 827
d2a0f284 828
bd81e77b
NC
829#ifdef PURIFY
830/* With -DPURFIY we allocate everything directly, and don't use arenas.
831 This seems a rather elegant way to simplify some of the code below. */
832#define HASARENA FALSE
833#else
834#define HASARENA TRUE
835#endif
836#define NOARENA FALSE
29489e7c 837
d2a0f284
JC
838/* Size the arenas to exactly fit a given number of bodies. A count
839 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
840 simplifying the default. If count > 0, the arena is sized to fit
841 only that many bodies, allowing arenas to be used for large, rare
842 bodies (XPVFM, XPVIO) without undue waste. The arena size is
843 limited by PERL_ARENA_SIZE, so we can safely oversize the
844 declarations.
845 */
95db5f15
MB
846#define FIT_ARENA0(body_size) \
847 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
848#define FIT_ARENAn(count,body_size) \
849 ( count * body_size <= PERL_ARENA_SIZE) \
850 ? count * body_size \
851 : FIT_ARENA0 (body_size)
852#define FIT_ARENA(count,body_size) \
853 count \
854 ? FIT_ARENAn (count, body_size) \
855 : FIT_ARENA0 (body_size)
d2a0f284 856
bd81e77b 857/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 858
bd81e77b
NC
859typedef struct {
860 STRLEN xpv_cur;
861 STRLEN xpv_len;
862} xpv_allocated;
29489e7c 863
bd81e77b 864to make its members accessible via a pointer to (say)
29489e7c 865
bd81e77b
NC
866struct xpv {
867 NV xnv_nv;
868 STRLEN xpv_cur;
869 STRLEN xpv_len;
870};
29489e7c 871
bd81e77b 872*/
29489e7c 873
bd81e77b
NC
874#define relative_STRUCT_OFFSET(longer, shorter, member) \
875 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 876
bd81e77b
NC
877/* Calculate the length to copy. Specifically work out the length less any
878 final padding the compiler needed to add. See the comment in sv_upgrade
879 for why copying the padding proved to be a bug. */
29489e7c 880
bd81e77b
NC
881#define copy_length(type, last_member) \
882 STRUCT_OFFSET(type, last_member) \
883 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 884
bd81e77b 885static const struct body_details bodies_by_type[] = {
10666ae3
NC
886 { sizeof(HE), 0, 0, SVt_NULL,
887 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284
JC
888
889 /* IVs are in the head, so the allocation size is 0.
890 However, the slot is overloaded for PTEs. */
891 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
892 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 893 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
894 NOARENA /* IVS don't need an arena */,
895 /* But PTEs need to know the size of their arena */
896 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
897 },
898
bd81e77b 899 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 900 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
901 FIT_ARENA(0, sizeof(NV)) },
902
903 /* RVs are in the head now. */
10666ae3 904 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 905
bd81e77b 906 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
907 { sizeof(xpv_allocated),
908 copy_length(XPV, xpv_len)
909 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
910 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 911 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 912
bd81e77b 913 /* 12 */
d2a0f284
JC
914 { sizeof(xpviv_allocated),
915 copy_length(XPVIV, xiv_u)
916 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
917 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 918 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 919
bd81e77b 920 /* 20 */
10666ae3 921 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
922 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
923
bd81e77b 924 /* 28 */
10666ae3 925 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
926 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
927
bd81e77b 928 /* 36 */
10666ae3 929 { sizeof(XPVBM), sizeof(XPVBM), 0, SVt_PVBM, TRUE, HADNV,
d2a0f284
JC
930 HASARENA, FIT_ARENA(0, sizeof(XPVBM)) },
931
bd81e77b 932 /* 48 */
10666ae3 933 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
934 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
935
bd81e77b 936 /* 64 */
10666ae3 937 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
938 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
939
940 { sizeof(xpvav_allocated),
941 copy_length(XPVAV, xmg_stash)
942 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
943 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 944 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
945
946 { sizeof(xpvhv_allocated),
947 copy_length(XPVHV, xmg_stash)
948 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
949 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 950 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 951
c84c4652 952 /* 56 */
4115f141 953 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 954 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 955 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 956
4115f141 957 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 958 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 959 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
960
961 /* XPVIO is 84 bytes, fits 48x */
10666ae3 962 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 963 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 964};
29489e7c 965
d2a0f284
JC
966#define new_body_type(sv_type) \
967 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 968
bd81e77b
NC
969#define del_body_type(p, sv_type) \
970 del_body(p, &PL_body_roots[sv_type])
29489e7c 971
29489e7c 972
bd81e77b 973#define new_body_allocated(sv_type) \
d2a0f284 974 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 975 - bodies_by_type[sv_type].offset)
29489e7c 976
bd81e77b
NC
977#define del_body_allocated(p, sv_type) \
978 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 979
29489e7c 980
bd81e77b
NC
981#define my_safemalloc(s) (void*)safemalloc(s)
982#define my_safecalloc(s) (void*)safecalloc(s, 1)
983#define my_safefree(p) safefree((char*)p)
29489e7c 984
bd81e77b 985#ifdef PURIFY
29489e7c 986
bd81e77b
NC
987#define new_XNV() my_safemalloc(sizeof(XPVNV))
988#define del_XNV(p) my_safefree(p)
29489e7c 989
bd81e77b
NC
990#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
991#define del_XPVNV(p) my_safefree(p)
29489e7c 992
bd81e77b
NC
993#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
994#define del_XPVAV(p) my_safefree(p)
29489e7c 995
bd81e77b
NC
996#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
997#define del_XPVHV(p) my_safefree(p)
29489e7c 998
bd81e77b
NC
999#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1000#define del_XPVMG(p) my_safefree(p)
29489e7c 1001
bd81e77b
NC
1002#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1003#define del_XPVGV(p) my_safefree(p)
29489e7c 1004
bd81e77b 1005#else /* !PURIFY */
29489e7c 1006
bd81e77b
NC
1007#define new_XNV() new_body_type(SVt_NV)
1008#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1009
bd81e77b
NC
1010#define new_XPVNV() new_body_type(SVt_PVNV)
1011#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1012
bd81e77b
NC
1013#define new_XPVAV() new_body_allocated(SVt_PVAV)
1014#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1015
bd81e77b
NC
1016#define new_XPVHV() new_body_allocated(SVt_PVHV)
1017#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1018
bd81e77b
NC
1019#define new_XPVMG() new_body_type(SVt_PVMG)
1020#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1021
bd81e77b
NC
1022#define new_XPVGV() new_body_type(SVt_PVGV)
1023#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1024
bd81e77b 1025#endif /* PURIFY */
93e68bfb 1026
bd81e77b 1027/* no arena for you! */
93e68bfb 1028
bd81e77b 1029#define new_NOARENA(details) \
d2a0f284 1030 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1031#define new_NOARENAZ(details) \
d2a0f284
JC
1032 my_safecalloc((details)->body_size + (details)->offset)
1033
10666ae3
NC
1034#ifdef DEBUGGING
1035static bool done_sanity_check;
1036#endif
1037
d2a0f284
JC
1038STATIC void *
1039S_more_bodies (pTHX_ svtype sv_type)
1040{
1041 dVAR;
1042 void ** const root = &PL_body_roots[sv_type];
96a5add6 1043 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1044 const size_t body_size = bdp->body_size;
1045 char *start;
1046 const char *end;
1047
1048 assert(bdp->arena_size);
10666ae3
NC
1049
1050#ifdef DEBUGGING
1051 if (!done_sanity_check) {
ea471437 1052 unsigned int i = SVt_LAST;
10666ae3
NC
1053
1054 done_sanity_check = TRUE;
1055
1056 while (i--)
1057 assert (bodies_by_type[i].type == i);
1058 }
1059#endif
1060
d2a0f284
JC
1061 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size);
1062
1063 end = start + bdp->arena_size - body_size;
1064
d2a0f284
JC
1065 /* computed count doesnt reflect the 1st slot reservation */
1066 DEBUG_m(PerlIO_printf(Perl_debug_log,
1067 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1068 start, end, bdp->arena_size, sv_type, body_size,
1069 bdp->arena_size / body_size));
d2a0f284
JC
1070
1071 *root = (void *)start;
1072
1073 while (start < end) {
1074 char * const next = start + body_size;
1075 *(void**) start = (void *)next;
1076 start = next;
1077 }
1078 *(void **)start = 0;
1079
1080 return *root;
1081}
1082
1083/* grab a new thing from the free list, allocating more if necessary.
1084 The inline version is used for speed in hot routines, and the
1085 function using it serves the rest (unless PURIFY).
1086*/
1087#define new_body_inline(xpv, sv_type) \
1088 STMT_START { \
1089 void ** const r3wt = &PL_body_roots[sv_type]; \
1090 LOCK_SV_MUTEX; \
1091 xpv = *((void **)(r3wt)) \
d4c19fe8 1092 ? *((void **)(r3wt)) : more_bodies(sv_type); \
d2a0f284
JC
1093 *(r3wt) = *(void**)(xpv); \
1094 UNLOCK_SV_MUTEX; \
1095 } STMT_END
1096
1097#ifndef PURIFY
1098
1099STATIC void *
1100S_new_body(pTHX_ svtype sv_type)
1101{
1102 dVAR;
1103 void *xpv;
1104 new_body_inline(xpv, sv_type);
1105 return xpv;
1106}
1107
1108#endif
93e68bfb 1109
bd81e77b
NC
1110/*
1111=for apidoc sv_upgrade
93e68bfb 1112
bd81e77b
NC
1113Upgrade an SV to a more complex form. Generally adds a new body type to the
1114SV, then copies across as much information as possible from the old body.
1115You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1116
bd81e77b 1117=cut
93e68bfb 1118*/
93e68bfb 1119
bd81e77b
NC
1120void
1121Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type)
cac9b346 1122{
97aff369 1123 dVAR;
bd81e77b
NC
1124 void* old_body;
1125 void* new_body;
1126 const U32 old_type = SvTYPE(sv);
d2a0f284 1127 const struct body_details *new_type_details;
bd81e77b
NC
1128 const struct body_details *const old_type_details
1129 = bodies_by_type + old_type;
cac9b346 1130
bd81e77b
NC
1131 if (new_type != SVt_PV && SvIsCOW(sv)) {
1132 sv_force_normal_flags(sv, 0);
1133 }
cac9b346 1134
bd81e77b
NC
1135 if (old_type == new_type)
1136 return;
cac9b346 1137
bd81e77b
NC
1138 if (old_type > new_type)
1139 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1140 (int)old_type, (int)new_type);
cac9b346 1141
cac9b346 1142
bd81e77b 1143 old_body = SvANY(sv);
de042e1d 1144
bd81e77b
NC
1145 /* Copying structures onto other structures that have been neatly zeroed
1146 has a subtle gotcha. Consider XPVMG
cac9b346 1147
bd81e77b
NC
1148 +------+------+------+------+------+-------+-------+
1149 | NV | CUR | LEN | IV | MAGIC | STASH |
1150 +------+------+------+------+------+-------+-------+
1151 0 4 8 12 16 20 24 28
645c22ef 1152
bd81e77b
NC
1153 where NVs are aligned to 8 bytes, so that sizeof that structure is
1154 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1155
bd81e77b
NC
1156 +------+------+------+------+------+-------+-------+------+
1157 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1158 +------+------+------+------+------+-------+-------+------+
1159 0 4 8 12 16 20 24 28 32
08742458 1160
bd81e77b 1161 so what happens if you allocate memory for this structure:
30f9da9e 1162
bd81e77b
NC
1163 +------+------+------+------+------+-------+-------+------+------+...
1164 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1165 +------+------+------+------+------+-------+-------+------+------+...
1166 0 4 8 12 16 20 24 28 32 36
bfc44f79 1167
bd81e77b
NC
1168 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1169 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1170 started out as zero once, but it's quite possible that it isn't. So now,
1171 rather than a nicely zeroed GP, you have it pointing somewhere random.
1172 Bugs ensue.
bfc44f79 1173
bd81e77b
NC
1174 (In fact, GP ends up pointing at a previous GP structure, because the
1175 principle cause of the padding in XPVMG getting garbage is a copy of
1176 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 1177
bd81e77b
NC
1178 So we are careful and work out the size of used parts of all the
1179 structures. */
bfc44f79 1180
bd81e77b
NC
1181 switch (old_type) {
1182 case SVt_NULL:
1183 break;
1184 case SVt_IV:
1185 if (new_type < SVt_PVIV) {
1186 new_type = (new_type == SVt_NV)
1187 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1188 }
1189 break;
1190 case SVt_NV:
1191 if (new_type < SVt_PVNV) {
1192 new_type = SVt_PVNV;
bd81e77b
NC
1193 }
1194 break;
1195 case SVt_RV:
1196 break;
1197 case SVt_PV:
1198 assert(new_type > SVt_PV);
1199 assert(SVt_IV < SVt_PV);
1200 assert(SVt_NV < SVt_PV);
1201 break;
1202 case SVt_PVIV:
1203 break;
1204 case SVt_PVNV:
1205 break;
1206 case SVt_PVMG:
1207 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1208 there's no way that it can be safely upgraded, because perl.c
1209 expects to Safefree(SvANY(PL_mess_sv)) */
1210 assert(sv != PL_mess_sv);
1211 /* This flag bit is used to mean other things in other scalar types.
1212 Given that it only has meaning inside the pad, it shouldn't be set
1213 on anything that can get upgraded. */
00b1698f 1214 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1215 break;
1216 default:
1217 if (old_type_details->cant_upgrade)
c81225bc
NC
1218 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1219 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1220 }
2fa1109b 1221 new_type_details = bodies_by_type + new_type;
645c22ef 1222
bd81e77b
NC
1223 SvFLAGS(sv) &= ~SVTYPEMASK;
1224 SvFLAGS(sv) |= new_type;
932e9ff9 1225
ab4416c0
NC
1226 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1227 the return statements above will have triggered. */
1228 assert (new_type != SVt_NULL);
bd81e77b 1229 switch (new_type) {
bd81e77b
NC
1230 case SVt_IV:
1231 assert(old_type == SVt_NULL);
1232 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1233 SvIV_set(sv, 0);
1234 return;
1235 case SVt_NV:
1236 assert(old_type == SVt_NULL);
1237 SvANY(sv) = new_XNV();
1238 SvNV_set(sv, 0);
1239 return;
1240 case SVt_RV:
1241 assert(old_type == SVt_NULL);
1242 SvANY(sv) = &sv->sv_u.svu_rv;
1243 SvRV_set(sv, 0);
1244 return;
1245 case SVt_PVHV:
bd81e77b 1246 case SVt_PVAV:
d2a0f284 1247 assert(new_type_details->body_size);
c1ae03ae
NC
1248
1249#ifndef PURIFY
1250 assert(new_type_details->arena);
d2a0f284 1251 assert(new_type_details->arena_size);
c1ae03ae 1252 /* This points to the start of the allocated area. */
d2a0f284
JC
1253 new_body_inline(new_body, new_type);
1254 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1255 new_body = ((char *)new_body) - new_type_details->offset;
1256#else
1257 /* We always allocated the full length item with PURIFY. To do this
1258 we fake things so that arena is false for all 16 types.. */
1259 new_body = new_NOARENAZ(new_type_details);
1260#endif
1261 SvANY(sv) = new_body;
1262 if (new_type == SVt_PVAV) {
1263 AvMAX(sv) = -1;
1264 AvFILLp(sv) = -1;
1265 AvREAL_only(sv);
1266 }
aeb18a1e 1267
bd81e77b
NC
1268 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1269 The target created by newSVrv also is, and it can have magic.
1270 However, it never has SvPVX set.
1271 */
1272 if (old_type >= SVt_RV) {
1273 assert(SvPVX_const(sv) == 0);
1274 }
aeb18a1e 1275
bd81e77b
NC
1276 /* Could put this in the else clause below, as PVMG must have SvPVX
1277 0 already (the assertion above) */
6136c704 1278 SvPV_set(sv, NULL);
93e68bfb 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);
bd81e77b
NC
1283 }
1284 break;
93e68bfb 1285
93e68bfb 1286
bd81e77b
NC
1287 case SVt_PVIV:
1288 /* XXX Is this still needed? Was it ever needed? Surely as there is
1289 no route from NV to PVIV, NOK can never be true */
1290 assert(!SvNOKp(sv));
1291 assert(!SvNOK(sv));
1292 case SVt_PVIO:
1293 case SVt_PVFM:
1294 case SVt_PVBM:
1295 case SVt_PVGV:
1296 case SVt_PVCV:
1297 case SVt_PVLV:
1298 case SVt_PVMG:
1299 case SVt_PVNV:
1300 case SVt_PV:
93e68bfb 1301
d2a0f284 1302 assert(new_type_details->body_size);
bd81e77b
NC
1303 /* We always allocated the full length item with PURIFY. To do this
1304 we fake things so that arena is false for all 16 types.. */
1305 if(new_type_details->arena) {
1306 /* This points to the start of the allocated area. */
d2a0f284
JC
1307 new_body_inline(new_body, new_type);
1308 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1309 new_body = ((char *)new_body) - new_type_details->offset;
1310 } else {
1311 new_body = new_NOARENAZ(new_type_details);
1312 }
1313 SvANY(sv) = new_body;
5e2fc214 1314
bd81e77b 1315 if (old_type_details->copy) {
f9ba3d20
NC
1316 /* There is now the potential for an upgrade from something without
1317 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1318 int offset = old_type_details->offset;
1319 int length = old_type_details->copy;
1320
1321 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1322 const int difference
f9ba3d20
NC
1323 = new_type_details->offset - old_type_details->offset;
1324 offset += difference;
1325 length -= difference;
1326 }
1327 assert (length >= 0);
1328
1329 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1330 char);
bd81e77b
NC
1331 }
1332
1333#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1334 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1335 * correct 0.0 for us. Otherwise, if the old body didn't have an
1336 * NV slot, but the new one does, then we need to initialise the
1337 * freshly created NV slot with whatever the correct bit pattern is
1338 * for 0.0 */
1339 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1340 SvNV_set(sv, 0);
82048762 1341#endif
5e2fc214 1342
bd81e77b 1343 if (new_type == SVt_PVIO)
f2524eef 1344 IoPAGE_LEN(sv) = 60;
bd81e77b 1345 if (old_type < SVt_RV)
6136c704 1346 SvPV_set(sv, NULL);
bd81e77b
NC
1347 break;
1348 default:
afd78fd5
JH
1349 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1350 (unsigned long)new_type);
bd81e77b 1351 }
73171d91 1352
d2a0f284
JC
1353 if (old_type_details->arena) {
1354 /* If there was an old body, then we need to free it.
1355 Note that there is an assumption that all bodies of types that
1356 can be upgraded came from arenas. Only the more complex non-
1357 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1358#ifdef PURIFY
1359 my_safefree(old_body);
1360#else
1361 del_body((void*)((char*)old_body + old_type_details->offset),
1362 &PL_body_roots[old_type]);
1363#endif
1364 }
1365}
73171d91 1366
bd81e77b
NC
1367/*
1368=for apidoc sv_backoff
73171d91 1369
bd81e77b
NC
1370Remove any string offset. You should normally use the C<SvOOK_off> macro
1371wrapper instead.
73171d91 1372
bd81e77b 1373=cut
73171d91
NC
1374*/
1375
bd81e77b
NC
1376int
1377Perl_sv_backoff(pTHX_ register SV *sv)
1378{
96a5add6 1379 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1380 assert(SvOOK(sv));
1381 assert(SvTYPE(sv) != SVt_PVHV);
1382 assert(SvTYPE(sv) != SVt_PVAV);
1383 if (SvIVX(sv)) {
1384 const char * const s = SvPVX_const(sv);
1385 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1386 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1387 SvIV_set(sv, 0);
1388 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1389 }
1390 SvFLAGS(sv) &= ~SVf_OOK;
1391 return 0;
1392}
73171d91 1393
bd81e77b
NC
1394/*
1395=for apidoc sv_grow
73171d91 1396
bd81e77b
NC
1397Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1398upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1399Use the C<SvGROW> wrapper instead.
93e68bfb 1400
bd81e77b
NC
1401=cut
1402*/
93e68bfb 1403
bd81e77b
NC
1404char *
1405Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1406{
1407 register char *s;
93e68bfb 1408
5db06880
NC
1409 if (PL_madskills && newlen >= 0x100000) {
1410 PerlIO_printf(Perl_debug_log,
1411 "Allocation too large: %"UVxf"\n", (UV)newlen);
1412 }
bd81e77b
NC
1413#ifdef HAS_64K_LIMIT
1414 if (newlen >= 0x10000) {
1415 PerlIO_printf(Perl_debug_log,
1416 "Allocation too large: %"UVxf"\n", (UV)newlen);
1417 my_exit(1);
1418 }
1419#endif /* HAS_64K_LIMIT */
1420 if (SvROK(sv))
1421 sv_unref(sv);
1422 if (SvTYPE(sv) < SVt_PV) {
1423 sv_upgrade(sv, SVt_PV);
1424 s = SvPVX_mutable(sv);
1425 }
1426 else if (SvOOK(sv)) { /* pv is offset? */
1427 sv_backoff(sv);
1428 s = SvPVX_mutable(sv);
1429 if (newlen > SvLEN(sv))
1430 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1431#ifdef HAS_64K_LIMIT
1432 if (newlen >= 0x10000)
1433 newlen = 0xFFFF;
1434#endif
1435 }
1436 else
1437 s = SvPVX_mutable(sv);
aeb18a1e 1438
bd81e77b
NC
1439 if (newlen > SvLEN(sv)) { /* need more room? */
1440 newlen = PERL_STRLEN_ROUNDUP(newlen);
1441 if (SvLEN(sv) && s) {
1442#ifdef MYMALLOC
1443 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1444 if (newlen <= l) {
1445 SvLEN_set(sv, l);
1446 return s;
1447 } else
1448#endif
1449 s = saferealloc(s, newlen);
1450 }
1451 else {
1452 s = safemalloc(newlen);
1453 if (SvPVX_const(sv) && SvCUR(sv)) {
1454 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1455 }
1456 }
1457 SvPV_set(sv, s);
1458 SvLEN_set(sv, newlen);
1459 }
1460 return s;
1461}
aeb18a1e 1462
bd81e77b
NC
1463/*
1464=for apidoc sv_setiv
932e9ff9 1465
bd81e77b
NC
1466Copies an integer into the given SV, upgrading first if necessary.
1467Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1468
bd81e77b
NC
1469=cut
1470*/
463ee0b2 1471
bd81e77b
NC
1472void
1473Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1474{
97aff369 1475 dVAR;
bd81e77b
NC
1476 SV_CHECK_THINKFIRST_COW_DROP(sv);
1477 switch (SvTYPE(sv)) {
1478 case SVt_NULL:
1479 sv_upgrade(sv, SVt_IV);
1480 break;
1481 case SVt_NV:
1482 sv_upgrade(sv, SVt_PVNV);
1483 break;
1484 case SVt_RV:
1485 case SVt_PV:
1486 sv_upgrade(sv, SVt_PVIV);
1487 break;
463ee0b2 1488
bd81e77b
NC
1489 case SVt_PVGV:
1490 case SVt_PVAV:
1491 case SVt_PVHV:
1492 case SVt_PVCV:
1493 case SVt_PVFM:
1494 case SVt_PVIO:
1495 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1496 OP_DESC(PL_op));
1497 }
1498 (void)SvIOK_only(sv); /* validate number */
1499 SvIV_set(sv, i);
1500 SvTAINT(sv);
1501}
932e9ff9 1502
bd81e77b
NC
1503/*
1504=for apidoc sv_setiv_mg
d33b2eba 1505
bd81e77b 1506Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1507
bd81e77b
NC
1508=cut
1509*/
d33b2eba 1510
bd81e77b
NC
1511void
1512Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1513{
1514 sv_setiv(sv,i);
1515 SvSETMAGIC(sv);
1516}
727879eb 1517
bd81e77b
NC
1518/*
1519=for apidoc sv_setuv
d33b2eba 1520
bd81e77b
NC
1521Copies an unsigned integer into the given SV, upgrading first if necessary.
1522Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1523
bd81e77b
NC
1524=cut
1525*/
d33b2eba 1526
bd81e77b
NC
1527void
1528Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1529{
1530 /* With these two if statements:
1531 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1532
bd81e77b
NC
1533 without
1534 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1535
bd81e77b
NC
1536 If you wish to remove them, please benchmark to see what the effect is
1537 */
1538 if (u <= (UV)IV_MAX) {
1539 sv_setiv(sv, (IV)u);
1540 return;
1541 }
1542 sv_setiv(sv, 0);
1543 SvIsUV_on(sv);
1544 SvUV_set(sv, u);
1545}
d33b2eba 1546
bd81e77b
NC
1547/*
1548=for apidoc sv_setuv_mg
727879eb 1549
bd81e77b 1550Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1551
bd81e77b
NC
1552=cut
1553*/
5e2fc214 1554
bd81e77b
NC
1555void
1556Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1557{
1558 sv_setiv(sv, 0);
1559 SvIsUV_on(sv);
1560 sv_setuv(sv,u);
1561 SvSETMAGIC(sv);
1562}
5e2fc214 1563
954c1994 1564/*
bd81e77b 1565=for apidoc sv_setnv
954c1994 1566
bd81e77b
NC
1567Copies a double into the given SV, upgrading first if necessary.
1568Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1569
1570=cut
1571*/
1572
63f97190 1573void
bd81e77b 1574Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1575{
97aff369 1576 dVAR;
bd81e77b
NC
1577 SV_CHECK_THINKFIRST_COW_DROP(sv);
1578 switch (SvTYPE(sv)) {
79072805 1579 case SVt_NULL:
79072805 1580 case SVt_IV:
bd81e77b 1581 sv_upgrade(sv, SVt_NV);
79072805 1582 break;
ed6116ce 1583 case SVt_RV:
79072805 1584 case SVt_PV:
79072805 1585 case SVt_PVIV:
bd81e77b 1586 sv_upgrade(sv, SVt_PVNV);
79072805 1587 break;
bd4b1eb5 1588
bd4b1eb5 1589 case SVt_PVGV:
bd81e77b
NC
1590 case SVt_PVAV:
1591 case SVt_PVHV:
79072805 1592 case SVt_PVCV:
bd81e77b
NC
1593 case SVt_PVFM:
1594 case SVt_PVIO:
1595 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1596 OP_NAME(PL_op));
2068cd4d 1597 }
bd81e77b
NC
1598 SvNV_set(sv, num);
1599 (void)SvNOK_only(sv); /* validate number */
1600 SvTAINT(sv);
79072805
LW
1601}
1602
645c22ef 1603/*
bd81e77b 1604=for apidoc sv_setnv_mg
645c22ef 1605
bd81e77b 1606Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1607
1608=cut
1609*/
1610
bd81e77b
NC
1611void
1612Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1613{
bd81e77b
NC
1614 sv_setnv(sv,num);
1615 SvSETMAGIC(sv);
79072805
LW
1616}
1617
bd81e77b
NC
1618/* Print an "isn't numeric" warning, using a cleaned-up,
1619 * printable version of the offending string
1620 */
954c1994 1621
bd81e77b
NC
1622STATIC void
1623S_not_a_number(pTHX_ SV *sv)
79072805 1624{
97aff369 1625 dVAR;
bd81e77b
NC
1626 SV *dsv;
1627 char tmpbuf[64];
1628 const char *pv;
94463019
JH
1629
1630 if (DO_UTF8(sv)) {
396482e1 1631 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1632 pv = sv_uni_display(dsv, sv, 10, 0);
1633 } else {
1634 char *d = tmpbuf;
551405c4 1635 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1636 /* each *s can expand to 4 chars + "...\0",
1637 i.e. need room for 8 chars */
ecdeb87c 1638
00b6aa41
AL
1639 const char *s = SvPVX_const(sv);
1640 const char * const end = s + SvCUR(sv);
1641 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1642 int ch = *s & 0xFF;
1643 if (ch & 128 && !isPRINT_LC(ch)) {
1644 *d++ = 'M';
1645 *d++ = '-';
1646 ch &= 127;
1647 }
1648 if (ch == '\n') {
1649 *d++ = '\\';
1650 *d++ = 'n';
1651 }
1652 else if (ch == '\r') {
1653 *d++ = '\\';
1654 *d++ = 'r';
1655 }
1656 else if (ch == '\f') {
1657 *d++ = '\\';
1658 *d++ = 'f';
1659 }
1660 else if (ch == '\\') {
1661 *d++ = '\\';
1662 *d++ = '\\';
1663 }
1664 else if (ch == '\0') {
1665 *d++ = '\\';
1666 *d++ = '0';
1667 }
1668 else if (isPRINT_LC(ch))
1669 *d++ = ch;
1670 else {
1671 *d++ = '^';
1672 *d++ = toCTRL(ch);
1673 }
1674 }
1675 if (s < end) {
1676 *d++ = '.';
1677 *d++ = '.';
1678 *d++ = '.';
1679 }
1680 *d = '\0';
1681 pv = tmpbuf;
a0d0e21e 1682 }
a0d0e21e 1683
533c011a 1684 if (PL_op)
9014280d 1685 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1686 "Argument \"%s\" isn't numeric in %s", pv,
1687 OP_DESC(PL_op));
a0d0e21e 1688 else
9014280d 1689 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1690 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1691}
1692
c2988b20
NC
1693/*
1694=for apidoc looks_like_number
1695
645c22ef
DM
1696Test if the content of an SV looks like a number (or is a number).
1697C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1698non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1699
1700=cut
1701*/
1702
1703I32
1704Perl_looks_like_number(pTHX_ SV *sv)
1705{
a3b680e6 1706 register const char *sbegin;
c2988b20
NC
1707 STRLEN len;
1708
1709 if (SvPOK(sv)) {
3f7c398e 1710 sbegin = SvPVX_const(sv);
c2988b20
NC
1711 len = SvCUR(sv);
1712 }
1713 else if (SvPOKp(sv))
83003860 1714 sbegin = SvPV_const(sv, len);
c2988b20 1715 else
e0ab1c0e 1716 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1717 return grok_number(sbegin, len, NULL);
1718}
25da4f38 1719
19f6321d
NC
1720STATIC bool
1721S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1722{
1723 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1724 SV *const buffer = sv_newmortal();
1725
1726 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1727 is on. */
1728 SvFAKE_off(gv);
1729 gv_efullname3(buffer, gv, "*");
1730 SvFLAGS(gv) |= wasfake;
1731
675c862f
AL
1732 /* We know that all GVs stringify to something that is not-a-number,
1733 so no need to test that. */
1734 if (ckWARN(WARN_NUMERIC))
1735 not_a_number(buffer);
1736 /* We just want something true to return, so that S_sv_2iuv_common
1737 can tail call us and return true. */
19f6321d 1738 return TRUE;
675c862f
AL
1739}
1740
1741STATIC char *
19f6321d 1742S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1743{
1744 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1745 SV *const buffer = sv_newmortal();
1746
1747 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1748 is on. */
1749 SvFAKE_off(gv);
1750 gv_efullname3(buffer, gv, "*");
1751 SvFLAGS(gv) |= wasfake;
1752
1753 assert(SvPOK(buffer));
1754 *len = SvCUR(buffer);
1755 return SvPVX(buffer);
180488f8
NC
1756}
1757
25da4f38
IZ
1758/* Actually, ISO C leaves conversion of UV to IV undefined, but
1759 until proven guilty, assume that things are not that bad... */
1760
645c22ef
DM
1761/*
1762 NV_PRESERVES_UV:
1763
1764 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1765 an IV (an assumption perl has been based on to date) it becomes necessary
1766 to remove the assumption that the NV always carries enough precision to
1767 recreate the IV whenever needed, and that the NV is the canonical form.
1768 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1769 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1770 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1771 1) to distinguish between IV/UV/NV slots that have cached a valid
1772 conversion where precision was lost and IV/UV/NV slots that have a
1773 valid conversion which has lost no precision
645c22ef 1774 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1775 would lose precision, the precise conversion (or differently
1776 imprecise conversion) is also performed and cached, to prevent
1777 requests for different numeric formats on the same SV causing
1778 lossy conversion chains. (lossless conversion chains are perfectly
1779 acceptable (still))
1780
1781
1782 flags are used:
1783 SvIOKp is true if the IV slot contains a valid value
1784 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1785 SvNOKp is true if the NV slot contains a valid value
1786 SvNOK is true only if the NV value is accurate
1787
1788 so
645c22ef 1789 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1790 IV(or UV) would lose accuracy over a direct conversion from PV to
1791 IV(or UV). If it would, cache both conversions, return NV, but mark
1792 SV as IOK NOKp (ie not NOK).
1793
645c22ef 1794 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1795 NV would lose accuracy over a direct conversion from PV to NV. If it
1796 would, cache both conversions, flag similarly.
1797
1798 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1799 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1800 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1801 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1802 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1803
645c22ef
DM
1804 The benefit of this is that operations such as pp_add know that if
1805 SvIOK is true for both left and right operands, then integer addition
1806 can be used instead of floating point (for cases where the result won't
1807 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1808 loss of precision compared with integer addition.
1809
1810 * making IV and NV equal status should make maths accurate on 64 bit
1811 platforms
1812 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1813 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1814 looking for SvIOK and checking for overflow will not outweigh the
1815 fp to integer speedup)
1816 * will slow down integer operations (callers of SvIV) on "inaccurate"
1817 values, as the change from SvIOK to SvIOKp will cause a call into
1818 sv_2iv each time rather than a macro access direct to the IV slot
1819 * should speed up number->string conversion on integers as IV is
645c22ef 1820 favoured when IV and NV are equally accurate
28e5dec8
JH
1821
1822 ####################################################################
645c22ef
DM
1823 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1824 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1825 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1826 ####################################################################
1827
645c22ef 1828 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1829 performance ratio.
1830*/
1831
1832#ifndef NV_PRESERVES_UV
645c22ef
DM
1833# define IS_NUMBER_UNDERFLOW_IV 1
1834# define IS_NUMBER_UNDERFLOW_UV 2
1835# define IS_NUMBER_IV_AND_UV 2
1836# define IS_NUMBER_OVERFLOW_IV 4
1837# define IS_NUMBER_OVERFLOW_UV 5
1838
1839/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1840
1841/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1842STATIC int
645c22ef 1843S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1844{
97aff369 1845 dVAR;
3f7c398e 1846 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
1847 if (SvNVX(sv) < (NV)IV_MIN) {
1848 (void)SvIOKp_on(sv);
1849 (void)SvNOK_on(sv);
45977657 1850 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1851 return IS_NUMBER_UNDERFLOW_IV;
1852 }
1853 if (SvNVX(sv) > (NV)UV_MAX) {
1854 (void)SvIOKp_on(sv);
1855 (void)SvNOK_on(sv);
1856 SvIsUV_on(sv);
607fa7f2 1857 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1858 return IS_NUMBER_OVERFLOW_UV;
1859 }
c2988b20
NC
1860 (void)SvIOKp_on(sv);
1861 (void)SvNOK_on(sv);
1862 /* Can't use strtol etc to convert this string. (See truth table in
1863 sv_2iv */
1864 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1865 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1866 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1867 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1868 } else {
1869 /* Integer is imprecise. NOK, IOKp */
1870 }
1871 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1872 }
1873 SvIsUV_on(sv);
607fa7f2 1874 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1875 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1876 if (SvUVX(sv) == UV_MAX) {
1877 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1878 possibly be preserved by NV. Hence, it must be overflow.
1879 NOK, IOKp */
1880 return IS_NUMBER_OVERFLOW_UV;
1881 }
1882 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1883 } else {
1884 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1885 }
c2988b20 1886 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1887}
645c22ef
DM
1888#endif /* !NV_PRESERVES_UV*/
1889
af359546
NC
1890STATIC bool
1891S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1892 dVAR;
af359546 1893 if (SvNOKp(sv)) {
28e5dec8
JH
1894 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1895 * without also getting a cached IV/UV from it at the same time
1896 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1897 * IV or UV at same time to avoid this. */
1898 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1899
1900 if (SvTYPE(sv) == SVt_NV)
1901 sv_upgrade(sv, SVt_PVNV);
1902
28e5dec8
JH
1903 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1904 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1905 certainly cast into the IV range at IV_MAX, whereas the correct
1906 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1907 cases go to UV */
cab190d4
JD
1908#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1909 if (Perl_isnan(SvNVX(sv))) {
1910 SvUV_set(sv, 0);
1911 SvIsUV_on(sv);
fdbe6d7c 1912 return FALSE;
cab190d4 1913 }
cab190d4 1914#endif
28e5dec8 1915 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1916 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1917 if (SvNVX(sv) == (NV) SvIVX(sv)
1918#ifndef NV_PRESERVES_UV
1919 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1920 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1921 /* Don't flag it as "accurately an integer" if the number
1922 came from a (by definition imprecise) NV operation, and
1923 we're outside the range of NV integer precision */
1924#endif
1925 ) {
1926 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1927 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1928 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1929 PTR2UV(sv),
1930 SvNVX(sv),
1931 SvIVX(sv)));
1932
1933 } else {
1934 /* IV not precise. No need to convert from PV, as NV
1935 conversion would already have cached IV if it detected
1936 that PV->IV would be better than PV->NV->IV
1937 flags already correct - don't set public IOK. */
1938 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1939 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1940 PTR2UV(sv),
1941 SvNVX(sv),
1942 SvIVX(sv)));
1943 }
1944 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1945 but the cast (NV)IV_MIN rounds to a the value less (more
1946 negative) than IV_MIN which happens to be equal to SvNVX ??
1947 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1948 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1949 (NV)UVX == NVX are both true, but the values differ. :-(
1950 Hopefully for 2s complement IV_MIN is something like
1951 0x8000000000000000 which will be exact. NWC */
d460ef45 1952 }
25da4f38 1953 else {
607fa7f2 1954 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1955 if (
1956 (SvNVX(sv) == (NV) SvUVX(sv))
1957#ifndef NV_PRESERVES_UV
1958 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1959 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1960 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1961 /* Don't flag it as "accurately an integer" if the number
1962 came from a (by definition imprecise) NV operation, and
1963 we're outside the range of NV integer precision */
1964#endif
1965 )
1966 SvIOK_on(sv);
25da4f38 1967 SvIsUV_on(sv);
1c846c1f 1968 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1969 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1970 PTR2UV(sv),
57def98f
JH
1971 SvUVX(sv),
1972 SvUVX(sv)));
25da4f38 1973 }
748a9306
LW
1974 }
1975 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1976 UV value;
504618e9 1977 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1978 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1979 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1980 the same as the direct translation of the initial string
1981 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1982 be careful to ensure that the value with the .456 is around if the
1983 NV value is requested in the future).
1c846c1f 1984
af359546 1985 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1986 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1987 cache the NV if we are sure it's not needed.
25da4f38 1988 */
16b7a9a4 1989
c2988b20
NC
1990 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1991 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1992 == IS_NUMBER_IN_UV) {
5e045b90 1993 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1994 if (SvTYPE(sv) < SVt_PVIV)
1995 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1996 (void)SvIOK_on(sv);
c2988b20
NC
1997 } else if (SvTYPE(sv) < SVt_PVNV)
1998 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1999
f2524eef 2000 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2001 we aren't going to call atof() below. If NVs don't preserve UVs
2002 then the value returned may have more precision than atof() will
2003 return, even though value isn't perfectly accurate. */
2004 if ((numtype & (IS_NUMBER_IN_UV
2005#ifdef NV_PRESERVES_UV
2006 | IS_NUMBER_NOT_INT
2007#endif
2008 )) == IS_NUMBER_IN_UV) {
2009 /* This won't turn off the public IOK flag if it was set above */
2010 (void)SvIOKp_on(sv);
2011
2012 if (!(numtype & IS_NUMBER_NEG)) {
2013 /* positive */;
2014 if (value <= (UV)IV_MAX) {
45977657 2015 SvIV_set(sv, (IV)value);
c2988b20 2016 } else {
af359546 2017 /* it didn't overflow, and it was positive. */
607fa7f2 2018 SvUV_set(sv, value);
c2988b20
NC
2019 SvIsUV_on(sv);
2020 }
2021 } else {
2022 /* 2s complement assumption */
2023 if (value <= (UV)IV_MIN) {
45977657 2024 SvIV_set(sv, -(IV)value);
c2988b20
NC
2025 } else {
2026 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2027 I'm assuming it will be rare. */
c2988b20
NC
2028 if (SvTYPE(sv) < SVt_PVNV)
2029 sv_upgrade(sv, SVt_PVNV);
2030 SvNOK_on(sv);
2031 SvIOK_off(sv);
2032 SvIOKp_on(sv);
9d6ce603 2033 SvNV_set(sv, -(NV)value);
45977657 2034 SvIV_set(sv, IV_MIN);
c2988b20
NC
2035 }
2036 }
2037 }
2038 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2039 will be in the previous block to set the IV slot, and the next
2040 block to set the NV slot. So no else here. */
2041
2042 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2043 != IS_NUMBER_IN_UV) {
2044 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2045 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2046
c2988b20
NC
2047 if (! numtype && ckWARN(WARN_NUMERIC))
2048 not_a_number(sv);
28e5dec8 2049
65202027 2050#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2051 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2052 PTR2UV(sv), SvNVX(sv)));
65202027 2053#else
1779d84d 2054 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2055 PTR2UV(sv), SvNVX(sv)));
65202027 2056#endif
28e5dec8 2057
28e5dec8 2058#ifdef NV_PRESERVES_UV
af359546
NC
2059 (void)SvIOKp_on(sv);
2060 (void)SvNOK_on(sv);
2061 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2062 SvIV_set(sv, I_V(SvNVX(sv)));
2063 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2064 SvIOK_on(sv);
2065 } else {
6f207bd3 2066 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2067 }
2068 /* UV will not work better than IV */
2069 } else {
2070 if (SvNVX(sv) > (NV)UV_MAX) {
2071 SvIsUV_on(sv);
2072 /* Integer is inaccurate. NOK, IOKp, is UV */
2073 SvUV_set(sv, UV_MAX);
af359546
NC
2074 } else {
2075 SvUV_set(sv, U_V(SvNVX(sv)));
2076 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2077 NV preservse UV so can do correct comparison. */
2078 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2079 SvIOK_on(sv);
af359546 2080 } else {
6f207bd3 2081 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2082 }
2083 }
4b0c9573 2084 SvIsUV_on(sv);
af359546 2085 }
28e5dec8 2086#else /* NV_PRESERVES_UV */
c2988b20
NC
2087 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2088 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2089 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2090 grok_number above. The NV slot has just been set using
2091 Atof. */
560b0c46 2092 SvNOK_on(sv);
c2988b20
NC
2093 assert (SvIOKp(sv));
2094 } else {
2095 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2096 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2097 /* Small enough to preserve all bits. */
2098 (void)SvIOKp_on(sv);
2099 SvNOK_on(sv);
45977657 2100 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2101 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2102 SvIOK_on(sv);
2103 /* Assumption: first non-preserved integer is < IV_MAX,
2104 this NV is in the preserved range, therefore: */
2105 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2106 < (UV)IV_MAX)) {
32fdb065 2107 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
2108 }
2109 } else {
2110 /* IN_UV NOT_INT
2111 0 0 already failed to read UV.
2112 0 1 already failed to read UV.
2113 1 0 you won't get here in this case. IV/UV
2114 slot set, public IOK, Atof() unneeded.
2115 1 1 already read UV.
2116 so there's no point in sv_2iuv_non_preserve() attempting
2117 to use atol, strtol, strtoul etc. */
40a17c4c 2118 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2119 }
2120 }
28e5dec8 2121#endif /* NV_PRESERVES_UV */
25da4f38 2122 }
af359546
NC
2123 }
2124 else {
675c862f 2125 if (isGV_with_GP(sv))
a0933d07 2126 return glob_2number((GV *)sv);
180488f8 2127
af359546
NC
2128 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2129 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2130 report_uninit(sv);
2131 }
25da4f38
IZ
2132 if (SvTYPE(sv) < SVt_IV)
2133 /* Typically the caller expects that sv_any is not NULL now. */
2134 sv_upgrade(sv, SVt_IV);
af359546
NC
2135 /* Return 0 from the caller. */
2136 return TRUE;
2137 }
2138 return FALSE;
2139}
2140
2141/*
2142=for apidoc sv_2iv_flags
2143
2144Return the integer value of an SV, doing any necessary string
2145conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2146Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2147
2148=cut
2149*/
2150
2151IV
2152Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2153{
97aff369 2154 dVAR;
af359546 2155 if (!sv)
a0d0e21e 2156 return 0;
af359546
NC
2157 if (SvGMAGICAL(sv)) {
2158 if (flags & SV_GMAGIC)
2159 mg_get(sv);
2160 if (SvIOKp(sv))
2161 return SvIVX(sv);
2162 if (SvNOKp(sv)) {
2163 return I_V(SvNVX(sv));
2164 }
71c558c3
NC
2165 if (SvPOKp(sv) && SvLEN(sv)) {
2166 UV value;
2167 const int numtype
2168 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2169
2170 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2171 == IS_NUMBER_IN_UV) {
2172 /* It's definitely an integer */
2173 if (numtype & IS_NUMBER_NEG) {
2174 if (value < (UV)IV_MIN)
2175 return -(IV)value;
2176 } else {
2177 if (value < (UV)IV_MAX)
2178 return (IV)value;
2179 }
2180 }
2181 if (!numtype) {
2182 if (ckWARN(WARN_NUMERIC))
2183 not_a_number(sv);
2184 }
2185 return I_V(Atof(SvPVX_const(sv)));
2186 }
1c7ff15e
NC
2187 if (SvROK(sv)) {
2188 goto return_rok;
af359546 2189 }
1c7ff15e
NC
2190 assert(SvTYPE(sv) >= SVt_PVMG);
2191 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2192 } else if (SvTHINKFIRST(sv)) {
af359546 2193 if (SvROK(sv)) {
1c7ff15e 2194 return_rok:
af359546
NC
2195 if (SvAMAGIC(sv)) {
2196 SV * const tmpstr=AMG_CALLun(sv,numer);
2197 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2198 return SvIV(tmpstr);
2199 }
2200 }
2201 return PTR2IV(SvRV(sv));
2202 }
2203 if (SvIsCOW(sv)) {
2204 sv_force_normal_flags(sv, 0);
2205 }
2206 if (SvREADONLY(sv) && !SvOK(sv)) {
2207 if (ckWARN(WARN_UNINITIALIZED))
2208 report_uninit(sv);
2209 return 0;
2210 }
2211 }
2212 if (!SvIOKp(sv)) {
2213 if (S_sv_2iuv_common(aTHX_ sv))
2214 return 0;
79072805 2215 }
1d7c1841
GS
2216 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2217 PTR2UV(sv),SvIVX(sv)));
25da4f38 2218 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2219}
2220
645c22ef 2221/*
891f9566 2222=for apidoc sv_2uv_flags
645c22ef
DM
2223
2224Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2225conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2226Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2227
2228=cut
2229*/
2230
ff68c719 2231UV
891f9566 2232Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2233{
97aff369 2234 dVAR;
ff68c719 2235 if (!sv)
2236 return 0;
2237 if (SvGMAGICAL(sv)) {
891f9566
YST
2238 if (flags & SV_GMAGIC)
2239 mg_get(sv);
ff68c719 2240 if (SvIOKp(sv))
2241 return SvUVX(sv);
2242 if (SvNOKp(sv))
2243 return U_V(SvNVX(sv));
71c558c3
NC
2244 if (SvPOKp(sv) && SvLEN(sv)) {
2245 UV value;
2246 const int numtype
2247 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2248
2249 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2250 == IS_NUMBER_IN_UV) {
2251 /* It's definitely an integer */
2252 if (!(numtype & IS_NUMBER_NEG))
2253 return value;
2254 }
2255 if (!numtype) {
2256 if (ckWARN(WARN_NUMERIC))
2257 not_a_number(sv);
2258 }
2259 return U_V(Atof(SvPVX_const(sv)));
2260 }
1c7ff15e
NC
2261 if (SvROK(sv)) {
2262 goto return_rok;
3fe9a6f1 2263 }
1c7ff15e
NC
2264 assert(SvTYPE(sv) >= SVt_PVMG);
2265 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2266 } else if (SvTHINKFIRST(sv)) {
ff68c719 2267 if (SvROK(sv)) {
1c7ff15e 2268 return_rok:
deb46114
NC
2269 if (SvAMAGIC(sv)) {
2270 SV *const tmpstr = AMG_CALLun(sv,numer);
2271 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2272 return SvUV(tmpstr);
2273 }
2274 }
2275 return PTR2UV(SvRV(sv));
ff68c719 2276 }
765f542d
NC
2277 if (SvIsCOW(sv)) {
2278 sv_force_normal_flags(sv, 0);
8a818333 2279 }
0336b60e 2280 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2281 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2282 report_uninit(sv);
ff68c719 2283 return 0;
2284 }
2285 }
af359546
NC
2286 if (!SvIOKp(sv)) {
2287 if (S_sv_2iuv_common(aTHX_ sv))
2288 return 0;
ff68c719 2289 }
25da4f38 2290
1d7c1841
GS
2291 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2292 PTR2UV(sv),SvUVX(sv)));
25da4f38 2293 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2294}
2295
645c22ef
DM
2296/*
2297=for apidoc sv_2nv
2298
2299Return the num value of an SV, doing any necessary string or integer
2300conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2301macros.
2302
2303=cut
2304*/
2305
65202027 2306NV
864dbfa3 2307Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2308{
97aff369 2309 dVAR;
79072805
LW
2310 if (!sv)
2311 return 0.0;
8990e307 2312 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2313 mg_get(sv);
2314 if (SvNOKp(sv))
2315 return SvNVX(sv);
0aa395f8 2316 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2317 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2318 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2319 not_a_number(sv);
3f7c398e 2320 return Atof(SvPVX_const(sv));
a0d0e21e 2321 }
25da4f38 2322 if (SvIOKp(sv)) {
1c846c1f 2323 if (SvIsUV(sv))
65202027 2324 return (NV)SvUVX(sv);
25da4f38 2325 else
65202027 2326 return (NV)SvIVX(sv);
47a72cb8
NC
2327 }
2328 if (SvROK(sv)) {
2329 goto return_rok;
2330 }
2331 assert(SvTYPE(sv) >= SVt_PVMG);
2332 /* This falls through to the report_uninit near the end of the
2333 function. */
2334 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2335 if (SvROK(sv)) {
47a72cb8 2336 return_rok:
deb46114
NC
2337 if (SvAMAGIC(sv)) {
2338 SV *const tmpstr = AMG_CALLun(sv,numer);
2339 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2340 return SvNV(tmpstr);
2341 }
2342 }
2343 return PTR2NV(SvRV(sv));
a0d0e21e 2344 }
765f542d
NC
2345 if (SvIsCOW(sv)) {
2346 sv_force_normal_flags(sv, 0);
8a818333 2347 }
0336b60e 2348 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2349 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2350 report_uninit(sv);
ed6116ce
LW
2351 return 0.0;
2352 }
79072805
LW
2353 }
2354 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2355 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2356 sv_upgrade(sv, SVt_NV);
906f284f 2357#ifdef USE_LONG_DOUBLE
097ee67d 2358 DEBUG_c({
f93f4e46 2359 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2360 PerlIO_printf(Perl_debug_log,
2361 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2362 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2363 RESTORE_NUMERIC_LOCAL();
2364 });
65202027 2365#else
572bbb43 2366 DEBUG_c({
f93f4e46 2367 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2368 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2369 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2370 RESTORE_NUMERIC_LOCAL();
2371 });
572bbb43 2372#endif
79072805
LW
2373 }
2374 else if (SvTYPE(sv) < SVt_PVNV)
2375 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2376 if (SvNOKp(sv)) {
2377 return SvNVX(sv);
61604483 2378 }
59d8ce62 2379 if (SvIOKp(sv)) {
9d6ce603 2380 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2381#ifdef NV_PRESERVES_UV
2382 SvNOK_on(sv);
2383#else
2384 /* Only set the public NV OK flag if this NV preserves the IV */
2385 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2386 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2387 : (SvIVX(sv) == I_V(SvNVX(sv))))
2388 SvNOK_on(sv);
2389 else
2390 SvNOKp_on(sv);
2391#endif
93a17b20 2392 }
748a9306 2393 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2394 UV value;
3f7c398e 2395 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2396 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2397 not_a_number(sv);
28e5dec8 2398#ifdef NV_PRESERVES_UV
c2988b20
NC
2399 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2400 == IS_NUMBER_IN_UV) {
5e045b90 2401 /* It's definitely an integer */
9d6ce603 2402 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2403 } else
3f7c398e 2404 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2405 SvNOK_on(sv);
2406#else
3f7c398e 2407 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2408 /* Only set the public NV OK flag if this NV preserves the value in
2409 the PV at least as well as an IV/UV would.
2410 Not sure how to do this 100% reliably. */
2411 /* if that shift count is out of range then Configure's test is
2412 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2413 UV_BITS */
2414 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2415 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2416 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2417 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2418 /* Can't use strtol etc to convert this string, so don't try.
2419 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2420 SvNOK_on(sv);
2421 } else {
2422 /* value has been set. It may not be precise. */
2423 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2424 /* 2s complement assumption for (UV)IV_MIN */
2425 SvNOK_on(sv); /* Integer is too negative. */
2426 } else {
2427 SvNOKp_on(sv);
2428 SvIOKp_on(sv);
6fa402ec 2429
c2988b20 2430 if (numtype & IS_NUMBER_NEG) {
45977657 2431 SvIV_set(sv, -(IV)value);
c2988b20 2432 } else if (value <= (UV)IV_MAX) {
45977657 2433 SvIV_set(sv, (IV)value);
c2988b20 2434 } else {
607fa7f2 2435 SvUV_set(sv, value);
c2988b20
NC
2436 SvIsUV_on(sv);
2437 }
2438
2439 if (numtype & IS_NUMBER_NOT_INT) {
2440 /* I believe that even if the original PV had decimals,
2441 they are lost beyond the limit of the FP precision.
2442 However, neither is canonical, so both only get p
2443 flags. NWC, 2000/11/25 */
2444 /* Both already have p flags, so do nothing */
2445 } else {
66a1b24b 2446 const NV nv = SvNVX(sv);
c2988b20
NC
2447 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2448 if (SvIVX(sv) == I_V(nv)) {
2449 SvNOK_on(sv);
c2988b20 2450 } else {
c2988b20
NC
2451 /* It had no "." so it must be integer. */
2452 }
00b6aa41 2453 SvIOK_on(sv);
c2988b20
NC
2454 } else {
2455 /* between IV_MAX and NV(UV_MAX).
2456 Could be slightly > UV_MAX */
6fa402ec 2457
c2988b20
NC
2458 if (numtype & IS_NUMBER_NOT_INT) {
2459 /* UV and NV both imprecise. */
2460 } else {
66a1b24b 2461 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2462
2463 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2464 SvNOK_on(sv);
c2988b20 2465 }
00b6aa41 2466 SvIOK_on(sv);
c2988b20
NC
2467 }
2468 }
2469 }
2470 }
2471 }
28e5dec8 2472#endif /* NV_PRESERVES_UV */
93a17b20 2473 }
79072805 2474 else {
f7877b28 2475 if (isGV_with_GP(sv)) {
19f6321d 2476 glob_2number((GV *)sv);
180488f8
NC
2477 return 0.0;
2478 }
2479
041457d9 2480 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2481 report_uninit(sv);
7e25a7e9
NC
2482 assert (SvTYPE(sv) >= SVt_NV);
2483 /* Typically the caller expects that sv_any is not NULL now. */
2484 /* XXX Ilya implies that this is a bug in callers that assume this
2485 and ideally should be fixed. */
a0d0e21e 2486 return 0.0;
79072805 2487 }
572bbb43 2488#if defined(USE_LONG_DOUBLE)
097ee67d 2489 DEBUG_c({
f93f4e46 2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2492 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2493 RESTORE_NUMERIC_LOCAL();
2494 });
65202027 2495#else
572bbb43 2496 DEBUG_c({
f93f4e46 2497 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2498 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2499 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2500 RESTORE_NUMERIC_LOCAL();
2501 });
572bbb43 2502#endif
463ee0b2 2503 return SvNVX(sv);
79072805
LW
2504}
2505
645c22ef
DM
2506/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2507 * UV as a string towards the end of buf, and return pointers to start and
2508 * end of it.
2509 *
2510 * We assume that buf is at least TYPE_CHARS(UV) long.
2511 */
2512
864dbfa3 2513static char *
aec46f14 2514S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2515{
25da4f38 2516 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2517 char * const ebuf = ptr;
25da4f38 2518 int sign;
25da4f38
IZ
2519
2520 if (is_uv)
2521 sign = 0;
2522 else if (iv >= 0) {
2523 uv = iv;
2524 sign = 0;
2525 } else {
2526 uv = -iv;
2527 sign = 1;
2528 }
2529 do {
eb160463 2530 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2531 } while (uv /= 10);
2532 if (sign)
2533 *--ptr = '-';
2534 *peob = ebuf;
2535 return ptr;
2536}
2537
9af30d34
NC
2538/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2539 * a regexp to its stringified form.
2540 */
2541
2542static char *
2543S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
97aff369 2544 dVAR;
00b6aa41 2545 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2546
2547 if (!mg->mg_ptr) {
2548 const char *fptr = "msix";
2549 char reflags[6];
2550 char ch;
2551 int left = 0;
2552 int right = 4;
00b6aa41 2553 bool need_newline = 0;
9af30d34
NC
2554 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2555
2556 while((ch = *fptr++)) {
2557 if(reganch & 1) {
2558 reflags[left++] = ch;
2559 }
2560 else {
2561 reflags[right--] = ch;
2562 }
2563 reganch >>= 1;
2564 }
2565 if(left != 4) {
2566 reflags[left] = '-';
2567 left = 5;
2568 }
2569
2570 mg->mg_len = re->prelen + 4 + left;
2571 /*
2572 * If /x was used, we have to worry about a regex ending with a
2573 * comment later being embedded within another regex. If so, we don't
2574 * want this regex's "commentization" to leak out to the right part of
2575 * the enclosing regex, we must cap it with a newline.
2576 *
2577 * So, if /x was used, we scan backwards from the end of the regex. If
2578 * we find a '#' before we find a newline, we need to add a newline
2579 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2580 * we don't need to add anything. -jfriedl
2581 */
2582 if (PMf_EXTENDED & re->reganch) {
2583 const char *endptr = re->precomp + re->prelen;
2584 while (endptr >= re->precomp) {
2585 const char c = *(endptr--);
2586 if (c == '\n')
2587 break; /* don't need another */
2588 if (c == '#') {
2589 /* we end while in a comment, so we need a newline */
2590 mg->mg_len++; /* save space for it */
2591 need_newline = 1; /* note to add it */
2592 break;
2593 }
2594 }
2595 }
2596
2597 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2598 mg->mg_ptr[0] = '(';
2599 mg->mg_ptr[1] = '?';
2600 Copy(reflags, mg->mg_ptr+2, left, char);
2601 *(mg->mg_ptr+left+2) = ':';
2602 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2603 if (need_newline)
2604 mg->mg_ptr[mg->mg_len - 2] = '\n';
2605 mg->mg_ptr[mg->mg_len - 1] = ')';
2606 mg->mg_ptr[mg->mg_len] = 0;
2607 }
2608 PL_reginterp_cnt += re->program[0].next_off;
2609
2610 if (re->reganch & ROPT_UTF8)
2611 SvUTF8_on(sv);
2612 else
2613 SvUTF8_off(sv);
2614 if (lp)
2615 *lp = mg->mg_len;
2616 return mg->mg_ptr;
2617}
2618
645c22ef
DM
2619/*
2620=for apidoc sv_2pv_flags
2621
ff276b08 2622Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2623If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2624if necessary.
2625Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2626usually end up here too.
2627
2628=cut
2629*/
2630
8d6d96c1
HS
2631char *
2632Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2633{
97aff369 2634 dVAR;
79072805 2635 register char *s;
79072805 2636
463ee0b2 2637 if (!sv) {
cdb061a3
NC
2638 if (lp)
2639 *lp = 0;
73d840c0 2640 return (char *)"";
463ee0b2 2641 }
8990e307 2642 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2643 if (flags & SV_GMAGIC)
2644 mg_get(sv);
463ee0b2 2645 if (SvPOKp(sv)) {
cdb061a3
NC
2646 if (lp)
2647 *lp = SvCUR(sv);
10516c54
NC
2648 if (flags & SV_MUTABLE_RETURN)
2649 return SvPVX_mutable(sv);
4d84ee25
NC
2650 if (flags & SV_CONST_RETURN)
2651 return (char *)SvPVX_const(sv);
463ee0b2
LW
2652 return SvPVX(sv);
2653 }
75dfc8ec
NC
2654 if (SvIOKp(sv) || SvNOKp(sv)) {
2655 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2656 STRLEN len;
2657
2658 if (SvIOKp(sv)) {
e8ada2d0
NC
2659 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2660 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2661 } else {
e8ada2d0
NC
2662 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2663 len = strlen(tbuf);
75dfc8ec 2664 }
b5b886f0
NC
2665 assert(!SvROK(sv));
2666 {
75dfc8ec
NC
2667 dVAR;
2668
2669#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2670 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2671 tbuf[0] = '0';
2672 tbuf[1] = 0;
75dfc8ec
NC
2673 len = 1;
2674 }
2675#endif
2676 SvUPGRADE(sv, SVt_PV);
2677 if (lp)
2678 *lp = len;
2679 s = SvGROW_mutable(sv, len + 1);
2680 SvCUR_set(sv, len);
2681 SvPOKp_on(sv);
e8ada2d0 2682 return memcpy(s, tbuf, len + 1);
75dfc8ec 2683 }
463ee0b2 2684 }
1c7ff15e
NC
2685 if (SvROK(sv)) {
2686 goto return_rok;
2687 }
2688 assert(SvTYPE(sv) >= SVt_PVMG);
2689 /* This falls through to the report_uninit near the end of the
2690 function. */
2691 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2692 if (SvROK(sv)) {
1c7ff15e 2693 return_rok:
deb46114
NC
2694 if (SvAMAGIC(sv)) {
2695 SV *const tmpstr = AMG_CALLun(sv,string);
2696 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2697 /* Unwrap this: */
2698 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2699 */
2700
2701 char *pv;
2702 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2703 if (flags & SV_CONST_RETURN) {
2704 pv = (char *) SvPVX_const(tmpstr);
2705 } else {
2706 pv = (flags & SV_MUTABLE_RETURN)
2707 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2708 }
2709 if (lp)
2710 *lp = SvCUR(tmpstr);
50adf7d2 2711 } else {
deb46114 2712 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2713 }
deb46114
NC
2714 if (SvUTF8(tmpstr))
2715 SvUTF8_on(sv);
2716 else
2717 SvUTF8_off(sv);
2718 return pv;
50adf7d2 2719 }
deb46114
NC
2720 }
2721 {
75dfc8ec 2722 SV *tsv;
f9277f47 2723 MAGIC *mg;
d8eae41e
NC
2724 const SV *const referent = (SV*)SvRV(sv);
2725
2726 if (!referent) {
396482e1 2727 tsv = sv_2mortal(newSVpvs("NULLREF"));
042dae7a
NC
2728 } else if (SvTYPE(referent) == SVt_PVMG
2729 && ((SvFLAGS(referent) &
2730 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2731 == (SVs_OBJECT|SVs_SMG))
2732 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
c445ea15 2733 return stringify_regexp(sv, mg, lp);
d8eae41e
NC
2734 } else {
2735 const char *const typestr = sv_reftype(referent, 0);
2736
2737 tsv = sv_newmortal();
2738 if (SvOBJECT(referent)) {
2739 const char *const name = HvNAME_get(SvSTASH(referent));
2740 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2741 name ? name : "__ANON__" , typestr,
2742 PTR2UV(referent));
2743 }
2744 else
2745 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2746 PTR2UV(referent));
c080367d 2747 }
042dae7a
NC
2748 if (lp)
2749 *lp = SvCUR(tsv);
2750 return SvPVX(tsv);
463ee0b2 2751 }
79072805 2752 }
0336b60e 2753 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2754 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2755 report_uninit(sv);
cdb061a3
NC
2756 if (lp)
2757 *lp = 0;
73d840c0 2758 return (char *)"";
79072805 2759 }
79072805 2760 }
28e5dec8
JH
2761 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2762 /* I'm assuming that if both IV and NV are equally valid then
2763 converting the IV is going to be more efficient */
e1ec3a88
AL
2764 const U32 isIOK = SvIOK(sv);
2765 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2766 char buf[TYPE_CHARS(UV)];
2767 char *ebuf, *ptr;
2768
2769 if (SvTYPE(sv) < SVt_PVIV)
2770 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2771 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2772 /* inlined from sv_setpvn */
2773 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2774 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2775 SvCUR_set(sv, ebuf - ptr);
2776 s = SvEND(sv);
2777 *s = '\0';
2778 if (isIOK)
2779 SvIOK_on(sv);
2780 else
2781 SvIOKp_on(sv);
2782 if (isUIOK)
2783 SvIsUV_on(sv);
2784 }
2785 else if (SvNOKp(sv)) {
c81271c3 2786 const int olderrno = errno;
79072805
LW
2787 if (SvTYPE(sv) < SVt_PVNV)
2788 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2789 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2790 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2791 /* some Xenix systems wipe out errno here */
79072805 2792#ifdef apollo
463ee0b2 2793 if (SvNVX(sv) == 0.0)
79072805
LW
2794 (void)strcpy(s,"0");
2795 else
2796#endif /*apollo*/
bbce6d69 2797 {
2d4389e4 2798 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2799 }
79072805 2800 errno = olderrno;
a0d0e21e
LW
2801#ifdef FIXNEGATIVEZERO
2802 if (*s == '-' && s[1] == '0' && !s[2])
2803 strcpy(s,"0");
2804#endif
79072805
LW
2805 while (*s) s++;
2806#ifdef hcx
2807 if (s[-1] == '.')
46fc3d4c 2808 *--s = '\0';
79072805
LW
2809#endif
2810 }
79072805 2811 else {
675c862f 2812 if (isGV_with_GP(sv))
19f6321d 2813 return glob_2pv((GV *)sv, lp);
180488f8 2814
041457d9 2815 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2816 report_uninit(sv);
cdb061a3 2817 if (lp)
00b6aa41 2818 *lp = 0;
25da4f38
IZ
2819 if (SvTYPE(sv) < SVt_PV)
2820 /* Typically the caller expects that sv_any is not NULL now. */
2821 sv_upgrade(sv, SVt_PV);
73d840c0 2822 return (char *)"";
79072805 2823 }
cdb061a3 2824 {
823a54a3 2825 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2826 if (lp)
2827 *lp = len;
2828 SvCUR_set(sv, len);
2829 }
79072805 2830 SvPOK_on(sv);
1d7c1841 2831 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2832 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2833 if (flags & SV_CONST_RETURN)
2834 return (char *)SvPVX_const(sv);
10516c54
NC
2835 if (flags & SV_MUTABLE_RETURN)
2836 return SvPVX_mutable(sv);
463ee0b2
LW
2837 return SvPVX(sv);
2838}
2839
645c22ef 2840/*
6050d10e
JP
2841=for apidoc sv_copypv
2842
2843Copies a stringified representation of the source SV into the
2844destination SV. Automatically performs any necessary mg_get and
54f0641b 2845coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2846UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2847sv_2pv[_flags] but operates directly on an SV instead of just the
2848string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2849would lose the UTF-8'ness of the PV.
2850
2851=cut
2852*/
2853
2854void
2855Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2856{
446eaa42 2857 STRLEN len;
53c1dcc0 2858 const char * const s = SvPV_const(ssv,len);
cb50f42d 2859 sv_setpvn(dsv,s,len);
446eaa42 2860 if (SvUTF8(ssv))
cb50f42d 2861 SvUTF8_on(dsv);
446eaa42 2862 else
cb50f42d 2863 SvUTF8_off(dsv);
6050d10e
JP
2864}
2865
2866/*
645c22ef
DM
2867=for apidoc sv_2pvbyte
2868
2869Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2870to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2871side-effect.
2872
2873Usually accessed via the C<SvPVbyte> macro.
2874
2875=cut
2876*/
2877
7340a771
GS
2878char *
2879Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2880{
0875d2fe 2881 sv_utf8_downgrade(sv,0);
97972285 2882 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2883}
2884
645c22ef 2885/*
035cbb0e
RGS
2886=for apidoc sv_2pvutf8
2887
2888Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2889to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2890
2891Usually accessed via the C<SvPVutf8> macro.
2892
2893=cut
2894*/
645c22ef 2895
7340a771
GS
2896char *
2897Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2898{
035cbb0e
RGS
2899 sv_utf8_upgrade(sv);
2900 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2901}
1c846c1f 2902
7ee2227d 2903
645c22ef
DM
2904/*
2905=for apidoc sv_2bool
2906
2907This function is only called on magical items, and is only used by
8cf8f3d1 2908sv_true() or its macro equivalent.
645c22ef
DM
2909
2910=cut
2911*/
2912
463ee0b2 2913bool
864dbfa3 2914Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2915{
97aff369 2916 dVAR;
5b295bef 2917 SvGETMAGIC(sv);
463ee0b2 2918
a0d0e21e
LW
2919 if (!SvOK(sv))
2920 return 0;
2921 if (SvROK(sv)) {
fabdb6c0
AL
2922 if (SvAMAGIC(sv)) {
2923 SV * const tmpsv = AMG_CALLun(sv,bool_);
2924 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2925 return (bool)SvTRUE(tmpsv);
2926 }
2927 return SvRV(sv) != 0;
a0d0e21e 2928 }
463ee0b2 2929 if (SvPOKp(sv)) {
53c1dcc0
AL
2930 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2931 if (Xpvtmp &&
339049b0 2932 (*sv->sv_u.svu_pv > '0' ||
11343788 2933 Xpvtmp->xpv_cur > 1 ||
339049b0 2934 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2935 return 1;
2936 else
2937 return 0;
2938 }
2939 else {
2940 if (SvIOKp(sv))
2941 return SvIVX(sv) != 0;
2942 else {
2943 if (SvNOKp(sv))
2944 return SvNVX(sv) != 0.0;
180488f8 2945 else {
f7877b28 2946 if (isGV_with_GP(sv))
180488f8
NC
2947 return TRUE;
2948 else
2949 return FALSE;
2950 }
463ee0b2
LW
2951 }
2952 }
79072805
LW
2953}
2954
c461cf8f
JH
2955/*
2956=for apidoc sv_utf8_upgrade
2957
78ea37eb 2958Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2959Forces the SV to string form if it is not already.
4411f3b6
NIS
2960Always sets the SvUTF8 flag to avoid future validity checks even
2961if all the bytes have hibit clear.
c461cf8f 2962
13a6c0e0
JH
2963This is not as a general purpose byte encoding to Unicode interface:
2964use the Encode extension for that.
2965
8d6d96c1
HS
2966=for apidoc sv_utf8_upgrade_flags
2967
78ea37eb 2968Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2969Forces the SV to string form if it is not already.
8d6d96c1
HS
2970Always sets the SvUTF8 flag to avoid future validity checks even
2971if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2972will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2973C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2974
13a6c0e0
JH
2975This is not as a general purpose byte encoding to Unicode interface:
2976use the Encode extension for that.
2977
8d6d96c1
HS
2978=cut
2979*/
2980
2981STRLEN
2982Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2983{
97aff369 2984 dVAR;
808c356f
RGS
2985 if (sv == &PL_sv_undef)
2986 return 0;
e0e62c2a
NIS
2987 if (!SvPOK(sv)) {
2988 STRLEN len = 0;
d52b7888
NC
2989 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2990 (void) sv_2pv_flags(sv,&len, flags);
2991 if (SvUTF8(sv))
2992 return len;
2993 } else {
2994 (void) SvPV_force(sv,len);
2995 }
e0e62c2a 2996 }
4411f3b6 2997
f5cee72b 2998 if (SvUTF8(sv)) {
5fec3b1d 2999 return SvCUR(sv);
f5cee72b 3000 }
5fec3b1d 3001
765f542d
NC
3002 if (SvIsCOW(sv)) {
3003 sv_force_normal_flags(sv, 0);
db42d148
NIS
3004 }
3005
88632417 3006 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 3007 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 3008 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
3009 /* This function could be much more efficient if we
3010 * had a FLAG in SVs to signal if there are any hibit
3011 * chars in the PV. Given that there isn't such a flag
3012 * make the loop as fast as possible. */
00b6aa41 3013 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3014 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3015 const U8 *t = s;
c4e7c712
NC
3016
3017 while (t < e) {
53c1dcc0 3018 const U8 ch = *t++;
00b6aa41
AL
3019 /* Check for hi bit */
3020 if (!NATIVE_IS_INVARIANT(ch)) {
3021 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3022 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3023
3024 SvPV_free(sv); /* No longer using what was there before. */
3025 SvPV_set(sv, (char*)recoded);
3026 SvCUR_set(sv, len - 1);
3027 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3028 break;
00b6aa41 3029 }
c4e7c712
NC
3030 }
3031 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3032 SvUTF8_on(sv);
560a288e 3033 }
4411f3b6 3034 return SvCUR(sv);
560a288e
GS
3035}
3036
c461cf8f
JH
3037/*
3038=for apidoc sv_utf8_downgrade
3039
78ea37eb
TS
3040Attempts to convert the PV of an SV from characters to bytes.
3041If the PV contains a character beyond byte, this conversion will fail;
3042in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3043true, croaks.
3044
13a6c0e0
JH
3045This is not as a general purpose Unicode to byte encoding interface:
3046use the Encode extension for that.
3047
c461cf8f
JH
3048=cut
3049*/
3050
560a288e
GS
3051bool
3052Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3053{
97aff369 3054 dVAR;
78ea37eb 3055 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3056 if (SvCUR(sv)) {
03cfe0ae 3057 U8 *s;
652088fc 3058 STRLEN len;
fa301091 3059
765f542d
NC
3060 if (SvIsCOW(sv)) {
3061 sv_force_normal_flags(sv, 0);
3062 }
03cfe0ae
NIS
3063 s = (U8 *) SvPV(sv, len);
3064 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3065 if (fail_ok)
3066 return FALSE;
3067 else {
3068 if (PL_op)
3069 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3070 OP_DESC(PL_op));
fa301091
JH
3071 else
3072 Perl_croak(aTHX_ "Wide character");
3073 }
4b3603a4 3074 }
b162af07 3075 SvCUR_set(sv, len);
67e989fb 3076 }
560a288e 3077 }
ffebcc3e 3078 SvUTF8_off(sv);
560a288e
GS
3079 return TRUE;
3080}
3081
c461cf8f
JH
3082/*
3083=for apidoc sv_utf8_encode
3084
78ea37eb
TS
3085Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3086flag off so that it looks like octets again.
c461cf8f
JH
3087
3088=cut
3089*/
3090
560a288e
GS
3091void
3092Perl_sv_utf8_encode(pTHX_ register SV *sv)
3093{
4411f3b6 3094 (void) sv_utf8_upgrade(sv);
4c94c214
NC
3095 if (SvIsCOW(sv)) {
3096 sv_force_normal_flags(sv, 0);
3097 }
3098 if (SvREADONLY(sv)) {
3099 Perl_croak(aTHX_ PL_no_modify);
3100 }
560a288e
GS
3101 SvUTF8_off(sv);
3102}
3103
4411f3b6
NIS
3104/*
3105=for apidoc sv_utf8_decode
3106
78ea37eb
TS
3107If the PV of the SV is an octet sequence in UTF-8
3108and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3109so that it looks like a character. If the PV contains only single-byte
3110characters, the C<SvUTF8> flag stays being off.
3111Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3112
3113=cut
3114*/
3115
560a288e
GS
3116bool
3117Perl_sv_utf8_decode(pTHX_ register SV *sv)
3118{
78ea37eb 3119 if (SvPOKp(sv)) {
93524f2b
NC
3120 const U8 *c;
3121 const U8 *e;
9cbac4c7 3122
645c22ef
DM
3123 /* The octets may have got themselves encoded - get them back as
3124 * bytes
3125 */
3126 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3127 return FALSE;
3128
3129 /* it is actually just a matter of turning the utf8 flag on, but
3130 * we want to make sure everything inside is valid utf8 first.
3131 */
93524f2b 3132 c = (const U8 *) SvPVX_const(sv);
63cd0674 3133 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3134 return FALSE;
93524f2b 3135 e = (const U8 *) SvEND(sv);
511c2ff0 3136 while (c < e) {
b64e5050 3137 const U8 ch = *c++;
c4d5f83a 3138 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3139 SvUTF8_on(sv);
3140 break;
3141 }
560a288e 3142 }
560a288e
GS
3143 }
3144 return TRUE;
3145}
3146
954c1994
GS
3147/*
3148=for apidoc sv_setsv
3149
645c22ef
DM
3150Copies the contents of the source SV C<ssv> into the destination SV
3151C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3152function if the source SV needs to be reused. Does not handle 'set' magic.
3153Loosely speaking, it performs a copy-by-value, obliterating any previous
3154content of the destination.
3155
3156You probably want to use one of the assortment of wrappers, such as
3157C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3158C<SvSetMagicSV_nosteal>.
3159
8d6d96c1
HS
3160=for apidoc sv_setsv_flags
3161
645c22ef
DM
3162Copies the contents of the source SV C<ssv> into the destination SV
3163C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3164function if the source SV needs to be reused. Does not handle 'set' magic.
3165Loosely speaking, it performs a copy-by-value, obliterating any previous
3166content of the destination.
3167If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3168C<ssv> if appropriate, else not. If the C<flags> parameter has the
3169C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3170and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3171
3172You probably want to use one of the assortment of wrappers, such as
3173C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3174C<SvSetMagicSV_nosteal>.
3175
3176This is the primary function for copying scalars, and most other
3177copy-ish functions and macros use this underneath.
8d6d96c1
HS
3178
3179=cut
3180*/
3181
5d0301b7 3182static void
2eb42952 3183S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3184{
3185 if (dtype != SVt_PVGV) {
3186 const char * const name = GvNAME(sstr);
3187 const STRLEN len = GvNAMELEN(sstr);
3188 /* don't upgrade SVt_PVLV: it can hold a glob */
f7877b28
NC
3189 if (dtype != SVt_PVLV) {
3190 if (dtype >= SVt_PV) {
3191 SvPV_free(dstr);
3192 SvPV_set(dstr, 0);
3193 SvLEN_set(dstr, 0);
3194 SvCUR_set(dstr, 0);
3195 }
5d0301b7 3196 sv_upgrade(dstr, SVt_PVGV);
dedf8e73
NC
3197 (void)SvOK_off(dstr);
3198 SvSCREAM_on(dstr);
f7877b28 3199 }
5d0301b7
NC
3200 GvSTASH(dstr) = GvSTASH(sstr);
3201 if (GvSTASH(dstr))
3202 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3203 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3204 SvFAKE_on(dstr); /* can coerce to non-glob */
3205 }
3206
3207#ifdef GV_UNIQUE_CHECK
3208 if (GvUNIQUE((GV*)dstr)) {
3209 Perl_croak(aTHX_ PL_no_modify);
3210 }
3211#endif
3212
f7877b28
NC
3213 gp_free((GV*)dstr);
3214 SvSCREAM_off(dstr);
5d0301b7 3215 (void)SvOK_off(dstr);
f7877b28 3216 SvSCREAM_on(dstr);
dedf8e73 3217 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3218 GvGP(dstr) = gp_ref(GvGP(sstr));
3219 if (SvTAINTED(sstr))
3220 SvTAINT(dstr);
3221 if (GvIMPORTED(dstr) != GVf_IMPORTED
3222 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3223 {
3224 GvIMPORTED_on(dstr);
3225 }
3226 GvMULTI_on(dstr);
3227 return;
3228}
3229
b8473700 3230static void
2eb42952 3231S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3232 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3233 SV *dref = NULL;
3234 const int intro = GvINTRO(dstr);
2440974c 3235 SV **location;
3386d083 3236 U8 import_flag = 0;
27242d61
NC
3237 const U32 stype = SvTYPE(sref);
3238
b8473700
NC
3239
3240#ifdef GV_UNIQUE_CHECK
3241 if (GvUNIQUE((GV*)dstr)) {
3242 Perl_croak(aTHX_ PL_no_modify);
3243 }
3244#endif
3245
3246 if (intro) {
3247 GvINTRO_off(dstr); /* one-shot flag */
3248 GvLINE(dstr) = CopLINE(PL_curcop);
3249 GvEGV(dstr) = (GV*)dstr;
3250 }
3251 GvMULTI_on(dstr);
27242d61 3252 switch (stype) {
b8473700 3253 case SVt_PVCV:
27242d61
NC
3254 location = (SV **) &GvCV(dstr);
3255 import_flag = GVf_IMPORTED_CV;
3256 goto common;
3257 case SVt_PVHV:
3258 location = (SV **) &GvHV(dstr);
3259 import_flag = GVf_IMPORTED_HV;
3260 goto common;
3261 case SVt_PVAV:
3262 location = (SV **) &GvAV(dstr);
3263 import_flag = GVf_IMPORTED_AV;
3264 goto common;
3265 case SVt_PVIO:
3266 location = (SV **) &GvIOp(dstr);
3267 goto common;
3268 case SVt_PVFM:
3269 location = (SV **) &GvFORM(dstr);
3270 default:
3271 location = &GvSV(dstr);
3272 import_flag = GVf_IMPORTED_SV;
3273 common:
b8473700 3274 if (intro) {
27242d61
NC
3275 if (stype == SVt_PVCV) {
3276 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3277 SvREFCNT_dec(GvCV(dstr));
3278 GvCV(dstr) = NULL;
3279 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3280 PL_sub_generation++;
3281 }
b8473700 3282 }
27242d61 3283 SAVEGENERICSV(*location);
b8473700
NC
3284 }
3285 else
27242d61
NC
3286 dref = *location;
3287 if (stype == SVt_PVCV && *location != sref) {
3288 CV* const cv = (CV*)*location;
b8473700
NC
3289 if (cv) {
3290 if (!GvCVGEN((GV*)dstr) &&
3291 (CvROOT(cv) || CvXSUB(cv)))
3292 {
3293 /* Redefining a sub - warning is mandatory if
3294 it was a const and its value changed. */
3295 if (CvCONST(cv) && CvCONST((CV*)sref)
3296 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3297 NOOP;
b8473700
NC
3298 /* They are 2 constant subroutines generated from
3299 the same constant. This probably means that
3300 they are really the "same" proxy subroutine
3301 instantiated in 2 places. Most likely this is
3302 when a constant is exported twice. Don't warn.
3303 */
3304 }
3305 else if (ckWARN(WARN_REDEFINE)
3306 || (CvCONST(cv)
3307 && (!CvCONST((CV*)sref)
3308 || sv_cmp(cv_const_sv(cv),
3309 cv_const_sv((CV*)sref))))) {
3310 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3311 CvCONST(cv)
3312 ? "Constant subroutine %s::%s redefined"
3313 : "Subroutine %s::%s redefined",
3314 HvNAME_get(GvSTASH((GV*)dstr)),
3315 GvENAME((GV*)dstr));
3316 }
3317 }
3318 if (!intro)
cbf82dd0
NC
3319 cv_ckproto_len(cv, (GV*)dstr,
3320 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3321 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3322 }
b8473700
NC
3323 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3324 GvASSUMECV_on(dstr);
3325 PL_sub_generation++;
3326 }
2440974c 3327 *location = sref;
3386d083
NC
3328 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3329 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3330 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3331 }
3332 break;
3333 }
b37c2d43 3334 SvREFCNT_dec(dref);
b8473700
NC
3335 if (SvTAINTED(sstr))
3336 SvTAINT(dstr);
3337 return;
3338}
3339
8d6d96c1
HS
3340void
3341Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3342{
97aff369 3343 dVAR;
8990e307
LW
3344 register U32 sflags;
3345 register int dtype;
3346 register int stype;
463ee0b2 3347
79072805
LW
3348 if (sstr == dstr)
3349 return;
765f542d 3350 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3351 if (!sstr)
3280af22 3352 sstr = &PL_sv_undef;
8990e307
LW
3353 stype = SvTYPE(sstr);
3354 dtype = SvTYPE(dstr);
79072805 3355
a0d0e21e 3356 SvAMAGIC_off(dstr);
7a5fa8a2 3357 if ( SvVOK(dstr) )
ece467f9
JP
3358 {
3359 /* need to nuke the magic */
3360 mg_free(dstr);
3361 SvRMAGICAL_off(dstr);
3362 }
9e7bc3e8 3363
463ee0b2 3364 /* There's a lot of redundancy below but we're going for speed here */
79072805 3365
8990e307 3366 switch (stype) {
79072805 3367 case SVt_NULL:
aece5585 3368 undef_sstr:
20408e3c
GS
3369 if (dtype != SVt_PVGV) {
3370 (void)SvOK_off(dstr);
3371 return;
3372 }
3373 break;
463ee0b2 3374 case SVt_IV:
aece5585
GA
3375 if (SvIOK(sstr)) {
3376 switch (dtype) {
3377 case SVt_NULL:
8990e307 3378 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3379 break;
3380 case SVt_NV:
aece5585
GA
3381 case SVt_RV:
3382 case SVt_PV:
a0d0e21e 3383 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
3384 break;
3385 }
3386 (void)SvIOK_only(dstr);
45977657 3387 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3388 if (SvIsUV(sstr))
3389 SvIsUV_on(dstr);
37c25af0
NC
3390 /* SvTAINTED can only be true if the SV has taint magic, which in
3391 turn means that the SV type is PVMG (or greater). This is the
3392 case statement for SVt_IV, so this cannot be true (whatever gcov
3393 may say). */
3394 assert(!SvTAINTED(sstr));
aece5585 3395 return;
8990e307 3396 }
aece5585
GA
3397 goto undef_sstr;
3398
463ee0b2 3399 case SVt_NV:
aece5585
GA
3400 if (SvNOK(sstr)) {
3401 switch (dtype) {
3402 case SVt_NULL:
3403 case SVt_IV:
8990e307 3404 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3405 break;
3406 case SVt_RV:
3407 case SVt_PV:
3408 case SVt_PVIV:
a0d0e21e 3409 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
3410 break;
3411 }
9d6ce603 3412 SvNV_set(dstr, SvNVX(sstr));
aece5585 3413 (void)SvNOK_only(dstr);
37c25af0
NC
3414 /* SvTAINTED can only be true if the SV has taint magic, which in
3415 turn means that the SV type is PVMG (or greater). This is the
3416 case statement for SVt_NV, so this cannot be true (whatever gcov
3417 may say). */
3418 assert(!SvTAINTED(sstr));
aece5585 3419 return;
8990e307 3420 }
aece5585
GA
3421 goto undef_sstr;
3422
ed6116ce 3423 case SVt_RV:
8990e307 3424 if (dtype < SVt_RV)
ed6116ce 3425 sv_upgrade(dstr, SVt_RV);
ed6116ce 3426 break;
fc36a67e 3427 case SVt_PVFM:
f8c7b90f 3428#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3429 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3430 if (dtype < SVt_PVIV)
3431 sv_upgrade(dstr, SVt_PVIV);
3432 break;
3433 }
3434 /* Fall through */
3435#endif
3436 case SVt_PV:
8990e307 3437 if (dtype < SVt_PV)
463ee0b2 3438 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3439 break;
3440 case SVt_PVIV:
8990e307 3441 if (dtype < SVt_PVIV)
463ee0b2 3442 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3443 break;
3444 case SVt_PVNV:
8990e307 3445 if (dtype < SVt_PVNV)
463ee0b2 3446 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3447 break;
489f7bfe 3448 default:
a3b680e6
AL
3449 {
3450 const char * const type = sv_reftype(sstr,0);
533c011a 3451 if (PL_op)
a3b680e6 3452 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3453 else
a3b680e6
AL
3454 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3455 }
4633a7c4
LW
3456 break;
3457
79072805 3458 case SVt_PVGV:
8990e307 3459 if (dtype <= SVt_PVGV) {
d4c19fe8 3460 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3461 return;
79072805 3462 }
5f66b61c 3463 /*FALLTHROUGH*/
79072805 3464
489f7bfe
NC
3465 case SVt_PVMG:
3466 case SVt_PVLV:
3467 case SVt_PVBM:
8d6d96c1 3468 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3469 mg_get(sstr);
eb160463 3470 if ((int)SvTYPE(sstr) != stype) {
973f89ab 3471 stype = SvTYPE(sstr);
b8c701c1 3472 if (stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3473 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3474 return;
3475 }
973f89ab
CS
3476 }
3477 }
ded42b9f 3478 if (stype == SVt_PVLV)
862a34c6 3479 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3480 else
862a34c6 3481 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3482 }
3483
ff920335
NC
3484 /* dstr may have been upgraded. */
3485 dtype = SvTYPE(dstr);
8990e307
LW
3486 sflags = SvFLAGS(sstr);
3487
3488 if (sflags & SVf_ROK) {
acaa9288
NC
3489 if (dtype == SVt_PVGV &&
3490 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
3491 sstr = SvRV(sstr);
3492 if (sstr == dstr) {
3493 if (GvIMPORTED(dstr) != GVf_IMPORTED
3494 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3495 {
3496 GvIMPORTED_on(dstr);
3497 }
3498 GvMULTI_on(dstr);
3499 return;
3500 }
d4c19fe8 3501 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3502 return;
3503 }
3504
8990e307 3505 if (dtype >= SVt_PV) {
b8c701c1 3506 if (dtype == SVt_PVGV) {
d4c19fe8 3507 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3508 return;
3509 }
3f7c398e 3510 if (SvPVX_const(dstr)) {
8bd4d4c5 3511 SvPV_free(dstr);
b162af07
SP
3512 SvLEN_set(dstr, 0);
3513 SvCUR_set(dstr, 0);
a0d0e21e 3514 }
8990e307 3515 }
a0d0e21e 3516 (void)SvOK_off(dstr);
b162af07 3517 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
dfd48732
NC
3518 SvFLAGS(dstr) |= sflags & (SVf_ROK|SVf_AMAGIC);
3519 assert(!(sflags & SVp_NOK));
3520 assert(!(sflags & SVp_IOK));
3521 assert(!(sflags & SVf_NOK));
3522 assert(!(sflags & SVf_IOK));
ed6116ce 3523 }
c0c44674
NC
3524 else if (dtype == SVt_PVGV) {
3525 if (!(sflags & SVf_OK)) {
3526 if (ckWARN(WARN_MISC))
3527 Perl_warner(aTHX_ packWARN(WARN_MISC),
3528 "Undefined value assigned to typeglob");
3529 }
3530 else {
3531 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3532 if (dstr != (SV*)gv) {
3533 if (GvGP(dstr))
3534 gp_free((GV*)dstr);
3535 GvGP(dstr) = gp_ref(GvGP(gv));
3536 }
3537 }
3538 }
8990e307 3539 else if (sflags & SVp_POK) {
765f542d 3540 bool isSwipe = 0;
79072805
LW
3541
3542 /*
3543 * Check to see if we can just swipe the string. If so, it's a
3544 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3545 * It might even be a win on short strings if SvPVX_const(dstr)
3546 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3547 */
3548
120fac95
NC
3549 /* Whichever path we take through the next code, we want this true,
3550 and doing it now facilitates the COW check. */
3551 (void)SvPOK_only(dstr);
3552
765f542d 3553 if (
b8f9541a
NC
3554 /* We're not already COW */
3555 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3556#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3557 /* or we are, but dstr isn't a suitable target. */
3558 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3559#endif
3560 )
765f542d 3561 &&
765f542d
NC
3562 !(isSwipe =
3563 (sflags & SVs_TEMP) && /* slated for free anyway? */
3564 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3565 (!(flags & SV_NOSTEAL)) &&
3566 /* and we're allowed to steal temps */
765f542d
NC
3567 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3568 SvLEN(sstr) && /* and really is a string */
645c22ef 3569 /* and won't be needed again, potentially */
765f542d 3570 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3571#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3572 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3573 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3574 && SvTYPE(sstr) >= SVt_PVIV)
3575#endif
3576 ) {
3577 /* Failed the swipe test, and it's not a shared hash key either.
3578 Have to copy the string. */
3579 STRLEN len = SvCUR(sstr);
3580 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3581 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3582 SvCUR_set(dstr, len);
3583 *SvEND(dstr) = '\0';
765f542d 3584 } else {
f8c7b90f 3585 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3586 be true in here. */
765f542d
NC
3587 /* Either it's a shared hash key, or it's suitable for
3588 copy-on-write or we can swipe the string. */
46187eeb 3589 if (DEBUG_C_TEST) {
ed252734 3590 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3591 sv_dump(sstr);
3592 sv_dump(dstr);
46187eeb 3593 }
f8c7b90f 3594#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3595 if (!isSwipe) {
3596 /* I believe I should acquire a global SV mutex if
3597 it's a COW sv (not a shared hash key) to stop
3598 it going un copy-on-write.
3599 If the source SV has gone un copy on write between up there
3600 and down here, then (assert() that) it is of the correct
3601 form to make it copy on write again */
3602 if ((sflags & (SVf_FAKE | SVf_READONLY))
3603 != (SVf_FAKE | SVf_READONLY)) {
3604 SvREADONLY_on(sstr);
3605 SvFAKE_on(sstr);
3606 /* Make the source SV into a loop of 1.
3607 (about to become 2) */
a29f6d03 3608 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3609 }
3610 }
3611#endif
3612 /* Initial code is common. */
94010e71
NC
3613 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3614 SvPV_free(dstr);
79072805 3615 }
765f542d 3616
765f542d
NC
3617 if (!isSwipe) {
3618 /* making another shared SV. */
3619 STRLEN cur = SvCUR(sstr);
3620 STRLEN len = SvLEN(sstr);
f8c7b90f 3621#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3622 if (len) {
b8f9541a 3623 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3624 /* SvIsCOW_normal */
3625 /* splice us in between source and next-after-source. */
a29f6d03
NC
3626 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3627 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3628 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3629 } else
3630#endif
3631 {
765f542d 3632 /* SvIsCOW_shared_hash */
46187eeb
NC
3633 DEBUG_C(PerlIO_printf(Perl_debug_log,
3634 "Copy on write: Sharing hash\n"));
b8f9541a 3635
bdd68bc3 3636 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3637 SvPV_set(dstr,
d1db91c6 3638 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3639 }
87a1ef3d
SP
3640 SvLEN_set(dstr, len);
3641 SvCUR_set(dstr, cur);
765f542d
NC
3642 SvREADONLY_on(dstr);
3643 SvFAKE_on(dstr);
3644 /* Relesase a global SV mutex. */
3645 }
3646 else
765f542d 3647 { /* Passes the swipe test. */
78d1e721 3648 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3649 SvLEN_set(dstr, SvLEN(sstr));
3650 SvCUR_set(dstr, SvCUR(sstr));
3651
3652 SvTEMP_off(dstr);
3653 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3654 SvPV_set(sstr, NULL);
765f542d
NC
3655 SvLEN_set(sstr, 0);
3656 SvCUR_set(sstr, 0);
3657 SvTEMP_off(sstr);
3658 }
3659 }
8990e307 3660 if (sflags & SVp_NOK) {
9d6ce603 3661 SvNV_set(dstr, SvNVX(sstr));
79072805 3662 }
8990e307 3663 if (sflags & SVp_IOK) {
23525414
NC
3664 SvRELEASE_IVX(dstr);
3665 SvIV_set(dstr, SvIVX(sstr));
3666 /* Must do this otherwise some other overloaded use of 0x80000000
3667 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3668 if (sflags & SVf_IVisUV)
25da4f38 3669 SvIsUV_on(dstr);
79072805 3670 }
dd2eae66
NC
3671 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8
3672 |SVf_AMAGIC);
4f2da183
NC
3673 {
3674 const MAGIC * const smg = SvVOK(sstr);
3675 if (smg) {
3676 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3677 smg->mg_ptr, smg->mg_len);
3678 SvRMAGICAL_on(dstr);
3679 }
7a5fa8a2 3680 }
79072805 3681 }
5d581361 3682 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3683 (void)SvOK_off(dstr);
dd2eae66
NC
3684 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK
3685 |SVf_AMAGIC);
5d581361
NC
3686 if (sflags & SVp_IOK) {
3687 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3688 SvIV_set(dstr, SvIVX(sstr));
3689 }
3332b3c1 3690 if (sflags & SVp_NOK) {
9d6ce603 3691 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3692 }
3693 }
79072805 3694 else {
f7877b28 3695 if (isGV_with_GP(sstr)) {
180488f8
NC
3696 /* This stringification rule for globs is spread in 3 places.
3697 This feels bad. FIXME. */
3698 const U32 wasfake = sflags & SVf_FAKE;
3699
3700 /* FAKE globs can get coerced, so need to turn this off
3701 temporarily if it is on. */
3702 SvFAKE_off(sstr);
3703 gv_efullname3(dstr, (GV *)sstr, "*");
3704 SvFLAGS(sstr) |= wasfake;
dd2eae66 3705 SvFLAGS(dstr) |= sflags & SVf_AMAGIC;
180488f8 3706 }
20408e3c
GS
3707 else
3708 (void)SvOK_off(dstr);
a0d0e21e 3709 }
27c9684d
AP
3710 if (SvTAINTED(sstr))
3711 SvTAINT(dstr);
79072805
LW
3712}
3713
954c1994
GS
3714/*
3715=for apidoc sv_setsv_mg
3716
3717Like C<sv_setsv>, but also handles 'set' magic.
3718
3719=cut
3720*/
3721
79072805 3722void
864dbfa3 3723Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3724{
3725 sv_setsv(dstr,sstr);
3726 SvSETMAGIC(dstr);
3727}
3728
f8c7b90f 3729#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3730SV *
3731Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3732{
3733 STRLEN cur = SvCUR(sstr);
3734 STRLEN len = SvLEN(sstr);
3735 register char *new_pv;
3736
3737 if (DEBUG_C_TEST) {
3738 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3739 sstr, dstr);
3740 sv_dump(sstr);
3741 if (dstr)
3742 sv_dump(dstr);
3743 }
3744
3745 if (dstr) {
3746 if (SvTHINKFIRST(dstr))
3747 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3748 else if (SvPVX_const(dstr))
3749 Safefree(SvPVX_const(dstr));
ed252734
NC
3750 }
3751 else
3752 new_SV(dstr);
862a34c6 3753 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3754
3755 assert (SvPOK(sstr));
3756 assert (SvPOKp(sstr));
3757 assert (!SvIOK(sstr));
3758 assert (!SvIOKp(sstr));
3759 assert (!SvNOK(sstr));
3760 assert (!SvNOKp(sstr));
3761
3762 if (SvIsCOW(sstr)) {
3763
3764 if (SvLEN(sstr) == 0) {
3765 /* source is a COW shared hash key. */
ed252734
NC
3766 DEBUG_C(PerlIO_printf(Perl_debug_log,
3767 "Fast copy on write: Sharing hash\n"));
d1db91c6 3768 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3769 goto common_exit;
3770 }
3771 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3772 } else {
3773 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3774 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3775 SvREADONLY_on(sstr);
3776 SvFAKE_on(sstr);
3777 DEBUG_C(PerlIO_printf(Perl_debug_log,
3778 "Fast copy on write: Converting sstr to COW\n"));
3779 SV_COW_NEXT_SV_SET(dstr, sstr);
3780 }
3781 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3782 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3783
3784 common_exit:
3785 SvPV_set(dstr, new_pv);
3786 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3787 if (SvUTF8(sstr))
3788 SvUTF8_on(dstr);
87a1ef3d
SP
3789 SvLEN_set(dstr, len);
3790 SvCUR_set(dstr, cur);
ed252734
NC
3791 if (DEBUG_C_TEST) {
3792 sv_dump(dstr);
3793 }
3794 return dstr;
3795}
3796#endif
3797
954c1994
GS
3798/*
3799=for apidoc sv_setpvn
3800
3801Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3802bytes to be copied. If the C<ptr> argument is NULL the SV will become
3803undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3804
3805=cut
3806*/
3807
ef50df4b 3808void
864dbfa3 3809Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3810{
97aff369 3811 dVAR;
c6f8c383 3812 register char *dptr;
22c522df 3813
765f542d 3814 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3815 if (!ptr) {
a0d0e21e 3816 (void)SvOK_off(sv);
463ee0b2
LW
3817 return;
3818 }
22c522df
JH
3819 else {
3820 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3821 const IV iv = len;
9c5ffd7c
JH
3822 if (iv < 0)
3823 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3824 }
862a34c6 3825 SvUPGRADE(sv, SVt_PV);
c6f8c383 3826
5902b6a9 3827 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3828 Move(ptr,dptr,len,char);
3829 dptr[len] = '\0';
79072805 3830 SvCUR_set(sv, len);
1aa99e6b 3831 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3832 SvTAINT(sv);
79072805
LW
3833}
3834
954c1994
GS
3835/*
3836=for apidoc sv_setpvn_mg
3837
3838Like C<sv_setpvn>, but also handles 'set' magic.
3839
3840=cut
3841*/
3842
79072805 3843void
864dbfa3 3844Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3845{
3846 sv_setpvn(sv,ptr,len);
3847 SvSETMAGIC(sv);
3848}
3849
954c1994
GS
3850/*
3851=for apidoc sv_setpv
3852
3853Copies a string into an SV. The string must be null-terminated. Does not
3854handle 'set' magic. See C<sv_setpv_mg>.
3855
3856=cut
3857*/
3858
ef50df4b 3859void
864dbfa3 3860Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3861{
97aff369 3862 dVAR;
79072805
LW
3863 register STRLEN len;
3864
765f542d 3865 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3866 if (!ptr) {
a0d0e21e 3867 (void)SvOK_off(sv);
463ee0b2
LW
3868 return;
3869 }
79072805 3870 len = strlen(ptr);
862a34c6 3871 SvUPGRADE(sv, SVt_PV);
c6f8c383 3872
79072805 3873 SvGROW(sv, len + 1);
463ee0b2 3874 Move(ptr,SvPVX(sv),len+1,char);
79072805 3875 SvCUR_set(sv, len);
1aa99e6b 3876 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3877 SvTAINT(sv);
3878}
3879
954c1994
GS
3880/*
3881=for apidoc sv_setpv_mg
3882
3883Like C<sv_setpv>, but also handles 'set' magic.
3884
3885=cut
3886*/
3887
463ee0b2 3888void
864dbfa3 3889Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3890{
3891 sv_setpv(sv,ptr);
3892 SvSETMAGIC(sv);
3893}
3894
954c1994 3895/*
47518d95 3896=for apidoc sv_usepvn_flags
954c1994 3897
794a0d33
JH
3898Tells an SV to use C<ptr> to find its string value. Normally the
3899string is stored inside the SV but sv_usepvn allows the SV to use an
3900outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
3901by C<malloc>. The string length, C<len>, must be supplied. By default
3902this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
3903so that pointer should not be freed or used by the programmer after
3904giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
3905that pointer (e.g. ptr + 1) be used.
3906
3907If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3908SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 3909will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 3910C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3911
3912=cut
3913*/
3914
ef50df4b 3915void
47518d95 3916Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3917{
97aff369 3918 dVAR;
1936d2a7 3919 STRLEN allocate;
765f542d 3920 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3921 SvUPGRADE(sv, SVt_PV);
463ee0b2 3922 if (!ptr) {
a0d0e21e 3923 (void)SvOK_off(sv);
47518d95
NC
3924 if (flags & SV_SMAGIC)
3925 SvSETMAGIC(sv);
463ee0b2
LW
3926 return;
3927 }
3f7c398e 3928 if (SvPVX_const(sv))
8bd4d4c5 3929 SvPV_free(sv);
1936d2a7 3930
2e90b4cd
NC
3931 if (flags & SV_HAS_TRAILING_NUL)
3932 assert(ptr[len] == '\0');
3933
c1c21316 3934 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3935 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
cbf82dd0
NC
3936 if (flags & SV_HAS_TRAILING_NUL) {
3937 /* It's long enough - do nothing.
3938 Specfically Perl_newCONSTSUB is relying on this. */
3939 } else {
69d25b4f 3940#ifdef DEBUGGING
69d25b4f
NC
3941 /* Force a move to shake out bugs in callers. */
3942 char *new_ptr = safemalloc(allocate);
3943 Copy(ptr, new_ptr, len, char);
3944 PoisonFree(ptr,len,char);
3945 Safefree(ptr);
3946 ptr = new_ptr;
69d25b4f 3947#else
c1c21316 3948 ptr = saferealloc (ptr, allocate);
69d25b4f 3949#endif
cbf82dd0 3950 }
f880fe2f 3951 SvPV_set(sv, ptr);
463ee0b2 3952 SvCUR_set(sv, len);
1936d2a7 3953 SvLEN_set(sv, allocate);
c1c21316
NC
3954 if (!(flags & SV_HAS_TRAILING_NUL)) {
3955 *SvEND(sv) = '\0';
3956 }
1aa99e6b 3957 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3958 SvTAINT(sv);
47518d95
NC
3959 if (flags & SV_SMAGIC)
3960 SvSETMAGIC(sv);
ef50df4b
GS
3961}
3962
f8c7b90f 3963#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3964/* Need to do this *after* making the SV normal, as we need the buffer
3965 pointer to remain valid until after we've copied it. If we let go too early,
3966 another thread could invalidate it by unsharing last of the same hash key
3967 (which it can do by means other than releasing copy-on-write Svs)
3968 or by changing the other copy-on-write SVs in the loop. */
3969STATIC void
bdd68bc3 3970S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3971{
3972 if (len) { /* this SV was SvIsCOW_normal(sv) */
3973 /* we need to find the SV pointing to us. */
cf5629ad 3974 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3975
765f542d
NC
3976 if (current == sv) {
3977 /* The SV we point to points back to us (there were only two of us
3978 in the loop.)
3979 Hence other SV is no longer copy on write either. */
3980 SvFAKE_off(after);
3981 SvREADONLY_off(after);
3982 } else {
3983 /* We need to follow the pointers around the loop. */
3984 SV *next;
3985 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3986 assert (next);
3987 current = next;
3988 /* don't loop forever if the structure is bust, and we have
3989 a pointer into a closed loop. */
3990 assert (current != after);
3f7c398e 3991 assert (SvPVX_const(current) == pvx);
765f542d
NC
3992 }
3993 /* Make the SV before us point to the SV after us. */
a29f6d03 3994 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3995 }
3996 } else {
bdd68bc3 3997 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3998 }
3999}
4000
4001int
4002Perl_sv_release_IVX(pTHX_ register SV *sv)
4003{
4004 if (SvIsCOW(sv))
4005 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
4006 SvOOK_off(sv);
4007 return 0;
765f542d
NC
4008}
4009#endif
645c22ef
DM
4010/*
4011=for apidoc sv_force_normal_flags
4012
4013Undo various types of fakery on an SV: if the PV is a shared string, make
4014a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4015an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4016we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4017then a copy-on-write scalar drops its PV buffer (if any) and becomes
4018SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4019set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4020C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4021with flags set to 0.
645c22ef
DM
4022
4023=cut
4024*/
4025
6fc92669 4026void
840a7b70 4027Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4028{
97aff369 4029 dVAR;
f8c7b90f 4030#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4031 if (SvREADONLY(sv)) {
4032 /* At this point I believe I should acquire a global SV mutex. */
4033 if (SvFAKE(sv)) {
b64e5050 4034 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4035 const STRLEN len = SvLEN(sv);
4036 const STRLEN cur = SvCUR(sv);
a28509cc 4037 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4038 if (DEBUG_C_TEST) {
4039 PerlIO_printf(Perl_debug_log,
4040 "Copy on write: Force normal %ld\n",
4041 (long) flags);
e419cbc5 4042 sv_dump(sv);
46187eeb 4043 }
765f542d
NC
4044 SvFAKE_off(sv);
4045 SvREADONLY_off(sv);
9f653bb5 4046 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4047 SvPV_set(sv, NULL);
87a1ef3d 4048 SvLEN_set(sv, 0);
765f542d
NC
4049 if (flags & SV_COW_DROP_PV) {
4050 /* OK, so we don't need to copy our buffer. */
4051 SvPOK_off(sv);
4052 } else {
4053 SvGROW(sv, cur + 1);
4054 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4055 SvCUR_set(sv, cur);
765f542d
NC
4056 *SvEND(sv) = '\0';
4057 }
bdd68bc3 4058 sv_release_COW(sv, pvx, len, next);
46187eeb 4059 if (DEBUG_C_TEST) {
e419cbc5 4060 sv_dump(sv);
46187eeb 4061 }
765f542d 4062 }
923e4eb5 4063 else if (IN_PERL_RUNTIME)
765f542d
NC
4064 Perl_croak(aTHX_ PL_no_modify);
4065 /* At this point I believe that I can drop the global SV mutex. */
4066 }
4067#else
2213622d 4068 if (SvREADONLY(sv)) {
1c846c1f 4069 if (SvFAKE(sv)) {
b64e5050 4070 const char * const pvx = SvPVX_const(sv);
66a1b24b 4071 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4072 SvFAKE_off(sv);
4073 SvREADONLY_off(sv);
bd61b366 4074 SvPV_set(sv, NULL);
66a1b24b 4075 SvLEN_set(sv, 0);
1c846c1f 4076 SvGROW(sv, len + 1);
706aa1c9 4077 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4078 *SvEND(sv) = '\0';
bdd68bc3 4079 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4080 }
923e4eb5 4081 else if (IN_PERL_RUNTIME)
cea2e8a9 4082 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4083 }
765f542d 4084#endif
2213622d 4085 if (SvROK(sv))
840a7b70 4086 sv_unref_flags(sv, flags);
6fc92669
GS
4087 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4088 sv_unglob(sv);
0f15f207 4089}
1c846c1f 4090
645c22ef 4091/*
954c1994
GS
4092=for apidoc sv_chop
4093
1c846c1f 4094Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4095SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4096the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4097string. Uses the "OOK hack".
3f7c398e 4098Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4099refer to the same chunk of data.
954c1994
GS
4100
4101=cut
4102*/
4103
79072805 4104void
f54cb97a 4105Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4106{
4107 register STRLEN delta;
a0d0e21e 4108 if (!ptr || !SvPOKp(sv))
79072805 4109 return;
3f7c398e 4110 delta = ptr - SvPVX_const(sv);
2213622d 4111 SV_CHECK_THINKFIRST(sv);
79072805
LW
4112 if (SvTYPE(sv) < SVt_PVIV)
4113 sv_upgrade(sv,SVt_PVIV);
4114
4115 if (!SvOOK(sv)) {
50483b2c 4116 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4117 const char *pvx = SvPVX_const(sv);
a28509cc 4118 const STRLEN len = SvCUR(sv);
50483b2c 4119 SvGROW(sv, len + 1);
706aa1c9 4120 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4121 *SvEND(sv) = '\0';
4122 }
45977657 4123 SvIV_set(sv, 0);
a4bfb290
AB
4124 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4125 and we do that anyway inside the SvNIOK_off
4126 */
7a5fa8a2 4127 SvFLAGS(sv) |= SVf_OOK;
79072805 4128 }
a4bfb290 4129 SvNIOK_off(sv);
b162af07
SP
4130 SvLEN_set(sv, SvLEN(sv) - delta);
4131 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4132 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4133 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4134}
4135
954c1994
GS
4136/*
4137=for apidoc sv_catpvn
4138
4139Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4140C<len> indicates number of bytes to copy. If the SV has the UTF-8
4141status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4142Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4143
8d6d96c1
HS
4144=for apidoc sv_catpvn_flags
4145
4146Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4147C<len> indicates number of bytes to copy. If the SV has the UTF-8
4148status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4149If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4150appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4151in terms of this function.
4152
4153=cut
4154*/
4155
4156void
4157Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4158{
97aff369 4159 dVAR;
8d6d96c1 4160 STRLEN dlen;
fabdb6c0 4161 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4162
8d6d96c1
HS
4163 SvGROW(dsv, dlen + slen + 1);
4164 if (sstr == dstr)
3f7c398e 4165 sstr = SvPVX_const(dsv);
8d6d96c1 4166 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4167 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4168 *SvEND(dsv) = '\0';
4169 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4170 SvTAINT(dsv);
bddd5118
NC
4171 if (flags & SV_SMAGIC)
4172 SvSETMAGIC(dsv);
79072805
LW
4173}
4174
954c1994 4175/*
954c1994
GS
4176=for apidoc sv_catsv
4177
13e8c8e3
JH
4178Concatenates the string from SV C<ssv> onto the end of the string in
4179SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4180not 'set' magic. See C<sv_catsv_mg>.
954c1994 4181
8d6d96c1
HS
4182=for apidoc sv_catsv_flags
4183
4184Concatenates the string from SV C<ssv> onto the end of the string in
4185SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4186bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4187and C<sv_catsv_nomg> are implemented in terms of this function.
4188
4189=cut */
4190
ef50df4b 4191void
8d6d96c1 4192Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4193{
97aff369 4194 dVAR;
bddd5118 4195 if (ssv) {
00b6aa41
AL
4196 STRLEN slen;
4197 const char *spv = SvPV_const(ssv, slen);
4198 if (spv) {
bddd5118
NC
4199 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4200 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4201 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4202 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4203 dsv->sv_flags doesn't have that bit set.
4fd84b44 4204 Andy Dougherty 12 Oct 2001
bddd5118
NC
4205 */
4206 const I32 sutf8 = DO_UTF8(ssv);
4207 I32 dutf8;
13e8c8e3 4208
bddd5118
NC
4209 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4210 mg_get(dsv);
4211 dutf8 = DO_UTF8(dsv);
8d6d96c1 4212
bddd5118
NC
4213 if (dutf8 != sutf8) {
4214 if (dutf8) {
4215 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4216 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4217
bddd5118
NC
4218 sv_utf8_upgrade(csv);
4219 spv = SvPV_const(csv, slen);
4220 }
4221 else
4222 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4223 }
bddd5118 4224 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4225 }
560a288e 4226 }
bddd5118
NC
4227 if (flags & SV_SMAGIC)
4228 SvSETMAGIC(dsv);
79072805
LW
4229}
4230
954c1994 4231/*
954c1994
GS
4232=for apidoc sv_catpv
4233
4234Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4235If the SV has the UTF-8 status set, then the bytes appended should be
4236valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4237
d5ce4a7c 4238=cut */
954c1994 4239
ef50df4b 4240void
0c981600 4241Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4242{
97aff369 4243 dVAR;
79072805 4244 register STRLEN len;
463ee0b2 4245 STRLEN tlen;
748a9306 4246 char *junk;
79072805 4247
0c981600 4248 if (!ptr)
79072805 4249 return;
748a9306 4250 junk = SvPV_force(sv, tlen);
0c981600 4251 len = strlen(ptr);
463ee0b2 4252 SvGROW(sv, tlen + len + 1);
0c981600 4253 if (ptr == junk)
3f7c398e 4254 ptr = SvPVX_const(sv);
0c981600 4255 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4256 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4257 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4258 SvTAINT(sv);
79072805
LW
4259}
4260
954c1994
GS
4261/*
4262=for apidoc sv_catpv_mg
4263
4264Like C<sv_catpv>, but also handles 'set' magic.
4265
4266=cut
4267*/
4268
ef50df4b 4269void
0c981600 4270Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4271{
0c981600 4272 sv_catpv(sv,ptr);
ef50df4b
GS
4273 SvSETMAGIC(sv);
4274}
4275
645c22ef
DM
4276/*
4277=for apidoc newSV
4278
561b68a9
SH
4279Creates a new SV. A non-zero C<len> parameter indicates the number of
4280bytes of preallocated string space the SV should have. An extra byte for a
4281trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4282space is allocated.) The reference count for the new SV is set to 1.
4283
4284In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4285parameter, I<x>, a debug aid which allowed callers to identify themselves.
4286This aid has been superseded by a new build option, PERL_MEM_LOG (see
4287L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4288modules supporting older perls.
645c22ef
DM
4289
4290=cut
4291*/
4292
79072805 4293SV *
864dbfa3 4294Perl_newSV(pTHX_ STRLEN len)
79072805 4295{
97aff369 4296 dVAR;
79072805 4297 register SV *sv;
1c846c1f 4298
4561caa4 4299 new_SV(sv);
79072805
LW
4300 if (len) {
4301 sv_upgrade(sv, SVt_PV);
4302 SvGROW(sv, len + 1);
4303 }
4304 return sv;
4305}
954c1994 4306/*
92110913 4307=for apidoc sv_magicext
954c1994 4308
68795e93 4309Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4310supplied vtable and returns a pointer to the magic added.
92110913 4311
2d8d5d5a
SH
4312Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4313In particular, you can add magic to SvREADONLY SVs, and add more than
4314one instance of the same 'how'.
645c22ef 4315
2d8d5d5a
SH
4316If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4317stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4318special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4319to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4320
2d8d5d5a 4321(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4322
4323=cut
4324*/
92110913 4325MAGIC *
92e67595 4326Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4327 const char* name, I32 namlen)
79072805 4328{
97aff369 4329 dVAR;
79072805 4330 MAGIC* mg;
68795e93 4331
92110913 4332 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4333 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4334 }
a02a5408 4335 Newxz(mg, 1, MAGIC);
79072805 4336 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4337 SvMAGIC_set(sv, mg);
75f9d97a 4338
05f95b08
SB
4339 /* Sometimes a magic contains a reference loop, where the sv and
4340 object refer to each other. To prevent a reference loop that
4341 would prevent such objects being freed, we look for such loops
4342 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4343
4344 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4345 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4346
4347 */
14befaf4
DM
4348 if (!obj || obj == sv ||
4349 how == PERL_MAGIC_arylen ||
4350 how == PERL_MAGIC_qr ||
8d2f4536 4351 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4352 (SvTYPE(obj) == SVt_PVGV &&
4353 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4354 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4355 GvFORM(obj) == (CV*)sv)))
75f9d97a 4356 {
8990e307 4357 mg->mg_obj = obj;
75f9d97a 4358 }
85e6fe83 4359 else {
b37c2d43 4360 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4361 mg->mg_flags |= MGf_REFCOUNTED;
4362 }
b5ccf5f2
YST
4363
4364 /* Normal self-ties simply pass a null object, and instead of
4365 using mg_obj directly, use the SvTIED_obj macro to produce a
4366 new RV as needed. For glob "self-ties", we are tieing the PVIO
4367 with an RV obj pointing to the glob containing the PVIO. In
4368 this case, to avoid a reference loop, we need to weaken the
4369 reference.
4370 */
4371
4372 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4373 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4374 {
4375 sv_rvweaken(obj);
4376 }
4377
79072805 4378 mg->mg_type = how;
565764a8 4379 mg->mg_len = namlen;
9cbac4c7 4380 if (name) {
92110913 4381 if (namlen > 0)
1edc1566 4382 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4383 else if (namlen == HEf_SVKEY)
b37c2d43 4384 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4385 else
92110913 4386 mg->mg_ptr = (char *) name;
9cbac4c7 4387 }
92110913 4388 mg->mg_virtual = vtable;
68795e93 4389
92110913
NIS
4390 mg_magical(sv);
4391 if (SvGMAGICAL(sv))
4392 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4393 return mg;
4394}
4395
4396/*
4397=for apidoc sv_magic
1c846c1f 4398
92110913
NIS
4399Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4400then adds a new magic item of type C<how> to the head of the magic list.
4401
2d8d5d5a
SH
4402See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4403handling of the C<name> and C<namlen> arguments.
4404
4509d3fb
SB
4405You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4406to add more than one instance of the same 'how'.
4407
92110913
NIS
4408=cut
4409*/
4410
4411void
4412Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4413{
97aff369 4414 dVAR;
92e67595 4415 MGVTBL *vtable;
92110913 4416 MAGIC* mg;
92110913 4417
f8c7b90f 4418#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4419 if (SvIsCOW(sv))
4420 sv_force_normal_flags(sv, 0);
4421#endif
92110913 4422 if (SvREADONLY(sv)) {
d8084ca5
DM
4423 if (
4424 /* its okay to attach magic to shared strings; the subsequent
4425 * upgrade to PVMG will unshare the string */
4426 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4427
4428 && IN_PERL_RUNTIME
92110913
NIS
4429 && how != PERL_MAGIC_regex_global
4430 && how != PERL_MAGIC_bm
4431 && how != PERL_MAGIC_fm
4432 && how != PERL_MAGIC_sv
e6469971 4433 && how != PERL_MAGIC_backref
92110913
NIS
4434 )
4435 {
4436 Perl_croak(aTHX_ PL_no_modify);
4437 }
4438 }
4439 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4440 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4441 /* sv_magic() refuses to add a magic of the same 'how' as an
4442 existing one
92110913 4443 */
2a509ed3 4444 if (how == PERL_MAGIC_taint) {
92110913 4445 mg->mg_len |= 1;
2a509ed3
NC
4446 /* Any scalar which already had taint magic on which someone
4447 (erroneously?) did SvIOK_on() or similar will now be
4448 incorrectly sporting public "OK" flags. */
4449 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4450 }
92110913
NIS
4451 return;
4452 }
4453 }
68795e93 4454
79072805 4455 switch (how) {
14befaf4 4456 case PERL_MAGIC_sv:
92110913 4457 vtable = &PL_vtbl_sv;
79072805 4458 break;
14befaf4 4459 case PERL_MAGIC_overload:
92110913 4460 vtable = &PL_vtbl_amagic;
a0d0e21e 4461 break;
14befaf4 4462 case PERL_MAGIC_overload_elem:
92110913 4463 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4464 break;
14befaf4 4465 case PERL_MAGIC_overload_table:
92110913 4466 vtable = &PL_vtbl_ovrld;
a0d0e21e 4467 break;
14befaf4 4468 case PERL_MAGIC_bm:
92110913 4469 vtable = &PL_vtbl_bm;
79072805 4470 break;
14befaf4 4471 case PERL_MAGIC_regdata:
92110913 4472 vtable = &PL_vtbl_regdata;
6cef1e77 4473 break;
14befaf4 4474 case PERL_MAGIC_regdatum:
92110913 4475 vtable = &PL_vtbl_regdatum;
6cef1e77 4476 break;
14befaf4 4477 case PERL_MAGIC_env:
92110913 4478 vtable = &PL_vtbl_env;
79072805 4479 break;
14befaf4 4480 case PERL_MAGIC_fm:
92110913 4481 vtable = &PL_vtbl_fm;
55497cff 4482 break;
14befaf4 4483 case PERL_MAGIC_envelem:
92110913 4484 vtable = &PL_vtbl_envelem;
79072805 4485 break;
14befaf4 4486 case PERL_MAGIC_regex_global:
92110913 4487 vtable = &PL_vtbl_mglob;
93a17b20 4488 break;
14befaf4 4489 case PERL_MAGIC_isa:
92110913 4490 vtable = &PL_vtbl_isa;
463ee0b2 4491 break;
14befaf4 4492 case PERL_MAGIC_isaelem:
92110913 4493 vtable = &PL_vtbl_isaelem;
463ee0b2 4494 break;
14befaf4 4495 case PERL_MAGIC_nkeys:
92110913 4496 vtable = &PL_vtbl_nkeys;
16660edb 4497 break;
14befaf4 4498 case PERL_MAGIC_dbfile:
aec46f14 4499 vtable = NULL;
93a17b20 4500 break;
14befaf4 4501 case PERL_MAGIC_dbline:
92110913 4502 vtable = &PL_vtbl_dbline;
79072805 4503 break;
36477c24 4504#ifdef USE_LOCALE_COLLATE
14befaf4 4505 case PERL_MAGIC_collxfrm:
92110913 4506 vtable = &PL_vtbl_collxfrm;
bbce6d69 4507 break;
36477c24 4508#endif /* USE_LOCALE_COLLATE */
14befaf4 4509 case PERL_MAGIC_tied:
92110913 4510 vtable = &PL_vtbl_pack;
463ee0b2 4511 break;
14befaf4
DM
4512 case PERL_MAGIC_tiedelem:
4513 case PERL_MAGIC_tiedscalar:
92110913 4514 vtable = &PL_vtbl_packelem;
463ee0b2 4515 break;
14befaf4 4516 case PERL_MAGIC_qr:
92110913 4517 vtable = &PL_vtbl_regexp;
c277df42 4518 break;
b3ca2e83
NC
4519 case PERL_MAGIC_hints:
4520 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4521 case PERL_MAGIC_sig:
92110913 4522 vtable = &PL_vtbl_sig;
79072805 4523 break;
14befaf4 4524 case PERL_MAGIC_sigelem:
92110913 4525 vtable = &PL_vtbl_sigelem;
79072805 4526 break;
14befaf4 4527 case PERL_MAGIC_taint:
92110913 4528 vtable = &PL_vtbl_taint;
463ee0b2 4529 break;
14befaf4 4530 case PERL_MAGIC_uvar:
92110913 4531 vtable = &PL_vtbl_uvar;
79072805 4532 break;
14befaf4 4533 case PERL_MAGIC_vec:
92110913 4534 vtable = &PL_vtbl_vec;
79072805 4535 break;
a3874608 4536 case PERL_MAGIC_arylen_p:
bfcb3514 4537 case PERL_MAGIC_rhash:
8d2f4536 4538 case PERL_MAGIC_symtab:
ece467f9 4539 case PERL_MAGIC_vstring:
aec46f14 4540 vtable = NULL;
ece467f9 4541 break;
7e8c5dac
HS
4542 case PERL_MAGIC_utf8:
4543 vtable = &PL_vtbl_utf8;
4544 break;
14befaf4 4545 case PERL_MAGIC_substr:
92110913 4546 vtable = &PL_vtbl_substr;
79072805 4547 break;
14befaf4 4548 case PERL_MAGIC_defelem:
92110913 4549 vtable = &PL_vtbl_defelem;
5f05dabc 4550 break;
14befaf4 4551 case PERL_MAGIC_arylen:
92110913 4552 vtable = &PL_vtbl_arylen;
79072805 4553 break;
14befaf4 4554 case PERL_MAGIC_pos:
92110913 4555 vtable = &PL_vtbl_pos;
a0d0e21e 4556 break;
14befaf4 4557 case PERL_MAGIC_backref:
92110913 4558 vtable = &PL_vtbl_backref;
810b8aa5 4559 break;
b3ca2e83
NC
4560 case PERL_MAGIC_hintselem:
4561 vtable = &PL_vtbl_hintselem;
4562 break;
14befaf4
DM
4563 case PERL_MAGIC_ext:
4564 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4565 /* Useful for attaching extension internal data to perl vars. */
4566 /* Note that multiple extensions may clash if magical scalars */
4567 /* etc holding private data from one are passed to another. */
aec46f14 4568 vtable = NULL;
a0d0e21e 4569 break;
79072805 4570 default:
14befaf4 4571 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4572 }
68795e93 4573
92110913 4574 /* Rest of work is done else where */
aec46f14 4575 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4576
92110913
NIS
4577 switch (how) {
4578 case PERL_MAGIC_taint:
4579 mg->mg_len = 1;
4580 break;
4581 case PERL_MAGIC_ext:
4582 case PERL_MAGIC_dbfile:
4583 SvRMAGICAL_on(sv);
4584 break;
4585 }
463ee0b2
LW
4586}
4587
c461cf8f
JH
4588/*
4589=for apidoc sv_unmagic
4590
645c22ef 4591Removes all magic of type C<type> from an SV.
c461cf8f
JH
4592
4593=cut
4594*/
4595
463ee0b2 4596int
864dbfa3 4597Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4598{
4599 MAGIC* mg;
4600 MAGIC** mgp;
91bba347 4601 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4602 return 0;
064cf529 4603 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4604 for (mg = *mgp; mg; mg = *mgp) {
4605 if (mg->mg_type == type) {
e1ec3a88 4606 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4607 *mgp = mg->mg_moremagic;
1d7c1841 4608 if (vtbl && vtbl->svt_free)
fc0dc3b3 4609 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4610 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4611 if (mg->mg_len > 0)
1edc1566 4612 Safefree(mg->mg_ptr);
565764a8 4613 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4614 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4615 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4616 Safefree(mg->mg_ptr);
9cbac4c7 4617 }
a0d0e21e
LW
4618 if (mg->mg_flags & MGf_REFCOUNTED)
4619 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4620 Safefree(mg);
4621 }
4622 else
4623 mgp = &mg->mg_moremagic;
79072805 4624 }
91bba347 4625 if (!SvMAGIC(sv)) {
463ee0b2 4626 SvMAGICAL_off(sv);
c268c2a6 4627 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4628 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4629 }
4630
4631 return 0;
79072805
LW
4632}
4633
c461cf8f
JH
4634/*
4635=for apidoc sv_rvweaken
4636
645c22ef
DM
4637Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4638referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4639push a back-reference to this RV onto the array of backreferences
4640associated with that magic.
c461cf8f
JH
4641
4642=cut
4643*/
4644
810b8aa5 4645SV *
864dbfa3 4646Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4647{
4648 SV *tsv;
4649 if (!SvOK(sv)) /* let undefs pass */
4650 return sv;
4651 if (!SvROK(sv))
cea2e8a9 4652 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4653 else if (SvWEAKREF(sv)) {
810b8aa5 4654 if (ckWARN(WARN_MISC))
9014280d 4655 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4656 return sv;
4657 }
4658 tsv = SvRV(sv);
e15faf7d 4659 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4660 SvWEAKREF_on(sv);
1c846c1f 4661 SvREFCNT_dec(tsv);
810b8aa5
GS
4662 return sv;
4663}
4664
645c22ef
DM
4665/* Give tsv backref magic if it hasn't already got it, then push a
4666 * back-reference to sv onto the array associated with the backref magic.
4667 */
4668
e15faf7d
NC
4669void
4670Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4671{
97aff369 4672 dVAR;
810b8aa5 4673 AV *av;
86f55936
NC
4674
4675 if (SvTYPE(tsv) == SVt_PVHV) {
4676 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4677
4678 av = *avp;
4679 if (!av) {
4680 /* There is no AV in the offical place - try a fixup. */
4681 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4682
4683 if (mg) {
4684 /* Aha. They've got it stowed in magic. Bring it back. */
4685 av = (AV*)mg->mg_obj;
4686 /* Stop mg_free decreasing the refernce count. */
4687 mg->mg_obj = NULL;
4688 /* Stop mg_free even calling the destructor, given that
4689 there's no AV to free up. */
4690 mg->mg_virtual = 0;
4691 sv_unmagic(tsv, PERL_MAGIC_backref);
4692 } else {
4693 av = newAV();
4694 AvREAL_off(av);
b37c2d43 4695 SvREFCNT_inc_simple_void(av);
86f55936
NC
4696 }
4697 *avp = av;
4698 }
4699 } else {
4700 const MAGIC *const mg
4701 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4702 if (mg)
4703 av = (AV*)mg->mg_obj;
4704 else {
4705 av = newAV();
4706 AvREAL_off(av);
4707 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4708 /* av now has a refcnt of 2, which avoids it getting freed
4709 * before us during global cleanup. The extra ref is removed
4710 * by magic_killbackrefs() when tsv is being freed */
4711 }
810b8aa5 4712 }
d91d49e8 4713 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4714 av_extend(av, AvFILLp(av)+1);
4715 }
4716 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4717}
4718
645c22ef
DM
4719/* delete a back-reference to ourselves from the backref magic associated
4720 * with the SV we point to.
4721 */
4722
1c846c1f 4723STATIC void
e15faf7d 4724S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4725{
97aff369 4726 dVAR;
86f55936 4727 AV *av = NULL;
810b8aa5
GS
4728 SV **svp;
4729 I32 i;
86f55936
NC
4730
4731 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4732 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4733 /* We mustn't attempt to "fix up" the hash here by moving the
4734 backreference array back to the hv_aux structure, as that is stored
4735 in the main HvARRAY(), and hfreentries assumes that no-one
4736 reallocates HvARRAY() while it is running. */
86f55936
NC
4737 }
4738 if (!av) {
4739 const MAGIC *const mg
4740 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4741 if (mg)
4742 av = (AV *)mg->mg_obj;
4743 }
4744 if (!av) {
e15faf7d
NC
4745 if (PL_in_clean_all)
4746 return;
cea2e8a9 4747 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4748 }
4749
4750 if (SvIS_FREED(av))
4751 return;
4752
810b8aa5 4753 svp = AvARRAY(av);
6a76db8b
NC
4754 /* We shouldn't be in here more than once, but for paranoia reasons lets
4755 not assume this. */
4756 for (i = AvFILLp(av); i >= 0; i--) {
4757 if (svp[i] == sv) {
4758 const SSize_t fill = AvFILLp(av);
4759 if (i != fill) {
4760 /* We weren't the last entry.
4761 An unordered list has this property that you can take the
4762 last element off the end to fill the hole, and it's still
4763 an unordered list :-)
4764 */
4765 svp[i] = svp[fill];
4766 }
a0714e2c 4767 svp[fill] = NULL;
6a76db8b
NC
4768 AvFILLp(av) = fill - 1;
4769 }
4770 }
810b8aa5
GS
4771}
4772
86f55936
NC
4773int
4774Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4775{
4776 SV **svp = AvARRAY(av);
4777
4778 PERL_UNUSED_ARG(sv);
4779
4780 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4781 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4782 if (svp && !SvIS_FREED(av)) {
4783 SV *const *const last = svp + AvFILLp(av);
4784
4785 while (svp <= last) {
4786 if (*svp) {
4787 SV *const referrer = *svp;
4788 if (SvWEAKREF(referrer)) {
4789 /* XXX Should we check that it hasn't changed? */
4790 SvRV_set(referrer, 0);
4791 SvOK_off(referrer);
4792 SvWEAKREF_off(referrer);
4793 } else if (SvTYPE(referrer) == SVt_PVGV ||
4794 SvTYPE(referrer) == SVt_PVLV) {
4795 /* You lookin' at me? */
4796 assert(GvSTASH(referrer));
4797 assert(GvSTASH(referrer) == (HV*)sv);
4798 GvSTASH(referrer) = 0;
4799 } else {
4800 Perl_croak(aTHX_
4801 "panic: magic_killbackrefs (flags=%"UVxf")",
4802 (UV)SvFLAGS(referrer));
4803 }
4804
a0714e2c 4805 *svp = NULL;
86f55936
NC
4806 }
4807 svp++;
4808 }
4809 }
4810 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4811 return 0;
4812}
4813
954c1994
GS
4814/*
4815=for apidoc sv_insert
4816
4817Inserts a string at the specified offset/length within the SV. Similar to
4818the Perl substr() function.
4819
4820=cut
4821*/
4822
79072805 4823void
e1ec3a88 4824Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4825{
97aff369 4826 dVAR;
79072805
LW
4827 register char *big;
4828 register char *mid;
4829 register char *midend;
4830 register char *bigend;
4831 register I32 i;
6ff81951 4832 STRLEN curlen;
1c846c1f 4833
79072805 4834
8990e307 4835 if (!bigstr)
cea2e8a9 4836 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4837 SvPV_force(bigstr, curlen);
60fa28ff 4838 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4839 if (offset + len > curlen) {
4840 SvGROW(bigstr, offset+len+1);
93524f2b 4841 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4842 SvCUR_set(bigstr, offset+len);
4843 }
79072805 4844
69b47968 4845 SvTAINT(bigstr);
79072805
LW
4846 i = littlelen - len;
4847 if (i > 0) { /* string might grow */
a0d0e21e 4848 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4849 mid = big + offset + len;
4850 midend = bigend = big + SvCUR(bigstr);
4851 bigend += i;
4852 *bigend = '\0';
4853 while (midend > mid) /* shove everything down */
4854 *--bigend = *--midend;
4855 Move(little,big+offset,littlelen,char);
b162af07 4856 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4857 SvSETMAGIC(bigstr);
4858 return;
4859 }
4860 else if (i == 0) {
463ee0b2 4861 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4862 SvSETMAGIC(bigstr);
4863 return;
4864 }
4865
463ee0b2 4866 big = SvPVX(bigstr);
79072805
LW
4867 mid = big + offset;
4868 midend = mid + len;
4869 bigend = big + SvCUR(bigstr);
4870
4871 if (midend > bigend)
cea2e8a9 4872 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4873
4874 if (mid - big > bigend - midend) { /* faster to shorten from end */
4875 if (littlelen) {
4876 Move(little, mid, littlelen,char);
4877 mid += littlelen;
4878 }
4879 i = bigend - midend;
4880 if (i > 0) {
4881 Move(midend, mid, i,char);
4882 mid += i;
4883 }
4884 *mid = '\0';
4885 SvCUR_set(bigstr, mid - big);
4886 }
155aba94 4887 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4888 midend -= littlelen;
4889 mid = midend;
4890 sv_chop(bigstr,midend-i);
4891 big += i;
4892 while (i--)
4893 *--midend = *--big;
4894 if (littlelen)
4895 Move(little, mid, littlelen,char);
4896 }
4897 else if (littlelen) {
4898 midend -= littlelen;
4899 sv_chop(bigstr,midend);
4900 Move(little,midend,littlelen,char);
4901 }
4902 else {
4903 sv_chop(bigstr,midend);
4904 }
4905 SvSETMAGIC(bigstr);
4906}
4907
c461cf8f
JH
4908/*
4909=for apidoc sv_replace
4910
4911Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4912The target SV physically takes over ownership of the body of the source SV
4913and inherits its flags; however, the target keeps any magic it owns,
4914and any magic in the source is discarded.
ff276b08 4915Note that this is a rather specialist SV copying operation; most of the
645c22ef 4916time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4917
4918=cut
4919*/
79072805
LW
4920
4921void
864dbfa3 4922Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4923{
97aff369 4924 dVAR;
a3b680e6 4925 const U32 refcnt = SvREFCNT(sv);
765f542d 4926 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4927 if (SvREFCNT(nsv) != 1) {
7437becc 4928 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4929 UVuf " != 1)", (UV) SvREFCNT(nsv));
4930 }
93a17b20 4931 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4932 if (SvMAGICAL(nsv))
4933 mg_free(nsv);
4934 else
4935 sv_upgrade(nsv, SVt_PVMG);
b162af07 4936 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4937 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4938 SvMAGICAL_off(sv);
b162af07 4939 SvMAGIC_set(sv, NULL);
93a17b20 4940 }
79072805
LW
4941 SvREFCNT(sv) = 0;
4942 sv_clear(sv);
477f5d66 4943 assert(!SvREFCNT(sv));
fd0854ff
DM
4944#ifdef DEBUG_LEAKING_SCALARS
4945 sv->sv_flags = nsv->sv_flags;
4946 sv->sv_any = nsv->sv_any;
4947 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4948 sv->sv_u = nsv->sv_u;
fd0854ff 4949#else
79072805 4950 StructCopy(nsv,sv,SV);
fd0854ff 4951#endif
7b2c381c
NC
4952 /* Currently could join these into one piece of pointer arithmetic, but
4953 it would be unclear. */
4954 if(SvTYPE(sv) == SVt_IV)
4955 SvANY(sv)
339049b0 4956 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4957 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4958 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4959 }
4960
fd0854ff 4961
f8c7b90f 4962#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4963 if (SvIsCOW_normal(nsv)) {
4964 /* We need to follow the pointers around the loop to make the
4965 previous SV point to sv, rather than nsv. */
4966 SV *next;
4967 SV *current = nsv;
4968 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4969 assert(next);
4970 current = next;
3f7c398e 4971 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4972 }
4973 /* Make the SV before us point to the SV after us. */
4974 if (DEBUG_C_TEST) {
4975 PerlIO_printf(Perl_debug_log, "previous is\n");
4976 sv_dump(current);
a29f6d03
NC
4977 PerlIO_printf(Perl_debug_log,
4978 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4979 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4980 }
a29f6d03 4981 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4982 }
4983#endif
79072805 4984 SvREFCNT(sv) = refcnt;
1edc1566 4985 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4986 SvREFCNT(nsv) = 0;
463ee0b2 4987 del_SV(nsv);
79072805
LW
4988}
4989
c461cf8f
JH
4990/*
4991=for apidoc sv_clear
4992
645c22ef
DM
4993Clear an SV: call any destructors, free up any memory used by the body,
4994and free the body itself. The SV's head is I<not> freed, although
4995its type is set to all 1's so that it won't inadvertently be assumed
4996to be live during global destruction etc.
4997This function should only be called when REFCNT is zero. Most of the time
4998you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4999instead.
c461cf8f
JH
5000
5001=cut
5002*/
5003
79072805 5004void
864dbfa3 5005Perl_sv_clear(pTHX_ register SV *sv)
79072805 5006{
27da23d5 5007 dVAR;
82bb6deb 5008 const U32 type = SvTYPE(sv);
8edfc514
NC
5009 const struct body_details *const sv_type_details
5010 = bodies_by_type + type;
82bb6deb 5011
79072805
LW
5012 assert(sv);
5013 assert(SvREFCNT(sv) == 0);
5014
d2a0f284
JC
5015 if (type <= SVt_IV) {
5016 /* See the comment in sv.h about the collusion between this early
5017 return and the overloading of the NULL and IV slots in the size
5018 table. */
82bb6deb 5019 return;
d2a0f284 5020 }
82bb6deb 5021
ed6116ce 5022 if (SvOBJECT(sv)) {
3280af22 5023 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5024 dSP;
893645bd 5025 HV* stash;
d460ef45 5026 do {
b464bac0 5027 CV* destructor;
4e8e7886 5028 stash = SvSTASH(sv);
32251b26 5029 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5030 if (destructor) {
1b6737cc 5031 SV* const tmpref = newRV(sv);
5cc433a6 5032 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5033 ENTER;
e788e7d3 5034 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5035 EXTEND(SP, 2);
5036 PUSHMARK(SP);
5cc433a6 5037 PUSHs(tmpref);
4e8e7886 5038 PUTBACK;
44389ee9 5039 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5040
5041
d3acc0f7 5042 POPSTACK;
3095d977 5043 SPAGAIN;
4e8e7886 5044 LEAVE;
5cc433a6
AB
5045 if(SvREFCNT(tmpref) < 2) {
5046 /* tmpref is not kept alive! */
5047 SvREFCNT(sv)--;
b162af07 5048 SvRV_set(tmpref, NULL);
5cc433a6
AB
5049 SvROK_off(tmpref);
5050 }
5051 SvREFCNT_dec(tmpref);
4e8e7886
GS
5052 }
5053 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5054
6f44e0a4
JP
5055
5056 if (SvREFCNT(sv)) {
5057 if (PL_in_clean_objs)
cea2e8a9 5058 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5059 HvNAME_get(stash));
6f44e0a4
JP
5060 /* DESTROY gave object new lease on life */
5061 return;
5062 }
a0d0e21e 5063 }
4e8e7886 5064
a0d0e21e 5065 if (SvOBJECT(sv)) {
4e8e7886 5066 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5067 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5068 if (type != SVt_PVIO)
3280af22 5069 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5070 }
463ee0b2 5071 }
82bb6deb 5072 if (type >= SVt_PVMG) {
e736a858
NC
5073 HV *ourstash;
5074 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5075 (ourstash = OURSTASH(sv))) {
5076 SvREFCNT_dec(ourstash);
5077 } else if (SvMAGIC(sv))
524189f1 5078 mg_free(sv);
00b1698f 5079 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5080 SvREFCNT_dec(SvSTASH(sv));
5081 }
82bb6deb 5082 switch (type) {
8990e307 5083 case SVt_PVIO:
df0bd2f4
GS
5084 if (IoIFP(sv) &&
5085 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5086 IoIFP(sv) != PerlIO_stdout() &&
5087 IoIFP(sv) != PerlIO_stderr())
93578b34 5088 {
f2b5be74 5089 io_close((IO*)sv, FALSE);
93578b34 5090 }
1d7c1841 5091 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5092 PerlDir_close(IoDIRP(sv));
1d7c1841 5093 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5094 Safefree(IoTOP_NAME(sv));
5095 Safefree(IoFMT_NAME(sv));
5096 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5097 goto freescalar;
79072805 5098 case SVt_PVBM:
a0d0e21e 5099 goto freescalar;
79072805 5100 case SVt_PVCV:
748a9306 5101 case SVt_PVFM:
85e6fe83 5102 cv_undef((CV*)sv);
a0d0e21e 5103 goto freescalar;
79072805 5104 case SVt_PVHV:
86f55936 5105 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5106 hv_undef((HV*)sv);
a0d0e21e 5107 break;
79072805 5108 case SVt_PVAV:
85e6fe83 5109 av_undef((AV*)sv);
a0d0e21e 5110 break;
02270b4e 5111 case SVt_PVLV:
dd28f7bb
DM
5112 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5113 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5114 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5115 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5116 }
5117 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5118 SvREFCNT_dec(LvTARG(sv));
02270b4e 5119 goto freescalar;
a0d0e21e 5120 case SVt_PVGV:
1edc1566 5121 gp_free((GV*)sv);
acda4c6a
NC
5122 if (GvNAME_HEK(sv)) {
5123 unshare_hek(GvNAME_HEK(sv));
5124 }
893645bd
NC
5125 /* If we're in a stash, we don't own a reference to it. However it does
5126 have a back reference to us, which needs to be cleared. */
5127 if (GvSTASH(sv))
5128 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5129 case SVt_PVMG:
79072805
LW
5130 case SVt_PVNV:
5131 case SVt_PVIV:
a0d0e21e 5132 freescalar:
5228ca4e
NC
5133 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5134 if (SvOOK(sv)) {
93524f2b 5135 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5136 /* Don't even bother with turning off the OOK flag. */
5137 }
79072805 5138 case SVt_PV:
a0d0e21e 5139 case SVt_RV:
810b8aa5 5140 if (SvROK(sv)) {
b37c2d43 5141 SV * const target = SvRV(sv);
810b8aa5 5142 if (SvWEAKREF(sv))
e15faf7d 5143 sv_del_backref(target, sv);
810b8aa5 5144 else
e15faf7d 5145 SvREFCNT_dec(target);
810b8aa5 5146 }
f8c7b90f 5147#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5148 else if (SvPVX_const(sv)) {
765f542d
NC
5149 if (SvIsCOW(sv)) {
5150 /* I believe I need to grab the global SV mutex here and
5151 then recheck the COW status. */
46187eeb
NC
5152 if (DEBUG_C_TEST) {
5153 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5154 sv_dump(sv);
46187eeb 5155 }
bdd68bc3
NC
5156 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5157 SV_COW_NEXT_SV(sv));
765f542d
NC
5158 /* And drop it here. */
5159 SvFAKE_off(sv);
5160 } else if (SvLEN(sv)) {
3f7c398e 5161 Safefree(SvPVX_const(sv));
765f542d
NC
5162 }
5163 }
5164#else
3f7c398e 5165 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5166 Safefree(SvPVX_mutable(sv));
3f7c398e 5167 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5168 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5169 SvFAKE_off(sv);
5170 }
765f542d 5171#endif
79072805
LW
5172 break;
5173 case SVt_NV:
79072805
LW
5174 break;
5175 }
5176
893645bd
NC
5177 SvFLAGS(sv) &= SVf_BREAK;
5178 SvFLAGS(sv) |= SVTYPEMASK;
5179
8edfc514 5180 if (sv_type_details->arena) {
b9502f15 5181 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5182 &PL_body_roots[type]);
5183 }
d2a0f284 5184 else if (sv_type_details->body_size) {
8edfc514
NC
5185 my_safefree(SvANY(sv));
5186 }
79072805
LW
5187}
5188
645c22ef
DM
5189/*
5190=for apidoc sv_newref
5191
5192Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5193instead.
5194
5195=cut
5196*/
5197
79072805 5198SV *
864dbfa3 5199Perl_sv_newref(pTHX_ SV *sv)
79072805 5200{
96a5add6 5201 PERL_UNUSED_CONTEXT;
463ee0b2 5202 if (sv)
4db098f4 5203 (SvREFCNT(sv))++;
79072805
LW
5204 return sv;
5205}
5206
c461cf8f
JH
5207/*
5208=for apidoc sv_free
5209
645c22ef
DM
5210Decrement an SV's reference count, and if it drops to zero, call
5211C<sv_clear> to invoke destructors and free up any memory used by
5212the body; finally, deallocate the SV's head itself.
5213Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5214
5215=cut
5216*/
5217
79072805 5218void
864dbfa3 5219Perl_sv_free(pTHX_ SV *sv)
79072805 5220{
27da23d5 5221 dVAR;
79072805
LW
5222 if (!sv)
5223 return;
a0d0e21e
LW
5224 if (SvREFCNT(sv) == 0) {
5225 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5226 /* this SV's refcnt has been artificially decremented to
5227 * trigger cleanup */
a0d0e21e 5228 return;
3280af22 5229 if (PL_in_clean_all) /* All is fair */
1edc1566 5230 return;
d689ffdd
JP
5231 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5232 /* make sure SvREFCNT(sv)==0 happens very seldom */
5233 SvREFCNT(sv) = (~(U32)0)/2;
5234 return;
5235 }
41e4abd8 5236 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5237 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5238 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5239 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5240#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5241 Perl_dump_sv_child(aTHX_ sv);
5242#endif
5243 }
79072805
LW
5244 return;
5245 }
4db098f4 5246 if (--(SvREFCNT(sv)) > 0)
8990e307 5247 return;
8c4d3c90
NC
5248 Perl_sv_free2(aTHX_ sv);
5249}
5250
5251void
5252Perl_sv_free2(pTHX_ SV *sv)
5253{
27da23d5 5254 dVAR;
463ee0b2
LW
5255#ifdef DEBUGGING
5256 if (SvTEMP(sv)) {
0453d815 5257 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5258 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5259 "Attempt to free temp prematurely: SV 0x%"UVxf
5260 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5261 return;
79072805 5262 }
463ee0b2 5263#endif
d689ffdd
JP
5264 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5265 /* make sure SvREFCNT(sv)==0 happens very seldom */
5266 SvREFCNT(sv) = (~(U32)0)/2;
5267 return;
5268 }
79072805 5269 sv_clear(sv);
477f5d66
CS
5270 if (! SvREFCNT(sv))
5271 del_SV(sv);
79072805
LW
5272}
5273
954c1994
GS
5274/*
5275=for apidoc sv_len
5276
645c22ef
DM
5277Returns the length of the string in the SV. Handles magic and type
5278coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5279
5280=cut
5281*/
5282
79072805 5283STRLEN
864dbfa3 5284Perl_sv_len(pTHX_ register SV *sv)
79072805 5285{
463ee0b2 5286 STRLEN len;
79072805
LW
5287
5288 if (!sv)
5289 return 0;
5290
8990e307 5291 if (SvGMAGICAL(sv))
565764a8 5292 len = mg_length(sv);
8990e307 5293 else
4d84ee25 5294 (void)SvPV_const(sv, len);
463ee0b2 5295 return len;
79072805
LW
5296}
5297
c461cf8f
JH
5298/*
5299=for apidoc sv_len_utf8
5300
5301Returns the number of characters in the string in an SV, counting wide
1e54db1a 5302UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5303
5304=cut
5305*/
5306
7e8c5dac
HS
5307/*
5308 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5309 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5310 * (Note that the mg_len is not the length of the mg_ptr field.
5311 * This allows the cache to store the character length of the string without
5312 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5313 *
7e8c5dac
HS
5314 */
5315
a0ed51b3 5316STRLEN
864dbfa3 5317Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5318{
a0ed51b3
LW
5319 if (!sv)
5320 return 0;
5321
a0ed51b3 5322 if (SvGMAGICAL(sv))
b76347f2 5323 return mg_length(sv);
a0ed51b3 5324 else
b76347f2 5325 {
26346457 5326 STRLEN len;
e62f0680 5327 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5328
26346457
NC
5329 if (PL_utf8cache) {
5330 STRLEN ulen;
5331 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5332
5333 if (mg && mg->mg_len != -1) {
5334 ulen = mg->mg_len;
5335 if (PL_utf8cache < 0) {
5336 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5337 if (real != ulen) {
5338 /* Need to turn the assertions off otherwise we may
5339 recurse infinitely while printing error messages.
5340 */
5341 SAVEI8(PL_utf8cache);
5342 PL_utf8cache = 0;
5343 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5344 " real %"UVf" for %"SVf,
26346457
NC
5345 (UV) ulen, (UV) real, sv);
5346 }
5347 }
5348 }
5349 else {
5350 ulen = Perl_utf8_length(aTHX_ s, s + len);
5351 if (!SvREADONLY(sv)) {
5352 if (!mg) {
5353 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5354 &PL_vtbl_utf8, 0, 0);
5355 }
cb9e20bb 5356 assert(mg);
26346457 5357 mg->mg_len = ulen;
cb9e20bb 5358 }
cb9e20bb 5359 }
26346457 5360 return ulen;
7e8c5dac 5361 }
26346457 5362 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5363 }
5364}
5365
9564a3bd
NC
5366/* Walk forwards to find the byte corresponding to the passed in UTF-8
5367 offset. */
bdf30dd6 5368static STRLEN
721e86b6 5369S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
bdf30dd6
NC
5370 STRLEN uoffset)
5371{
5372 const U8 *s = start;
5373
5374 while (s < send && uoffset--)
5375 s += UTF8SKIP(s);
5376 if (s > send) {
5377 /* This is the existing behaviour. Possibly it should be a croak, as
5378 it's actually a bounds error */
5379 s = send;
5380 }
5381 return s - start;
5382}
5383
9564a3bd
NC
5384/* Given the length of the string in both bytes and UTF-8 characters, decide
5385 whether to walk forwards or backwards to find the byte corresponding to
5386 the passed in UTF-8 offset. */
c336ad0b 5387static STRLEN
721e86b6 5388S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
c336ad0b
NC
5389 STRLEN uoffset, STRLEN uend)
5390{
5391 STRLEN backw = uend - uoffset;
5392 if (uoffset < 2 * backw) {
25a8a4ef 5393 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5394 forward (that's where the 2 * backw comes from).
5395 (The real figure of course depends on the UTF-8 data.) */
721e86b6 5396 return sv_pos_u2b_forwards(start, send, uoffset);
c336ad0b
NC
5397 }
5398
5399 while (backw--) {
5400 send--;
5401 while (UTF8_IS_CONTINUATION(*send))
5402 send--;
5403 }
5404 return send - start;
5405}
5406
9564a3bd
NC
5407/* For the string representation of the given scalar, find the byte
5408 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5409 give another position in the string, *before* the sought offset, which
5410 (which is always true, as 0, 0 is a valid pair of positions), which should
5411 help reduce the amount of linear searching.
5412 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5413 will be used to reduce the amount of linear searching. The cache will be
5414 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5415static STRLEN
5416S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5417 const U8 *const send, STRLEN uoffset,
5418 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5419 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5420 bool found = FALSE;
5421
75c33c12
NC
5422 assert (uoffset >= uoffset0);
5423
c336ad0b 5424 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5425 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5426 if ((*mgp)->mg_ptr) {
5427 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5428 if (cache[0] == uoffset) {
5429 /* An exact match. */
5430 return cache[1];
5431 }
ab455f60
NC
5432 if (cache[2] == uoffset) {
5433 /* An exact match. */
5434 return cache[3];
5435 }
668af93f
NC
5436
5437 if (cache[0] < uoffset) {
d8b2e1f9
NC
5438 /* The cache already knows part of the way. */
5439 if (cache[0] > uoffset0) {
5440 /* The cache knows more than the passed in pair */
5441 uoffset0 = cache[0];
5442 boffset0 = cache[1];
5443 }
5444 if ((*mgp)->mg_len != -1) {
5445 /* And we know the end too. */
5446 boffset = boffset0
721e86b6 5447 + sv_pos_u2b_midway(start + boffset0, send,
d8b2e1f9
NC
5448 uoffset - uoffset0,
5449 (*mgp)->mg_len - uoffset0);
5450 } else {
5451 boffset = boffset0
721e86b6 5452 + sv_pos_u2b_forwards(start + boffset0,
d8b2e1f9
NC
5453 send, uoffset - uoffset0);
5454 }
dd7c5fd3
NC
5455 }
5456 else if (cache[2] < uoffset) {
5457 /* We're between the two cache entries. */
5458 if (cache[2] > uoffset0) {
5459 /* and the cache knows more than the passed in pair */
5460 uoffset0 = cache[2];
5461 boffset0 = cache[3];
5462 }
5463
668af93f 5464 boffset = boffset0
721e86b6 5465 + sv_pos_u2b_midway(start + boffset0,
668af93f
NC
5466 start + cache[1],
5467 uoffset - uoffset0,
5468 cache[0] - uoffset0);
dd7c5fd3
NC
5469 } else {
5470 boffset = boffset0
721e86b6 5471 + sv_pos_u2b_midway(start + boffset0,
dd7c5fd3
NC
5472 start + cache[3],
5473 uoffset - uoffset0,
5474 cache[2] - uoffset0);
d8b2e1f9 5475 }
668af93f 5476 found = TRUE;
d8b2e1f9
NC
5477 }
5478 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5479 /* If we can take advantage of a passed in offset, do so. */
5480 /* In fact, offset0 is either 0, or less than offset, so don't
5481 need to worry about the other possibility. */
5482 boffset = boffset0
721e86b6 5483 + sv_pos_u2b_midway(start + boffset0, send,
75c33c12
NC
5484 uoffset - uoffset0,
5485 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5486 found = TRUE;
5487 }
28ccbf94 5488 }
c336ad0b
NC
5489
5490 if (!found || PL_utf8cache < 0) {
75c33c12 5491 const STRLEN real_boffset
721e86b6 5492 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
75c33c12
NC
5493 send, uoffset - uoffset0);
5494
c336ad0b
NC
5495 if (found && PL_utf8cache < 0) {
5496 if (real_boffset != boffset) {
5497 /* Need to turn the assertions off otherwise we may recurse
5498 infinitely while printing error messages. */
5499 SAVEI8(PL_utf8cache);
5500 PL_utf8cache = 0;
5501 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5502 " real %"UVf" for %"SVf,
5503 (UV) boffset, (UV) real_boffset, sv);
5504 }
5505 }
5506 boffset = real_boffset;
28ccbf94 5507 }
0905937d 5508
ab455f60 5509 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5510 return boffset;
5511}
5512
9564a3bd
NC
5513
5514/*
5515=for apidoc sv_pos_u2b
5516
5517Converts the value pointed to by offsetp from a count of UTF-8 chars from
5518the start of the string, to a count of the equivalent number of bytes; if
5519lenp is non-zero, it does the same to lenp, but this time starting from
5520the offset, rather than from the start of the string. Handles magic and
5521type coercion.
5522
5523=cut
5524*/
5525
5526/*
5527 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5528 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5529 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5530 *
5531 */
5532
a0ed51b3 5533void
864dbfa3 5534Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5535{
245d4a47 5536 const U8 *start;
a0ed51b3
LW
5537 STRLEN len;
5538
5539 if (!sv)
5540 return;
5541
245d4a47 5542 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5543 if (len) {
bdf30dd6
NC
5544 STRLEN uoffset = (STRLEN) *offsetp;
5545 const U8 * const send = start + len;
0905937d 5546 MAGIC *mg = NULL;
721e86b6 5547 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
28ccbf94 5548 uoffset, 0, 0);
bdf30dd6
NC
5549
5550 *offsetp = (I32) boffset;
5551
5552 if (lenp) {
28ccbf94 5553 /* Convert the relative offset to absolute. */
721e86b6
AL
5554 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5555 const STRLEN boffset2
5556 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
28ccbf94 5557 uoffset, boffset) - boffset;
bdf30dd6 5558
28ccbf94 5559 *lenp = boffset2;
bdf30dd6 5560 }
7e8c5dac
HS
5561 }
5562 else {
5563 *offsetp = 0;
5564 if (lenp)
5565 *lenp = 0;
a0ed51b3 5566 }
e23c8137 5567
a0ed51b3
LW
5568 return;
5569}
5570
9564a3bd
NC
5571/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5572 byte length pairing. The (byte) length of the total SV is passed in too,
5573 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5574 may not have updated SvCUR, so we can't rely on reading it directly.
5575
5576 The proffered utf8/byte length pairing isn't used if the cache already has
5577 two pairs, and swapping either for the proffered pair would increase the
5578 RMS of the intervals between known byte offsets.
5579
5580 The cache itself consists of 4 STRLEN values
5581 0: larger UTF-8 offset
5582 1: corresponding byte offset
5583 2: smaller UTF-8 offset
5584 3: corresponding byte offset
5585
5586 Unused cache pairs have the value 0, 0.
5587 Keeping the cache "backwards" means that the invariant of
5588 cache[0] >= cache[2] is maintained even with empty slots, which means that
5589 the code that uses it doesn't need to worry if only 1 entry has actually
5590 been set to non-zero. It also makes the "position beyond the end of the
5591 cache" logic much simpler, as the first slot is always the one to start
5592 from.
645c22ef 5593*/
ec07b5e0 5594static void
ab455f60
NC
5595S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5596 STRLEN blen)
ec07b5e0
NC
5597{
5598 STRLEN *cache;
5599 if (SvREADONLY(sv))
5600 return;
5601
5602 if (!*mgp) {
5603 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5604 0);
5605 (*mgp)->mg_len = -1;
5606 }
5607 assert(*mgp);
5608
5609 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5610 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5611 (*mgp)->mg_ptr = (char *) cache;
5612 }
5613 assert(cache);
5614
5615 if (PL_utf8cache < 0) {
ef816a78 5616 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5617 const U8 *const end = start + byte;
5618 STRLEN realutf8 = 0;
5619
5620 while (start < end) {
5621 start += UTF8SKIP(start);
5622 realutf8++;
5623 }
5624
5625 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5626 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5627 doesn't? I don't know whether this difference was introduced with
5628 the caching code in 5.8.1. */
ec07b5e0
NC
5629
5630 if (realutf8 != utf8) {
5631 /* Need to turn the assertions off otherwise we may recurse
5632 infinitely while printing error messages. */
5633 SAVEI8(PL_utf8cache);
5634 PL_utf8cache = 0;
5635 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5636 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5637 }
5638 }
ab455f60
NC
5639
5640 /* Cache is held with the later position first, to simplify the code
5641 that deals with unbounded ends. */
5642
5643 ASSERT_UTF8_CACHE(cache);
5644 if (cache[1] == 0) {
5645 /* Cache is totally empty */
5646 cache[0] = utf8;
5647 cache[1] = byte;
5648 } else if (cache[3] == 0) {
5649 if (byte > cache[1]) {
5650 /* New one is larger, so goes first. */
5651 cache[2] = cache[0];
5652 cache[3] = cache[1];
5653 cache[0] = utf8;
5654 cache[1] = byte;
5655 } else {
5656 cache[2] = utf8;
5657 cache[3] = byte;
5658 }
5659 } else {
5660#define THREEWAY_SQUARE(a,b,c,d) \
5661 ((float)((d) - (c))) * ((float)((d) - (c))) \
5662 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5663 + ((float)((b) - (a))) * ((float)((b) - (a)))
5664
5665 /* Cache has 2 slots in use, and we know three potential pairs.
5666 Keep the two that give the lowest RMS distance. Do the
5667 calcualation in bytes simply because we always know the byte
5668 length. squareroot has the same ordering as the positive value,
5669 so don't bother with the actual square root. */
5670 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5671 if (byte > cache[1]) {
5672 /* New position is after the existing pair of pairs. */
5673 const float keep_earlier
5674 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5675 const float keep_later
5676 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5677
5678 if (keep_later < keep_earlier) {
5679 if (keep_later < existing) {
5680 cache[2] = cache[0];
5681 cache[3] = cache[1];
5682 cache[0] = utf8;
5683 cache[1] = byte;
5684 }
5685 }
5686 else {
5687 if (keep_earlier < existing) {
5688 cache[0] = utf8;
5689 cache[1] = byte;
5690 }
5691 }
5692 }
57d7fbf1
NC
5693 else if (byte > cache[3]) {
5694 /* New position is between the existing pair of pairs. */
5695 const float keep_earlier
5696 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5697 const float keep_later
5698 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5699
5700 if (keep_later < keep_earlier) {
5701 if (keep_later < existing) {
5702 cache[2] = utf8;
5703 cache[3] = byte;
5704 }
5705 }
5706 else {
5707 if (keep_earlier < existing) {
5708 cache[0] = utf8;
5709 cache[1] = byte;
5710 }
5711 }
5712 }
5713 else {
5714 /* New position is before the existing pair of pairs. */
5715 const float keep_earlier
5716 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5717 const float keep_later
5718 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5719
5720 if (keep_later < keep_earlier) {
5721 if (keep_later < existing) {
5722 cache[2] = utf8;
5723 cache[3] = byte;
5724 }
5725 }
5726 else {
5727 if (keep_earlier < existing) {
5728 cache[0] = cache[2];
5729 cache[1] = cache[3];
5730 cache[2] = utf8;
5731 cache[3] = byte;
5732 }
5733 }
5734 }
ab455f60 5735 }
0905937d 5736 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5737}
5738
5739/* If we don't know the character offset of the end of a region, our only
5740 option is to walk forwards to the target byte offset. */
5741static STRLEN
5742S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5743{
5744 STRLEN len = 0;
5745 while (s < target) {
5746 STRLEN n = 1;
5747
5748 /* Call utf8n_to_uvchr() to validate the sequence
5749 * (unless a simple non-UTF character) */
5750 if (!UTF8_IS_INVARIANT(*s))
5751 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5752 if (n > 0) {
5753 s += n;
5754 len++;
5755 }
5756 else
5757 break;
5758 }
5759 return len;
5760}
5761
5762/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5763 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5764 backward is half the speed of walking forward. */
ec07b5e0
NC
5765static STRLEN
5766S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5767 STRLEN endu)
5768{
5769 const STRLEN forw = target - s;
5770 STRLEN backw = end - target;
5771
5772 if (forw < 2 * backw) {
5773 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5774 }
5775
5776 while (end > target) {
5777 end--;
5778 while (UTF8_IS_CONTINUATION(*end)) {
5779 end--;
5780 }
5781 endu--;
5782 }
5783 return endu;
5784}
5785
9564a3bd
NC
5786/*
5787=for apidoc sv_pos_b2u
5788
5789Converts the value pointed to by offsetp from a count of bytes from the
5790start of the string, to a count of the equivalent number of UTF-8 chars.
5791Handles magic and type coercion.
5792
5793=cut
5794*/
5795
5796/*
5797 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5798 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5799 * byte offsets.
5800 *
5801 */
a0ed51b3 5802void
7e8c5dac 5803Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5804{
83003860 5805 const U8* s;
ec07b5e0 5806 const STRLEN byte = *offsetp;
7087a21c 5807 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5808 STRLEN blen;
ec07b5e0
NC
5809 MAGIC* mg = NULL;
5810 const U8* send;
a922f900 5811 bool found = FALSE;
a0ed51b3
LW
5812
5813 if (!sv)
5814 return;
5815
ab455f60 5816 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5817
ab455f60 5818 if (blen < byte)
ec07b5e0 5819 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5820
ec07b5e0 5821 send = s + byte;
a67d7df9 5822
ffca234a
NC
5823 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5824 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5825 if (mg->mg_ptr) {
d4c19fe8 5826 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5827 if (cache[1] == byte) {
ec07b5e0
NC
5828 /* An exact match. */
5829 *offsetp = cache[0];
ec07b5e0 5830 return;
7e8c5dac 5831 }
ab455f60
NC
5832 if (cache[3] == byte) {
5833 /* An exact match. */
5834 *offsetp = cache[2];
5835 return;
5836 }
668af93f
NC
5837
5838 if (cache[1] < byte) {
ec07b5e0 5839 /* We already know part of the way. */
b9f984a5
NC
5840 if (mg->mg_len != -1) {
5841 /* Actually, we know the end too. */
5842 len = cache[0]
5843 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5844 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5845 } else {
5846 len = cache[0]
5847 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5848 }
7e8c5dac 5849 }
9f985e4c
NC
5850 else if (cache[3] < byte) {
5851 /* We're between the two cached pairs, so we do the calculation
5852 offset by the byte/utf-8 positions for the earlier pair,
5853 then add the utf-8 characters from the string start to
5854 there. */
5855 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5856 s + cache[1], cache[0] - cache[2])
5857 + cache[2];
5858
5859 }
5860 else { /* cache[3] > byte */
5861 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5862 cache[2]);
7e8c5dac 5863
7e8c5dac 5864 }
ec07b5e0 5865 ASSERT_UTF8_CACHE(cache);
a922f900 5866 found = TRUE;
ffca234a 5867 } else if (mg->mg_len != -1) {
ab455f60 5868 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5869 found = TRUE;
7e8c5dac 5870 }
a0ed51b3 5871 }
a922f900
NC
5872 if (!found || PL_utf8cache < 0) {
5873 const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5874
5875 if (found && PL_utf8cache < 0) {
5876 if (len != real_len) {
5877 /* Need to turn the assertions off otherwise we may recurse
5878 infinitely while printing error messages. */
5879 SAVEI8(PL_utf8cache);
5880 PL_utf8cache = 0;
5881 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5882 " real %"UVf" for %"SVf,
5883 (UV) len, (UV) real_len, sv);
5884 }
5885 }
5886 len = real_len;
ec07b5e0
NC
5887 }
5888 *offsetp = len;
5889
ab455f60 5890 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5891}
5892
954c1994
GS
5893/*
5894=for apidoc sv_eq
5895
5896Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5897identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5898coerce its args to strings if necessary.
954c1994
GS
5899
5900=cut
5901*/
5902
79072805 5903I32
e01b9e88 5904Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5905{
97aff369 5906 dVAR;
e1ec3a88 5907 const char *pv1;
463ee0b2 5908 STRLEN cur1;
e1ec3a88 5909 const char *pv2;
463ee0b2 5910 STRLEN cur2;
e01b9e88 5911 I32 eq = 0;
bd61b366 5912 char *tpv = NULL;
a0714e2c 5913 SV* svrecode = NULL;
79072805 5914
e01b9e88 5915 if (!sv1) {
79072805
LW
5916 pv1 = "";
5917 cur1 = 0;
5918 }
463ee0b2 5919 else
4d84ee25 5920 pv1 = SvPV_const(sv1, cur1);
79072805 5921
e01b9e88
SC
5922 if (!sv2){
5923 pv2 = "";
5924 cur2 = 0;
92d29cee 5925 }
e01b9e88 5926 else
4d84ee25 5927 pv2 = SvPV_const(sv2, cur2);
79072805 5928
cf48d248 5929 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5930 /* Differing utf8ness.
5931 * Do not UTF8size the comparands as a side-effect. */
5932 if (PL_encoding) {
5933 if (SvUTF8(sv1)) {
553e1bcc
AT
5934 svrecode = newSVpvn(pv2, cur2);
5935 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5936 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5937 }
5938 else {
553e1bcc
AT
5939 svrecode = newSVpvn(pv1, cur1);
5940 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5941 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5942 }
5943 /* Now both are in UTF-8. */
0a1bd7ac
DM
5944 if (cur1 != cur2) {
5945 SvREFCNT_dec(svrecode);
799ef3cb 5946 return FALSE;
0a1bd7ac 5947 }
799ef3cb
JH
5948 }
5949 else {
5950 bool is_utf8 = TRUE;
5951
5952 if (SvUTF8(sv1)) {
5953 /* sv1 is the UTF-8 one,
5954 * if is equal it must be downgrade-able */
9d4ba2ae 5955 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5956 &cur1, &is_utf8);
5957 if (pv != pv1)
553e1bcc 5958 pv1 = tpv = pv;
799ef3cb
JH
5959 }
5960 else {
5961 /* sv2 is the UTF-8 one,
5962 * if is equal it must be downgrade-able */
9d4ba2ae 5963 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5964 &cur2, &is_utf8);
5965 if (pv != pv2)
553e1bcc 5966 pv2 = tpv = pv;
799ef3cb
JH
5967 }
5968 if (is_utf8) {
5969 /* Downgrade not possible - cannot be eq */
bf694877 5970 assert (tpv == 0);
799ef3cb
JH
5971 return FALSE;
5972 }
5973 }
cf48d248
JH
5974 }
5975
5976 if (cur1 == cur2)
765f542d 5977 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5978
b37c2d43 5979 SvREFCNT_dec(svrecode);
553e1bcc
AT
5980 if (tpv)
5981 Safefree(tpv);
cf48d248 5982
e01b9e88 5983 return eq;
79072805
LW
5984}
5985
954c1994
GS
5986/*
5987=for apidoc sv_cmp
5988
5989Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5990string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5991C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5992coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5993
5994=cut
5995*/
5996
79072805 5997I32
e01b9e88 5998Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5999{
97aff369 6000 dVAR;
560a288e 6001 STRLEN cur1, cur2;
e1ec3a88 6002 const char *pv1, *pv2;
bd61b366 6003 char *tpv = NULL;
cf48d248 6004 I32 cmp;
a0714e2c 6005 SV *svrecode = NULL;
560a288e 6006
e01b9e88
SC
6007 if (!sv1) {
6008 pv1 = "";
560a288e
GS
6009 cur1 = 0;
6010 }
e01b9e88 6011 else
4d84ee25 6012 pv1 = SvPV_const(sv1, cur1);
560a288e 6013
553e1bcc 6014 if (!sv2) {
e01b9e88 6015 pv2 = "";
560a288e
GS
6016 cur2 = 0;
6017 }
e01b9e88 6018 else
4d84ee25 6019 pv2 = SvPV_const(sv2, cur2);
79072805 6020
cf48d248 6021 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6022 /* Differing utf8ness.
6023 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6024 if (SvUTF8(sv1)) {
799ef3cb 6025 if (PL_encoding) {
553e1bcc
AT
6026 svrecode = newSVpvn(pv2, cur2);
6027 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6028 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6029 }
6030 else {
e1ec3a88 6031 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6032 }
cf48d248
JH
6033 }
6034 else {
799ef3cb 6035 if (PL_encoding) {
553e1bcc
AT
6036 svrecode = newSVpvn(pv1, cur1);
6037 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6038 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6039 }
6040 else {
e1ec3a88 6041 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6042 }
cf48d248
JH
6043 }
6044 }
6045
e01b9e88 6046 if (!cur1) {
cf48d248 6047 cmp = cur2 ? -1 : 0;
e01b9e88 6048 } else if (!cur2) {
cf48d248
JH
6049 cmp = 1;
6050 } else {
e1ec3a88 6051 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6052
6053 if (retval) {
cf48d248 6054 cmp = retval < 0 ? -1 : 1;
e01b9e88 6055 } else if (cur1 == cur2) {
cf48d248
JH
6056 cmp = 0;
6057 } else {
6058 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6059 }
cf48d248 6060 }
16660edb 6061
b37c2d43 6062 SvREFCNT_dec(svrecode);
553e1bcc
AT
6063 if (tpv)
6064 Safefree(tpv);
cf48d248
JH
6065
6066 return cmp;
bbce6d69 6067}
16660edb 6068
c461cf8f
JH
6069/*
6070=for apidoc sv_cmp_locale
6071
645c22ef
DM
6072Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6073'use bytes' aware, handles get magic, and will coerce its args to strings
6074if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6075
6076=cut
6077*/
6078
bbce6d69 6079I32
864dbfa3 6080Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6081{
97aff369 6082 dVAR;
36477c24 6083#ifdef USE_LOCALE_COLLATE
16660edb 6084
bbce6d69 6085 char *pv1, *pv2;
6086 STRLEN len1, len2;
6087 I32 retval;
16660edb 6088
3280af22 6089 if (PL_collation_standard)
bbce6d69 6090 goto raw_compare;
16660edb 6091
bbce6d69 6092 len1 = 0;
8ac85365 6093 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6094 len2 = 0;
8ac85365 6095 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6096
bbce6d69 6097 if (!pv1 || !len1) {
6098 if (pv2 && len2)
6099 return -1;
6100 else
6101 goto raw_compare;
6102 }
6103 else {
6104 if (!pv2 || !len2)
6105 return 1;
6106 }
16660edb 6107
bbce6d69 6108 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6109
bbce6d69 6110 if (retval)
16660edb 6111 return retval < 0 ? -1 : 1;
6112
bbce6d69 6113 /*
6114 * When the result of collation is equality, that doesn't mean
6115 * that there are no differences -- some locales exclude some
6116 * characters from consideration. So to avoid false equalities,
6117 * we use the raw string as a tiebreaker.
6118 */
16660edb 6119
bbce6d69 6120 raw_compare:
5f66b61c 6121 /*FALLTHROUGH*/
16660edb 6122
36477c24 6123#endif /* USE_LOCALE_COLLATE */
16660edb 6124
bbce6d69 6125 return sv_cmp(sv1, sv2);
6126}
79072805 6127
645c22ef 6128
36477c24 6129#ifdef USE_LOCALE_COLLATE
645c22ef 6130
7a4c00b4 6131/*
645c22ef
DM
6132=for apidoc sv_collxfrm
6133
6134Add Collate Transform magic to an SV if it doesn't already have it.
6135
6136Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6137scalar data of the variable, but transformed to such a format that a normal
6138memory comparison can be used to compare the data according to the locale
6139settings.
6140
6141=cut
6142*/
6143
bbce6d69 6144char *
864dbfa3 6145Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6146{
97aff369 6147 dVAR;
7a4c00b4 6148 MAGIC *mg;
16660edb 6149
14befaf4 6150 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6151 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6152 const char *s;
6153 char *xf;
bbce6d69 6154 STRLEN len, xlen;
6155
7a4c00b4 6156 if (mg)
6157 Safefree(mg->mg_ptr);
93524f2b 6158 s = SvPV_const(sv, len);
bbce6d69 6159 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6160 if (SvREADONLY(sv)) {
6161 SAVEFREEPV(xf);
6162 *nxp = xlen;
3280af22 6163 return xf + sizeof(PL_collation_ix);
ff0cee69 6164 }
7a4c00b4 6165 if (! mg) {
d83f0a82
NC
6166#ifdef PERL_OLD_COPY_ON_WRITE
6167 if (SvIsCOW(sv))
6168 sv_force_normal_flags(sv, 0);
6169#endif
6170 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6171 0, 0);
7a4c00b4 6172 assert(mg);
bbce6d69 6173 }
7a4c00b4 6174 mg->mg_ptr = xf;
565764a8 6175 mg->mg_len = xlen;
7a4c00b4 6176 }
6177 else {
ff0cee69 6178 if (mg) {
6179 mg->mg_ptr = NULL;
565764a8 6180 mg->mg_len = -1;
ff0cee69 6181 }
bbce6d69 6182 }
6183 }
7a4c00b4 6184 if (mg && mg->mg_ptr) {
565764a8 6185 *nxp = mg->mg_len;
3280af22 6186 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6187 }
6188 else {
6189 *nxp = 0;
6190 return NULL;
16660edb 6191 }
79072805
LW
6192}
6193
36477c24 6194#endif /* USE_LOCALE_COLLATE */
bbce6d69 6195
c461cf8f
JH
6196/*
6197=for apidoc sv_gets
6198
6199Get a line from the filehandle and store it into the SV, optionally
6200appending to the currently-stored string.
6201
6202=cut
6203*/
6204
79072805 6205char *
864dbfa3 6206Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6207{
97aff369 6208 dVAR;
e1ec3a88 6209 const char *rsptr;
c07a80fd 6210 STRLEN rslen;
6211 register STDCHAR rslast;
6212 register STDCHAR *bp;
6213 register I32 cnt;
9c5ffd7c 6214 I32 i = 0;
8bfdd7d9 6215 I32 rspara = 0;
c07a80fd 6216
bc44a8a2
NC
6217 if (SvTHINKFIRST(sv))
6218 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6219 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6220 from <>.
6221 However, perlbench says it's slower, because the existing swipe code
6222 is faster than copy on write.
6223 Swings and roundabouts. */
862a34c6 6224 SvUPGRADE(sv, SVt_PV);
99491443 6225
ff68c719 6226 SvSCREAM_off(sv);
efd8b2ba
AE
6227
6228 if (append) {
6229 if (PerlIO_isutf8(fp)) {
6230 if (!SvUTF8(sv)) {
6231 sv_utf8_upgrade_nomg(sv);
6232 sv_pos_u2b(sv,&append,0);
6233 }
6234 } else if (SvUTF8(sv)) {
561b68a9 6235 SV * const tsv = newSV(0);
efd8b2ba
AE
6236 sv_gets(tsv, fp, 0);
6237 sv_utf8_upgrade_nomg(tsv);
6238 SvCUR_set(sv,append);
6239 sv_catsv(sv,tsv);
6240 sv_free(tsv);
6241 goto return_string_or_null;
6242 }
6243 }
6244
6245 SvPOK_only(sv);
6246 if (PerlIO_isutf8(fp))
6247 SvUTF8_on(sv);
c07a80fd 6248
923e4eb5 6249 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6250 /* we always read code in line mode */
6251 rsptr = "\n";
6252 rslen = 1;
6253 }
6254 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6255 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6256 of amount we are going to read -- may result in mallocing
6257 more memory than we really need if the layers below reduce
6258 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6259 */
e311fd51 6260 Stat_t st;
e468d35b 6261 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6262 const Off_t offset = PerlIO_tell(fp);
58f1856e 6263 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6264 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6265 }
6266 }
c07a80fd 6267 rsptr = NULL;
6268 rslen = 0;
6269 }
3280af22 6270 else if (RsRECORD(PL_rs)) {
e311fd51 6271 I32 bytesread;
5b2b9c68 6272 char *buffer;
acbd132f 6273 U32 recsize;
5b2b9c68
HM
6274
6275 /* Grab the size of the record we're getting */
acbd132f 6276 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6277 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6278 /* Go yank in */
6279#ifdef VMS
6280 /* VMS wants read instead of fread, because fread doesn't respect */
6281 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6282 /* doing, but we've got no other real choice - except avoid stdio
6283 as implementation - perhaps write a :vms layer ?
6284 */
5b2b9c68
HM
6285 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6286#else
6287 bytesread = PerlIO_read(fp, buffer, recsize);
6288#endif
27e6ca2d
AE
6289 if (bytesread < 0)
6290 bytesread = 0;
e311fd51 6291 SvCUR_set(sv, bytesread += append);
e670df4e 6292 buffer[bytesread] = '\0';
efd8b2ba 6293 goto return_string_or_null;
5b2b9c68 6294 }
3280af22 6295 else if (RsPARA(PL_rs)) {
c07a80fd 6296 rsptr = "\n\n";
6297 rslen = 2;
8bfdd7d9 6298 rspara = 1;
c07a80fd 6299 }
7d59b7e4
NIS
6300 else {
6301 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6302 if (PerlIO_isutf8(fp)) {
6303 rsptr = SvPVutf8(PL_rs, rslen);
6304 }
6305 else {
6306 if (SvUTF8(PL_rs)) {
6307 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6308 Perl_croak(aTHX_ "Wide character in $/");
6309 }
6310 }
93524f2b 6311 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6312 }
6313 }
6314
c07a80fd 6315 rslast = rslen ? rsptr[rslen - 1] : '\0';
6316
8bfdd7d9 6317 if (rspara) { /* have to do this both before and after */
79072805 6318 do { /* to make sure file boundaries work right */
760ac839 6319 if (PerlIO_eof(fp))
a0d0e21e 6320 return 0;
760ac839 6321 i = PerlIO_getc(fp);
79072805 6322 if (i != '\n') {
a0d0e21e
LW
6323 if (i == -1)
6324 return 0;
760ac839 6325 PerlIO_ungetc(fp,i);
79072805
LW
6326 break;
6327 }
6328 } while (i != EOF);
6329 }
c07a80fd 6330
760ac839
LW
6331 /* See if we know enough about I/O mechanism to cheat it ! */
6332
6333 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6334 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6335 enough here - and may even be a macro allowing compile
6336 time optimization.
6337 */
6338
6339 if (PerlIO_fast_gets(fp)) {
6340
6341 /*
6342 * We're going to steal some values from the stdio struct
6343 * and put EVERYTHING in the innermost loop into registers.
6344 */
6345 register STDCHAR *ptr;
6346 STRLEN bpx;
6347 I32 shortbuffered;
6348
16660edb 6349#if defined(VMS) && defined(PERLIO_IS_STDIO)
6350 /* An ungetc()d char is handled separately from the regular
6351 * buffer, so we getc() it back out and stuff it in the buffer.
6352 */
6353 i = PerlIO_getc(fp);
6354 if (i == EOF) return 0;
6355 *(--((*fp)->_ptr)) = (unsigned char) i;
6356 (*fp)->_cnt++;
6357#endif
c07a80fd 6358
c2960299 6359 /* Here is some breathtakingly efficient cheating */
c07a80fd 6360
a20bf0c3 6361 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6362 /* make sure we have the room */
7a5fa8a2 6363 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6364 /* Not room for all of it
7a5fa8a2 6365 if we are looking for a separator and room for some
e468d35b
NIS
6366 */
6367 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6368 /* just process what we have room for */
79072805
LW
6369 shortbuffered = cnt - SvLEN(sv) + append + 1;
6370 cnt -= shortbuffered;
6371 }
6372 else {
6373 shortbuffered = 0;
bbce6d69 6374 /* remember that cnt can be negative */
eb160463 6375 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6376 }
6377 }
7a5fa8a2 6378 else
79072805 6379 shortbuffered = 0;
3f7c398e 6380 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6381 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6382 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6383 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6384 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6385 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6386 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6387 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6388 for (;;) {
6389 screamer:
93a17b20 6390 if (cnt > 0) {
c07a80fd 6391 if (rslen) {
760ac839
LW
6392 while (cnt > 0) { /* this | eat */
6393 cnt--;
c07a80fd 6394 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6395 goto thats_all_folks; /* screams | sed :-) */
6396 }
6397 }
6398 else {
1c846c1f
NIS
6399 Copy(ptr, bp, cnt, char); /* this | eat */
6400 bp += cnt; /* screams | dust */
c07a80fd 6401 ptr += cnt; /* louder | sed :-) */
a5f75d66 6402 cnt = 0;
93a17b20 6403 }
79072805
LW
6404 }
6405
748a9306 6406 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6407 cnt = shortbuffered;
6408 shortbuffered = 0;
3f7c398e 6409 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6410 SvCUR_set(sv, bpx);
6411 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6412 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6413 continue;
6414 }
6415
16660edb 6416 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6417 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6418 PTR2UV(ptr),(long)cnt));
cc00df79 6419 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6420#if 0
16660edb 6421 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6422 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6423 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6424 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6425#endif
1c846c1f 6426 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6427 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6428 another abstraction. */
760ac839 6429 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6430#if 0
16660edb 6431 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6432 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6433 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6434 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6435#endif
a20bf0c3
JH
6436 cnt = PerlIO_get_cnt(fp);
6437 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6438 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6439 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6440
748a9306
LW
6441 if (i == EOF) /* all done for ever? */
6442 goto thats_really_all_folks;
6443
3f7c398e 6444 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6445 SvCUR_set(sv, bpx);
6446 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6447 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6448
eb160463 6449 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6450
c07a80fd 6451 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6452 goto thats_all_folks;
79072805
LW
6453 }
6454
6455thats_all_folks:
3f7c398e 6456 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6457 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6458 goto screamer; /* go back to the fray */
79072805
LW
6459thats_really_all_folks:
6460 if (shortbuffered)
6461 cnt += shortbuffered;
16660edb 6462 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6463 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6464 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6465 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6466 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6467 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6468 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6469 *bp = '\0';
3f7c398e 6470 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6471 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6472 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6473 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6474 }
6475 else
79072805 6476 {
6edd2cd5 6477 /*The big, slow, and stupid way. */
27da23d5 6478#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6479 STDCHAR *buf = NULL;
a02a5408 6480 Newx(buf, 8192, STDCHAR);
6edd2cd5 6481 assert(buf);
4d2c4e07 6482#else
6edd2cd5 6483 STDCHAR buf[8192];
4d2c4e07 6484#endif
79072805 6485
760ac839 6486screamer2:
c07a80fd 6487 if (rslen) {
00b6aa41 6488 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6489 bp = buf;
eb160463 6490 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6491 ; /* keep reading */
6492 cnt = bp - buf;
c07a80fd 6493 }
6494 else {
760ac839 6495 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6496 /* Accomodate broken VAXC compiler, which applies U8 cast to
6497 * both args of ?: operator, causing EOF to change into 255
6498 */
37be0adf 6499 if (cnt > 0)
cbe9e203
JH
6500 i = (U8)buf[cnt - 1];
6501 else
37be0adf 6502 i = EOF;
c07a80fd 6503 }
79072805 6504
cbe9e203
JH
6505 if (cnt < 0)
6506 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6507 if (append)
6508 sv_catpvn(sv, (char *) buf, cnt);
6509 else
6510 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6511
6512 if (i != EOF && /* joy */
6513 (!rslen ||
6514 SvCUR(sv) < rslen ||
3f7c398e 6515 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6516 {
6517 append = -1;
63e4d877
CS
6518 /*
6519 * If we're reading from a TTY and we get a short read,
6520 * indicating that the user hit his EOF character, we need
6521 * to notice it now, because if we try to read from the TTY
6522 * again, the EOF condition will disappear.
6523 *
6524 * The comparison of cnt to sizeof(buf) is an optimization
6525 * that prevents unnecessary calls to feof().
6526 *
6527 * - jik 9/25/96
6528 */
bb7a0f54 6529 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 6530 goto screamer2;
79072805 6531 }
6edd2cd5 6532
27da23d5 6533#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6534 Safefree(buf);
6535#endif
79072805
LW
6536 }
6537
8bfdd7d9 6538 if (rspara) { /* have to do this both before and after */
c07a80fd 6539 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6540 i = PerlIO_getc(fp);
79072805 6541 if (i != '\n') {
760ac839 6542 PerlIO_ungetc(fp,i);
79072805
LW
6543 break;
6544 }
6545 }
6546 }
c07a80fd 6547
efd8b2ba 6548return_string_or_null:
bd61b366 6549 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6550}
6551
954c1994
GS
6552/*
6553=for apidoc sv_inc
6554
645c22ef
DM
6555Auto-increment of the value in the SV, doing string to numeric conversion
6556if necessary. Handles 'get' magic.
954c1994
GS
6557
6558=cut
6559*/
6560
79072805 6561void
864dbfa3 6562Perl_sv_inc(pTHX_ register SV *sv)
79072805 6563{
97aff369 6564 dVAR;
79072805 6565 register char *d;
463ee0b2 6566 int flags;
79072805
LW
6567
6568 if (!sv)
6569 return;
5b295bef 6570 SvGETMAGIC(sv);
ed6116ce 6571 if (SvTHINKFIRST(sv)) {
765f542d
NC
6572 if (SvIsCOW(sv))
6573 sv_force_normal_flags(sv, 0);
0f15f207 6574 if (SvREADONLY(sv)) {
923e4eb5 6575 if (IN_PERL_RUNTIME)
cea2e8a9 6576 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6577 }
a0d0e21e 6578 if (SvROK(sv)) {
b5be31e9 6579 IV i;
9e7bc3e8
JD
6580 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6581 return;
56431972 6582 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6583 sv_unref(sv);
6584 sv_setiv(sv, i);
a0d0e21e 6585 }
ed6116ce 6586 }
8990e307 6587 flags = SvFLAGS(sv);
28e5dec8
JH
6588 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6589 /* It's (privately or publicly) a float, but not tested as an
6590 integer, so test it to see. */
d460ef45 6591 (void) SvIV(sv);
28e5dec8
JH
6592 flags = SvFLAGS(sv);
6593 }
6594 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6595 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6596#ifdef PERL_PRESERVE_IVUV
28e5dec8 6597 oops_its_int:
59d8ce62 6598#endif
25da4f38
IZ
6599 if (SvIsUV(sv)) {
6600 if (SvUVX(sv) == UV_MAX)
a1e868e7 6601 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6602 else
6603 (void)SvIOK_only_UV(sv);
607fa7f2 6604 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6605 } else {
6606 if (SvIVX(sv) == IV_MAX)
28e5dec8 6607 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6608 else {
6609 (void)SvIOK_only(sv);
45977657 6610 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6611 }
55497cff 6612 }
79072805
LW
6613 return;
6614 }
28e5dec8
JH
6615 if (flags & SVp_NOK) {
6616 (void)SvNOK_only(sv);
9d6ce603 6617 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6618 return;
6619 }
6620
3f7c398e 6621 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6622 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6623 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6624 (void)SvIOK_only(sv);
45977657 6625 SvIV_set(sv, 1);
79072805
LW
6626 return;
6627 }
463ee0b2 6628 d = SvPVX(sv);
79072805
LW
6629 while (isALPHA(*d)) d++;
6630 while (isDIGIT(*d)) d++;
6631 if (*d) {
28e5dec8 6632#ifdef PERL_PRESERVE_IVUV
d1be9408 6633 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6634 warnings. Probably ought to make the sv_iv_please() that does
6635 the conversion if possible, and silently. */
504618e9 6636 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6637 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6638 /* Need to try really hard to see if it's an integer.
6639 9.22337203685478e+18 is an integer.
6640 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6641 so $a="9.22337203685478e+18"; $a+0; $a++
6642 needs to be the same as $a="9.22337203685478e+18"; $a++
6643 or we go insane. */
d460ef45 6644
28e5dec8
JH
6645 (void) sv_2iv(sv);
6646 if (SvIOK(sv))
6647 goto oops_its_int;
6648
6649 /* sv_2iv *should* have made this an NV */
6650 if (flags & SVp_NOK) {
6651 (void)SvNOK_only(sv);
9d6ce603 6652 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6653 return;
6654 }
6655 /* I don't think we can get here. Maybe I should assert this
6656 And if we do get here I suspect that sv_setnv will croak. NWC
6657 Fall through. */
6658#if defined(USE_LONG_DOUBLE)
6659 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 6660 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6661#else
1779d84d 6662 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 6663 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6664#endif
6665 }
6666#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6667 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6668 return;
6669 }
6670 d--;
3f7c398e 6671 while (d >= SvPVX_const(sv)) {
79072805
LW
6672 if (isDIGIT(*d)) {
6673 if (++*d <= '9')
6674 return;
6675 *(d--) = '0';
6676 }
6677 else {
9d116dd7
JH
6678#ifdef EBCDIC
6679 /* MKS: The original code here died if letters weren't consecutive.
6680 * at least it didn't have to worry about non-C locales. The
6681 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6682 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6683 * [A-Za-z] are accepted by isALPHA in the C locale.
6684 */
6685 if (*d != 'z' && *d != 'Z') {
6686 do { ++*d; } while (!isALPHA(*d));
6687 return;
6688 }
6689 *(d--) -= 'z' - 'a';
6690#else
79072805
LW
6691 ++*d;
6692 if (isALPHA(*d))
6693 return;
6694 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6695#endif
79072805
LW
6696 }
6697 }
6698 /* oh,oh, the number grew */
6699 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6700 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6701 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6702 *d = d[-1];
6703 if (isDIGIT(d[1]))
6704 *d = '1';
6705 else
6706 *d = d[1];
6707}
6708
954c1994
GS
6709/*
6710=for apidoc sv_dec
6711
645c22ef
DM
6712Auto-decrement of the value in the SV, doing string to numeric conversion
6713if necessary. Handles 'get' magic.
954c1994
GS
6714
6715=cut
6716*/
6717
79072805 6718void
864dbfa3 6719Perl_sv_dec(pTHX_ register SV *sv)
79072805 6720{
97aff369 6721 dVAR;
463ee0b2
LW
6722 int flags;
6723
79072805
LW
6724 if (!sv)
6725 return;
5b295bef 6726 SvGETMAGIC(sv);
ed6116ce 6727 if (SvTHINKFIRST(sv)) {
765f542d
NC
6728 if (SvIsCOW(sv))
6729 sv_force_normal_flags(sv, 0);
0f15f207 6730 if (SvREADONLY(sv)) {
923e4eb5 6731 if (IN_PERL_RUNTIME)
cea2e8a9 6732 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6733 }
a0d0e21e 6734 if (SvROK(sv)) {
b5be31e9 6735 IV i;
9e7bc3e8
JD
6736 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6737 return;
56431972 6738 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6739 sv_unref(sv);
6740 sv_setiv(sv, i);
a0d0e21e 6741 }
ed6116ce 6742 }
28e5dec8
JH
6743 /* Unlike sv_inc we don't have to worry about string-never-numbers
6744 and keeping them magic. But we mustn't warn on punting */
8990e307 6745 flags = SvFLAGS(sv);
28e5dec8
JH
6746 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6747 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6748#ifdef PERL_PRESERVE_IVUV
28e5dec8 6749 oops_its_int:
59d8ce62 6750#endif
25da4f38
IZ
6751 if (SvIsUV(sv)) {
6752 if (SvUVX(sv) == 0) {
6753 (void)SvIOK_only(sv);
45977657 6754 SvIV_set(sv, -1);
25da4f38
IZ
6755 }
6756 else {
6757 (void)SvIOK_only_UV(sv);
f4eee32f 6758 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6759 }
25da4f38
IZ
6760 } else {
6761 if (SvIVX(sv) == IV_MIN)
65202027 6762 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6763 else {
6764 (void)SvIOK_only(sv);
45977657 6765 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6766 }
55497cff 6767 }
6768 return;
6769 }
28e5dec8 6770 if (flags & SVp_NOK) {
9d6ce603 6771 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6772 (void)SvNOK_only(sv);
6773 return;
6774 }
8990e307 6775 if (!(flags & SVp_POK)) {
ef088171
NC
6776 if ((flags & SVTYPEMASK) < SVt_PVIV)
6777 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6778 SvIV_set(sv, -1);
6779 (void)SvIOK_only(sv);
79072805
LW
6780 return;
6781 }
28e5dec8
JH
6782#ifdef PERL_PRESERVE_IVUV
6783 {
504618e9 6784 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6785 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6786 /* Need to try really hard to see if it's an integer.
6787 9.22337203685478e+18 is an integer.
6788 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6789 so $a="9.22337203685478e+18"; $a+0; $a--
6790 needs to be the same as $a="9.22337203685478e+18"; $a--
6791 or we go insane. */
d460ef45 6792
28e5dec8
JH
6793 (void) sv_2iv(sv);
6794 if (SvIOK(sv))
6795 goto oops_its_int;
6796
6797 /* sv_2iv *should* have made this an NV */
6798 if (flags & SVp_NOK) {
6799 (void)SvNOK_only(sv);
9d6ce603 6800 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6801 return;
6802 }
6803 /* I don't think we can get here. Maybe I should assert this
6804 And if we do get here I suspect that sv_setnv will croak. NWC
6805 Fall through. */
6806#if defined(USE_LONG_DOUBLE)
6807 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 6808 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6809#else
1779d84d 6810 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 6811 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6812#endif
6813 }
6814 }
6815#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6816 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6817}
6818
954c1994
GS
6819/*
6820=for apidoc sv_mortalcopy
6821
645c22ef 6822Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6823The new SV is marked as mortal. It will be destroyed "soon", either by an
6824explicit call to FREETMPS, or by an implicit call at places such as
6825statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6826
6827=cut
6828*/
6829
79072805
LW
6830/* Make a string that will exist for the duration of the expression
6831 * evaluation. Actually, it may have to last longer than that, but
6832 * hopefully we won't free it until it has been assigned to a
6833 * permanent location. */
6834
6835SV *
864dbfa3 6836Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6837{
97aff369 6838 dVAR;
463ee0b2 6839 register SV *sv;
b881518d 6840
4561caa4 6841 new_SV(sv);
79072805 6842 sv_setsv(sv,oldstr);
677b06e3
GS
6843 EXTEND_MORTAL(1);
6844 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6845 SvTEMP_on(sv);
6846 return sv;
6847}
6848
954c1994
GS
6849/*
6850=for apidoc sv_newmortal
6851
645c22ef 6852Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6853set to 1. It will be destroyed "soon", either by an explicit call to
6854FREETMPS, or by an implicit call at places such as statement boundaries.
6855See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6856
6857=cut
6858*/
6859
8990e307 6860SV *
864dbfa3 6861Perl_sv_newmortal(pTHX)
8990e307 6862{
97aff369 6863 dVAR;
8990e307
LW
6864 register SV *sv;
6865
4561caa4 6866 new_SV(sv);
8990e307 6867 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6868 EXTEND_MORTAL(1);
6869 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6870 return sv;
6871}
6872
954c1994
GS
6873/*
6874=for apidoc sv_2mortal
6875
d4236ebc
DM
6876Marks an existing SV as mortal. The SV will be destroyed "soon", either
6877by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6878statement boundaries. SvTEMP() is turned on which means that the SV's
6879string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6880and C<sv_mortalcopy>.
954c1994
GS
6881
6882=cut
6883*/
6884
79072805 6885SV *
864dbfa3 6886Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6887{
27da23d5 6888 dVAR;
79072805 6889 if (!sv)
7a5b473e 6890 return NULL;
d689ffdd 6891 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6892 return sv;
677b06e3
GS
6893 EXTEND_MORTAL(1);
6894 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6895 SvTEMP_on(sv);
79072805
LW
6896 return sv;
6897}
6898
954c1994
GS
6899/*
6900=for apidoc newSVpv
6901
6902Creates a new SV and copies a string into it. The reference count for the
6903SV is set to 1. If C<len> is zero, Perl will compute the length using
6904strlen(). For efficiency, consider using C<newSVpvn> instead.
6905
6906=cut
6907*/
6908
79072805 6909SV *
864dbfa3 6910Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6911{
97aff369 6912 dVAR;
463ee0b2 6913 register SV *sv;
79072805 6914
4561caa4 6915 new_SV(sv);
616d8c9c 6916 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6917 return sv;
6918}
6919
954c1994
GS
6920/*
6921=for apidoc newSVpvn
6922
6923Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6924SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6925string. You are responsible for ensuring that the source string is at least
9e09f5f2 6926C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6927
6928=cut
6929*/
6930
9da1e3b5 6931SV *
864dbfa3 6932Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6933{
97aff369 6934 dVAR;
9da1e3b5
MUN
6935 register SV *sv;
6936
6937 new_SV(sv);
9da1e3b5
MUN
6938 sv_setpvn(sv,s,len);
6939 return sv;
6940}
6941
bd08039b
NC
6942
6943/*
926f8064 6944=for apidoc newSVhek
bd08039b
NC
6945
6946Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6947point to the shared string table where possible. Returns a new (undefined)
6948SV if the hek is NULL.
bd08039b
NC
6949
6950=cut
6951*/
6952
6953SV *
c1b02ed8 6954Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6955{
97aff369 6956 dVAR;
5aaec2b4
NC
6957 if (!hek) {
6958 SV *sv;
6959
6960 new_SV(sv);
6961 return sv;
6962 }
6963
bd08039b
NC
6964 if (HEK_LEN(hek) == HEf_SVKEY) {
6965 return newSVsv(*(SV**)HEK_KEY(hek));
6966 } else {
6967 const int flags = HEK_FLAGS(hek);
6968 if (flags & HVhek_WASUTF8) {
6969 /* Trouble :-)
6970 Andreas would like keys he put in as utf8 to come back as utf8
6971 */
6972 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6973 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6974 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6975
6976 SvUTF8_on (sv);
6977 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6978 return sv;
6979 } else if (flags & HVhek_REHASH) {
6980 /* We don't have a pointer to the hv, so we have to replicate the
6981 flag into every HEK. This hv is using custom a hasing
6982 algorithm. Hence we can't return a shared string scalar, as
6983 that would contain the (wrong) hash value, and might get passed
6984 into an hv routine with a regular hash */
6985
b64e5050 6986 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6987 if (HEK_UTF8(hek))
6988 SvUTF8_on (sv);
6989 return sv;
6990 }
6991 /* This will be overwhelminly the most common case. */
409dfe77
NC
6992 {
6993 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6994 more efficient than sharepvn(). */
6995 SV *sv;
6996
6997 new_SV(sv);
6998 sv_upgrade(sv, SVt_PV);
6999 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7000 SvCUR_set(sv, HEK_LEN(hek));
7001 SvLEN_set(sv, 0);
7002 SvREADONLY_on(sv);
7003 SvFAKE_on(sv);
7004 SvPOK_on(sv);
7005 if (HEK_UTF8(hek))
7006 SvUTF8_on(sv);
7007 return sv;
7008 }
bd08039b
NC
7009 }
7010}
7011
1c846c1f
NIS
7012/*
7013=for apidoc newSVpvn_share
7014
3f7c398e 7015Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7016table. If the string does not already exist in the table, it is created
7017first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7018slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7019otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7020is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7021hash lookup will avoid string compare.
1c846c1f
NIS
7022
7023=cut
7024*/
7025
7026SV *
c3654f1a 7027Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7028{
97aff369 7029 dVAR;
1c846c1f 7030 register SV *sv;
c3654f1a 7031 bool is_utf8 = FALSE;
a51caccf
NC
7032 const char *const orig_src = src;
7033
c3654f1a 7034 if (len < 0) {
77caf834 7035 STRLEN tmplen = -len;
c3654f1a 7036 is_utf8 = TRUE;
75a54232 7037 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7038 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7039 len = tmplen;
7040 }
1c846c1f 7041 if (!hash)
5afd6d42 7042 PERL_HASH(hash, src, len);
1c846c1f 7043 new_SV(sv);
bdd68bc3 7044 sv_upgrade(sv, SVt_PV);
f880fe2f 7045 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7046 SvCUR_set(sv, len);
b162af07 7047 SvLEN_set(sv, 0);
1c846c1f
NIS
7048 SvREADONLY_on(sv);
7049 SvFAKE_on(sv);
7050 SvPOK_on(sv);
c3654f1a
IH
7051 if (is_utf8)
7052 SvUTF8_on(sv);
a51caccf
NC
7053 if (src != orig_src)
7054 Safefree(src);
1c846c1f
NIS
7055 return sv;
7056}
7057
645c22ef 7058
cea2e8a9 7059#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7060
7061/* pTHX_ magic can't cope with varargs, so this is a no-context
7062 * version of the main function, (which may itself be aliased to us).
7063 * Don't access this version directly.
7064 */
7065
46fc3d4c 7066SV *
cea2e8a9 7067Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7068{
cea2e8a9 7069 dTHX;
46fc3d4c 7070 register SV *sv;
7071 va_list args;
46fc3d4c 7072 va_start(args, pat);
c5be433b 7073 sv = vnewSVpvf(pat, &args);
46fc3d4c 7074 va_end(args);
7075 return sv;
7076}
cea2e8a9 7077#endif
46fc3d4c 7078
954c1994
GS
7079/*
7080=for apidoc newSVpvf
7081
645c22ef 7082Creates a new SV and initializes it with the string formatted like
954c1994
GS
7083C<sprintf>.
7084
7085=cut
7086*/
7087
cea2e8a9
GS
7088SV *
7089Perl_newSVpvf(pTHX_ const char* pat, ...)
7090{
7091 register SV *sv;
7092 va_list args;
cea2e8a9 7093 va_start(args, pat);
c5be433b 7094 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7095 va_end(args);
7096 return sv;
7097}
46fc3d4c 7098
645c22ef
DM
7099/* backend for newSVpvf() and newSVpvf_nocontext() */
7100
79072805 7101SV *
c5be433b
GS
7102Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7103{
97aff369 7104 dVAR;
c5be433b
GS
7105 register SV *sv;
7106 new_SV(sv);
4608196e 7107 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7108 return sv;
7109}
7110
954c1994
GS
7111/*
7112=for apidoc newSVnv
7113
7114Creates a new SV and copies a floating point value into it.
7115The reference count for the SV is set to 1.
7116
7117=cut
7118*/
7119
c5be433b 7120SV *
65202027 7121Perl_newSVnv(pTHX_ NV n)
79072805 7122{
97aff369 7123 dVAR;
463ee0b2 7124 register SV *sv;
79072805 7125
4561caa4 7126 new_SV(sv);
79072805
LW
7127 sv_setnv(sv,n);
7128 return sv;
7129}
7130
954c1994
GS
7131/*
7132=for apidoc newSViv
7133
7134Creates a new SV and copies an integer into it. The reference count for the
7135SV is set to 1.
7136
7137=cut
7138*/
7139
79072805 7140SV *
864dbfa3 7141Perl_newSViv(pTHX_ IV i)
79072805 7142{
97aff369 7143 dVAR;
463ee0b2 7144 register SV *sv;
79072805 7145
4561caa4 7146 new_SV(sv);
79072805
LW
7147 sv_setiv(sv,i);
7148 return sv;
7149}
7150
954c1994 7151/*
1a3327fb
JH
7152=for apidoc newSVuv
7153
7154Creates a new SV and copies an unsigned integer into it.
7155The reference count for the SV is set to 1.
7156
7157=cut
7158*/
7159
7160SV *
7161Perl_newSVuv(pTHX_ UV u)
7162{
97aff369 7163 dVAR;
1a3327fb
JH
7164 register SV *sv;
7165
7166 new_SV(sv);
7167 sv_setuv(sv,u);
7168 return sv;
7169}
7170
7171/*
954c1994
GS
7172=for apidoc newRV_noinc
7173
7174Creates an RV wrapper for an SV. The reference count for the original
7175SV is B<not> incremented.
7176
7177=cut
7178*/
7179
2304df62 7180SV *
864dbfa3 7181Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7182{
97aff369 7183 dVAR;
2304df62
AD
7184 register SV *sv;
7185
4561caa4 7186 new_SV(sv);
2304df62 7187 sv_upgrade(sv, SVt_RV);
76e3520e 7188 SvTEMP_off(tmpRef);
b162af07 7189 SvRV_set(sv, tmpRef);
2304df62 7190 SvROK_on(sv);
2304df62
AD
7191 return sv;
7192}
7193
ff276b08 7194/* newRV_inc is the official function name to use now.
645c22ef
DM
7195 * newRV_inc is in fact #defined to newRV in sv.h
7196 */
7197
5f05dabc 7198SV *
7f466ec7 7199Perl_newRV(pTHX_ SV *sv)
5f05dabc 7200{
97aff369 7201 dVAR;
7f466ec7 7202 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7203}
5f05dabc 7204
954c1994
GS
7205/*
7206=for apidoc newSVsv
7207
7208Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7209(Uses C<sv_setsv>).
954c1994
GS
7210
7211=cut
7212*/
7213
79072805 7214SV *
864dbfa3 7215Perl_newSVsv(pTHX_ register SV *old)
79072805 7216{
97aff369 7217 dVAR;
463ee0b2 7218 register SV *sv;
79072805
LW
7219
7220 if (!old)
7a5b473e 7221 return NULL;
8990e307 7222 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7223 if (ckWARN_d(WARN_INTERNAL))
9014280d 7224 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7225 return NULL;
79072805 7226 }
4561caa4 7227 new_SV(sv);
e90aabeb
NC
7228 /* SV_GMAGIC is the default for sv_setv()
7229 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7230 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7231 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7232 return sv;
79072805
LW
7233}
7234
645c22ef
DM
7235/*
7236=for apidoc sv_reset
7237
7238Underlying implementation for the C<reset> Perl function.
7239Note that the perl-level function is vaguely deprecated.
7240
7241=cut
7242*/
7243
79072805 7244void
e1ec3a88 7245Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7246{
27da23d5 7247 dVAR;
4802d5d7 7248 char todo[PERL_UCHAR_MAX+1];
79072805 7249
49d8d3a1
MB
7250 if (!stash)
7251 return;
7252
79072805 7253 if (!*s) { /* reset ?? searches */
aec46f14 7254 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7255 if (mg) {
7256 PMOP *pm = (PMOP *) mg->mg_obj;
7257 while (pm) {
7258 pm->op_pmdynflags &= ~PMdf_USED;
7259 pm = pm->op_pmnext;
7260 }
79072805
LW
7261 }
7262 return;
7263 }
7264
7265 /* reset variables */
7266
7267 if (!HvARRAY(stash))
7268 return;
463ee0b2
LW
7269
7270 Zero(todo, 256, char);
79072805 7271 while (*s) {
b464bac0
AL
7272 I32 max;
7273 I32 i = (unsigned char)*s;
79072805
LW
7274 if (s[1] == '-') {
7275 s += 2;
7276 }
4802d5d7 7277 max = (unsigned char)*s++;
79072805 7278 for ( ; i <= max; i++) {
463ee0b2
LW
7279 todo[i] = 1;
7280 }
a0d0e21e 7281 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7282 HE *entry;
79072805 7283 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7284 entry;
7285 entry = HeNEXT(entry))
7286 {
b464bac0
AL
7287 register GV *gv;
7288 register SV *sv;
7289
1edc1566 7290 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7291 continue;
1edc1566 7292 gv = (GV*)HeVAL(entry);
79072805 7293 sv = GvSV(gv);
e203899d
NC
7294 if (sv) {
7295 if (SvTHINKFIRST(sv)) {
7296 if (!SvREADONLY(sv) && SvROK(sv))
7297 sv_unref(sv);
7298 /* XXX Is this continue a bug? Why should THINKFIRST
7299 exempt us from resetting arrays and hashes? */
7300 continue;
7301 }
7302 SvOK_off(sv);
7303 if (SvTYPE(sv) >= SVt_PV) {
7304 SvCUR_set(sv, 0);
bd61b366 7305 if (SvPVX_const(sv) != NULL)
e203899d
NC
7306 *SvPVX(sv) = '\0';
7307 SvTAINT(sv);
7308 }
79072805
LW
7309 }
7310 if (GvAV(gv)) {
7311 av_clear(GvAV(gv));
7312 }
bfcb3514 7313 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7314#if defined(VMS)
7315 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7316#else /* ! VMS */
463ee0b2 7317 hv_clear(GvHV(gv));
b0269e46
AB
7318# if defined(USE_ENVIRON_ARRAY)
7319 if (gv == PL_envgv)
7320 my_clearenv();
7321# endif /* USE_ENVIRON_ARRAY */
7322#endif /* VMS */
79072805
LW
7323 }
7324 }
7325 }
7326 }
7327}
7328
645c22ef
DM
7329/*
7330=for apidoc sv_2io
7331
7332Using various gambits, try to get an IO from an SV: the IO slot if its a
7333GV; or the recursive result if we're an RV; or the IO slot of the symbol
7334named after the PV if we're a string.
7335
7336=cut
7337*/
7338
46fc3d4c 7339IO*
864dbfa3 7340Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7341{
7342 IO* io;
7343 GV* gv;
7344
7345 switch (SvTYPE(sv)) {
7346 case SVt_PVIO:
7347 io = (IO*)sv;
7348 break;
7349 case SVt_PVGV:
7350 gv = (GV*)sv;
7351 io = GvIO(gv);
7352 if (!io)
cea2e8a9 7353 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7354 break;
7355 default:
7356 if (!SvOK(sv))
cea2e8a9 7357 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7358 if (SvROK(sv))
7359 return sv_2io(SvRV(sv));
f776e3cd 7360 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7361 if (gv)
7362 io = GvIO(gv);
7363 else
7364 io = 0;
7365 if (!io)
35c1215d 7366 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7367 break;
7368 }
7369 return io;
7370}
7371
645c22ef
DM
7372/*
7373=for apidoc sv_2cv
7374
7375Using various gambits, try to get a CV from an SV; in addition, try if
7376possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7377The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7378
7379=cut
7380*/
7381
79072805 7382CV *
864dbfa3 7383Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7384{
27da23d5 7385 dVAR;
a0714e2c 7386 GV *gv = NULL;
601f1833 7387 CV *cv = NULL;
79072805 7388
85dec29a
NC
7389 if (!sv) {
7390 *st = NULL;
7391 *gvp = NULL;
7392 return NULL;
7393 }
79072805 7394 switch (SvTYPE(sv)) {
79072805
LW
7395 case SVt_PVCV:
7396 *st = CvSTASH(sv);
a0714e2c 7397 *gvp = NULL;
79072805
LW
7398 return (CV*)sv;
7399 case SVt_PVHV:
7400 case SVt_PVAV:
ef58ba18 7401 *st = NULL;
a0714e2c 7402 *gvp = NULL;
601f1833 7403 return NULL;
8990e307
LW
7404 case SVt_PVGV:
7405 gv = (GV*)sv;
a0d0e21e 7406 *gvp = gv;
8990e307
LW
7407 *st = GvESTASH(gv);
7408 goto fix_gv;
7409
79072805 7410 default:
5b295bef 7411 SvGETMAGIC(sv);
a0d0e21e 7412 if (SvROK(sv)) {
823a54a3 7413 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7414 tryAMAGICunDEREF(to_cv);
7415
62f274bf
GS
7416 sv = SvRV(sv);
7417 if (SvTYPE(sv) == SVt_PVCV) {
7418 cv = (CV*)sv;
a0714e2c 7419 *gvp = NULL;
62f274bf
GS
7420 *st = CvSTASH(cv);
7421 return cv;
7422 }
7423 else if(isGV(sv))
7424 gv = (GV*)sv;
7425 else
cea2e8a9 7426 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7427 }
62f274bf 7428 else if (isGV(sv))
79072805
LW
7429 gv = (GV*)sv;
7430 else
7a5fd60d 7431 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7432 *gvp = gv;
ef58ba18
NC
7433 if (!gv) {
7434 *st = NULL;
601f1833 7435 return NULL;
ef58ba18 7436 }
e26df76a
NC
7437 /* Some flags to gv_fetchsv mean don't really create the GV */
7438 if (SvTYPE(gv) != SVt_PVGV) {
7439 *st = NULL;
7440 return NULL;
7441 }
79072805 7442 *st = GvESTASH(gv);
8990e307 7443 fix_gv:
8ebc5c01 7444 if (lref && !GvCVu(gv)) {
4633a7c4 7445 SV *tmpsv;
748a9306 7446 ENTER;
561b68a9 7447 tmpsv = newSV(0);
bd61b366 7448 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7449 /* XXX this is probably not what they think they're getting.
7450 * It has the same effect as "sub name;", i.e. just a forward
7451 * declaration! */
774d564b 7452 newSUB(start_subparse(FALSE, 0),
4633a7c4 7453 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7454 NULL, NULL);
748a9306 7455 LEAVE;
8ebc5c01 7456 if (!GvCVu(gv))
35c1215d
NC
7457 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7458 sv);
8990e307 7459 }
8ebc5c01 7460 return GvCVu(gv);
79072805
LW
7461 }
7462}
7463
c461cf8f
JH
7464/*
7465=for apidoc sv_true
7466
7467Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7468Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7469instead use an in-line version.
c461cf8f
JH
7470
7471=cut
7472*/
7473
79072805 7474I32
864dbfa3 7475Perl_sv_true(pTHX_ register SV *sv)
79072805 7476{
8990e307
LW
7477 if (!sv)
7478 return 0;
79072805 7479 if (SvPOK(sv)) {
823a54a3
AL
7480 register const XPV* const tXpv = (XPV*)SvANY(sv);
7481 if (tXpv &&
c2f1de04 7482 (tXpv->xpv_cur > 1 ||
339049b0 7483 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7484 return 1;
7485 else
7486 return 0;
7487 }
7488 else {
7489 if (SvIOK(sv))
463ee0b2 7490 return SvIVX(sv) != 0;
79072805
LW
7491 else {
7492 if (SvNOK(sv))
463ee0b2 7493 return SvNVX(sv) != 0.0;
79072805 7494 else
463ee0b2 7495 return sv_2bool(sv);
79072805
LW
7496 }
7497 }
7498}
79072805 7499
645c22ef 7500/*
c461cf8f
JH
7501=for apidoc sv_pvn_force
7502
7503Get a sensible string out of the SV somehow.
645c22ef
DM
7504A private implementation of the C<SvPV_force> macro for compilers which
7505can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7506
8d6d96c1
HS
7507=for apidoc sv_pvn_force_flags
7508
7509Get a sensible string out of the SV somehow.
7510If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7511appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7512implemented in terms of this function.
645c22ef
DM
7513You normally want to use the various wrapper macros instead: see
7514C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7515
7516=cut
7517*/
7518
7519char *
7520Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7521{
97aff369 7522 dVAR;
6fc92669 7523 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7524 sv_force_normal_flags(sv, 0);
1c846c1f 7525
a0d0e21e 7526 if (SvPOK(sv)) {
13c5b33c
NC
7527 if (lp)
7528 *lp = SvCUR(sv);
a0d0e21e
LW
7529 }
7530 else {
a3b680e6 7531 char *s;
13c5b33c
NC
7532 STRLEN len;
7533
4d84ee25 7534 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7535 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7536 if (PL_op)
7537 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7538 ref, OP_NAME(PL_op));
4d84ee25 7539 else
b64e5050 7540 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7541 }
b64e5050 7542 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7543 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7544 OP_NAME(PL_op));
b64e5050 7545 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7546 if (lp)
7547 *lp = len;
7548
3f7c398e 7549 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7550 if (SvROK(sv))
7551 sv_unref(sv);
862a34c6 7552 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7553 SvGROW(sv, len + 1);
706aa1c9 7554 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7555 SvCUR_set(sv, len);
7556 *SvEND(sv) = '\0';
7557 }
7558 if (!SvPOK(sv)) {
7559 SvPOK_on(sv); /* validate pointer */
7560 SvTAINT(sv);
1d7c1841 7561 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7562 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7563 }
7564 }
4d84ee25 7565 return SvPVX_mutable(sv);
a0d0e21e
LW
7566}
7567
645c22ef 7568/*
645c22ef
DM
7569=for apidoc sv_pvbyten_force
7570
0feed65a 7571The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7572
7573=cut
7574*/
7575
7340a771
GS
7576char *
7577Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7578{
46ec2f14 7579 sv_pvn_force(sv,lp);
ffebcc3e 7580 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7581 *lp = SvCUR(sv);
7582 return SvPVX(sv);
7340a771
GS
7583}
7584
645c22ef 7585/*
c461cf8f
JH
7586=for apidoc sv_pvutf8n_force
7587
0feed65a 7588The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7589
7590=cut
7591*/
7592
7340a771
GS
7593char *
7594Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7595{
46ec2f14 7596 sv_pvn_force(sv,lp);
560a288e 7597 sv_utf8_upgrade(sv);
46ec2f14
TS
7598 *lp = SvCUR(sv);
7599 return SvPVX(sv);
7340a771
GS
7600}
7601
c461cf8f
JH
7602/*
7603=for apidoc sv_reftype
7604
7605Returns a string describing what the SV is a reference to.
7606
7607=cut
7608*/
7609
1cb0ed9b 7610char *
bfed75c6 7611Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7612{
07409e01
NC
7613 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7614 inside return suggests a const propagation bug in g++. */
c86bf373 7615 if (ob && SvOBJECT(sv)) {
1b6737cc 7616 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7617 return name ? name : (char *) "__ANON__";
c86bf373 7618 }
a0d0e21e
LW
7619 else {
7620 switch (SvTYPE(sv)) {
7621 case SVt_NULL:
7622 case SVt_IV:
7623 case SVt_NV:
7624 case SVt_RV:
7625 case SVt_PV:
7626 case SVt_PVIV:
7627 case SVt_PVNV:
7628 case SVt_PVMG:
7629 case SVt_PVBM:
1cb0ed9b 7630 if (SvVOK(sv))
439cb1c4 7631 return "VSTRING";
a0d0e21e
LW
7632 if (SvROK(sv))
7633 return "REF";
7634 else
7635 return "SCALAR";
1cb0ed9b 7636
07409e01 7637 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7638 /* tied lvalues should appear to be
7639 * scalars for backwards compatitbility */
7640 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7641 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7642 case SVt_PVAV: return "ARRAY";
7643 case SVt_PVHV: return "HASH";
7644 case SVt_PVCV: return "CODE";
7645 case SVt_PVGV: return "GLOB";
1d2dff63 7646 case SVt_PVFM: return "FORMAT";
27f9d8f3 7647 case SVt_PVIO: return "IO";
a0d0e21e
LW
7648 default: return "UNKNOWN";
7649 }
7650 }
7651}
7652
954c1994
GS
7653/*
7654=for apidoc sv_isobject
7655
7656Returns a boolean indicating whether the SV is an RV pointing to a blessed
7657object. If the SV is not an RV, or if the object is not blessed, then this
7658will return false.
7659
7660=cut
7661*/
7662
463ee0b2 7663int
864dbfa3 7664Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7665{
68dc0745 7666 if (!sv)
7667 return 0;
5b295bef 7668 SvGETMAGIC(sv);
85e6fe83
LW
7669 if (!SvROK(sv))
7670 return 0;
7671 sv = (SV*)SvRV(sv);
7672 if (!SvOBJECT(sv))
7673 return 0;
7674 return 1;
7675}
7676
954c1994
GS
7677/*
7678=for apidoc sv_isa
7679
7680Returns a boolean indicating whether the SV is blessed into the specified
7681class. This does not check for subtypes; use C<sv_derived_from> to verify
7682an inheritance relationship.
7683
7684=cut
7685*/
7686
85e6fe83 7687int
864dbfa3 7688Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7689{
bfcb3514 7690 const char *hvname;
68dc0745 7691 if (!sv)
7692 return 0;
5b295bef 7693 SvGETMAGIC(sv);
ed6116ce 7694 if (!SvROK(sv))
463ee0b2 7695 return 0;
ed6116ce
LW
7696 sv = (SV*)SvRV(sv);
7697 if (!SvOBJECT(sv))
463ee0b2 7698 return 0;
bfcb3514
NC
7699 hvname = HvNAME_get(SvSTASH(sv));
7700 if (!hvname)
e27ad1f2 7701 return 0;
463ee0b2 7702
bfcb3514 7703 return strEQ(hvname, name);
463ee0b2
LW
7704}
7705
954c1994
GS
7706/*
7707=for apidoc newSVrv
7708
7709Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7710it will be upgraded to one. If C<classname> is non-null then the new SV will
7711be blessed in the specified package. The new SV is returned and its
7712reference count is 1.
7713
7714=cut
7715*/
7716
463ee0b2 7717SV*
864dbfa3 7718Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7719{
97aff369 7720 dVAR;
463ee0b2
LW
7721 SV *sv;
7722
4561caa4 7723 new_SV(sv);
51cf62d8 7724
765f542d 7725 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7726 SvAMAGIC_off(rv);
51cf62d8 7727
0199fce9 7728 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7729 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7730 SvREFCNT(rv) = 0;
7731 sv_clear(rv);
7732 SvFLAGS(rv) = 0;
7733 SvREFCNT(rv) = refcnt;
0199fce9 7734
dc5494d2
NC
7735 sv_upgrade(rv, SVt_RV);
7736 } else if (SvROK(rv)) {
7737 SvREFCNT_dec(SvRV(rv));
7738 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7739 sv_upgrade(rv, SVt_RV);
7740 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7741 SvPV_free(rv);
0199fce9
JD
7742 SvCUR_set(rv, 0);
7743 SvLEN_set(rv, 0);
7744 }
51cf62d8 7745
0c34ef67 7746 SvOK_off(rv);
b162af07 7747 SvRV_set(rv, sv);
ed6116ce 7748 SvROK_on(rv);
463ee0b2 7749
a0d0e21e 7750 if (classname) {
1b6737cc 7751 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7752 (void)sv_bless(rv, stash);
7753 }
7754 return sv;
7755}
7756
954c1994
GS
7757/*
7758=for apidoc sv_setref_pv
7759
7760Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7761argument will be upgraded to an RV. That RV will be modified to point to
7762the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7763into the SV. The C<classname> argument indicates the package for the
bd61b366 7764blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7765will have a reference count of 1, and the RV will be returned.
954c1994
GS
7766
7767Do not use with other Perl types such as HV, AV, SV, CV, because those
7768objects will become corrupted by the pointer copy process.
7769
7770Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7771
7772=cut
7773*/
7774
a0d0e21e 7775SV*
864dbfa3 7776Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7777{
97aff369 7778 dVAR;
189b2af5 7779 if (!pv) {
3280af22 7780 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7781 SvSETMAGIC(rv);
7782 }
a0d0e21e 7783 else
56431972 7784 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7785 return rv;
7786}
7787
954c1994
GS
7788/*
7789=for apidoc sv_setref_iv
7790
7791Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7792argument will be upgraded to an RV. That RV will be modified to point to
7793the new SV. The C<classname> argument indicates the package for the
bd61b366 7794blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7795will have a reference count of 1, and the RV will be returned.
954c1994
GS
7796
7797=cut
7798*/
7799
a0d0e21e 7800SV*
864dbfa3 7801Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7802{
7803 sv_setiv(newSVrv(rv,classname), iv);
7804 return rv;
7805}
7806
954c1994 7807/*
e1c57cef
JH
7808=for apidoc sv_setref_uv
7809
7810Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7811argument will be upgraded to an RV. That RV will be modified to point to
7812the new SV. The C<classname> argument indicates the package for the
bd61b366 7813blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7814will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7815
7816=cut
7817*/
7818
7819SV*
7820Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7821{
7822 sv_setuv(newSVrv(rv,classname), uv);
7823 return rv;
7824}
7825
7826/*
954c1994
GS
7827=for apidoc sv_setref_nv
7828
7829Copies a double into a new SV, optionally blessing the SV. The C<rv>
7830argument will be upgraded to an RV. That RV will be modified to point to
7831the new SV. The C<classname> argument indicates the package for the
bd61b366 7832blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7833will have a reference count of 1, and the RV will be returned.
954c1994
GS
7834
7835=cut
7836*/
7837
a0d0e21e 7838SV*
65202027 7839Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7840{
7841 sv_setnv(newSVrv(rv,classname), nv);
7842 return rv;
7843}
463ee0b2 7844
954c1994
GS
7845/*
7846=for apidoc sv_setref_pvn
7847
7848Copies a string into a new SV, optionally blessing the SV. The length of the
7849string must be specified with C<n>. The C<rv> argument will be upgraded to
7850an RV. That RV will be modified to point to the new SV. The C<classname>
7851argument indicates the package for the blessing. Set C<classname> to
bd61b366 7852C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7853of 1, and the RV will be returned.
954c1994
GS
7854
7855Note that C<sv_setref_pv> copies the pointer while this copies the string.
7856
7857=cut
7858*/
7859
a0d0e21e 7860SV*
1b6737cc 7861Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7862{
7863 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7864 return rv;
7865}
7866
954c1994
GS
7867/*
7868=for apidoc sv_bless
7869
7870Blesses an SV into a specified package. The SV must be an RV. The package
7871must be designated by its stash (see C<gv_stashpv()>). The reference count
7872of the SV is unaffected.
7873
7874=cut
7875*/
7876
a0d0e21e 7877SV*
864dbfa3 7878Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7879{
97aff369 7880 dVAR;
76e3520e 7881 SV *tmpRef;
a0d0e21e 7882 if (!SvROK(sv))
cea2e8a9 7883 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7884 tmpRef = SvRV(sv);
7885 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7886 if (SvREADONLY(tmpRef))
cea2e8a9 7887 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7888 if (SvOBJECT(tmpRef)) {
7889 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7890 --PL_sv_objcount;
76e3520e 7891 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7892 }
a0d0e21e 7893 }
76e3520e
GS
7894 SvOBJECT_on(tmpRef);
7895 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7896 ++PL_sv_objcount;
862a34c6 7897 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7898 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7899
2e3febc6
CS
7900 if (Gv_AMG(stash))
7901 SvAMAGIC_on(sv);
7902 else
7903 SvAMAGIC_off(sv);
a0d0e21e 7904
1edbfb88
AB
7905 if(SvSMAGICAL(tmpRef))
7906 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7907 mg_set(tmpRef);
7908
7909
ecdeb87c 7910
a0d0e21e
LW
7911 return sv;
7912}
7913
645c22ef 7914/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7915 */
7916
76e3520e 7917STATIC void
cea2e8a9 7918S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7919{
97aff369 7920 dVAR;
850fabdf 7921 void *xpvmg;
b37c2d43 7922 SV * const temp = sv_newmortal();
850fabdf 7923
a0d0e21e
LW
7924 assert(SvTYPE(sv) == SVt_PVGV);
7925 SvFAKE_off(sv);
180488f8
NC
7926 gv_efullname3(temp, (GV *) sv, "*");
7927
f7877b28 7928 if (GvGP(sv)) {
1edc1566 7929 gp_free((GV*)sv);
f7877b28 7930 }
e826b3c7 7931 if (GvSTASH(sv)) {
e15faf7d 7932 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7933 GvSTASH(sv) = NULL;
e826b3c7 7934 }
a5f75d66 7935 GvMULTI_off(sv);
acda4c6a
NC
7936 if (GvNAME_HEK(sv)) {
7937 unshare_hek(GvNAME_HEK(sv));
7938 }
dedf8e73 7939 SvSCREAM_off(sv);
850fabdf
GS
7940
7941 /* need to keep SvANY(sv) in the right arena */
7942 xpvmg = new_XPVMG();
7943 StructCopy(SvANY(sv), xpvmg, XPVMG);
7944 del_XPVGV(SvANY(sv));
7945 SvANY(sv) = xpvmg;
7946
a0d0e21e
LW
7947 SvFLAGS(sv) &= ~SVTYPEMASK;
7948 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7949
7950 /* Intentionally not calling any local SET magic, as this isn't so much a
7951 set operation as merely an internal storage change. */
7952 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7953}
7954
954c1994 7955/*
840a7b70 7956=for apidoc sv_unref_flags
954c1994
GS
7957
7958Unsets the RV status of the SV, and decrements the reference count of
7959whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7960as a reversal of C<newSVrv>. The C<cflags> argument can contain
7961C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7962(otherwise the decrementing is conditional on the reference count being
7963different from one or the reference being a readonly SV).
7889fe52 7964See C<SvROK_off>.
954c1994
GS
7965
7966=cut
7967*/
7968
ed6116ce 7969void
e15faf7d 7970Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7971{
b64e5050 7972 SV* const target = SvRV(ref);
810b8aa5 7973
e15faf7d
NC
7974 if (SvWEAKREF(ref)) {
7975 sv_del_backref(target, ref);
7976 SvWEAKREF_off(ref);
7977 SvRV_set(ref, NULL);
810b8aa5
GS
7978 return;
7979 }
e15faf7d
NC
7980 SvRV_set(ref, NULL);
7981 SvROK_off(ref);
7982 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7983 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7984 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7985 SvREFCNT_dec(target);
840a7b70 7986 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7987 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7988}
8990e307 7989
840a7b70 7990/*
645c22ef
DM
7991=for apidoc sv_untaint
7992
7993Untaint an SV. Use C<SvTAINTED_off> instead.
7994=cut
7995*/
7996
bbce6d69 7997void
864dbfa3 7998Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7999{
13f57bf8 8000 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 8001 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 8002 if (mg)
565764a8 8003 mg->mg_len &= ~1;
36477c24 8004 }
bbce6d69 8005}
8006
645c22ef
DM
8007/*
8008=for apidoc sv_tainted
8009
8010Test an SV for taintedness. Use C<SvTAINTED> instead.
8011=cut
8012*/
8013
bbce6d69 8014bool
864dbfa3 8015Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8016{
13f57bf8 8017 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8018 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8019 if (mg && (mg->mg_len & 1) )
36477c24 8020 return TRUE;
8021 }
8022 return FALSE;
bbce6d69 8023}
8024
09540bc3
JH
8025/*
8026=for apidoc sv_setpviv
8027
8028Copies an integer into the given SV, also updating its string value.
8029Does not handle 'set' magic. See C<sv_setpviv_mg>.
8030
8031=cut
8032*/
8033
8034void
8035Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8036{
8037 char buf[TYPE_CHARS(UV)];
8038 char *ebuf;
b64e5050 8039 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8040
8041 sv_setpvn(sv, ptr, ebuf - ptr);
8042}
8043
8044/*
8045=for apidoc sv_setpviv_mg
8046
8047Like C<sv_setpviv>, but also handles 'set' magic.
8048
8049=cut
8050*/
8051
8052void
8053Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8054{
df7eb254 8055 sv_setpviv(sv, iv);
09540bc3
JH
8056 SvSETMAGIC(sv);
8057}
8058
cea2e8a9 8059#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8060
8061/* pTHX_ magic can't cope with varargs, so this is a no-context
8062 * version of the main function, (which may itself be aliased to us).
8063 * Don't access this version directly.
8064 */
8065
cea2e8a9
GS
8066void
8067Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8068{
8069 dTHX;
8070 va_list args;
8071 va_start(args, pat);
c5be433b 8072 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8073 va_end(args);
8074}
8075
645c22ef
DM
8076/* pTHX_ magic can't cope with varargs, so this is a no-context
8077 * version of the main function, (which may itself be aliased to us).
8078 * Don't access this version directly.
8079 */
cea2e8a9
GS
8080
8081void
8082Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8083{
8084 dTHX;
8085 va_list args;
8086 va_start(args, pat);
c5be433b 8087 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8088 va_end(args);
cea2e8a9
GS
8089}
8090#endif
8091
954c1994
GS
8092/*
8093=for apidoc sv_setpvf
8094
bffc3d17
SH
8095Works like C<sv_catpvf> but copies the text into the SV instead of
8096appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8097
8098=cut
8099*/
8100
46fc3d4c 8101void
864dbfa3 8102Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8103{
8104 va_list args;
46fc3d4c 8105 va_start(args, pat);
c5be433b 8106 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8107 va_end(args);
8108}
8109
bffc3d17
SH
8110/*
8111=for apidoc sv_vsetpvf
8112
8113Works like C<sv_vcatpvf> but copies the text into the SV instead of
8114appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8115
8116Usually used via its frontend C<sv_setpvf>.
8117
8118=cut
8119*/
645c22ef 8120
c5be433b
GS
8121void
8122Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8123{
4608196e 8124 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8125}
ef50df4b 8126
954c1994
GS
8127/*
8128=for apidoc sv_setpvf_mg
8129
8130Like C<sv_setpvf>, but also handles 'set' magic.
8131
8132=cut
8133*/
8134
ef50df4b 8135void
864dbfa3 8136Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8137{
8138 va_list args;
ef50df4b 8139 va_start(args, pat);
c5be433b 8140 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8141 va_end(args);
c5be433b
GS
8142}
8143
bffc3d17
SH
8144/*
8145=for apidoc sv_vsetpvf_mg
8146
8147Like C<sv_vsetpvf>, but also handles 'set' magic.
8148
8149Usually used via its frontend C<sv_setpvf_mg>.
8150
8151=cut
8152*/
645c22ef 8153
c5be433b
GS
8154void
8155Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8156{
4608196e 8157 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8158 SvSETMAGIC(sv);
8159}
8160
cea2e8a9 8161#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8162
8163/* pTHX_ magic can't cope with varargs, so this is a no-context
8164 * version of the main function, (which may itself be aliased to us).
8165 * Don't access this version directly.
8166 */
8167
cea2e8a9
GS
8168void
8169Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8170{
8171 dTHX;
8172 va_list args;
8173 va_start(args, pat);
c5be433b 8174 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8175 va_end(args);
8176}
8177
645c22ef
DM
8178/* pTHX_ magic can't cope with varargs, so this is a no-context
8179 * version of the main function, (which may itself be aliased to us).
8180 * Don't access this version directly.
8181 */
8182
cea2e8a9
GS
8183void
8184Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8185{
8186 dTHX;
8187 va_list args;
8188 va_start(args, pat);
c5be433b 8189 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8190 va_end(args);
cea2e8a9
GS
8191}
8192#endif
8193
954c1994
GS
8194/*
8195=for apidoc sv_catpvf
8196
d5ce4a7c
GA
8197Processes its arguments like C<sprintf> and appends the formatted
8198output to an SV. If the appended data contains "wide" characters
8199(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8200and characters >255 formatted with %c), the original SV might get
bffc3d17 8201upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8202C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8203valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8204
d5ce4a7c 8205=cut */
954c1994 8206
46fc3d4c 8207void
864dbfa3 8208Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8209{
8210 va_list args;
46fc3d4c 8211 va_start(args, pat);
c5be433b 8212 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8213 va_end(args);
8214}
8215
bffc3d17
SH
8216/*
8217=for apidoc sv_vcatpvf
8218
8219Processes its arguments like C<vsprintf> and appends the formatted output
8220to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8221
8222Usually used via its frontend C<sv_catpvf>.
8223
8224=cut
8225*/
645c22ef 8226
ef50df4b 8227void
c5be433b
GS
8228Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8229{
4608196e 8230 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8231}
8232
954c1994
GS
8233/*
8234=for apidoc sv_catpvf_mg
8235
8236Like C<sv_catpvf>, but also handles 'set' magic.
8237
8238=cut
8239*/
8240
c5be433b 8241void
864dbfa3 8242Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8243{
8244 va_list args;
ef50df4b 8245 va_start(args, pat);
c5be433b 8246 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8247 va_end(args);
c5be433b
GS
8248}
8249
bffc3d17
SH
8250/*
8251=for apidoc sv_vcatpvf_mg
8252
8253Like C<sv_vcatpvf>, but also handles 'set' magic.
8254
8255Usually used via its frontend C<sv_catpvf_mg>.
8256
8257=cut
8258*/
645c22ef 8259
c5be433b
GS
8260void
8261Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8262{
4608196e 8263 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8264 SvSETMAGIC(sv);
8265}
8266
954c1994
GS
8267/*
8268=for apidoc sv_vsetpvfn
8269
bffc3d17 8270Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8271appending it.
8272
bffc3d17 8273Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8274
954c1994
GS
8275=cut
8276*/
8277
46fc3d4c 8278void
7d5ea4e7 8279Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8280{
8281 sv_setpvn(sv, "", 0);
7d5ea4e7 8282 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8283}
8284
2d00ba3b 8285STATIC I32
9dd79c3f 8286S_expect_number(pTHX_ char** pattern)
211dfcf1 8287{
97aff369 8288 dVAR;
211dfcf1
HS
8289 I32 var = 0;
8290 switch (**pattern) {
8291 case '1': case '2': case '3':
8292 case '4': case '5': case '6':
8293 case '7': case '8': case '9':
2fba7546
GA
8294 var = *(*pattern)++ - '0';
8295 while (isDIGIT(**pattern)) {
5f66b61c 8296 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8297 if (tmp < var)
8298 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8299 var = tmp;
8300 }
211dfcf1
HS
8301 }
8302 return var;
8303}
211dfcf1 8304
c445ea15
AL
8305STATIC char *
8306S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8307{
a3b680e6 8308 const int neg = nv < 0;
4151a5fe 8309 UV uv;
4151a5fe
IZ
8310
8311 if (neg)
8312 nv = -nv;
8313 if (nv < UV_MAX) {
b464bac0 8314 char *p = endbuf;
4151a5fe 8315 nv += 0.5;
028f8eaa 8316 uv = (UV)nv;
4151a5fe
IZ
8317 if (uv & 1 && uv == nv)
8318 uv--; /* Round to even */
8319 do {
a3b680e6 8320 const unsigned dig = uv % 10;
4151a5fe
IZ
8321 *--p = '0' + dig;
8322 } while (uv /= 10);
8323 if (neg)
8324 *--p = '-';
8325 *len = endbuf - p;
8326 return p;
8327 }
bd61b366 8328 return NULL;
4151a5fe
IZ
8329}
8330
8331
954c1994
GS
8332/*
8333=for apidoc sv_vcatpvfn
8334
8335Processes its arguments like C<vsprintf> and appends the formatted output
8336to an SV. Uses an array of SVs if the C style variable argument list is
8337missing (NULL). When running with taint checks enabled, indicates via
8338C<maybe_tainted> if results are untrustworthy (often due to the use of
8339locales).
8340
bffc3d17 8341Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8342
954c1994
GS
8343=cut
8344*/
8345
8896765a
RB
8346
8347#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8348 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8349 vec_utf8 = DO_UTF8(vecsv);
8350
1ef29b0e
RGS
8351/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8352
46fc3d4c 8353void
7d5ea4e7 8354Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8355{
97aff369 8356 dVAR;
46fc3d4c 8357 char *p;
8358 char *q;
a3b680e6 8359 const char *patend;
fc36a67e 8360 STRLEN origlen;
46fc3d4c 8361 I32 svix = 0;
27da23d5 8362 static const char nullstr[] = "(null)";
a0714e2c 8363 SV *argsv = NULL;
b464bac0
AL
8364 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8365 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8366 SV *nsv = NULL;
4151a5fe
IZ
8367 /* Times 4: a decimal digit takes more than 3 binary digits.
8368 * NV_DIG: mantissa takes than many decimal digits.
8369 * Plus 32: Playing safe. */
8370 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8371 /* large enough for "%#.#f" --chip */
8372 /* what about long double NVs? --jhi */
db79b45b 8373
53c1dcc0
AL
8374 PERL_UNUSED_ARG(maybe_tainted);
8375
46fc3d4c 8376 /* no matter what, this is a string now */
fc36a67e 8377 (void)SvPV_force(sv, origlen);
46fc3d4c 8378
8896765a 8379 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8380 if (patlen == 0)
8381 return;
0dbb1585 8382 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8383 if (args) {
8384 const char * const s = va_arg(*args, char*);
8385 sv_catpv(sv, s ? s : nullstr);
8386 }
8387 else if (svix < svmax) {
8388 sv_catsv(sv, *svargs);
2d03de9c
AL
8389 }
8390 return;
0dbb1585 8391 }
8896765a
RB
8392 if (args && patlen == 3 && pat[0] == '%' &&
8393 pat[1] == '-' && pat[2] == 'p') {
8394 argsv = va_arg(*args, SV*);
8395 sv_catsv(sv, argsv);
8896765a 8396 return;
46fc3d4c 8397 }
8398
1d917b39 8399#ifndef USE_LONG_DOUBLE
4151a5fe 8400 /* special-case "%.<number>[gf]" */
7af36d83 8401 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8402 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8403 unsigned digits = 0;
8404 const char *pp;
8405
8406 pp = pat + 2;
8407 while (*pp >= '0' && *pp <= '9')
8408 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8409 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8410 NV nv;
8411
7af36d83 8412 if (svix < svmax)
4151a5fe
IZ
8413 nv = SvNV(*svargs);
8414 else
8415 return;
8416 if (*pp == 'g') {
2873255c
NC
8417 /* Add check for digits != 0 because it seems that some
8418 gconverts are buggy in this case, and we don't yet have
8419 a Configure test for this. */
8420 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8421 /* 0, point, slack */
2e59c212 8422 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8423 sv_catpv(sv, ebuf);
8424 if (*ebuf) /* May return an empty string for digits==0 */
8425 return;
8426 }
8427 } else if (!digits) {
8428 STRLEN l;
8429
8430 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8431 sv_catpvn(sv, p, l);
8432 return;
8433 }
8434 }
8435 }
8436 }
1d917b39 8437#endif /* !USE_LONG_DOUBLE */
4151a5fe 8438
2cf2cfc6 8439 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8440 has_utf8 = TRUE;
2cf2cfc6 8441
46fc3d4c 8442 patend = (char*)pat + patlen;
8443 for (p = (char*)pat; p < patend; p = q) {
8444 bool alt = FALSE;
8445 bool left = FALSE;
b22c7a20 8446 bool vectorize = FALSE;
211dfcf1 8447 bool vectorarg = FALSE;
2cf2cfc6 8448 bool vec_utf8 = FALSE;
46fc3d4c 8449 char fill = ' ';
8450 char plus = 0;
8451 char intsize = 0;
8452 STRLEN width = 0;
fc36a67e 8453 STRLEN zeros = 0;
46fc3d4c 8454 bool has_precis = FALSE;
8455 STRLEN precis = 0;
c445ea15 8456 const I32 osvix = svix;
2cf2cfc6 8457 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8458#ifdef HAS_LDBL_SPRINTF_BUG
8459 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8460 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8461 bool fix_ldbl_sprintf_bug = FALSE;
8462#endif
205f51d8 8463
46fc3d4c 8464 char esignbuf[4];
89ebb4a3 8465 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8466 STRLEN esignlen = 0;
8467
bd61b366 8468 const char *eptr = NULL;
fc36a67e 8469 STRLEN elen = 0;
a0714e2c 8470 SV *vecsv = NULL;
4608196e 8471 const U8 *vecstr = NULL;
b22c7a20 8472 STRLEN veclen = 0;
934abaf1 8473 char c = 0;
46fc3d4c 8474 int i;
9c5ffd7c 8475 unsigned base = 0;
8c8eb53c
RB
8476 IV iv = 0;
8477 UV uv = 0;
9e5b023a
JH
8478 /* we need a long double target in case HAS_LONG_DOUBLE but
8479 not USE_LONG_DOUBLE
8480 */
35fff930 8481#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8482 long double nv;
8483#else
65202027 8484 NV nv;
9e5b023a 8485#endif
46fc3d4c 8486 STRLEN have;
8487 STRLEN need;
8488 STRLEN gap;
7af36d83 8489 const char *dotstr = ".";
b22c7a20 8490 STRLEN dotstrlen = 1;
211dfcf1 8491 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8492 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8493 I32 epix = 0; /* explicit precision index */
8494 I32 evix = 0; /* explicit vector index */
eb3fce90 8495 bool asterisk = FALSE;
46fc3d4c 8496
211dfcf1 8497 /* echo everything up to the next format specification */
46fc3d4c 8498 for (q = p; q < patend && *q != '%'; ++q) ;
8499 if (q > p) {
db79b45b
JH
8500 if (has_utf8 && !pat_utf8)
8501 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8502 else
8503 sv_catpvn(sv, p, q - p);
46fc3d4c 8504 p = q;
8505 }
8506 if (q++ >= patend)
8507 break;
8508
211dfcf1
HS
8509/*
8510 We allow format specification elements in this order:
8511 \d+\$ explicit format parameter index
8512 [-+ 0#]+ flags
a472f209 8513 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8514 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8515 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8516 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8517 [hlqLV] size
8896765a
RB
8518 [%bcdefginopsuxDFOUX] format (mandatory)
8519*/
8520
8521 if (args) {
8522/*
8523 As of perl5.9.3, printf format checking is on by default.
8524 Internally, perl uses %p formats to provide an escape to
8525 some extended formatting. This block deals with those
8526 extensions: if it does not match, (char*)q is reset and
8527 the normal format processing code is used.
8528
8529 Currently defined extensions are:
8530 %p include pointer address (standard)
8531 %-p (SVf) include an SV (previously %_)
8532 %-<num>p include an SV with precision <num>
8533 %1p (VDf) include a v-string (as %vd)
8534 %<num>p reserved for future extensions
8535
8536 Robin Barker 2005-07-14
211dfcf1 8537*/
8896765a
RB
8538 char* r = q;
8539 bool sv = FALSE;
8540 STRLEN n = 0;
8541 if (*q == '-')
8542 sv = *q++;
c445ea15 8543 n = expect_number(&q);
8896765a
RB
8544 if (*q++ == 'p') {
8545 if (sv) { /* SVf */
8546 if (n) {
8547 precis = n;
8548 has_precis = TRUE;
8549 }
8550 argsv = va_arg(*args, SV*);
8551 eptr = SvPVx_const(argsv, elen);
8552 if (DO_UTF8(argsv))
8553 is_utf8 = TRUE;
8554 goto string;
8555 }
8556#if vdNUMBER
8557 else if (n == vdNUMBER) { /* VDf */
8558 vectorize = TRUE;
8559 VECTORIZE_ARGS
8560 goto format_vd;
8561 }
8562#endif
8563 else if (n) {
8564 if (ckWARN_d(WARN_INTERNAL))
8565 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8566 "internal %%<num>p might conflict with future printf extensions");
8567 }
8568 }
8569 q = r;
8570 }
8571
c445ea15 8572 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8573 if (*q == '$') {
8574 ++q;
8575 efix = width;
8576 } else {
8577 goto gotwidth;
8578 }
8579 }
8580
fc36a67e 8581 /* FLAGS */
8582
46fc3d4c 8583 while (*q) {
8584 switch (*q) {
8585 case ' ':
8586 case '+':
8587 plus = *q++;
8588 continue;
8589
8590 case '-':
8591 left = TRUE;
8592 q++;
8593 continue;
8594
8595 case '0':
8596 fill = *q++;
8597 continue;
8598
8599 case '#':
8600 alt = TRUE;
8601 q++;
8602 continue;
8603
fc36a67e 8604 default:
8605 break;
8606 }
8607 break;
8608 }
46fc3d4c 8609
211dfcf1 8610 tryasterisk:
eb3fce90 8611 if (*q == '*') {
211dfcf1 8612 q++;
c445ea15 8613 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8614 if (*q++ != '$')
8615 goto unknown;
eb3fce90 8616 asterisk = TRUE;
211dfcf1
HS
8617 }
8618 if (*q == 'v') {
eb3fce90 8619 q++;
211dfcf1
HS
8620 if (vectorize)
8621 goto unknown;
9cbac4c7 8622 if ((vectorarg = asterisk)) {
211dfcf1
HS
8623 evix = ewix;
8624 ewix = 0;
8625 asterisk = FALSE;
8626 }
8627 vectorize = TRUE;
8628 goto tryasterisk;
eb3fce90
JH
8629 }
8630
211dfcf1 8631 if (!asterisk)
858a90f9 8632 {
7a5fa8a2 8633 if( *q == '0' )
f3583277 8634 fill = *q++;
c445ea15 8635 width = expect_number(&q);
858a90f9 8636 }
211dfcf1
HS
8637
8638 if (vectorize) {
8639 if (vectorarg) {
8640 if (args)
8641 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8642 else if (evix) {
8643 vecsv = (evix > 0 && evix <= svmax)
8644 ? svargs[evix-1] : &PL_sv_undef;
8645 } else {
8646 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8647 }
245d4a47 8648 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8649 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8650 bad with tied or overloaded values that return UTF8. */
211dfcf1 8651 if (DO_UTF8(vecsv))
2cf2cfc6 8652 is_utf8 = TRUE;
640283f5
NC
8653 else if (has_utf8) {
8654 vecsv = sv_mortalcopy(vecsv);
8655 sv_utf8_upgrade(vecsv);
8656 dotstr = SvPV_const(vecsv, dotstrlen);
8657 is_utf8 = TRUE;
8658 }
211dfcf1
HS
8659 }
8660 if (args) {
8896765a 8661 VECTORIZE_ARGS
eb3fce90 8662 }
7ad96abb 8663 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8664 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8665 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8666 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8667
8668 /* if this is a version object, we need to convert
8669 * back into v-string notation and then let the
8670 * vectorize happen normally
d7aa5382 8671 */
96b8f7ce
JP
8672 if (sv_derived_from(vecsv, "version")) {
8673 char *version = savesvpv(vecsv);
34ba6322
SP
8674 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8675 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8676 "vector argument not supported with alpha versions");
8677 goto unknown;
8678 }
96b8f7ce
JP
8679 vecsv = sv_newmortal();
8680 /* scan_vstring is expected to be called during
8681 * tokenization, so we need to fake up the end
8682 * of the buffer for it
8683 */
8684 PL_bufend = version + veclen;
8685 scan_vstring(version, vecsv);
8686 vecstr = (U8*)SvPV_const(vecsv, veclen);
8687 vec_utf8 = DO_UTF8(vecsv);
8688 Safefree(version);
d7aa5382 8689 }
211dfcf1
HS
8690 }
8691 else {
8692 vecstr = (U8*)"";
8693 veclen = 0;
8694 }
eb3fce90 8695 }
fc36a67e 8696
eb3fce90 8697 if (asterisk) {
fc36a67e 8698 if (args)
8699 i = va_arg(*args, int);
8700 else
eb3fce90
JH
8701 i = (ewix ? ewix <= svmax : svix < svmax) ?
8702 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8703 left |= (i < 0);
8704 width = (i < 0) ? -i : i;
fc36a67e 8705 }
211dfcf1 8706 gotwidth:
fc36a67e 8707
8708 /* PRECISION */
46fc3d4c 8709
fc36a67e 8710 if (*q == '.') {
8711 q++;
8712 if (*q == '*') {
211dfcf1 8713 q++;
c445ea15 8714 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8715 goto unknown;
8716 /* XXX: todo, support specified precision parameter */
8717 if (epix)
211dfcf1 8718 goto unknown;
46fc3d4c 8719 if (args)
8720 i = va_arg(*args, int);
8721 else
eb3fce90
JH
8722 i = (ewix ? ewix <= svmax : svix < svmax)
8723 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8724 precis = (i < 0) ? 0 : i;
fc36a67e 8725 }
8726 else {
8727 precis = 0;
8728 while (isDIGIT(*q))
8729 precis = precis * 10 + (*q++ - '0');
8730 }
8731 has_precis = TRUE;
8732 }
46fc3d4c 8733
fc36a67e 8734 /* SIZE */
46fc3d4c 8735
fc36a67e 8736 switch (*q) {
c623ac67
GS
8737#ifdef WIN32
8738 case 'I': /* Ix, I32x, and I64x */
8739# ifdef WIN64
8740 if (q[1] == '6' && q[2] == '4') {
8741 q += 3;
8742 intsize = 'q';
8743 break;
8744 }
8745# endif
8746 if (q[1] == '3' && q[2] == '2') {
8747 q += 3;
8748 break;
8749 }
8750# ifdef WIN64
8751 intsize = 'q';
8752# endif
8753 q++;
8754 break;
8755#endif
9e5b023a 8756#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8757 case 'L': /* Ld */
5f66b61c 8758 /*FALLTHROUGH*/
e5c81feb 8759#ifdef HAS_QUAD
6f9bb7fd 8760 case 'q': /* qd */
9e5b023a 8761#endif
6f9bb7fd
GS
8762 intsize = 'q';
8763 q++;
8764 break;
8765#endif
fc36a67e 8766 case 'l':
9e5b023a 8767#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8768 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8769 intsize = 'q';
8770 q += 2;
46fc3d4c 8771 break;
cf2093f6 8772 }
fc36a67e 8773#endif
5f66b61c 8774 /*FALLTHROUGH*/
fc36a67e 8775 case 'h':
5f66b61c 8776 /*FALLTHROUGH*/
fc36a67e 8777 case 'V':
8778 intsize = *q++;
46fc3d4c 8779 break;
8780 }
8781
fc36a67e 8782 /* CONVERSION */
8783
211dfcf1
HS
8784 if (*q == '%') {
8785 eptr = q++;
8786 elen = 1;
26372e71
GA
8787 if (vectorize) {
8788 c = '%';
8789 goto unknown;
8790 }
211dfcf1
HS
8791 goto string;
8792 }
8793
26372e71 8794 if (!vectorize && !args) {
86c51f8b
NC
8795 if (efix) {
8796 const I32 i = efix-1;
8797 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8798 } else {
8799 argsv = (svix >= 0 && svix < svmax)
8800 ? svargs[svix++] : &PL_sv_undef;
8801 }
863811b2 8802 }
211dfcf1 8803
46fc3d4c 8804 switch (c = *q++) {
8805
8806 /* STRINGS */
8807
46fc3d4c 8808 case 'c':
26372e71
GA
8809 if (vectorize)
8810 goto unknown;
8811 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8812 if ((uv > 255 ||
8813 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8814 && !IN_BYTES) {
dfe13c55 8815 eptr = (char*)utf8buf;
9041c2e3 8816 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8817 is_utf8 = TRUE;
7e2040f0
GS
8818 }
8819 else {
8820 c = (char)uv;
8821 eptr = &c;
8822 elen = 1;
a0ed51b3 8823 }
46fc3d4c 8824 goto string;
8825
46fc3d4c 8826 case 's':
26372e71
GA
8827 if (vectorize)
8828 goto unknown;
8829 if (args) {
fc36a67e 8830 eptr = va_arg(*args, char*);
c635e13b 8831 if (eptr)
1d7c1841
GS
8832#ifdef MACOS_TRADITIONAL
8833 /* On MacOS, %#s format is used for Pascal strings */
8834 if (alt)
8835 elen = *eptr++;
8836 else
8837#endif
c635e13b 8838 elen = strlen(eptr);
8839 else {
27da23d5 8840 eptr = (char *)nullstr;
c635e13b 8841 elen = sizeof nullstr - 1;
8842 }
46fc3d4c 8843 }
211dfcf1 8844 else {
4d84ee25 8845 eptr = SvPVx_const(argsv, elen);
7e2040f0 8846 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8847 if (has_precis && precis < elen) {
8848 I32 p = precis;
7e2040f0 8849 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8850 precis = p;
8851 }
8852 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8853 width += elen - sv_len_utf8(argsv);
a0ed51b3 8854 }
2cf2cfc6 8855 is_utf8 = TRUE;
a0ed51b3
LW
8856 }
8857 }
fc36a67e 8858
46fc3d4c 8859 string:
8860 if (has_precis && elen > precis)
8861 elen = precis;
8862 break;
8863
8864 /* INTEGERS */
8865
fc36a67e 8866 case 'p':
be75b157 8867 if (alt || vectorize)
c2e66d9e 8868 goto unknown;
211dfcf1 8869 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8870 base = 16;
8871 goto integer;
8872
46fc3d4c 8873 case 'D':
29fe7a80 8874#ifdef IV_IS_QUAD
22f3ae8c 8875 intsize = 'q';
29fe7a80 8876#else
46fc3d4c 8877 intsize = 'l';
29fe7a80 8878#endif
5f66b61c 8879 /*FALLTHROUGH*/
46fc3d4c 8880 case 'd':
8881 case 'i':
8896765a
RB
8882#if vdNUMBER
8883 format_vd:
8884#endif
b22c7a20 8885 if (vectorize) {
ba210ebe 8886 STRLEN ulen;
211dfcf1
HS
8887 if (!veclen)
8888 continue;
2cf2cfc6
A
8889 if (vec_utf8)
8890 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8891 UTF8_ALLOW_ANYUV);
b22c7a20 8892 else {
e83d50c9 8893 uv = *vecstr;
b22c7a20
GS
8894 ulen = 1;
8895 }
8896 vecstr += ulen;
8897 veclen -= ulen;
e83d50c9
JP
8898 if (plus)
8899 esignbuf[esignlen++] = plus;
b22c7a20
GS
8900 }
8901 else if (args) {
46fc3d4c 8902 switch (intsize) {
8903 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8904 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8905 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8906 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8907#ifdef HAS_QUAD
8908 case 'q': iv = va_arg(*args, Quad_t); break;
8909#endif
46fc3d4c 8910 }
8911 }
8912 else {
b10c0dba 8913 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8914 switch (intsize) {
b10c0dba
MHM
8915 case 'h': iv = (short)tiv; break;
8916 case 'l': iv = (long)tiv; break;
8917 case 'V':
8918 default: iv = tiv; break;
cf2093f6 8919#ifdef HAS_QUAD
b10c0dba 8920 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8921#endif
46fc3d4c 8922 }
8923 }
e83d50c9
JP
8924 if ( !vectorize ) /* we already set uv above */
8925 {
8926 if (iv >= 0) {
8927 uv = iv;
8928 if (plus)
8929 esignbuf[esignlen++] = plus;
8930 }
8931 else {
8932 uv = -iv;
8933 esignbuf[esignlen++] = '-';
8934 }
46fc3d4c 8935 }
8936 base = 10;
8937 goto integer;
8938
fc36a67e 8939 case 'U':
29fe7a80 8940#ifdef IV_IS_QUAD
22f3ae8c 8941 intsize = 'q';
29fe7a80 8942#else
fc36a67e 8943 intsize = 'l';
29fe7a80 8944#endif
5f66b61c 8945 /*FALLTHROUGH*/
fc36a67e 8946 case 'u':
8947 base = 10;
8948 goto uns_integer;
8949
4f19785b
WSI
8950 case 'b':
8951 base = 2;
8952 goto uns_integer;
8953
46fc3d4c 8954 case 'O':
29fe7a80 8955#ifdef IV_IS_QUAD
22f3ae8c 8956 intsize = 'q';
29fe7a80 8957#else
46fc3d4c 8958 intsize = 'l';
29fe7a80 8959#endif
5f66b61c 8960 /*FALLTHROUGH*/
46fc3d4c 8961 case 'o':
8962 base = 8;
8963 goto uns_integer;
8964
8965 case 'X':
46fc3d4c 8966 case 'x':
8967 base = 16;
46fc3d4c 8968
8969 uns_integer:
b22c7a20 8970 if (vectorize) {
ba210ebe 8971 STRLEN ulen;
b22c7a20 8972 vector:
211dfcf1
HS
8973 if (!veclen)
8974 continue;
2cf2cfc6
A
8975 if (vec_utf8)
8976 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8977 UTF8_ALLOW_ANYUV);
b22c7a20 8978 else {
a05b299f 8979 uv = *vecstr;
b22c7a20
GS
8980 ulen = 1;
8981 }
8982 vecstr += ulen;
8983 veclen -= ulen;
8984 }
8985 else if (args) {
46fc3d4c 8986 switch (intsize) {
8987 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8988 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8989 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8990 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8991#ifdef HAS_QUAD
9e3321a5 8992 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8993#endif
46fc3d4c 8994 }
8995 }
8996 else {
b10c0dba 8997 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8998 switch (intsize) {
b10c0dba
MHM
8999 case 'h': uv = (unsigned short)tuv; break;
9000 case 'l': uv = (unsigned long)tuv; break;
9001 case 'V':
9002 default: uv = tuv; break;
cf2093f6 9003#ifdef HAS_QUAD
b10c0dba 9004 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 9005#endif
46fc3d4c 9006 }
9007 }
9008
9009 integer:
4d84ee25
NC
9010 {
9011 char *ptr = ebuf + sizeof ebuf;
9012 switch (base) {
9013 unsigned dig;
9014 case 16:
9015 if (!uv)
9016 alt = FALSE;
9017 p = (char*)((c == 'X')
9018 ? "0123456789ABCDEF" : "0123456789abcdef");
9019 do {
9020 dig = uv & 15;
9021 *--ptr = p[dig];
9022 } while (uv >>= 4);
9023 if (alt) {
9024 esignbuf[esignlen++] = '0';
9025 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9026 }
9027 break;
9028 case 8:
9029 do {
9030 dig = uv & 7;
9031 *--ptr = '0' + dig;
9032 } while (uv >>= 3);
9033 if (alt && *ptr != '0')
9034 *--ptr = '0';
9035 break;
9036 case 2:
ed2b91d2
GA
9037 if (!uv)
9038 alt = FALSE;
4d84ee25
NC
9039 do {
9040 dig = uv & 1;
9041 *--ptr = '0' + dig;
9042 } while (uv >>= 1);
9043 if (alt) {
9044 esignbuf[esignlen++] = '0';
9045 esignbuf[esignlen++] = 'b';
9046 }
9047 break;
9048 default: /* it had better be ten or less */
9049 do {
9050 dig = uv % base;
9051 *--ptr = '0' + dig;
9052 } while (uv /= base);
9053 break;
46fc3d4c 9054 }
4d84ee25
NC
9055 elen = (ebuf + sizeof ebuf) - ptr;
9056 eptr = ptr;
9057 if (has_precis) {
9058 if (precis > elen)
9059 zeros = precis - elen;
9060 else if (precis == 0 && elen == 1 && *eptr == '0')
9061 elen = 0;
eda88b6d 9062 }
c10ed8b9 9063 }
46fc3d4c 9064 break;
9065
9066 /* FLOATING POINT */
9067
fc36a67e 9068 case 'F':
9069 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9070 /*FALLTHROUGH*/
46fc3d4c 9071 case 'e': case 'E':
fc36a67e 9072 case 'f':
46fc3d4c 9073 case 'g': case 'G':
26372e71
GA
9074 if (vectorize)
9075 goto unknown;
46fc3d4c 9076
9077 /* This is evil, but floating point is even more evil */
9078
9e5b023a
JH
9079 /* for SV-style calling, we can only get NV
9080 for C-style calling, we assume %f is double;
9081 for simplicity we allow any of %Lf, %llf, %qf for long double
9082 */
9083 switch (intsize) {
9084 case 'V':
9085#if defined(USE_LONG_DOUBLE)
9086 intsize = 'q';
9087#endif
9088 break;
8a2e3f14 9089/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9090 case 'l':
5f66b61c 9091 /*FALLTHROUGH*/
9e5b023a
JH
9092 default:
9093#if defined(USE_LONG_DOUBLE)
9094 intsize = args ? 0 : 'q';
9095#endif
9096 break;
9097 case 'q':
9098#if defined(HAS_LONG_DOUBLE)
9099 break;
9100#else
5f66b61c 9101 /*FALLTHROUGH*/
9e5b023a
JH
9102#endif
9103 case 'h':
9e5b023a
JH
9104 goto unknown;
9105 }
9106
9107 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9108 nv = (args) ?
35fff930
JH
9109#if LONG_DOUBLESIZE > DOUBLESIZE
9110 intsize == 'q' ?
205f51d8
AS
9111 va_arg(*args, long double) :
9112 va_arg(*args, double)
35fff930 9113#else
205f51d8 9114 va_arg(*args, double)
35fff930 9115#endif
9e5b023a 9116 : SvNVx(argsv);
fc36a67e 9117
9118 need = 0;
9119 if (c != 'e' && c != 'E') {
9120 i = PERL_INT_MIN;
9e5b023a
JH
9121 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9122 will cast our (long double) to (double) */
73b309ea 9123 (void)Perl_frexp(nv, &i);
fc36a67e 9124 if (i == PERL_INT_MIN)
cea2e8a9 9125 Perl_die(aTHX_ "panic: frexp");
c635e13b 9126 if (i > 0)
fc36a67e 9127 need = BIT_DIGITS(i);
9128 }
9129 need += has_precis ? precis : 6; /* known default */
20f6aaab 9130
fc36a67e 9131 if (need < width)
9132 need = width;
9133
20f6aaab
AS
9134#ifdef HAS_LDBL_SPRINTF_BUG
9135 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9136 with sfio - Allen <allens@cpan.org> */
9137
9138# ifdef DBL_MAX
9139# define MY_DBL_MAX DBL_MAX
9140# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9141# if DOUBLESIZE >= 8
9142# define MY_DBL_MAX 1.7976931348623157E+308L
9143# else
9144# define MY_DBL_MAX 3.40282347E+38L
9145# endif
9146# endif
9147
9148# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9149# define MY_DBL_MAX_BUG 1L
20f6aaab 9150# else
205f51d8 9151# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9152# endif
20f6aaab 9153
205f51d8
AS
9154# ifdef DBL_MIN
9155# define MY_DBL_MIN DBL_MIN
9156# else /* XXX guessing! -Allen */
9157# if DOUBLESIZE >= 8
9158# define MY_DBL_MIN 2.2250738585072014E-308L
9159# else
9160# define MY_DBL_MIN 1.17549435E-38L
9161# endif
9162# endif
20f6aaab 9163
205f51d8
AS
9164 if ((intsize == 'q') && (c == 'f') &&
9165 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9166 (need < DBL_DIG)) {
9167 /* it's going to be short enough that
9168 * long double precision is not needed */
9169
9170 if ((nv <= 0L) && (nv >= -0L))
9171 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9172 else {
9173 /* would use Perl_fp_class as a double-check but not
9174 * functional on IRIX - see perl.h comments */
9175
9176 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9177 /* It's within the range that a double can represent */
9178#if defined(DBL_MAX) && !defined(DBL_MIN)
9179 if ((nv >= ((long double)1/DBL_MAX)) ||
9180 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9181#endif
205f51d8 9182 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9183 }
205f51d8
AS
9184 }
9185 if (fix_ldbl_sprintf_bug == TRUE) {
9186 double temp;
9187
9188 intsize = 0;
9189 temp = (double)nv;
9190 nv = (NV)temp;
9191 }
20f6aaab 9192 }
205f51d8
AS
9193
9194# undef MY_DBL_MAX
9195# undef MY_DBL_MAX_BUG
9196# undef MY_DBL_MIN
9197
20f6aaab
AS
9198#endif /* HAS_LDBL_SPRINTF_BUG */
9199
46fc3d4c 9200 need += 20; /* fudge factor */
80252599
GS
9201 if (PL_efloatsize < need) {
9202 Safefree(PL_efloatbuf);
9203 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9204 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9205 PL_efloatbuf[0] = '\0';
46fc3d4c 9206 }
9207
4151a5fe
IZ
9208 if ( !(width || left || plus || alt) && fill != '0'
9209 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9210 /* See earlier comment about buggy Gconvert when digits,
9211 aka precis is 0 */
9212 if ( c == 'g' && precis) {
2e59c212 9213 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9214 /* May return an empty string for digits==0 */
9215 if (*PL_efloatbuf) {
9216 elen = strlen(PL_efloatbuf);
4151a5fe 9217 goto float_converted;
4150c189 9218 }
4151a5fe
IZ
9219 } else if ( c == 'f' && !precis) {
9220 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9221 break;
9222 }
9223 }
4d84ee25
NC
9224 {
9225 char *ptr = ebuf + sizeof ebuf;
9226 *--ptr = '\0';
9227 *--ptr = c;
9228 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9229#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9230 if (intsize == 'q') {
9231 /* Copy the one or more characters in a long double
9232 * format before the 'base' ([efgEFG]) character to
9233 * the format string. */
9234 static char const prifldbl[] = PERL_PRIfldbl;
9235 char const *p = prifldbl + sizeof(prifldbl) - 3;
9236 while (p >= prifldbl) { *--ptr = *p--; }
9237 }
65202027 9238#endif
4d84ee25
NC
9239 if (has_precis) {
9240 base = precis;
9241 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9242 *--ptr = '.';
9243 }
9244 if (width) {
9245 base = width;
9246 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9247 }
9248 if (fill == '0')
9249 *--ptr = fill;
9250 if (left)
9251 *--ptr = '-';
9252 if (plus)
9253 *--ptr = plus;
9254 if (alt)
9255 *--ptr = '#';
9256 *--ptr = '%';
9257
9258 /* No taint. Otherwise we are in the strange situation
9259 * where printf() taints but print($float) doesn't.
9260 * --jhi */
9e5b023a 9261#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9262 elen = ((intsize == 'q')
9263 ? my_sprintf(PL_efloatbuf, ptr, nv)
9264 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9265#else
4150c189 9266 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9267#endif
4d84ee25 9268 }
4151a5fe 9269 float_converted:
80252599 9270 eptr = PL_efloatbuf;
46fc3d4c 9271 break;
9272
fc36a67e 9273 /* SPECIAL */
9274
9275 case 'n':
26372e71
GA
9276 if (vectorize)
9277 goto unknown;
fc36a67e 9278 i = SvCUR(sv) - origlen;
26372e71 9279 if (args) {
c635e13b 9280 switch (intsize) {
9281 case 'h': *(va_arg(*args, short*)) = i; break;
9282 default: *(va_arg(*args, int*)) = i; break;
9283 case 'l': *(va_arg(*args, long*)) = i; break;
9284 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9285#ifdef HAS_QUAD
9286 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9287#endif
c635e13b 9288 }
fc36a67e 9289 }
9dd79c3f 9290 else
211dfcf1 9291 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9292 continue; /* not "break" */
9293
9294 /* UNKNOWN */
9295
46fc3d4c 9296 default:
fc36a67e 9297 unknown:
041457d9
DM
9298 if (!args
9299 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9300 && ckWARN(WARN_PRINTF))
9301 {
c4420975 9302 SV * const msg = sv_newmortal();
35c1215d
NC
9303 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9304 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9305 if (c) {
0f4b6630 9306 if (isPRINT(c))
1c846c1f 9307 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9308 "\"%%%c\"", c & 0xFF);
9309 else
9310 Perl_sv_catpvf(aTHX_ msg,
57def98f 9311 "\"%%\\%03"UVof"\"",
0f4b6630 9312 (UV)c & 0xFF);
0f4b6630 9313 } else
396482e1 9314 sv_catpvs(msg, "end of string");
9014280d 9315 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9316 }
fb73857a 9317
9318 /* output mangled stuff ... */
9319 if (c == '\0')
9320 --q;
46fc3d4c 9321 eptr = p;
9322 elen = q - p;
fb73857a 9323
9324 /* ... right here, because formatting flags should not apply */
9325 SvGROW(sv, SvCUR(sv) + elen + 1);
9326 p = SvEND(sv);
4459522c 9327 Copy(eptr, p, elen, char);
fb73857a 9328 p += elen;
9329 *p = '\0';
3f7c398e 9330 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9331 svix = osvix;
fb73857a 9332 continue; /* not "break" */
46fc3d4c 9333 }
9334
6c94ec8b
HS
9335 /* calculate width before utf8_upgrade changes it */
9336 have = esignlen + zeros + elen;
ed2b91d2
GA
9337 if (have < zeros)
9338 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9339
d2876be5
JH
9340 if (is_utf8 != has_utf8) {
9341 if (is_utf8) {
9342 if (SvCUR(sv))
9343 sv_utf8_upgrade(sv);
9344 }
9345 else {
53c1dcc0 9346 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9347 sv_utf8_upgrade(nsv);
93524f2b 9348 eptr = SvPVX_const(nsv);
d2876be5
JH
9349 elen = SvCUR(nsv);
9350 }
9351 SvGROW(sv, SvCUR(sv) + elen + 1);
9352 p = SvEND(sv);
9353 *p = '\0';
9354 }
6af65485 9355
46fc3d4c 9356 need = (have > width ? have : width);
9357 gap = need - have;
9358
d2641cbd
PC
9359 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9360 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9361 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9362 p = SvEND(sv);
9363 if (esignlen && fill == '0') {
53c1dcc0 9364 int i;
eb160463 9365 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9366 *p++ = esignbuf[i];
9367 }
9368 if (gap && !left) {
9369 memset(p, fill, gap);
9370 p += gap;
9371 }
9372 if (esignlen && fill != '0') {
53c1dcc0 9373 int i;
eb160463 9374 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9375 *p++ = esignbuf[i];
9376 }
fc36a67e 9377 if (zeros) {
53c1dcc0 9378 int i;
fc36a67e 9379 for (i = zeros; i; i--)
9380 *p++ = '0';
9381 }
46fc3d4c 9382 if (elen) {
4459522c 9383 Copy(eptr, p, elen, char);
46fc3d4c 9384 p += elen;
9385 }
9386 if (gap && left) {
9387 memset(p, ' ', gap);
9388 p += gap;
9389 }
b22c7a20
GS
9390 if (vectorize) {
9391 if (veclen) {
4459522c 9392 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9393 p += dotstrlen;
9394 }
9395 else
9396 vectorize = FALSE; /* done iterating over vecstr */
9397 }
2cf2cfc6
A
9398 if (is_utf8)
9399 has_utf8 = TRUE;
9400 if (has_utf8)
7e2040f0 9401 SvUTF8_on(sv);
46fc3d4c 9402 *p = '\0';
3f7c398e 9403 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9404 if (vectorize) {
9405 esignlen = 0;
9406 goto vector;
9407 }
46fc3d4c 9408 }
9409}
51371543 9410
645c22ef
DM
9411/* =========================================================================
9412
9413=head1 Cloning an interpreter
9414
9415All the macros and functions in this section are for the private use of
9416the main function, perl_clone().
9417
9418The foo_dup() functions make an exact copy of an existing foo thinngy.
9419During the course of a cloning, a hash table is used to map old addresses
9420to new addresses. The table is created and manipulated with the
9421ptr_table_* functions.
9422
9423=cut
9424
9425============================================================================*/
9426
9427
1d7c1841
GS
9428#if defined(USE_ITHREADS)
9429
d4c19fe8 9430/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9431#ifndef GpREFCNT_inc
9432# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9433#endif
9434
9435
a41cc44e
NC
9436/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
9437 that currently av_dup and hv_dup are the same as sv_dup. If this changes,
9438 please unmerge ss_dup. */
d2d73c3e 9439#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9440#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9441#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9442#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9443#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9444#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9445#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9446#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9447#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9448#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9449#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9450#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9451#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9452#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9453
d2d73c3e 9454
d2f185dc
AMS
9455/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9456 regcomp.c. AMS 20010712 */
645c22ef 9457
1d7c1841 9458REGEXP *
53c1dcc0 9459Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9460{
27da23d5 9461 dVAR;
d2f185dc
AMS
9462 REGEXP *ret;
9463 int i, len, npar;
9464 struct reg_substr_datum *s;
9465
9466 if (!r)
9467 return (REGEXP *)NULL;
9468
9469 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9470 return ret;
9471
9472 len = r->offsets[0];
9473 npar = r->nparens+1;
9474
a02a5408 9475 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9476 Copy(r->program, ret->program, len+1, regnode);
9477
a02a5408 9478 Newx(ret->startp, npar, I32);
d2f185dc 9479 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9480 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9481 Copy(r->startp, ret->startp, npar, I32);
9482
a02a5408 9483 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9484 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9485 s->min_offset = r->substrs->data[i].min_offset;
9486 s->max_offset = r->substrs->data[i].max_offset;
9487 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9488 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9489 }
9490
70612e96 9491 ret->regstclass = NULL;
d2f185dc
AMS
9492 if (r->data) {
9493 struct reg_data *d;
e1ec3a88 9494 const int count = r->data->count;
53c1dcc0 9495 int i;
d2f185dc 9496
a02a5408 9497 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9498 char, struct reg_data);
a02a5408 9499 Newx(d->what, count, U8);
d2f185dc
AMS
9500
9501 d->count = count;
9502 for (i = 0; i < count; i++) {
9503 d->what[i] = r->data->what[i];
9504 switch (d->what[i]) {
a3621e74
YO
9505 /* legal options are one of: sfpont
9506 see also regcomp.h and pregfree() */
d2f185dc
AMS
9507 case 's':
9508 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9509 break;
9510 case 'p':
9511 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9512 break;
9513 case 'f':
9514 /* This is cheating. */
a02a5408 9515 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9516 StructCopy(r->data->data[i], d->data[i],
9517 struct regnode_charclass_class);
70612e96 9518 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9519 break;
9520 case 'o':
33773810
AMS
9521 /* Compiled op trees are readonly, and can thus be
9522 shared without duplication. */
b34c0dd4 9523 OP_REFCNT_LOCK;
9b978d73 9524 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9525 OP_REFCNT_UNLOCK;
9b978d73 9526 break;
d2f185dc
AMS
9527 case 'n':
9528 d->data[i] = r->data->data[i];
9529 break;
a3621e74
YO
9530 case 't':
9531 d->data[i] = r->data->data[i];
9532 OP_REFCNT_LOCK;
9533 ((reg_trie_data*)d->data[i])->refcount++;
9534 OP_REFCNT_UNLOCK;
9535 break;
9536 default:
9537 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9538 }
9539 }
9540
9541 ret->data = d;
9542 }
9543 else
9544 ret->data = NULL;
9545
a02a5408 9546 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9547 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9548
e01c5899 9549 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9550 ret->refcnt = r->refcnt;
9551 ret->minlen = r->minlen;
9552 ret->prelen = r->prelen;
9553 ret->nparens = r->nparens;
9554 ret->lastparen = r->lastparen;
9555 ret->lastcloseparen = r->lastcloseparen;
9556 ret->reganch = r->reganch;
9557
70612e96
RG
9558 ret->sublen = r->sublen;
9559
9560 if (RX_MATCH_COPIED(ret))
e01c5899 9561 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9562 else
bd61b366 9563 ret->subbeg = NULL;
f8c7b90f 9564#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9565 ret->saved_copy = NULL;
9a26048b 9566#endif
70612e96 9567
d2f185dc
AMS
9568 ptr_table_store(PL_ptr_table, r, ret);
9569 return ret;
1d7c1841
GS
9570}
9571
d2d73c3e 9572/* duplicate a file handle */
645c22ef 9573
1d7c1841 9574PerlIO *
a8fc9800 9575Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9576{
9577 PerlIO *ret;
53c1dcc0
AL
9578
9579 PERL_UNUSED_ARG(type);
73d840c0 9580
1d7c1841
GS
9581 if (!fp)
9582 return (PerlIO*)NULL;
9583
9584 /* look for it in the table first */
9585 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9586 if (ret)
9587 return ret;
9588
9589 /* create anew and remember what it is */
ecdeb87c 9590 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9591 ptr_table_store(PL_ptr_table, fp, ret);
9592 return ret;
9593}
9594
645c22ef
DM
9595/* duplicate a directory handle */
9596
1d7c1841
GS
9597DIR *
9598Perl_dirp_dup(pTHX_ DIR *dp)
9599{
96a5add6 9600 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9601 if (!dp)
9602 return (DIR*)NULL;
9603 /* XXX TODO */
9604 return dp;
9605}
9606
ff276b08 9607/* duplicate a typeglob */
645c22ef 9608
1d7c1841 9609GP *
a8fc9800 9610Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9611{
9612 GP *ret;
b37c2d43 9613
1d7c1841
GS
9614 if (!gp)
9615 return (GP*)NULL;
9616 /* look for it in the table first */
9617 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9618 if (ret)
9619 return ret;
9620
9621 /* create anew and remember what it is */
a02a5408 9622 Newxz(ret, 1, GP);
1d7c1841
GS
9623 ptr_table_store(PL_ptr_table, gp, ret);
9624
9625 /* clone */
9626 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9627 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9628 ret->gp_io = io_dup_inc(gp->gp_io, param);
9629 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9630 ret->gp_av = av_dup_inc(gp->gp_av, param);
9631 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9632 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9633 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9634 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9635 ret->gp_line = gp->gp_line;
9636 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9637 return ret;
9638}
9639
645c22ef
DM
9640/* duplicate a chain of magic */
9641
1d7c1841 9642MAGIC *
a8fc9800 9643Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9644{
cb359b41
JH
9645 MAGIC *mgprev = (MAGIC*)NULL;
9646 MAGIC *mgret;
1d7c1841
GS
9647 if (!mg)
9648 return (MAGIC*)NULL;
9649 /* look for it in the table first */
9650 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9651 if (mgret)
9652 return mgret;
9653
9654 for (; mg; mg = mg->mg_moremagic) {
9655 MAGIC *nmg;
a02a5408 9656 Newxz(nmg, 1, MAGIC);
cb359b41 9657 if (mgprev)
1d7c1841 9658 mgprev->mg_moremagic = nmg;
cb359b41
JH
9659 else
9660 mgret = nmg;
1d7c1841
GS
9661 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9662 nmg->mg_private = mg->mg_private;
9663 nmg->mg_type = mg->mg_type;
9664 nmg->mg_flags = mg->mg_flags;
14befaf4 9665 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9666 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9667 }
05bd4103 9668 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9669 /* The backref AV has its reference count deliberately bumped by
9670 1. */
9671 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9672 }
8d2f4536
NC
9673 else if (mg->mg_type == PERL_MAGIC_symtab) {
9674 nmg->mg_obj = mg->mg_obj;
9675 }
1d7c1841
GS
9676 else {
9677 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9678 ? sv_dup_inc(mg->mg_obj, param)
9679 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9680 }
9681 nmg->mg_len = mg->mg_len;
9682 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9683 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9684 if (mg->mg_len > 0) {
1d7c1841 9685 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9686 if (mg->mg_type == PERL_MAGIC_overload_table &&
9687 AMT_AMAGIC((AMT*)mg->mg_ptr))
9688 {
c445ea15 9689 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9690 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9691 I32 i;
9692 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9693 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9694 }
9695 }
9696 }
9697 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9698 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9699 }
68795e93
NIS
9700 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9701 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9702 }
1d7c1841
GS
9703 mgprev = nmg;
9704 }
9705 return mgret;
9706}
9707
645c22ef
DM
9708/* create a new pointer-mapping table */
9709
1d7c1841
GS
9710PTR_TBL_t *
9711Perl_ptr_table_new(pTHX)
9712{
9713 PTR_TBL_t *tbl;
96a5add6
AL
9714 PERL_UNUSED_CONTEXT;
9715
a02a5408 9716 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9717 tbl->tbl_max = 511;
9718 tbl->tbl_items = 0;
a02a5408 9719 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9720 return tbl;
9721}
9722
7119fd33
NC
9723#define PTR_TABLE_HASH(ptr) \
9724 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9725
93e68bfb
JC
9726/*
9727 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9728 following define) and at call to new_body_inline made below in
9729 Perl_ptr_table_store()
9730 */
9731
9732#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9733
645c22ef
DM
9734/* map an existing pointer using a table */
9735
7bf61b54 9736STATIC PTR_TBL_ENT_t *
b0e6ae5b 9737S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9738 PTR_TBL_ENT_t *tblent;
4373e329 9739 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9740 assert(tbl);
9741 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9742 for (; tblent; tblent = tblent->next) {
9743 if (tblent->oldval == sv)
7bf61b54 9744 return tblent;
1d7c1841 9745 }
d4c19fe8 9746 return NULL;
7bf61b54
NC
9747}
9748
9749void *
9750Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9751{
b0e6ae5b 9752 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9753 PERL_UNUSED_CONTEXT;
d4c19fe8 9754 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9755}
9756
645c22ef
DM
9757/* add a new entry to a pointer-mapping table */
9758
1d7c1841 9759void
44f8325f 9760Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9761{
0c9fdfe0 9762 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9763 PERL_UNUSED_CONTEXT;
1d7c1841 9764
7bf61b54
NC
9765 if (tblent) {
9766 tblent->newval = newsv;
9767 } else {
9768 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9769
d2a0f284
JC
9770 new_body_inline(tblent, PTE_SVSLOT);
9771
7bf61b54
NC
9772 tblent->oldval = oldsv;
9773 tblent->newval = newsv;
9774 tblent->next = tbl->tbl_ary[entry];
9775 tbl->tbl_ary[entry] = tblent;
9776 tbl->tbl_items++;
9777 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9778 ptr_table_split(tbl);
1d7c1841 9779 }
1d7c1841
GS
9780}
9781
645c22ef
DM
9782/* double the hash bucket size of an existing ptr table */
9783
1d7c1841
GS
9784void
9785Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9786{
9787 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9788 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9789 UV newsize = oldsize * 2;
9790 UV i;
96a5add6 9791 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9792
9793 Renew(ary, newsize, PTR_TBL_ENT_t*);
9794 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9795 tbl->tbl_max = --newsize;
9796 tbl->tbl_ary = ary;
9797 for (i=0; i < oldsize; i++, ary++) {
9798 PTR_TBL_ENT_t **curentp, **entp, *ent;
9799 if (!*ary)
9800 continue;
9801 curentp = ary + oldsize;
9802 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9803 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9804 *entp = ent->next;
9805 ent->next = *curentp;
9806 *curentp = ent;
9807 continue;
9808 }
9809 else
9810 entp = &ent->next;
9811 }
9812 }
9813}
9814
645c22ef
DM
9815/* remove all the entries from a ptr table */
9816
a0739874
DM
9817void
9818Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9819{
d5cefff9 9820 if (tbl && tbl->tbl_items) {
c445ea15 9821 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9822 UV riter = tbl->tbl_max;
a0739874 9823
d5cefff9
NC
9824 do {
9825 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9826
d5cefff9 9827 while (entry) {
00b6aa41 9828 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9829 entry = entry->next;
9830 del_pte(oentry);
9831 }
9832 } while (riter--);
a0739874 9833
d5cefff9
NC
9834 tbl->tbl_items = 0;
9835 }
a0739874
DM
9836}
9837
645c22ef
DM
9838/* clear and free a ptr table */
9839
a0739874
DM
9840void
9841Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9842{
9843 if (!tbl) {
9844 return;
9845 }
9846 ptr_table_clear(tbl);
9847 Safefree(tbl->tbl_ary);
9848 Safefree(tbl);
9849}
9850
5bd07a3d 9851
83841fad 9852void
eb86f8b3 9853Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9854{
9855 if (SvROK(sstr)) {
b162af07
SP
9856 SvRV_set(dstr, SvWEAKREF(sstr)
9857 ? sv_dup(SvRV(sstr), param)
9858 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9859
83841fad 9860 }
3f7c398e 9861 else if (SvPVX_const(sstr)) {
83841fad
NIS
9862 /* Has something there */
9863 if (SvLEN(sstr)) {
68795e93 9864 /* Normal PV - clone whole allocated space */
3f7c398e 9865 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9866 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9867 /* Not that normal - actually sstr is copy on write.
9868 But we are a true, independant SV, so: */
9869 SvREADONLY_off(dstr);
9870 SvFAKE_off(dstr);
9871 }
68795e93 9872 }
83841fad
NIS
9873 else {
9874 /* Special case - not normally malloced for some reason */
f7877b28
NC
9875 if (isGV_with_GP(sstr)) {
9876 /* Don't need to do anything here. */
9877 }
9878 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9879 /* A "shared" PV - clone it as "shared" PV */
9880 SvPV_set(dstr,
9881 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9882 param)));
83841fad
NIS
9883 }
9884 else {
9885 /* Some other special case - random pointer */
f880fe2f 9886 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9887 }
83841fad
NIS
9888 }
9889 }
9890 else {
4608196e 9891 /* Copy the NULL */
f880fe2f 9892 if (SvTYPE(dstr) == SVt_RV)
b162af07 9893 SvRV_set(dstr, NULL);
f880fe2f 9894 else
6136c704 9895 SvPV_set(dstr, NULL);
83841fad
NIS
9896 }
9897}
9898
662fb8b2
NC
9899/* duplicate an SV of any type (including AV, HV etc) */
9900
1d7c1841 9901SV *
eb86f8b3 9902Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9903{
27da23d5 9904 dVAR;
1d7c1841
GS
9905 SV *dstr;
9906
9907 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9908 return NULL;
1d7c1841
GS
9909 /* look for it in the table first */
9910 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9911 if (dstr)
9912 return dstr;
9913
0405e91e
AB
9914 if(param->flags & CLONEf_JOIN_IN) {
9915 /** We are joining here so we don't want do clone
9916 something that is bad **/
eb86f8b3
AL
9917 if (SvTYPE(sstr) == SVt_PVHV) {
9918 const char * const hvname = HvNAME_get(sstr);
9919 if (hvname)
9920 /** don't clone stashes if they already exist **/
9921 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9922 }
9923 }
9924
1d7c1841
GS
9925 /* create anew and remember what it is */
9926 new_SV(dstr);
fd0854ff
DM
9927
9928#ifdef DEBUG_LEAKING_SCALARS
9929 dstr->sv_debug_optype = sstr->sv_debug_optype;
9930 dstr->sv_debug_line = sstr->sv_debug_line;
9931 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9932 dstr->sv_debug_cloned = 1;
fd0854ff 9933 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9934#endif
9935
1d7c1841
GS
9936 ptr_table_store(PL_ptr_table, sstr, dstr);
9937
9938 /* clone */
9939 SvFLAGS(dstr) = SvFLAGS(sstr);
9940 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9941 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9942
9943#ifdef DEBUGGING
3f7c398e 9944 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9945 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9946 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9947#endif
9948
9660f481
DM
9949 /* don't clone objects whose class has asked us not to */
9950 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9951 SvFLAGS(dstr) &= ~SVTYPEMASK;
9952 SvOBJECT_off(dstr);
9953 return dstr;
9954 }
9955
1d7c1841
GS
9956 switch (SvTYPE(sstr)) {
9957 case SVt_NULL:
9958 SvANY(dstr) = NULL;
9959 break;
9960 case SVt_IV:
339049b0 9961 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9962 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9963 break;
9964 case SVt_NV:
9965 SvANY(dstr) = new_XNV();
9d6ce603 9966 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9967 break;
9968 case SVt_RV:
339049b0 9969 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9970 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9971 break;
662fb8b2
NC
9972 default:
9973 {
9974 /* These are all the types that need complex bodies allocating. */
662fb8b2 9975 void *new_body;
2bcc16b3
NC
9976 const svtype sv_type = SvTYPE(sstr);
9977 const struct body_details *const sv_type_details
9978 = bodies_by_type + sv_type;
662fb8b2 9979
93e68bfb 9980 switch (sv_type) {
662fb8b2 9981 default:
bb263b4e 9982 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9983 break;
9984
662fb8b2
NC
9985 case SVt_PVGV:
9986 if (GvUNIQUE((GV*)sstr)) {
6f207bd3 9987 NOOP; /* Do sharing here, and fall through */
662fb8b2 9988 }
c22188b4
NC
9989 case SVt_PVIO:
9990 case SVt_PVFM:
9991 case SVt_PVHV:
9992 case SVt_PVAV:
93e68bfb 9993 case SVt_PVBM:
662fb8b2 9994 case SVt_PVCV:
662fb8b2 9995 case SVt_PVLV:
662fb8b2 9996 case SVt_PVMG:
662fb8b2 9997 case SVt_PVNV:
662fb8b2 9998 case SVt_PVIV:
662fb8b2 9999 case SVt_PV:
d2a0f284 10000 assert(sv_type_details->body_size);
c22188b4 10001 if (sv_type_details->arena) {
d2a0f284 10002 new_body_inline(new_body, sv_type);
c22188b4 10003 new_body
b9502f15 10004 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
10005 } else {
10006 new_body = new_NOARENA(sv_type_details);
10007 }
1d7c1841 10008 }
662fb8b2
NC
10009 assert(new_body);
10010 SvANY(dstr) = new_body;
10011
2bcc16b3 10012#ifndef PURIFY
b9502f15
NC
10013 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10014 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 10015 sv_type_details->copy, char);
2bcc16b3
NC
10016#else
10017 Copy(((char*)SvANY(sstr)),
10018 ((char*)SvANY(dstr)),
d2a0f284 10019 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10020#endif
662fb8b2 10021
f7877b28
NC
10022 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10023 && !isGV_with_GP(dstr))
662fb8b2
NC
10024 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10025
10026 /* The Copy above means that all the source (unduplicated) pointers
10027 are now in the destination. We can check the flags and the
10028 pointers in either, but it's possible that there's less cache
10029 missing by always going for the destination.
10030 FIXME - instrument and check that assumption */
f32993d6 10031 if (sv_type >= SVt_PVMG) {
e736a858 10032 HV *ourstash;
2892acdb 10033 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
10034 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
10035 } else if (SvMAGIC(dstr))
662fb8b2
NC
10036 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10037 if (SvSTASH(dstr))
10038 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10039 }
662fb8b2 10040
f32993d6
NC
10041 /* The cast silences a GCC warning about unhandled types. */
10042 switch ((int)sv_type) {
662fb8b2
NC
10043 case SVt_PV:
10044 break;
10045 case SVt_PVIV:
10046 break;
10047 case SVt_PVNV:
10048 break;
10049 case SVt_PVMG:
10050 break;
10051 case SVt_PVBM:
10052 break;
10053 case SVt_PVLV:
10054 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10055 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10056 LvTARG(dstr) = dstr;
10057 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10058 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10059 else
10060 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10061 break;
10062 case SVt_PVGV:
acda4c6a
NC
10063 if (GvNAME_HEK(dstr))
10064 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 10065
e15faf7d
NC
10066 /* Don't call sv_add_backref here as it's going to be created
10067 as part of the magic cloning of the symbol table. */
f7877b28
NC
10068 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10069 if(isGV_with_GP(sstr)) {
10070 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10071 at the point of this comment. */
10072 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10073 (void)GpREFCNT_inc(GvGP(dstr));
10074 } else
10075 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10076 break;
10077 case SVt_PVIO:
10078 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10079 if (IoOFP(dstr) == IoIFP(sstr))
10080 IoOFP(dstr) = IoIFP(dstr);
10081 else
10082 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10083 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10084 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10085 /* I have no idea why fake dirp (rsfps)
10086 should be treated differently but otherwise
10087 we end up with leaks -- sky*/
10088 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10089 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10090 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10091 } else {
10092 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10093 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10094 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10095 if (IoDIRP(dstr)) {
10096 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10097 } else {
6f207bd3 10098 NOOP;
100ce7e1
NC
10099 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10100 }
662fb8b2
NC
10101 }
10102 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10103 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10104 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10105 break;
10106 case SVt_PVAV:
10107 if (AvARRAY((AV*)sstr)) {
10108 SV **dst_ary, **src_ary;
10109 SSize_t items = AvFILLp((AV*)sstr) + 1;
10110
10111 src_ary = AvARRAY((AV*)sstr);
a02a5408 10112 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10113 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10114 SvPV_set(dstr, (char*)dst_ary);
10115 AvALLOC((AV*)dstr) = dst_ary;
10116 if (AvREAL((AV*)sstr)) {
10117 while (items-- > 0)
10118 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10119 }
10120 else {
10121 while (items-- > 0)
10122 *dst_ary++ = sv_dup(*src_ary++, param);
10123 }
10124 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10125 while (items-- > 0) {
10126 *dst_ary++ = &PL_sv_undef;
10127 }
bfcb3514 10128 }
662fb8b2 10129 else {
bd61b366 10130 SvPV_set(dstr, NULL);
662fb8b2 10131 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10132 }
662fb8b2
NC
10133 break;
10134 case SVt_PVHV:
10135 {
cbbf8932 10136 HEK *hvname = NULL;
662fb8b2
NC
10137
10138 if (HvARRAY((HV*)sstr)) {
10139 STRLEN i = 0;
10140 const bool sharekeys = !!HvSHAREKEYS(sstr);
10141 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10142 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10143 char *darray;
a02a5408 10144 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10145 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10146 char);
10147 HvARRAY(dstr) = (HE**)darray;
10148 while (i <= sxhv->xhv_max) {
5c4138a0 10149 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10150 HvARRAY(dstr)[i] = source
10151 ? he_dup(source, sharekeys, param) : 0;
10152 ++i;
10153 }
10154 if (SvOOK(sstr)) {
00b6aa41
AL
10155 struct xpvhv_aux * const saux = HvAUX(sstr);
10156 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10157 /* This flag isn't copied. */
10158 /* SvOOK_on(hv) attacks the IV flags. */
10159 SvFLAGS(dstr) |= SVf_OOK;
10160
10161 hvname = saux->xhv_name;
dd690478
NC
10162 daux->xhv_name
10163 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10164
10165 daux->xhv_riter = saux->xhv_riter;
10166 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10167 ? he_dup(saux->xhv_eiter,
10168 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10169 daux->xhv_backreferences = saux->xhv_backreferences
10170 ? (AV*) SvREFCNT_inc(
10171 sv_dup((SV*)saux->
10172 xhv_backreferences,
10173 param))
10174 : 0;
662fb8b2
NC
10175 }
10176 }
10177 else {
bd61b366 10178 SvPV_set(dstr, NULL);
662fb8b2
NC
10179 }
10180 /* Record stashes for possible cloning in Perl_clone(). */
10181 if(hvname)
10182 av_push(param->stashes, dstr);
10183 }
10184 break;
662fb8b2 10185 case SVt_PVCV:
bb172083
NC
10186 if (!(param->flags & CLONEf_COPY_STACKS)) {
10187 CvDEPTH(dstr) = 0;
10188 }
10189 case SVt_PVFM:
662fb8b2
NC
10190 /* NOTE: not refcounted */
10191 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10192 OP_REFCNT_LOCK;
d04ba589
NC
10193 if (!CvISXSUB(dstr))
10194 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10195 OP_REFCNT_UNLOCK;
cfae286e 10196 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10197 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10198 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10199 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10200 }
10201 /* don't dup if copying back - CvGV isn't refcounted, so the
10202 * duped GV may never be freed. A bit of a hack! DAPM */
10203 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10204 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10205 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10206 CvOUTSIDE(dstr) =
10207 CvWEAKOUTSIDE(sstr)
10208 ? cv_dup( CvOUTSIDE(dstr), param)
10209 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10210 if (!CvISXSUB(dstr))
662fb8b2
NC
10211 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10212 break;
bfcb3514 10213 }
1d7c1841 10214 }
1d7c1841
GS
10215 }
10216
10217 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10218 ++PL_sv_objcount;
10219
10220 return dstr;
d2d73c3e 10221 }
1d7c1841 10222
645c22ef
DM
10223/* duplicate a context */
10224
1d7c1841 10225PERL_CONTEXT *
a8fc9800 10226Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10227{
10228 PERL_CONTEXT *ncxs;
10229
10230 if (!cxs)
10231 return (PERL_CONTEXT*)NULL;
10232
10233 /* look for it in the table first */
10234 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10235 if (ncxs)
10236 return ncxs;
10237
10238 /* create anew and remember what it is */
a02a5408 10239 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10240 ptr_table_store(PL_ptr_table, cxs, ncxs);
10241
10242 while (ix >= 0) {
c445ea15
AL
10243 PERL_CONTEXT * const cx = &cxs[ix];
10244 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10245 ncx->cx_type = cx->cx_type;
10246 if (CxTYPE(cx) == CXt_SUBST) {
10247 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10248 }
10249 else {
10250 ncx->blk_oldsp = cx->blk_oldsp;
10251 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10252 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10253 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10254 ncx->blk_oldpm = cx->blk_oldpm;
10255 ncx->blk_gimme = cx->blk_gimme;
10256 switch (CxTYPE(cx)) {
10257 case CXt_SUB:
10258 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10259 ? cv_dup_inc(cx->blk_sub.cv, param)
10260 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10261 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10262 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10263 : NULL);
d2d73c3e 10264 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10265 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10266 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10267 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10268 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10269 break;
10270 case CXt_EVAL:
10271 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10272 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10273 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10274 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10275 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10276 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10277 break;
10278 case CXt_LOOP:
10279 ncx->blk_loop.label = cx->blk_loop.label;
10280 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10281 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10282 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10283 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10284 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10285 ? cx->blk_loop.iterdata
d2d73c3e 10286 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10287 ncx->blk_loop.oldcomppad
10288 = (PAD*)ptr_table_fetch(PL_ptr_table,
10289 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10290 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10291 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10292 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10293 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10294 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10295 break;
10296 case CXt_FORMAT:
d2d73c3e
AB
10297 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10298 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10299 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10300 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10301 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10302 break;
10303 case CXt_BLOCK:
10304 case CXt_NULL:
10305 break;
10306 }
10307 }
10308 --ix;
10309 }
10310 return ncxs;
10311}
10312
645c22ef
DM
10313/* duplicate a stack info structure */
10314
1d7c1841 10315PERL_SI *
a8fc9800 10316Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10317{
10318 PERL_SI *nsi;
10319
10320 if (!si)
10321 return (PERL_SI*)NULL;
10322
10323 /* look for it in the table first */
10324 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10325 if (nsi)
10326 return nsi;
10327
10328 /* create anew and remember what it is */
a02a5408 10329 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10330 ptr_table_store(PL_ptr_table, si, nsi);
10331
d2d73c3e 10332 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10333 nsi->si_cxix = si->si_cxix;
10334 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10335 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10336 nsi->si_type = si->si_type;
d2d73c3e
AB
10337 nsi->si_prev = si_dup(si->si_prev, param);
10338 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10339 nsi->si_markoff = si->si_markoff;
10340
10341 return nsi;
10342}
10343
10344#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10345#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10346#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10347#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10348#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10349#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10350#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10351#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10352#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10353#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10354#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10355#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10356#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10357#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10358
10359/* XXXXX todo */
10360#define pv_dup_inc(p) SAVEPV(p)
10361#define pv_dup(p) SAVEPV(p)
10362#define svp_dup_inc(p,pp) any_dup(p,pp)
10363
645c22ef
DM
10364/* map any object to the new equivent - either something in the
10365 * ptr table, or something in the interpreter structure
10366 */
10367
1d7c1841 10368void *
53c1dcc0 10369Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10370{
10371 void *ret;
10372
10373 if (!v)
10374 return (void*)NULL;
10375
10376 /* look for it in the table first */
10377 ret = ptr_table_fetch(PL_ptr_table, v);
10378 if (ret)
10379 return ret;
10380
10381 /* see if it is part of the interpreter structure */
10382 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10383 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10384 else {
1d7c1841 10385 ret = v;
05ec9bb3 10386 }
1d7c1841
GS
10387
10388 return ret;
10389}
10390
645c22ef
DM
10391/* duplicate the save stack */
10392
1d7c1841 10393ANY *
a8fc9800 10394Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10395{
53c1dcc0
AL
10396 ANY * const ss = proto_perl->Tsavestack;
10397 const I32 max = proto_perl->Tsavestack_max;
10398 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10399 ANY *nss;
10400 SV *sv;
10401 GV *gv;
10402 AV *av;
10403 HV *hv;
10404 void* ptr;
10405 int intval;
10406 long longval;
10407 GP *gp;
10408 IV iv;
c4e33207 10409 char *c = NULL;
1d7c1841 10410 void (*dptr) (void*);
acfe0abc 10411 void (*dxptr) (pTHX_ void*);
1d7c1841 10412
a02a5408 10413 Newxz(nss, max, ANY);
1d7c1841
GS
10414
10415 while (ix > 0) {
b464bac0 10416 I32 i = POPINT(ss,ix);
1d7c1841
GS
10417 TOPINT(nss,ix) = i;
10418 switch (i) {
10419 case SAVEt_ITEM: /* normal string */
a41cc44e 10420 case SAVEt_SV: /* scalar reference */
1d7c1841 10421 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10422 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10423 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10424 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10425 break;
05ec9bb3
NIS
10426 case SAVEt_SHARED_PVREF: /* char* in shared space */
10427 c = (char*)POPPTR(ss,ix);
10428 TOPPTR(nss,ix) = savesharedpv(c);
10429 ptr = POPPTR(ss,ix);
10430 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10431 break;
1d7c1841
GS
10432 case SAVEt_GENERIC_SVREF: /* generic sv */
10433 case SAVEt_SVREF: /* scalar reference */
10434 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10435 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10436 ptr = POPPTR(ss,ix);
10437 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10438 break;
a41cc44e 10439 case SAVEt_HV: /* hash reference */
1d7c1841 10440 case SAVEt_AV: /* array reference */
337d28f5
NC
10441 sv = POPPTR(ss,ix);
10442 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10443 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10444 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10445 break;
10446 case SAVEt_INT: /* int reference */
10447 ptr = POPPTR(ss,ix);
10448 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10449 intval = (int)POPINT(ss,ix);
10450 TOPINT(nss,ix) = intval;
10451 break;
10452 case SAVEt_LONG: /* long reference */
10453 ptr = POPPTR(ss,ix);
10454 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10455 longval = (long)POPLONG(ss,ix);
10456 TOPLONG(nss,ix) = longval;
10457 break;
10458 case SAVEt_I32: /* I32 reference */
10459 case SAVEt_I16: /* I16 reference */
10460 case SAVEt_I8: /* I8 reference */
88effcc9 10461 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10462 ptr = POPPTR(ss,ix);
10463 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10464 i = POPINT(ss,ix);
10465 TOPINT(nss,ix) = i;
10466 break;
10467 case SAVEt_IV: /* IV reference */
10468 ptr = POPPTR(ss,ix);
10469 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10470 iv = POPIV(ss,ix);
10471 TOPIV(nss,ix) = iv;
10472 break;
a41cc44e
NC
10473 case SAVEt_HPTR: /* HV* reference */
10474 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10475 case SAVEt_SPTR: /* SV* reference */
10476 ptr = POPPTR(ss,ix);
10477 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10478 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10479 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10480 break;
10481 case SAVEt_VPTR: /* random* reference */
10482 ptr = POPPTR(ss,ix);
10483 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10484 ptr = POPPTR(ss,ix);
10485 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10486 break;
b03d03b0 10487 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10488 case SAVEt_PPTR: /* char* reference */
10489 ptr = POPPTR(ss,ix);
10490 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10491 c = (char*)POPPTR(ss,ix);
10492 TOPPTR(nss,ix) = pv_dup(c);
10493 break;
1d7c1841
GS
10494 case SAVEt_NSTAB:
10495 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10496 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10497 break;
10498 case SAVEt_GP: /* scalar reference */
10499 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10500 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10501 (void)GpREFCNT_inc(gp);
10502 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10503 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10504 c = (char*)POPPTR(ss,ix);
10505 TOPPTR(nss,ix) = pv_dup(c);
10506 iv = POPIV(ss,ix);
10507 TOPIV(nss,ix) = iv;
10508 iv = POPIV(ss,ix);
10509 TOPIV(nss,ix) = iv;
10510 break;
10511 case SAVEt_FREESV:
26d9b02f 10512 case SAVEt_MORTALIZESV:
1d7c1841 10513 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10514 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10515 break;
10516 case SAVEt_FREEOP:
10517 ptr = POPPTR(ss,ix);
10518 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10519 /* these are assumed to be refcounted properly */
53c1dcc0 10520 OP *o;
1d7c1841
GS
10521 switch (((OP*)ptr)->op_type) {
10522 case OP_LEAVESUB:
10523 case OP_LEAVESUBLV:
10524 case OP_LEAVEEVAL:
10525 case OP_LEAVE:
10526 case OP_SCOPE:
10527 case OP_LEAVEWRITE:
e977893f
GS
10528 TOPPTR(nss,ix) = ptr;
10529 o = (OP*)ptr;
10530 OpREFCNT_inc(o);
1d7c1841
GS
10531 break;
10532 default:
5f66b61c 10533 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10534 break;
10535 }
10536 }
10537 else
5f66b61c 10538 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10539 break;
10540 case SAVEt_FREEPV:
10541 c = (char*)POPPTR(ss,ix);
10542 TOPPTR(nss,ix) = pv_dup_inc(c);
10543 break;
10544 case SAVEt_CLEARSV:
10545 longval = POPLONG(ss,ix);
10546 TOPLONG(nss,ix) = longval;
10547 break;
10548 case SAVEt_DELETE:
10549 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10550 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10551 c = (char*)POPPTR(ss,ix);
10552 TOPPTR(nss,ix) = pv_dup_inc(c);
10553 i = POPINT(ss,ix);
10554 TOPINT(nss,ix) = i;
10555 break;
10556 case SAVEt_DESTRUCTOR:
10557 ptr = POPPTR(ss,ix);
10558 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10559 dptr = POPDPTR(ss,ix);
8141890a
JH
10560 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10561 any_dup(FPTR2DPTR(void *, dptr),
10562 proto_perl));
1d7c1841
GS
10563 break;
10564 case SAVEt_DESTRUCTOR_X:
10565 ptr = POPPTR(ss,ix);
10566 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10567 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10568 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10569 any_dup(FPTR2DPTR(void *, dxptr),
10570 proto_perl));
1d7c1841
GS
10571 break;
10572 case SAVEt_REGCONTEXT:
10573 case SAVEt_ALLOC:
10574 i = POPINT(ss,ix);
10575 TOPINT(nss,ix) = i;
10576 ix -= i;
10577 break;
10578 case SAVEt_STACK_POS: /* Position on Perl stack */
10579 i = POPINT(ss,ix);
10580 TOPINT(nss,ix) = i;
10581 break;
10582 case SAVEt_AELEM: /* array element */
10583 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10584 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10585 i = POPINT(ss,ix);
10586 TOPINT(nss,ix) = i;
10587 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10588 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10589 break;
10590 case SAVEt_HELEM: /* hash element */
10591 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10592 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10593 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10594 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10595 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10596 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10597 break;
10598 case SAVEt_OP:
10599 ptr = POPPTR(ss,ix);
10600 TOPPTR(nss,ix) = ptr;
10601 break;
10602 case SAVEt_HINTS:
10603 i = POPINT(ss,ix);
10604 TOPINT(nss,ix) = i;
b3ca2e83 10605 ptr = POPPTR(ss,ix);
080ac856 10606 if (ptr) {
7b6dd8c3 10607 HINTS_REFCNT_LOCK;
080ac856 10608 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10609 HINTS_REFCNT_UNLOCK;
10610 }
cbb1fbea 10611 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10612 if (i & HINT_LOCALIZE_HH) {
10613 hv = (HV*)POPPTR(ss,ix);
10614 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10615 }
1d7c1841 10616 break;
c4410b1b
GS
10617 case SAVEt_COMPPAD:
10618 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10619 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10620 break;
c3564e5c
GS
10621 case SAVEt_PADSV:
10622 longval = (long)POPLONG(ss,ix);
10623 TOPLONG(nss,ix) = longval;
10624 ptr = POPPTR(ss,ix);
10625 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10626 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10627 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10628 break;
a1bb4754 10629 case SAVEt_BOOL:
38d8b13e 10630 ptr = POPPTR(ss,ix);
b9609c01 10631 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10632 longval = (long)POPBOOL(ss,ix);
b9609c01 10633 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10634 break;
8bd2680e
MHM
10635 case SAVEt_SET_SVFLAGS:
10636 i = POPINT(ss,ix);
10637 TOPINT(nss,ix) = i;
10638 i = POPINT(ss,ix);
10639 TOPINT(nss,ix) = i;
10640 sv = (SV*)POPPTR(ss,ix);
10641 TOPPTR(nss,ix) = sv_dup(sv, param);
10642 break;
5bfb7d0e
NC
10643 case SAVEt_RE_STATE:
10644 {
10645 const struct re_save_state *const old_state
10646 = (struct re_save_state *)
10647 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10648 struct re_save_state *const new_state
10649 = (struct re_save_state *)
10650 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10651
10652 Copy(old_state, new_state, 1, struct re_save_state);
10653 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10654
10655 new_state->re_state_bostr
10656 = pv_dup(old_state->re_state_bostr);
10657 new_state->re_state_reginput
10658 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10659 new_state->re_state_regeol
10660 = pv_dup(old_state->re_state_regeol);
10661 new_state->re_state_regstartp
10662 = any_dup(old_state->re_state_regstartp, proto_perl);
10663 new_state->re_state_regendp
10664 = any_dup(old_state->re_state_regendp, proto_perl);
10665 new_state->re_state_reglastparen
10666 = any_dup(old_state->re_state_reglastparen, proto_perl);
10667 new_state->re_state_reglastcloseparen
10668 = any_dup(old_state->re_state_reglastcloseparen,
10669 proto_perl);
5bfb7d0e
NC
10670 /* XXX This just has to be broken. The old save_re_context
10671 code did SAVEGENERICPV(PL_reg_start_tmp);
10672 PL_reg_start_tmp is char **.
10673 Look above to what the dup code does for
10674 SAVEt_GENERIC_PVREF
10675 It can never have worked.
10676 So this is merely a faithful copy of the exiting bug: */
10677 new_state->re_state_reg_start_tmp
10678 = (char **) pv_dup((char *)
10679 old_state->re_state_reg_start_tmp);
10680 /* I assume that it only ever "worked" because no-one called
10681 (pseudo)fork while the regexp engine had re-entered itself.
10682 */
5bfb7d0e
NC
10683#ifdef PERL_OLD_COPY_ON_WRITE
10684 new_state->re_state_nrs
10685 = sv_dup(old_state->re_state_nrs, param);
10686#endif
10687 new_state->re_state_reg_magic
10688 = any_dup(old_state->re_state_reg_magic, proto_perl);
10689 new_state->re_state_reg_oldcurpm
10690 = any_dup(old_state->re_state_reg_oldcurpm, proto_perl);
10691 new_state->re_state_reg_curpm
10692 = any_dup(old_state->re_state_reg_curpm, proto_perl);
10693 new_state->re_state_reg_oldsaved
10694 = pv_dup(old_state->re_state_reg_oldsaved);
10695 new_state->re_state_reg_poscache
10696 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
10697 new_state->re_state_reg_starttry
10698 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
10699 break;
10700 }
68da3b2f
NC
10701 case SAVEt_COMPILE_WARNINGS:
10702 ptr = POPPTR(ss,ix);
10703 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10704 break;
1d7c1841 10705 default:
ca05af4a 10706 Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i);
1d7c1841
GS
10707 }
10708 }
10709
bd81e77b
NC
10710 return nss;
10711}
10712
10713
10714/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10715 * flag to the result. This is done for each stash before cloning starts,
10716 * so we know which stashes want their objects cloned */
10717
10718static void
10719do_mark_cloneable_stash(pTHX_ SV *sv)
10720{
10721 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10722 if (hvname) {
10723 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10724 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10725 if (cloner && GvCV(cloner)) {
10726 dSP;
10727 UV status;
10728
10729 ENTER;
10730 SAVETMPS;
10731 PUSHMARK(SP);
10732 XPUSHs(sv_2mortal(newSVhek(hvname)));
10733 PUTBACK;
10734 call_sv((SV*)GvCV(cloner), G_SCALAR);
10735 SPAGAIN;
10736 status = POPu;
10737 PUTBACK;
10738 FREETMPS;
10739 LEAVE;
10740 if (status)
10741 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10742 }
10743 }
10744}
10745
10746
10747
10748/*
10749=for apidoc perl_clone
10750
10751Create and return a new interpreter by cloning the current one.
10752
10753perl_clone takes these flags as parameters:
10754
10755CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10756without it we only clone the data and zero the stacks,
10757with it we copy the stacks and the new perl interpreter is
10758ready to run at the exact same point as the previous one.
10759The pseudo-fork code uses COPY_STACKS while the
10760threads->new doesn't.
10761
10762CLONEf_KEEP_PTR_TABLE
10763perl_clone keeps a ptr_table with the pointer of the old
10764variable as a key and the new variable as a value,
10765this allows it to check if something has been cloned and not
10766clone it again but rather just use the value and increase the
10767refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10768the ptr_table using the function
10769C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10770reason to keep it around is if you want to dup some of your own
10771variable who are outside the graph perl scans, example of this
10772code is in threads.xs create
10773
10774CLONEf_CLONE_HOST
10775This is a win32 thing, it is ignored on unix, it tells perls
10776win32host code (which is c++) to clone itself, this is needed on
10777win32 if you want to run two threads at the same time,
10778if you just want to do some stuff in a separate perl interpreter
10779and then throw it away and return to the original one,
10780you don't need to do anything.
10781
10782=cut
10783*/
10784
10785/* XXX the above needs expanding by someone who actually understands it ! */
10786EXTERN_C PerlInterpreter *
10787perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10788
10789PerlInterpreter *
10790perl_clone(PerlInterpreter *proto_perl, UV flags)
10791{
10792 dVAR;
10793#ifdef PERL_IMPLICIT_SYS
10794
10795 /* perlhost.h so we need to call into it
10796 to clone the host, CPerlHost should have a c interface, sky */
10797
10798 if (flags & CLONEf_CLONE_HOST) {
10799 return perl_clone_host(proto_perl,flags);
10800 }
10801 return perl_clone_using(proto_perl, flags,
10802 proto_perl->IMem,
10803 proto_perl->IMemShared,
10804 proto_perl->IMemParse,
10805 proto_perl->IEnv,
10806 proto_perl->IStdIO,
10807 proto_perl->ILIO,
10808 proto_perl->IDir,
10809 proto_perl->ISock,
10810 proto_perl->IProc);
10811}
10812
10813PerlInterpreter *
10814perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10815 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10816 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10817 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10818 struct IPerlDir* ipD, struct IPerlSock* ipS,
10819 struct IPerlProc* ipP)
10820{
10821 /* XXX many of the string copies here can be optimized if they're
10822 * constants; they need to be allocated as common memory and just
10823 * their pointers copied. */
10824
10825 IV i;
10826 CLONE_PARAMS clone_params;
5f66b61c 10827 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10828
5f66b61c 10829 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10830 /* for each stash, determine whether its objects should be cloned */
10831 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10832 PERL_SET_THX(my_perl);
10833
10834# ifdef DEBUGGING
7e337ee0 10835 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10836 PL_op = NULL;
10837 PL_curcop = NULL;
bd81e77b
NC
10838 PL_markstack = 0;
10839 PL_scopestack = 0;
10840 PL_savestack = 0;
10841 PL_savestack_ix = 0;
10842 PL_savestack_max = -1;
10843 PL_sig_pending = 0;
10844 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10845# else /* !DEBUGGING */
10846 Zero(my_perl, 1, PerlInterpreter);
10847# endif /* DEBUGGING */
10848
10849 /* host pointers */
10850 PL_Mem = ipM;
10851 PL_MemShared = ipMS;
10852 PL_MemParse = ipMP;
10853 PL_Env = ipE;
10854 PL_StdIO = ipStd;
10855 PL_LIO = ipLIO;
10856 PL_Dir = ipD;
10857 PL_Sock = ipS;
10858 PL_Proc = ipP;
10859#else /* !PERL_IMPLICIT_SYS */
10860 IV i;
10861 CLONE_PARAMS clone_params;
10862 CLONE_PARAMS* param = &clone_params;
5f66b61c 10863 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10864 /* for each stash, determine whether its objects should be cloned */
10865 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10866 PERL_SET_THX(my_perl);
10867
10868# ifdef DEBUGGING
7e337ee0 10869 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10870 PL_op = NULL;
10871 PL_curcop = NULL;
bd81e77b
NC
10872 PL_markstack = 0;
10873 PL_scopestack = 0;
10874 PL_savestack = 0;
10875 PL_savestack_ix = 0;
10876 PL_savestack_max = -1;
10877 PL_sig_pending = 0;
10878 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10879# else /* !DEBUGGING */
10880 Zero(my_perl, 1, PerlInterpreter);
10881# endif /* DEBUGGING */
10882#endif /* PERL_IMPLICIT_SYS */
10883 param->flags = flags;
10884 param->proto_perl = proto_perl;
10885
7cb608b5
NC
10886 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10887
fdda85ca 10888 PL_body_arenas = NULL;
bd81e77b
NC
10889 Zero(&PL_body_roots, 1, PL_body_roots);
10890
10891 PL_nice_chunk = NULL;
10892 PL_nice_chunk_size = 0;
10893 PL_sv_count = 0;
10894 PL_sv_objcount = 0;
a0714e2c
SS
10895 PL_sv_root = NULL;
10896 PL_sv_arenaroot = NULL;
bd81e77b
NC
10897
10898 PL_debug = proto_perl->Idebug;
10899
10900 PL_hash_seed = proto_perl->Ihash_seed;
10901 PL_rehash_seed = proto_perl->Irehash_seed;
10902
10903#ifdef USE_REENTRANT_API
10904 /* XXX: things like -Dm will segfault here in perlio, but doing
10905 * PERL_SET_CONTEXT(proto_perl);
10906 * breaks too many other things
10907 */
10908 Perl_reentrant_init(aTHX);
10909#endif
10910
10911 /* create SV map for pointer relocation */
10912 PL_ptr_table = ptr_table_new();
10913
10914 /* initialize these special pointers as early as possible */
10915 SvANY(&PL_sv_undef) = NULL;
10916 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10917 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10918 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10919
10920 SvANY(&PL_sv_no) = new_XPVNV();
10921 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10922 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10923 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10924 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
10925 SvCUR_set(&PL_sv_no, 0);
10926 SvLEN_set(&PL_sv_no, 1);
10927 SvIV_set(&PL_sv_no, 0);
10928 SvNV_set(&PL_sv_no, 0);
10929 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10930
10931 SvANY(&PL_sv_yes) = new_XPVNV();
10932 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10933 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10934 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10935 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
10936 SvCUR_set(&PL_sv_yes, 1);
10937 SvLEN_set(&PL_sv_yes, 2);
10938 SvIV_set(&PL_sv_yes, 1);
10939 SvNV_set(&PL_sv_yes, 1);
10940 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10941
10942 /* create (a non-shared!) shared string table */
10943 PL_strtab = newHV();
10944 HvSHAREKEYS_off(PL_strtab);
10945 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10946 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10947
10948 PL_compiling = proto_perl->Icompiling;
10949
10950 /* These two PVs will be free'd special way so must set them same way op.c does */
10951 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10952 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10953
10954 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10955 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10956
10957 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10958 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
bd81e77b
NC
10959 if (!specialCopIO(PL_compiling.cop_io))
10960 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
cbb1fbea
NC
10961 if (PL_compiling.cop_hints) {
10962 HINTS_REFCNT_LOCK;
10963 PL_compiling.cop_hints->refcounted_he_refcnt++;
10964 HINTS_REFCNT_UNLOCK;
10965 }
bd81e77b
NC
10966 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10967
10968 /* pseudo environmental stuff */
10969 PL_origargc = proto_perl->Iorigargc;
10970 PL_origargv = proto_perl->Iorigargv;
10971
10972 param->stashes = newAV(); /* Setup array of objects to call clone on */
10973
10974 /* Set tainting stuff before PerlIO_debug can possibly get called */
10975 PL_tainting = proto_perl->Itainting;
10976 PL_taint_warn = proto_perl->Itaint_warn;
10977
10978#ifdef PERLIO_LAYERS
10979 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10980 PerlIO_clone(aTHX_ proto_perl, param);
10981#endif
10982
10983 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10984 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10985 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10986 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10987 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10988 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10989
10990 /* switches */
10991 PL_minus_c = proto_perl->Iminus_c;
10992 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10993 PL_localpatches = proto_perl->Ilocalpatches;
10994 PL_splitstr = proto_perl->Isplitstr;
10995 PL_preprocess = proto_perl->Ipreprocess;
10996 PL_minus_n = proto_perl->Iminus_n;
10997 PL_minus_p = proto_perl->Iminus_p;
10998 PL_minus_l = proto_perl->Iminus_l;
10999 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 11000 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
11001 PL_minus_F = proto_perl->Iminus_F;
11002 PL_doswitches = proto_perl->Idoswitches;
11003 PL_dowarn = proto_perl->Idowarn;
11004 PL_doextract = proto_perl->Idoextract;
11005 PL_sawampersand = proto_perl->Isawampersand;
11006 PL_unsafe = proto_perl->Iunsafe;
11007 PL_inplace = SAVEPV(proto_perl->Iinplace);
11008 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11009 PL_perldb = proto_perl->Iperldb;
11010 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11011 PL_exit_flags = proto_perl->Iexit_flags;
11012
11013 /* magical thingies */
11014 /* XXX time(&PL_basetime) when asked for? */
11015 PL_basetime = proto_perl->Ibasetime;
11016 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11017
11018 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
11019 PL_statusvalue = proto_perl->Istatusvalue;
11020#ifdef VMS
11021 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11022#else
11023 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11024#endif
11025 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11026
11027 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
11028 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
11029 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
11030
11031 /* Clone the regex array */
11032 PL_regex_padav = newAV();
11033 {
11034 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 11035 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 11036 IV i;
7f466ec7 11037 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 11038 for(i = 1; i <= len; i++) {
7a5b473e
AL
11039 const SV * const regex = regexen[i];
11040 SV * const sv =
11041 SvREPADTMP(regex)
11042 ? sv_dup_inc(regex, param)
11043 : SvREFCNT_inc(
11044 newSViv(PTR2IV(re_dup(
11045 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11046 ;
11047 av_push(PL_regex_padav, sv);
bd81e77b
NC
11048 }
11049 }
11050 PL_regex_pad = AvARRAY(PL_regex_padav);
11051
11052 /* shortcuts to various I/O objects */
11053 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11054 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11055 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11056 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11057 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11058 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11059
bd81e77b
NC
11060 /* shortcuts to regexp stuff */
11061 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11062
bd81e77b
NC
11063 /* shortcuts to misc objects */
11064 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11065
bd81e77b
NC
11066 /* shortcuts to debugging objects */
11067 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11068 PL_DBline = gv_dup(proto_perl->IDBline, param);
11069 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11070 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11071 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11072 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11073 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11074 PL_lineary = av_dup(proto_perl->Ilineary, param);
11075 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11076
bd81e77b
NC
11077 /* symbol tables */
11078 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11079 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11080 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11081 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11082 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11083
11084 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11085 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11086 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11087 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11088 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11089 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11090
11091 PL_sub_generation = proto_perl->Isub_generation;
11092
11093 /* funky return mechanisms */
11094 PL_forkprocess = proto_perl->Iforkprocess;
11095
11096 /* subprocess state */
11097 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11098
11099 /* internal state */
11100 PL_maxo = proto_perl->Imaxo;
11101 if (proto_perl->Iop_mask)
11102 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11103 else
bd61b366 11104 PL_op_mask = NULL;
bd81e77b
NC
11105 /* PL_asserting = proto_perl->Iasserting; */
11106
11107 /* current interpreter roots */
11108 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
11109 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
11110 PL_main_start = proto_perl->Imain_start;
11111 PL_eval_root = proto_perl->Ieval_root;
11112 PL_eval_start = proto_perl->Ieval_start;
11113
11114 /* runtime control stuff */
11115 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11116 PL_copline = proto_perl->Icopline;
11117
11118 PL_filemode = proto_perl->Ifilemode;
11119 PL_lastfd = proto_perl->Ilastfd;
11120 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11121 PL_Argv = NULL;
bd61b366 11122 PL_Cmd = NULL;
bd81e77b
NC
11123 PL_gensym = proto_perl->Igensym;
11124 PL_preambled = proto_perl->Ipreambled;
11125 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11126 PL_laststatval = proto_perl->Ilaststatval;
11127 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11128 PL_mess_sv = NULL;
bd81e77b
NC
11129
11130 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11131
11132 /* interpreter atexit processing */
11133 PL_exitlistlen = proto_perl->Iexitlistlen;
11134 if (PL_exitlistlen) {
11135 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11136 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11137 }
bd81e77b
NC
11138 else
11139 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11140
11141 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11142 if (PL_my_cxt_size) {
f16dd614
DM
11143 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11144 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11145 }
11146 else
11147 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11148 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11149 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11150 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11151
11152 PL_profiledata = NULL;
11153 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11154 /* PL_rsfp_filters entries have fake IoDIRP() */
11155 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11156
bd81e77b 11157 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11158
bd81e77b 11159 PAD_CLONE_VARS(proto_perl, param);
9660f481 11160
bd81e77b
NC
11161#ifdef HAVE_INTERP_INTERN
11162 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11163#endif
645c22ef 11164
bd81e77b
NC
11165 /* more statics moved here */
11166 PL_generation = proto_perl->Igeneration;
11167 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11168
bd81e77b
NC
11169 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11170 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11171
bd81e77b
NC
11172 PL_uid = proto_perl->Iuid;
11173 PL_euid = proto_perl->Ieuid;
11174 PL_gid = proto_perl->Igid;
11175 PL_egid = proto_perl->Iegid;
11176 PL_nomemok = proto_perl->Inomemok;
11177 PL_an = proto_perl->Ian;
11178 PL_evalseq = proto_perl->Ievalseq;
11179 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11180 PL_origalen = proto_perl->Iorigalen;
11181#ifdef PERL_USES_PL_PIDSTATUS
11182 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11183#endif
11184 PL_osname = SAVEPV(proto_perl->Iosname);
11185 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11186
bd81e77b 11187 PL_runops = proto_perl->Irunops;
6a78b4db 11188
bd81e77b 11189 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11190
bd81e77b
NC
11191#ifdef CSH
11192 PL_cshlen = proto_perl->Icshlen;
11193 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11194#endif
645c22ef 11195
bd81e77b
NC
11196 PL_lex_state = proto_perl->Ilex_state;
11197 PL_lex_defer = proto_perl->Ilex_defer;
11198 PL_lex_expect = proto_perl->Ilex_expect;
11199 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11200 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11201 PL_lex_starts = proto_perl->Ilex_starts;
11202 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11203 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11204 PL_lex_op = proto_perl->Ilex_op;
11205 PL_lex_inpat = proto_perl->Ilex_inpat;
11206 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11207 PL_lex_brackets = proto_perl->Ilex_brackets;
11208 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11209 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11210 PL_lex_casemods = proto_perl->Ilex_casemods;
11211 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11212 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11213
5db06880
NC
11214#ifdef PERL_MAD
11215 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11216 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11217 PL_realtokenstart = proto_perl->Irealtokenstart;
11218 PL_faketokens = proto_perl->Ifaketokens;
11219 PL_thismad = proto_perl->Ithismad;
11220 PL_thistoken = proto_perl->Ithistoken;
11221 PL_thisopen = proto_perl->Ithisopen;
11222 PL_thisstuff = proto_perl->Ithisstuff;
11223 PL_thisclose = proto_perl->Ithisclose;
11224 PL_thiswhite = proto_perl->Ithiswhite;
11225 PL_nextwhite = proto_perl->Inextwhite;
11226 PL_skipwhite = proto_perl->Iskipwhite;
11227 PL_endwhite = proto_perl->Iendwhite;
11228 PL_curforce = proto_perl->Icurforce;
5db06880 11229#else
bd81e77b
NC
11230 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11231 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11232 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11233#endif
c43294b8 11234
bd81e77b
NC
11235 /* XXX This is probably masking the deeper issue of why
11236 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11237 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11238 * (A little debugging with a watchpoint on it may help.)
11239 */
11240 if (SvANY(proto_perl->Ilinestr)) {
11241 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11242 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11243 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11244 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11245 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11246 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11247 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11248 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11249 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11250 }
11251 else {
561b68a9 11252 PL_linestr = newSV(79);
bd81e77b
NC
11253 sv_upgrade(PL_linestr,SVt_PVIV);
11254 sv_setpvn(PL_linestr,"",0);
11255 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11256 }
11257 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11258 PL_pending_ident = proto_perl->Ipending_ident;
11259 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11260
11261 PL_expect = proto_perl->Iexpect;
11262
11263 PL_multi_start = proto_perl->Imulti_start;
11264 PL_multi_end = proto_perl->Imulti_end;
11265 PL_multi_open = proto_perl->Imulti_open;
11266 PL_multi_close = proto_perl->Imulti_close;
11267
11268 PL_error_count = proto_perl->Ierror_count;
11269 PL_subline = proto_perl->Isubline;
11270 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11271
bd81e77b
NC
11272 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11273 if (SvANY(proto_perl->Ilinestr)) {
11274 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11275 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11276 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11277 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11278 PL_last_lop_op = proto_perl->Ilast_lop_op;
11279 }
11280 else {
11281 PL_last_uni = SvPVX(PL_linestr);
11282 PL_last_lop = SvPVX(PL_linestr);
11283 PL_last_lop_op = 0;
11284 }
11285 PL_in_my = proto_perl->Iin_my;
11286 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11287#ifdef FCRYPT
11288 PL_cryptseen = proto_perl->Icryptseen;
11289#endif
1d7c1841 11290
bd81e77b 11291 PL_hints = proto_perl->Ihints;
1d7c1841 11292
bd81e77b 11293 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11294
bd81e77b
NC
11295#ifdef USE_LOCALE_COLLATE
11296 PL_collation_ix = proto_perl->Icollation_ix;
11297 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11298 PL_collation_standard = proto_perl->Icollation_standard;
11299 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11300 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11301#endif /* USE_LOCALE_COLLATE */
1d7c1841 11302
bd81e77b
NC
11303#ifdef USE_LOCALE_NUMERIC
11304 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11305 PL_numeric_standard = proto_perl->Inumeric_standard;
11306 PL_numeric_local = proto_perl->Inumeric_local;
11307 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11308#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11309
bd81e77b
NC
11310 /* utf8 character classes */
11311 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11312 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11313 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11314 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11315 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11316 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11317 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11318 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11319 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11320 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11321 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11322 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11323 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11324 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11325 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11326 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11327 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11328 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11329 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11330 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11331
bd81e77b
NC
11332 /* Did the locale setup indicate UTF-8? */
11333 PL_utf8locale = proto_perl->Iutf8locale;
11334 /* Unicode features (see perlrun/-C) */
11335 PL_unicode = proto_perl->Iunicode;
1d7c1841 11336
bd81e77b
NC
11337 /* Pre-5.8 signals control */
11338 PL_signals = proto_perl->Isignals;
1d7c1841 11339
bd81e77b
NC
11340 /* times() ticks per second */
11341 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11342
bd81e77b
NC
11343 /* Recursion stopper for PerlIO_find_layer */
11344 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11345
bd81e77b
NC
11346 /* sort() routine */
11347 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11348
bd81e77b
NC
11349 /* Not really needed/useful since the reenrant_retint is "volatile",
11350 * but do it for consistency's sake. */
11351 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11352
bd81e77b
NC
11353 /* Hooks to shared SVs and locks. */
11354 PL_sharehook = proto_perl->Isharehook;
11355 PL_lockhook = proto_perl->Ilockhook;
11356 PL_unlockhook = proto_perl->Iunlockhook;
11357 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11358
bd81e77b
NC
11359 PL_runops_std = proto_perl->Irunops_std;
11360 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11361
bd81e77b
NC
11362#ifdef THREADS_HAVE_PIDS
11363 PL_ppid = proto_perl->Ippid;
11364#endif
1d7c1841 11365
bd81e77b 11366 /* swatch cache */
5c284bb0 11367 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11368 PL_last_swash_klen = 0;
11369 PL_last_swash_key[0]= '\0';
11370 PL_last_swash_tmps = (U8*)NULL;
11371 PL_last_swash_slen = 0;
1d7c1841 11372
bd81e77b
NC
11373 PL_glob_index = proto_perl->Iglob_index;
11374 PL_srand_called = proto_perl->Isrand_called;
11375 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11376 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11377
bd81e77b
NC
11378 if (proto_perl->Ipsig_pend) {
11379 Newxz(PL_psig_pend, SIG_SIZE, int);
11380 }
11381 else {
11382 PL_psig_pend = (int*)NULL;
11383 }
05ec9bb3 11384
bd81e77b
NC
11385 if (proto_perl->Ipsig_ptr) {
11386 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11387 Newxz(PL_psig_name, SIG_SIZE, SV*);
11388 for (i = 1; i < SIG_SIZE; i++) {
11389 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11390 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11391 }
11392 }
11393 else {
11394 PL_psig_ptr = (SV**)NULL;
11395 PL_psig_name = (SV**)NULL;
11396 }
05ec9bb3 11397
bd81e77b 11398 /* thrdvar.h stuff */
1d7c1841 11399
bd81e77b
NC
11400 if (flags & CLONEf_COPY_STACKS) {
11401 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11402 PL_tmps_ix = proto_perl->Ttmps_ix;
11403 PL_tmps_max = proto_perl->Ttmps_max;
11404 PL_tmps_floor = proto_perl->Ttmps_floor;
11405 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11406 i = 0;
11407 while (i <= PL_tmps_ix) {
11408 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11409 ++i;
11410 }
d2d73c3e 11411
bd81e77b
NC
11412 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11413 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11414 Newxz(PL_markstack, i, I32);
11415 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11416 - proto_perl->Tmarkstack);
11417 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11418 - proto_perl->Tmarkstack);
11419 Copy(proto_perl->Tmarkstack, PL_markstack,
11420 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11421
bd81e77b
NC
11422 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11423 * NOTE: unlike the others! */
11424 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11425 PL_scopestack_max = proto_perl->Tscopestack_max;
11426 Newxz(PL_scopestack, PL_scopestack_max, I32);
11427 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11428
bd81e77b
NC
11429 /* NOTE: si_dup() looks at PL_markstack */
11430 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11431
bd81e77b
NC
11432 /* PL_curstack = PL_curstackinfo->si_stack; */
11433 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11434 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11435
bd81e77b
NC
11436 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11437 PL_stack_base = AvARRAY(PL_curstack);
11438 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11439 - proto_perl->Tstack_base);
11440 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11441
bd81e77b
NC
11442 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11443 * NOTE: unlike the others! */
11444 PL_savestack_ix = proto_perl->Tsavestack_ix;
11445 PL_savestack_max = proto_perl->Tsavestack_max;
11446 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11447 PL_savestack = ss_dup(proto_perl, param);
11448 }
11449 else {
11450 init_stacks();
11451 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11452
11453 /* although we're not duplicating the tmps stack, we should still
11454 * add entries for any SVs on the tmps stack that got cloned by a
11455 * non-refcount means (eg a temp in @_); otherwise they will be
11456 * orphaned
11457 */
11458 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11459 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11460 proto_perl->Ttmps_stack[i]);
11461 if (nsv && !SvREFCNT(nsv)) {
11462 EXTEND_MORTAL(1);
b37c2d43 11463 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11464 }
11465 }
bd81e77b 11466 }
1d7c1841 11467
bd81e77b
NC
11468 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11469 PL_top_env = &PL_start_env;
1d7c1841 11470
bd81e77b 11471 PL_op = proto_perl->Top;
4a4c6fe3 11472
a0714e2c 11473 PL_Sv = NULL;
bd81e77b
NC
11474 PL_Xpv = (XPV*)NULL;
11475 PL_na = proto_perl->Tna;
1fcf4c12 11476
bd81e77b
NC
11477 PL_statbuf = proto_perl->Tstatbuf;
11478 PL_statcache = proto_perl->Tstatcache;
11479 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11480 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11481#ifdef HAS_TIMES
11482 PL_timesbuf = proto_perl->Ttimesbuf;
11483#endif
1d7c1841 11484
bd81e77b
NC
11485 PL_tainted = proto_perl->Ttainted;
11486 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11487 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11488 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11489 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11490 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11491 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11492 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11493 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11494 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11495
bd81e77b
NC
11496 PL_restartop = proto_perl->Trestartop;
11497 PL_in_eval = proto_perl->Tin_eval;
11498 PL_delaymagic = proto_perl->Tdelaymagic;
11499 PL_dirty = proto_perl->Tdirty;
11500 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11501
bd81e77b 11502 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11503 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11504 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11505 PL_lastgotoprobe = NULL;
bd81e77b 11506 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11507
bd81e77b
NC
11508 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11509 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11510 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11511 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11512 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11513 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11514
bd81e77b 11515 /* regex stuff */
1d7c1841 11516
bd81e77b
NC
11517 PL_screamfirst = NULL;
11518 PL_screamnext = NULL;
11519 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11520 PL_lastscream = NULL;
1d7c1841 11521
bd81e77b 11522 PL_watchaddr = NULL;
bd61b366 11523 PL_watchok = NULL;
1d7c1841 11524
bd81e77b 11525 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11526 PL_colorset = 0; /* reinits PL_colors[] */
11527 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11528
bd81e77b
NC
11529 /* RE engine - function pointers */
11530 PL_regcompp = proto_perl->Tregcompp;
11531 PL_regexecp = proto_perl->Tregexecp;
11532 PL_regint_start = proto_perl->Tregint_start;
11533 PL_regint_string = proto_perl->Tregint_string;
11534 PL_regfree = proto_perl->Tregfree;
46ab3289 11535 Zero(&PL_reg_state, 1, struct re_save_state);
bd81e77b 11536 PL_reginterp_cnt = 0;
5d9a96ca 11537 PL_regmatch_slab = NULL;
1d7c1841 11538
bd81e77b
NC
11539 /* Pluggable optimizer */
11540 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11541
bd81e77b 11542 PL_stashcache = newHV();
1d7c1841 11543
bd81e77b
NC
11544 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11545 ptr_table_free(PL_ptr_table);
11546 PL_ptr_table = NULL;
11547 }
1d7c1841 11548
bd81e77b
NC
11549 /* Call the ->CLONE method, if it exists, for each of the stashes
11550 identified by sv_dup() above.
11551 */
11552 while(av_len(param->stashes) != -1) {
11553 HV* const stash = (HV*) av_shift(param->stashes);
11554 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11555 if (cloner && GvCV(cloner)) {
11556 dSP;
11557 ENTER;
11558 SAVETMPS;
11559 PUSHMARK(SP);
11560 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11561 PUTBACK;
11562 call_sv((SV*)GvCV(cloner), G_DISCARD);
11563 FREETMPS;
11564 LEAVE;
11565 }
1d7c1841 11566 }
1d7c1841 11567
bd81e77b 11568 SvREFCNT_dec(param->stashes);
1d7c1841 11569
bd81e77b
NC
11570 /* orphaned? eg threads->new inside BEGIN or use */
11571 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11572 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11573 SAVEFREESV(PL_compcv);
11574 }
dd2155a4 11575
bd81e77b
NC
11576 return my_perl;
11577}
1d7c1841 11578
bd81e77b 11579#endif /* USE_ITHREADS */
1d7c1841 11580
bd81e77b
NC
11581/*
11582=head1 Unicode Support
1d7c1841 11583
bd81e77b 11584=for apidoc sv_recode_to_utf8
1d7c1841 11585
bd81e77b
NC
11586The encoding is assumed to be an Encode object, on entry the PV
11587of the sv is assumed to be octets in that encoding, and the sv
11588will be converted into Unicode (and UTF-8).
1d7c1841 11589
bd81e77b
NC
11590If the sv already is UTF-8 (or if it is not POK), or if the encoding
11591is not a reference, nothing is done to the sv. If the encoding is not
11592an C<Encode::XS> Encoding object, bad things will happen.
11593(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11594
bd81e77b 11595The PV of the sv is returned.
1d7c1841 11596
bd81e77b 11597=cut */
1d7c1841 11598
bd81e77b
NC
11599char *
11600Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11601{
11602 dVAR;
11603 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11604 SV *uni;
11605 STRLEN len;
11606 const char *s;
11607 dSP;
11608 ENTER;
11609 SAVETMPS;
11610 save_re_context();
11611 PUSHMARK(sp);
11612 EXTEND(SP, 3);
11613 XPUSHs(encoding);
11614 XPUSHs(sv);
11615/*
11616 NI-S 2002/07/09
11617 Passing sv_yes is wrong - it needs to be or'ed set of constants
11618 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11619 remove converted chars from source.
1d7c1841 11620
bd81e77b 11621 Both will default the value - let them.
1d7c1841 11622
bd81e77b
NC
11623 XPUSHs(&PL_sv_yes);
11624*/
11625 PUTBACK;
11626 call_method("decode", G_SCALAR);
11627 SPAGAIN;
11628 uni = POPs;
11629 PUTBACK;
11630 s = SvPV_const(uni, len);
11631 if (s != SvPVX_const(sv)) {
11632 SvGROW(sv, len + 1);
11633 Move(s, SvPVX(sv), len + 1, char);
11634 SvCUR_set(sv, len);
11635 }
11636 FREETMPS;
11637 LEAVE;
11638 SvUTF8_on(sv);
11639 return SvPVX(sv);
389edf32 11640 }
bd81e77b
NC
11641 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11642}
1d7c1841 11643
bd81e77b
NC
11644/*
11645=for apidoc sv_cat_decode
1d7c1841 11646
bd81e77b
NC
11647The encoding is assumed to be an Encode object, the PV of the ssv is
11648assumed to be octets in that encoding and decoding the input starts
11649from the position which (PV + *offset) pointed to. The dsv will be
11650concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11651when the string tstr appears in decoding output or the input ends on
11652the PV of the ssv. The value which the offset points will be modified
11653to the last input position on the ssv.
1d7c1841 11654
bd81e77b 11655Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11656
bd81e77b
NC
11657=cut */
11658
11659bool
11660Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11661 SV *ssv, int *offset, char *tstr, int tlen)
11662{
11663 dVAR;
11664 bool ret = FALSE;
11665 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11666 SV *offsv;
11667 dSP;
11668 ENTER;
11669 SAVETMPS;
11670 save_re_context();
11671 PUSHMARK(sp);
11672 EXTEND(SP, 6);
11673 XPUSHs(encoding);
11674 XPUSHs(dsv);
11675 XPUSHs(ssv);
11676 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11677 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11678 PUTBACK;
11679 call_method("cat_decode", G_SCALAR);
11680 SPAGAIN;
11681 ret = SvTRUE(TOPs);
11682 *offset = SvIV(offsv);
11683 PUTBACK;
11684 FREETMPS;
11685 LEAVE;
389edf32 11686 }
bd81e77b
NC
11687 else
11688 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11689 return ret;
1d7c1841 11690
bd81e77b 11691}
1d7c1841 11692
bd81e77b
NC
11693/* ---------------------------------------------------------------------
11694 *
11695 * support functions for report_uninit()
11696 */
1d7c1841 11697
bd81e77b
NC
11698/* the maxiumum size of array or hash where we will scan looking
11699 * for the undefined element that triggered the warning */
1d7c1841 11700
bd81e77b 11701#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11702
bd81e77b
NC
11703/* Look for an entry in the hash whose value has the same SV as val;
11704 * If so, return a mortal copy of the key. */
1d7c1841 11705
bd81e77b
NC
11706STATIC SV*
11707S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11708{
11709 dVAR;
11710 register HE **array;
11711 I32 i;
6c3182a5 11712
bd81e77b
NC
11713 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11714 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11715 return NULL;
6c3182a5 11716
bd81e77b 11717 array = HvARRAY(hv);
6c3182a5 11718
bd81e77b
NC
11719 for (i=HvMAX(hv); i>0; i--) {
11720 register HE *entry;
11721 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11722 if (HeVAL(entry) != val)
11723 continue;
11724 if ( HeVAL(entry) == &PL_sv_undef ||
11725 HeVAL(entry) == &PL_sv_placeholder)
11726 continue;
11727 if (!HeKEY(entry))
a0714e2c 11728 return NULL;
bd81e77b
NC
11729 if (HeKLEN(entry) == HEf_SVKEY)
11730 return sv_mortalcopy(HeKEY_sv(entry));
11731 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11732 }
11733 }
a0714e2c 11734 return NULL;
bd81e77b 11735}
6c3182a5 11736
bd81e77b
NC
11737/* Look for an entry in the array whose value has the same SV as val;
11738 * If so, return the index, otherwise return -1. */
6c3182a5 11739
bd81e77b
NC
11740STATIC I32
11741S_find_array_subscript(pTHX_ AV *av, SV* val)
11742{
97aff369 11743 dVAR;
bd81e77b
NC
11744 SV** svp;
11745 I32 i;
11746 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11747 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11748 return -1;
57c6e6d2 11749
bd81e77b
NC
11750 svp = AvARRAY(av);
11751 for (i=AvFILLp(av); i>=0; i--) {
11752 if (svp[i] == val && svp[i] != &PL_sv_undef)
11753 return i;
11754 }
11755 return -1;
11756}
15a5279a 11757
bd81e77b
NC
11758/* S_varname(): return the name of a variable, optionally with a subscript.
11759 * If gv is non-zero, use the name of that global, along with gvtype (one
11760 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11761 * targ. Depending on the value of the subscript_type flag, return:
11762 */
bce260cd 11763
bd81e77b
NC
11764#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11765#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11766#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11767#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11768
bd81e77b
NC
11769STATIC SV*
11770S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11771 SV* keyname, I32 aindex, int subscript_type)
11772{
1d7c1841 11773
bd81e77b
NC
11774 SV * const name = sv_newmortal();
11775 if (gv) {
11776 char buffer[2];
11777 buffer[0] = gvtype;
11778 buffer[1] = 0;
1d7c1841 11779
bd81e77b 11780 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11781
bd81e77b 11782 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11783
bd81e77b
NC
11784 if ((unsigned int)SvPVX(name)[1] <= 26) {
11785 buffer[0] = '^';
11786 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11787
bd81e77b
NC
11788 /* Swap the 1 unprintable control character for the 2 byte pretty
11789 version - ie substr($name, 1, 1) = $buffer; */
11790 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11791 }
bd81e77b
NC
11792 }
11793 else {
11794 U32 unused;
11795 CV * const cv = find_runcv(&unused);
11796 SV *sv;
11797 AV *av;
1d7c1841 11798
bd81e77b 11799 if (!cv || !CvPADLIST(cv))
a0714e2c 11800 return NULL;
bd81e77b
NC
11801 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11802 sv = *av_fetch(av, targ, FALSE);
11803 /* SvLEN in a pad name is not to be trusted */
11804 sv_setpv(name, SvPV_nolen_const(sv));
11805 }
1d7c1841 11806
bd81e77b 11807 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11808 SV * const sv = newSV(0);
bd81e77b
NC
11809 *SvPVX(name) = '$';
11810 Perl_sv_catpvf(aTHX_ name, "{%s}",
11811 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11812 SvREFCNT_dec(sv);
11813 }
11814 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11815 *SvPVX(name) = '$';
11816 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11817 }
11818 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11819 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11820
bd81e77b
NC
11821 return name;
11822}
1d7c1841 11823
1d7c1841 11824
bd81e77b
NC
11825/*
11826=for apidoc find_uninit_var
1d7c1841 11827
bd81e77b
NC
11828Find the name of the undefined variable (if any) that caused the operator o
11829to issue a "Use of uninitialized value" warning.
11830If match is true, only return a name if it's value matches uninit_sv.
11831So roughly speaking, if a unary operator (such as OP_COS) generates a
11832warning, then following the direct child of the op may yield an
11833OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11834other hand, with OP_ADD there are two branches to follow, so we only print
11835the variable name if we get an exact match.
1d7c1841 11836
bd81e77b 11837The name is returned as a mortal SV.
1d7c1841 11838
bd81e77b
NC
11839Assumes that PL_op is the op that originally triggered the error, and that
11840PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11841
bd81e77b
NC
11842=cut
11843*/
1d7c1841 11844
bd81e77b
NC
11845STATIC SV *
11846S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11847{
11848 dVAR;
11849 SV *sv;
11850 AV *av;
11851 GV *gv;
11852 OP *o, *o2, *kid;
1d7c1841 11853
bd81e77b
NC
11854 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11855 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11856 return NULL;
1d7c1841 11857
bd81e77b 11858 switch (obase->op_type) {
1d7c1841 11859
bd81e77b
NC
11860 case OP_RV2AV:
11861 case OP_RV2HV:
11862 case OP_PADAV:
11863 case OP_PADHV:
11864 {
11865 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11866 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11867 I32 index = 0;
a0714e2c 11868 SV *keysv = NULL;
bd81e77b 11869 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11870
bd81e77b
NC
11871 if (pad) { /* @lex, %lex */
11872 sv = PAD_SVl(obase->op_targ);
a0714e2c 11873 gv = NULL;
bd81e77b
NC
11874 }
11875 else {
11876 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11877 /* @global, %global */
11878 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11879 if (!gv)
11880 break;
11881 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11882 }
11883 else /* @{expr}, %{expr} */
11884 return find_uninit_var(cUNOPx(obase)->op_first,
11885 uninit_sv, match);
11886 }
1d7c1841 11887
bd81e77b
NC
11888 /* attempt to find a match within the aggregate */
11889 if (hash) {
d4c19fe8 11890 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11891 if (keysv)
11892 subscript_type = FUV_SUBSCRIPT_HASH;
11893 }
11894 else {
e15d5972 11895 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11896 if (index >= 0)
11897 subscript_type = FUV_SUBSCRIPT_ARRAY;
11898 }
1d7c1841 11899
bd81e77b
NC
11900 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11901 break;
1d7c1841 11902
bd81e77b
NC
11903 return varname(gv, hash ? '%' : '@', obase->op_targ,
11904 keysv, index, subscript_type);
11905 }
1d7c1841 11906
bd81e77b
NC
11907 case OP_PADSV:
11908 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11909 break;
a0714e2c
SS
11910 return varname(NULL, '$', obase->op_targ,
11911 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11912
bd81e77b
NC
11913 case OP_GVSV:
11914 gv = cGVOPx_gv(obase);
11915 if (!gv || (match && GvSV(gv) != uninit_sv))
11916 break;
a0714e2c 11917 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11918
bd81e77b
NC
11919 case OP_AELEMFAST:
11920 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11921 if (match) {
11922 SV **svp;
11923 av = (AV*)PAD_SV(obase->op_targ);
11924 if (!av || SvRMAGICAL(av))
11925 break;
11926 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11927 if (!svp || *svp != uninit_sv)
11928 break;
11929 }
a0714e2c
SS
11930 return varname(NULL, '$', obase->op_targ,
11931 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11932 }
11933 else {
11934 gv = cGVOPx_gv(obase);
11935 if (!gv)
11936 break;
11937 if (match) {
11938 SV **svp;
11939 av = GvAV(gv);
11940 if (!av || SvRMAGICAL(av))
11941 break;
11942 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11943 if (!svp || *svp != uninit_sv)
11944 break;
11945 }
11946 return varname(gv, '$', 0,
a0714e2c 11947 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11948 }
11949 break;
1d7c1841 11950
bd81e77b
NC
11951 case OP_EXISTS:
11952 o = cUNOPx(obase)->op_first;
11953 if (!o || o->op_type != OP_NULL ||
11954 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11955 break;
11956 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11957
bd81e77b
NC
11958 case OP_AELEM:
11959 case OP_HELEM:
11960 if (PL_op == obase)
11961 /* $a[uninit_expr] or $h{uninit_expr} */
11962 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11963
a0714e2c 11964 gv = NULL;
bd81e77b
NC
11965 o = cBINOPx(obase)->op_first;
11966 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11967
bd81e77b 11968 /* get the av or hv, and optionally the gv */
a0714e2c 11969 sv = NULL;
bd81e77b
NC
11970 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11971 sv = PAD_SV(o->op_targ);
11972 }
11973 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11974 && cUNOPo->op_first->op_type == OP_GV)
11975 {
11976 gv = cGVOPx_gv(cUNOPo->op_first);
11977 if (!gv)
11978 break;
11979 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11980 }
11981 if (!sv)
11982 break;
11983
11984 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11985 /* index is constant */
11986 if (match) {
11987 if (SvMAGICAL(sv))
11988 break;
11989 if (obase->op_type == OP_HELEM) {
11990 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11991 if (!he || HeVAL(he) != uninit_sv)
11992 break;
11993 }
11994 else {
00b6aa41 11995 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11996 if (!svp || *svp != uninit_sv)
11997 break;
11998 }
11999 }
12000 if (obase->op_type == OP_HELEM)
12001 return varname(gv, '%', o->op_targ,
12002 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12003 else
a0714e2c 12004 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 12005 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12006 }
12007 else {
12008 /* index is an expression;
12009 * attempt to find a match within the aggregate */
12010 if (obase->op_type == OP_HELEM) {
d4c19fe8 12011 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
12012 if (keysv)
12013 return varname(gv, '%', o->op_targ,
12014 keysv, 0, FUV_SUBSCRIPT_HASH);
12015 }
12016 else {
d4c19fe8 12017 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
12018 if (index >= 0)
12019 return varname(gv, '@', o->op_targ,
a0714e2c 12020 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12021 }
12022 if (match)
12023 break;
12024 return varname(gv,
12025 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12026 ? '@' : '%',
a0714e2c 12027 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 12028 }
bd81e77b 12029 break;
dc507217 12030
bd81e77b
NC
12031 case OP_AASSIGN:
12032 /* only examine RHS */
12033 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 12034
bd81e77b
NC
12035 case OP_OPEN:
12036 o = cUNOPx(obase)->op_first;
12037 if (o->op_type == OP_PUSHMARK)
12038 o = o->op_sibling;
1d7c1841 12039
bd81e77b
NC
12040 if (!o->op_sibling) {
12041 /* one-arg version of open is highly magical */
a0ae6670 12042
bd81e77b
NC
12043 if (o->op_type == OP_GV) { /* open FOO; */
12044 gv = cGVOPx_gv(o);
12045 if (match && GvSV(gv) != uninit_sv)
12046 break;
12047 return varname(gv, '$', 0,
a0714e2c 12048 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12049 }
12050 /* other possibilities not handled are:
12051 * open $x; or open my $x; should return '${*$x}'
12052 * open expr; should return '$'.expr ideally
12053 */
12054 break;
12055 }
12056 goto do_op;
ccfc67b7 12057
bd81e77b
NC
12058 /* ops where $_ may be an implicit arg */
12059 case OP_TRANS:
12060 case OP_SUBST:
12061 case OP_MATCH:
12062 if ( !(obase->op_flags & OPf_STACKED)) {
12063 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12064 ? PAD_SVl(obase->op_targ)
12065 : DEFSV))
12066 {
12067 sv = sv_newmortal();
12068 sv_setpvn(sv, "$_", 2);
12069 return sv;
12070 }
12071 }
12072 goto do_op;
9f4817db 12073
bd81e77b
NC
12074 case OP_PRTF:
12075 case OP_PRINT:
12076 /* skip filehandle as it can't produce 'undef' warning */
12077 o = cUNOPx(obase)->op_first;
12078 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12079 o = o->op_sibling->op_sibling;
12080 goto do_op2;
9f4817db 12081
9f4817db 12082
bd81e77b
NC
12083 case OP_RV2SV:
12084 case OP_CUSTOM:
12085 case OP_ENTERSUB:
12086 match = 1; /* XS or custom code could trigger random warnings */
12087 goto do_op;
9f4817db 12088
bd81e77b
NC
12089 case OP_SCHOMP:
12090 case OP_CHOMP:
12091 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12092 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12093 /*FALLTHROUGH*/
5d170f3a 12094
bd81e77b
NC
12095 default:
12096 do_op:
12097 if (!(obase->op_flags & OPf_KIDS))
12098 break;
12099 o = cUNOPx(obase)->op_first;
12100
12101 do_op2:
12102 if (!o)
12103 break;
f9893866 12104
bd81e77b
NC
12105 /* if all except one arg are constant, or have no side-effects,
12106 * or are optimized away, then it's unambiguous */
5f66b61c 12107 o2 = NULL;
bd81e77b 12108 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
12109 if (kid) {
12110 const OPCODE type = kid->op_type;
12111 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
12112 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12113 || (type == OP_PUSHMARK)
bd81e77b 12114 )
bd81e77b 12115 continue;
e15d5972 12116 }
bd81e77b 12117 if (o2) { /* more than one found */
5f66b61c 12118 o2 = NULL;
bd81e77b
NC
12119 break;
12120 }
12121 o2 = kid;
12122 }
12123 if (o2)
12124 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12125
bd81e77b
NC
12126 /* scan all args */
12127 while (o) {
12128 sv = find_uninit_var(o, uninit_sv, 1);
12129 if (sv)
12130 return sv;
12131 o = o->op_sibling;
d0063567 12132 }
bd81e77b 12133 break;
f9893866 12134 }
a0714e2c 12135 return NULL;
9f4817db
JH
12136}
12137
220e2d4e 12138
bd81e77b
NC
12139/*
12140=for apidoc report_uninit
68795e93 12141
bd81e77b 12142Print appropriate "Use of uninitialized variable" warning
220e2d4e 12143
bd81e77b
NC
12144=cut
12145*/
220e2d4e 12146
bd81e77b
NC
12147void
12148Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12149{
97aff369 12150 dVAR;
bd81e77b 12151 if (PL_op) {
a0714e2c 12152 SV* varname = NULL;
bd81e77b
NC
12153 if (uninit_sv) {
12154 varname = find_uninit_var(PL_op, uninit_sv,0);
12155 if (varname)
12156 sv_insert(varname, 0, 0, " ", 1);
12157 }
12158 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12159 varname ? SvPV_nolen_const(varname) : "",
12160 " in ", OP_DESC(PL_op));
220e2d4e 12161 }
a73e8557 12162 else
bd81e77b
NC
12163 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12164 "", "", "");
220e2d4e 12165}
f9893866 12166
241d1a3b
NC
12167/*
12168 * Local variables:
12169 * c-indentation-style: bsd
12170 * c-basic-offset: 4
12171 * indent-tabs-mode: t
12172 * End:
12173 *
37442d52
RGS
12174 * ex: set ts=8 sts=4 sw=4 noet:
12175 */