This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
True consting goodness in regexes
[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 */
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
JH
43#else
44#define ASSERT_UTF8_CACHE(cache) NOOP
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
180488f8
NC
1720STATIC char *
1721S_glob_2inpuv(pTHX_ GV *gv, STRLEN *len, bool want_number)
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
1732 if (want_number) {
1733 /* We know that all GVs stringify to something that is not-a-number,
1734 so no need to test that. */
1735 if (ckWARN(WARN_NUMERIC))
1736 not_a_number(buffer);
1737 /* We just want something true to return, so that S_sv_2iuv_common
1738 can tail call us and return true. */
1739 return (char *) 1;
1740 } else {
1741 return SvPV(buffer, *len);
1742 }
1743}
1744
25da4f38
IZ
1745/* Actually, ISO C leaves conversion of UV to IV undefined, but
1746 until proven guilty, assume that things are not that bad... */
1747
645c22ef
DM
1748/*
1749 NV_PRESERVES_UV:
1750
1751 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1752 an IV (an assumption perl has been based on to date) it becomes necessary
1753 to remove the assumption that the NV always carries enough precision to
1754 recreate the IV whenever needed, and that the NV is the canonical form.
1755 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1756 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1757 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1758 1) to distinguish between IV/UV/NV slots that have cached a valid
1759 conversion where precision was lost and IV/UV/NV slots that have a
1760 valid conversion which has lost no precision
645c22ef 1761 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1762 would lose precision, the precise conversion (or differently
1763 imprecise conversion) is also performed and cached, to prevent
1764 requests for different numeric formats on the same SV causing
1765 lossy conversion chains. (lossless conversion chains are perfectly
1766 acceptable (still))
1767
1768
1769 flags are used:
1770 SvIOKp is true if the IV slot contains a valid value
1771 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1772 SvNOKp is true if the NV slot contains a valid value
1773 SvNOK is true only if the NV value is accurate
1774
1775 so
645c22ef 1776 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1777 IV(or UV) would lose accuracy over a direct conversion from PV to
1778 IV(or UV). If it would, cache both conversions, return NV, but mark
1779 SV as IOK NOKp (ie not NOK).
1780
645c22ef 1781 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1782 NV would lose accuracy over a direct conversion from PV to NV. If it
1783 would, cache both conversions, flag similarly.
1784
1785 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1786 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1787 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1788 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1789 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1790
645c22ef
DM
1791 The benefit of this is that operations such as pp_add know that if
1792 SvIOK is true for both left and right operands, then integer addition
1793 can be used instead of floating point (for cases where the result won't
1794 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1795 loss of precision compared with integer addition.
1796
1797 * making IV and NV equal status should make maths accurate on 64 bit
1798 platforms
1799 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1800 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1801 looking for SvIOK and checking for overflow will not outweigh the
1802 fp to integer speedup)
1803 * will slow down integer operations (callers of SvIV) on "inaccurate"
1804 values, as the change from SvIOK to SvIOKp will cause a call into
1805 sv_2iv each time rather than a macro access direct to the IV slot
1806 * should speed up number->string conversion on integers as IV is
645c22ef 1807 favoured when IV and NV are equally accurate
28e5dec8
JH
1808
1809 ####################################################################
645c22ef
DM
1810 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1811 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1812 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1813 ####################################################################
1814
645c22ef 1815 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1816 performance ratio.
1817*/
1818
1819#ifndef NV_PRESERVES_UV
645c22ef
DM
1820# define IS_NUMBER_UNDERFLOW_IV 1
1821# define IS_NUMBER_UNDERFLOW_UV 2
1822# define IS_NUMBER_IV_AND_UV 2
1823# define IS_NUMBER_OVERFLOW_IV 4
1824# define IS_NUMBER_OVERFLOW_UV 5
1825
1826/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1827
1828/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1829STATIC int
645c22ef 1830S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1831{
97aff369 1832 dVAR;
3f7c398e 1833 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
1834 if (SvNVX(sv) < (NV)IV_MIN) {
1835 (void)SvIOKp_on(sv);
1836 (void)SvNOK_on(sv);
45977657 1837 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1838 return IS_NUMBER_UNDERFLOW_IV;
1839 }
1840 if (SvNVX(sv) > (NV)UV_MAX) {
1841 (void)SvIOKp_on(sv);
1842 (void)SvNOK_on(sv);
1843 SvIsUV_on(sv);
607fa7f2 1844 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1845 return IS_NUMBER_OVERFLOW_UV;
1846 }
c2988b20
NC
1847 (void)SvIOKp_on(sv);
1848 (void)SvNOK_on(sv);
1849 /* Can't use strtol etc to convert this string. (See truth table in
1850 sv_2iv */
1851 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1852 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1853 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1854 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1855 } else {
1856 /* Integer is imprecise. NOK, IOKp */
1857 }
1858 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1859 }
1860 SvIsUV_on(sv);
607fa7f2 1861 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1862 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1863 if (SvUVX(sv) == UV_MAX) {
1864 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1865 possibly be preserved by NV. Hence, it must be overflow.
1866 NOK, IOKp */
1867 return IS_NUMBER_OVERFLOW_UV;
1868 }
1869 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1870 } else {
1871 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1872 }
c2988b20 1873 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1874}
645c22ef
DM
1875#endif /* !NV_PRESERVES_UV*/
1876
af359546
NC
1877STATIC bool
1878S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1879 dVAR;
af359546 1880 if (SvNOKp(sv)) {
28e5dec8
JH
1881 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1882 * without also getting a cached IV/UV from it at the same time
1883 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1884 * IV or UV at same time to avoid this. */
1885 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1886
1887 if (SvTYPE(sv) == SVt_NV)
1888 sv_upgrade(sv, SVt_PVNV);
1889
28e5dec8
JH
1890 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1891 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1892 certainly cast into the IV range at IV_MAX, whereas the correct
1893 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1894 cases go to UV */
cab190d4
JD
1895#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1896 if (Perl_isnan(SvNVX(sv))) {
1897 SvUV_set(sv, 0);
1898 SvIsUV_on(sv);
fdbe6d7c 1899 return FALSE;
cab190d4 1900 }
cab190d4 1901#endif
28e5dec8 1902 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1903 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1904 if (SvNVX(sv) == (NV) SvIVX(sv)
1905#ifndef NV_PRESERVES_UV
1906 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1907 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1908 /* Don't flag it as "accurately an integer" if the number
1909 came from a (by definition imprecise) NV operation, and
1910 we're outside the range of NV integer precision */
1911#endif
1912 ) {
1913 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1914 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1915 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1916 PTR2UV(sv),
1917 SvNVX(sv),
1918 SvIVX(sv)));
1919
1920 } else {
1921 /* IV not precise. No need to convert from PV, as NV
1922 conversion would already have cached IV if it detected
1923 that PV->IV would be better than PV->NV->IV
1924 flags already correct - don't set public IOK. */
1925 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1926 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1927 PTR2UV(sv),
1928 SvNVX(sv),
1929 SvIVX(sv)));
1930 }
1931 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1932 but the cast (NV)IV_MIN rounds to a the value less (more
1933 negative) than IV_MIN which happens to be equal to SvNVX ??
1934 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1935 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1936 (NV)UVX == NVX are both true, but the values differ. :-(
1937 Hopefully for 2s complement IV_MIN is something like
1938 0x8000000000000000 which will be exact. NWC */
d460ef45 1939 }
25da4f38 1940 else {
607fa7f2 1941 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1942 if (
1943 (SvNVX(sv) == (NV) SvUVX(sv))
1944#ifndef NV_PRESERVES_UV
1945 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1946 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1947 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1948 /* Don't flag it as "accurately an integer" if the number
1949 came from a (by definition imprecise) NV operation, and
1950 we're outside the range of NV integer precision */
1951#endif
1952 )
1953 SvIOK_on(sv);
25da4f38 1954 SvIsUV_on(sv);
1c846c1f 1955 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1956 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1957 PTR2UV(sv),
57def98f
JH
1958 SvUVX(sv),
1959 SvUVX(sv)));
25da4f38 1960 }
748a9306
LW
1961 }
1962 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1963 UV value;
504618e9 1964 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1965 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1966 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1967 the same as the direct translation of the initial string
1968 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1969 be careful to ensure that the value with the .456 is around if the
1970 NV value is requested in the future).
1c846c1f 1971
af359546 1972 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1973 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1974 cache the NV if we are sure it's not needed.
25da4f38 1975 */
16b7a9a4 1976
c2988b20
NC
1977 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1978 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1979 == IS_NUMBER_IN_UV) {
5e045b90 1980 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1981 if (SvTYPE(sv) < SVt_PVIV)
1982 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1983 (void)SvIOK_on(sv);
c2988b20
NC
1984 } else if (SvTYPE(sv) < SVt_PVNV)
1985 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1986
f2524eef 1987 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
1988 we aren't going to call atof() below. If NVs don't preserve UVs
1989 then the value returned may have more precision than atof() will
1990 return, even though value isn't perfectly accurate. */
1991 if ((numtype & (IS_NUMBER_IN_UV
1992#ifdef NV_PRESERVES_UV
1993 | IS_NUMBER_NOT_INT
1994#endif
1995 )) == IS_NUMBER_IN_UV) {
1996 /* This won't turn off the public IOK flag if it was set above */
1997 (void)SvIOKp_on(sv);
1998
1999 if (!(numtype & IS_NUMBER_NEG)) {
2000 /* positive */;
2001 if (value <= (UV)IV_MAX) {
45977657 2002 SvIV_set(sv, (IV)value);
c2988b20 2003 } else {
af359546 2004 /* it didn't overflow, and it was positive. */
607fa7f2 2005 SvUV_set(sv, value);
c2988b20
NC
2006 SvIsUV_on(sv);
2007 }
2008 } else {
2009 /* 2s complement assumption */
2010 if (value <= (UV)IV_MIN) {
45977657 2011 SvIV_set(sv, -(IV)value);
c2988b20
NC
2012 } else {
2013 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2014 I'm assuming it will be rare. */
c2988b20
NC
2015 if (SvTYPE(sv) < SVt_PVNV)
2016 sv_upgrade(sv, SVt_PVNV);
2017 SvNOK_on(sv);
2018 SvIOK_off(sv);
2019 SvIOKp_on(sv);
9d6ce603 2020 SvNV_set(sv, -(NV)value);
45977657 2021 SvIV_set(sv, IV_MIN);
c2988b20
NC
2022 }
2023 }
2024 }
2025 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2026 will be in the previous block to set the IV slot, and the next
2027 block to set the NV slot. So no else here. */
2028
2029 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2030 != IS_NUMBER_IN_UV) {
2031 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2032 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2033
c2988b20
NC
2034 if (! numtype && ckWARN(WARN_NUMERIC))
2035 not_a_number(sv);
28e5dec8 2036
65202027 2037#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2038 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2039 PTR2UV(sv), SvNVX(sv)));
65202027 2040#else
1779d84d 2041 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2042 PTR2UV(sv), SvNVX(sv)));
65202027 2043#endif
28e5dec8 2044
28e5dec8 2045#ifdef NV_PRESERVES_UV
af359546
NC
2046 (void)SvIOKp_on(sv);
2047 (void)SvNOK_on(sv);
2048 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2049 SvIV_set(sv, I_V(SvNVX(sv)));
2050 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2051 SvIOK_on(sv);
2052 } else {
bb263b4e 2053 /*EMPTY*/; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2054 }
2055 /* UV will not work better than IV */
2056 } else {
2057 if (SvNVX(sv) > (NV)UV_MAX) {
2058 SvIsUV_on(sv);
2059 /* Integer is inaccurate. NOK, IOKp, is UV */
2060 SvUV_set(sv, UV_MAX);
af359546
NC
2061 } else {
2062 SvUV_set(sv, U_V(SvNVX(sv)));
2063 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2064 NV preservse UV so can do correct comparison. */
2065 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2066 SvIOK_on(sv);
af359546 2067 } else {
bb263b4e 2068 /*EMPTY*/; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2069 }
2070 }
4b0c9573 2071 SvIsUV_on(sv);
af359546 2072 }
28e5dec8 2073#else /* NV_PRESERVES_UV */
c2988b20
NC
2074 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2075 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2076 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2077 grok_number above. The NV slot has just been set using
2078 Atof. */
560b0c46 2079 SvNOK_on(sv);
c2988b20
NC
2080 assert (SvIOKp(sv));
2081 } else {
2082 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2083 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2084 /* Small enough to preserve all bits. */
2085 (void)SvIOKp_on(sv);
2086 SvNOK_on(sv);
45977657 2087 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2088 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2089 SvIOK_on(sv);
2090 /* Assumption: first non-preserved integer is < IV_MAX,
2091 this NV is in the preserved range, therefore: */
2092 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2093 < (UV)IV_MAX)) {
32fdb065 2094 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
2095 }
2096 } else {
2097 /* IN_UV NOT_INT
2098 0 0 already failed to read UV.
2099 0 1 already failed to read UV.
2100 1 0 you won't get here in this case. IV/UV
2101 slot set, public IOK, Atof() unneeded.
2102 1 1 already read UV.
2103 so there's no point in sv_2iuv_non_preserve() attempting
2104 to use atol, strtol, strtoul etc. */
40a17c4c 2105 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2106 }
2107 }
28e5dec8 2108#endif /* NV_PRESERVES_UV */
25da4f38 2109 }
af359546
NC
2110 }
2111 else {
f7877b28 2112 if (isGV_with_GP(sv)) {
b21a80ce 2113 return (bool)PTR2IV(glob_2inpuv((GV *)sv, NULL, TRUE));
180488f8 2114 }
180488f8 2115
af359546
NC
2116 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2117 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2118 report_uninit(sv);
2119 }
25da4f38
IZ
2120 if (SvTYPE(sv) < SVt_IV)
2121 /* Typically the caller expects that sv_any is not NULL now. */
2122 sv_upgrade(sv, SVt_IV);
af359546
NC
2123 /* Return 0 from the caller. */
2124 return TRUE;
2125 }
2126 return FALSE;
2127}
2128
2129/*
2130=for apidoc sv_2iv_flags
2131
2132Return the integer value of an SV, doing any necessary string
2133conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2134Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2135
2136=cut
2137*/
2138
2139IV
2140Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2141{
97aff369 2142 dVAR;
af359546 2143 if (!sv)
a0d0e21e 2144 return 0;
af359546
NC
2145 if (SvGMAGICAL(sv)) {
2146 if (flags & SV_GMAGIC)
2147 mg_get(sv);
2148 if (SvIOKp(sv))
2149 return SvIVX(sv);
2150 if (SvNOKp(sv)) {
2151 return I_V(SvNVX(sv));
2152 }
71c558c3
NC
2153 if (SvPOKp(sv) && SvLEN(sv)) {
2154 UV value;
2155 const int numtype
2156 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2157
2158 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2159 == IS_NUMBER_IN_UV) {
2160 /* It's definitely an integer */
2161 if (numtype & IS_NUMBER_NEG) {
2162 if (value < (UV)IV_MIN)
2163 return -(IV)value;
2164 } else {
2165 if (value < (UV)IV_MAX)
2166 return (IV)value;
2167 }
2168 }
2169 if (!numtype) {
2170 if (ckWARN(WARN_NUMERIC))
2171 not_a_number(sv);
2172 }
2173 return I_V(Atof(SvPVX_const(sv)));
2174 }
1c7ff15e
NC
2175 if (SvROK(sv)) {
2176 goto return_rok;
af359546 2177 }
1c7ff15e
NC
2178 assert(SvTYPE(sv) >= SVt_PVMG);
2179 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2180 } else if (SvTHINKFIRST(sv)) {
af359546 2181 if (SvROK(sv)) {
1c7ff15e 2182 return_rok:
af359546
NC
2183 if (SvAMAGIC(sv)) {
2184 SV * const tmpstr=AMG_CALLun(sv,numer);
2185 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2186 return SvIV(tmpstr);
2187 }
2188 }
2189 return PTR2IV(SvRV(sv));
2190 }
2191 if (SvIsCOW(sv)) {
2192 sv_force_normal_flags(sv, 0);
2193 }
2194 if (SvREADONLY(sv) && !SvOK(sv)) {
2195 if (ckWARN(WARN_UNINITIALIZED))
2196 report_uninit(sv);
2197 return 0;
2198 }
2199 }
2200 if (!SvIOKp(sv)) {
2201 if (S_sv_2iuv_common(aTHX_ sv))
2202 return 0;
79072805 2203 }
1d7c1841
GS
2204 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2205 PTR2UV(sv),SvIVX(sv)));
25da4f38 2206 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2207}
2208
645c22ef 2209/*
891f9566 2210=for apidoc sv_2uv_flags
645c22ef
DM
2211
2212Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2213conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2214Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2215
2216=cut
2217*/
2218
ff68c719 2219UV
891f9566 2220Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2221{
97aff369 2222 dVAR;
ff68c719 2223 if (!sv)
2224 return 0;
2225 if (SvGMAGICAL(sv)) {
891f9566
YST
2226 if (flags & SV_GMAGIC)
2227 mg_get(sv);
ff68c719 2228 if (SvIOKp(sv))
2229 return SvUVX(sv);
2230 if (SvNOKp(sv))
2231 return U_V(SvNVX(sv));
71c558c3
NC
2232 if (SvPOKp(sv) && SvLEN(sv)) {
2233 UV value;
2234 const int numtype
2235 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2236
2237 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2238 == IS_NUMBER_IN_UV) {
2239 /* It's definitely an integer */
2240 if (!(numtype & IS_NUMBER_NEG))
2241 return value;
2242 }
2243 if (!numtype) {
2244 if (ckWARN(WARN_NUMERIC))
2245 not_a_number(sv);
2246 }
2247 return U_V(Atof(SvPVX_const(sv)));
2248 }
1c7ff15e
NC
2249 if (SvROK(sv)) {
2250 goto return_rok;
3fe9a6f1 2251 }
1c7ff15e
NC
2252 assert(SvTYPE(sv) >= SVt_PVMG);
2253 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2254 } else if (SvTHINKFIRST(sv)) {
ff68c719 2255 if (SvROK(sv)) {
1c7ff15e 2256 return_rok:
deb46114
NC
2257 if (SvAMAGIC(sv)) {
2258 SV *const tmpstr = AMG_CALLun(sv,numer);
2259 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2260 return SvUV(tmpstr);
2261 }
2262 }
2263 return PTR2UV(SvRV(sv));
ff68c719 2264 }
765f542d
NC
2265 if (SvIsCOW(sv)) {
2266 sv_force_normal_flags(sv, 0);
8a818333 2267 }
0336b60e 2268 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2269 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2270 report_uninit(sv);
ff68c719 2271 return 0;
2272 }
2273 }
af359546
NC
2274 if (!SvIOKp(sv)) {
2275 if (S_sv_2iuv_common(aTHX_ sv))
2276 return 0;
ff68c719 2277 }
25da4f38 2278
1d7c1841
GS
2279 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2280 PTR2UV(sv),SvUVX(sv)));
25da4f38 2281 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2282}
2283
645c22ef
DM
2284/*
2285=for apidoc sv_2nv
2286
2287Return the num value of an SV, doing any necessary string or integer
2288conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2289macros.
2290
2291=cut
2292*/
2293
65202027 2294NV
864dbfa3 2295Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2296{
97aff369 2297 dVAR;
79072805
LW
2298 if (!sv)
2299 return 0.0;
8990e307 2300 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2301 mg_get(sv);
2302 if (SvNOKp(sv))
2303 return SvNVX(sv);
0aa395f8 2304 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2305 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2306 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2307 not_a_number(sv);
3f7c398e 2308 return Atof(SvPVX_const(sv));
a0d0e21e 2309 }
25da4f38 2310 if (SvIOKp(sv)) {
1c846c1f 2311 if (SvIsUV(sv))
65202027 2312 return (NV)SvUVX(sv);
25da4f38 2313 else
65202027 2314 return (NV)SvIVX(sv);
47a72cb8
NC
2315 }
2316 if (SvROK(sv)) {
2317 goto return_rok;
2318 }
2319 assert(SvTYPE(sv) >= SVt_PVMG);
2320 /* This falls through to the report_uninit near the end of the
2321 function. */
2322 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2323 if (SvROK(sv)) {
47a72cb8 2324 return_rok:
deb46114
NC
2325 if (SvAMAGIC(sv)) {
2326 SV *const tmpstr = AMG_CALLun(sv,numer);
2327 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2328 return SvNV(tmpstr);
2329 }
2330 }
2331 return PTR2NV(SvRV(sv));
a0d0e21e 2332 }
765f542d
NC
2333 if (SvIsCOW(sv)) {
2334 sv_force_normal_flags(sv, 0);
8a818333 2335 }
0336b60e 2336 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2337 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2338 report_uninit(sv);
ed6116ce
LW
2339 return 0.0;
2340 }
79072805
LW
2341 }
2342 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2343 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2344 sv_upgrade(sv, SVt_NV);
906f284f 2345#ifdef USE_LONG_DOUBLE
097ee67d 2346 DEBUG_c({
f93f4e46 2347 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2348 PerlIO_printf(Perl_debug_log,
2349 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2350 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2351 RESTORE_NUMERIC_LOCAL();
2352 });
65202027 2353#else
572bbb43 2354 DEBUG_c({
f93f4e46 2355 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2356 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2357 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2358 RESTORE_NUMERIC_LOCAL();
2359 });
572bbb43 2360#endif
79072805
LW
2361 }
2362 else if (SvTYPE(sv) < SVt_PVNV)
2363 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2364 if (SvNOKp(sv)) {
2365 return SvNVX(sv);
61604483 2366 }
59d8ce62 2367 if (SvIOKp(sv)) {
9d6ce603 2368 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2369#ifdef NV_PRESERVES_UV
2370 SvNOK_on(sv);
2371#else
2372 /* Only set the public NV OK flag if this NV preserves the IV */
2373 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2374 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2375 : (SvIVX(sv) == I_V(SvNVX(sv))))
2376 SvNOK_on(sv);
2377 else
2378 SvNOKp_on(sv);
2379#endif
93a17b20 2380 }
748a9306 2381 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2382 UV value;
3f7c398e 2383 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2384 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2385 not_a_number(sv);
28e5dec8 2386#ifdef NV_PRESERVES_UV
c2988b20
NC
2387 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2388 == IS_NUMBER_IN_UV) {
5e045b90 2389 /* It's definitely an integer */
9d6ce603 2390 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2391 } else
3f7c398e 2392 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2393 SvNOK_on(sv);
2394#else
3f7c398e 2395 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2396 /* Only set the public NV OK flag if this NV preserves the value in
2397 the PV at least as well as an IV/UV would.
2398 Not sure how to do this 100% reliably. */
2399 /* if that shift count is out of range then Configure's test is
2400 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2401 UV_BITS */
2402 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2403 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2404 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2405 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2406 /* Can't use strtol etc to convert this string, so don't try.
2407 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2408 SvNOK_on(sv);
2409 } else {
2410 /* value has been set. It may not be precise. */
2411 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2412 /* 2s complement assumption for (UV)IV_MIN */
2413 SvNOK_on(sv); /* Integer is too negative. */
2414 } else {
2415 SvNOKp_on(sv);
2416 SvIOKp_on(sv);
6fa402ec 2417
c2988b20 2418 if (numtype & IS_NUMBER_NEG) {
45977657 2419 SvIV_set(sv, -(IV)value);
c2988b20 2420 } else if (value <= (UV)IV_MAX) {
45977657 2421 SvIV_set(sv, (IV)value);
c2988b20 2422 } else {
607fa7f2 2423 SvUV_set(sv, value);
c2988b20
NC
2424 SvIsUV_on(sv);
2425 }
2426
2427 if (numtype & IS_NUMBER_NOT_INT) {
2428 /* I believe that even if the original PV had decimals,
2429 they are lost beyond the limit of the FP precision.
2430 However, neither is canonical, so both only get p
2431 flags. NWC, 2000/11/25 */
2432 /* Both already have p flags, so do nothing */
2433 } else {
66a1b24b 2434 const NV nv = SvNVX(sv);
c2988b20
NC
2435 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2436 if (SvIVX(sv) == I_V(nv)) {
2437 SvNOK_on(sv);
c2988b20 2438 } else {
c2988b20
NC
2439 /* It had no "." so it must be integer. */
2440 }
00b6aa41 2441 SvIOK_on(sv);
c2988b20
NC
2442 } else {
2443 /* between IV_MAX and NV(UV_MAX).
2444 Could be slightly > UV_MAX */
6fa402ec 2445
c2988b20
NC
2446 if (numtype & IS_NUMBER_NOT_INT) {
2447 /* UV and NV both imprecise. */
2448 } else {
66a1b24b 2449 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2450
2451 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2452 SvNOK_on(sv);
c2988b20 2453 }
00b6aa41 2454 SvIOK_on(sv);
c2988b20
NC
2455 }
2456 }
2457 }
2458 }
2459 }
28e5dec8 2460#endif /* NV_PRESERVES_UV */
93a17b20 2461 }
79072805 2462 else {
f7877b28 2463 if (isGV_with_GP(sv)) {
180488f8
NC
2464 glob_2inpuv((GV *)sv, NULL, TRUE);
2465 return 0.0;
2466 }
2467
041457d9 2468 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2469 report_uninit(sv);
7e25a7e9
NC
2470 assert (SvTYPE(sv) >= SVt_NV);
2471 /* Typically the caller expects that sv_any is not NULL now. */
2472 /* XXX Ilya implies that this is a bug in callers that assume this
2473 and ideally should be fixed. */
a0d0e21e 2474 return 0.0;
79072805 2475 }
572bbb43 2476#if defined(USE_LONG_DOUBLE)
097ee67d 2477 DEBUG_c({
f93f4e46 2478 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2479 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2480 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2481 RESTORE_NUMERIC_LOCAL();
2482 });
65202027 2483#else
572bbb43 2484 DEBUG_c({
f93f4e46 2485 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2486 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2487 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2488 RESTORE_NUMERIC_LOCAL();
2489 });
572bbb43 2490#endif
463ee0b2 2491 return SvNVX(sv);
79072805
LW
2492}
2493
645c22ef
DM
2494/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2495 * UV as a string towards the end of buf, and return pointers to start and
2496 * end of it.
2497 *
2498 * We assume that buf is at least TYPE_CHARS(UV) long.
2499 */
2500
864dbfa3 2501static char *
aec46f14 2502S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2503{
25da4f38 2504 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2505 char * const ebuf = ptr;
25da4f38 2506 int sign;
25da4f38
IZ
2507
2508 if (is_uv)
2509 sign = 0;
2510 else if (iv >= 0) {
2511 uv = iv;
2512 sign = 0;
2513 } else {
2514 uv = -iv;
2515 sign = 1;
2516 }
2517 do {
eb160463 2518 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2519 } while (uv /= 10);
2520 if (sign)
2521 *--ptr = '-';
2522 *peob = ebuf;
2523 return ptr;
2524}
2525
9af30d34
NC
2526/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2527 * a regexp to its stringified form.
2528 */
2529
2530static char *
2531S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
97aff369 2532 dVAR;
00b6aa41 2533 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2534
2535 if (!mg->mg_ptr) {
2536 const char *fptr = "msix";
2537 char reflags[6];
2538 char ch;
2539 int left = 0;
2540 int right = 4;
00b6aa41 2541 bool need_newline = 0;
9af30d34
NC
2542 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2543
2544 while((ch = *fptr++)) {
2545 if(reganch & 1) {
2546 reflags[left++] = ch;
2547 }
2548 else {
2549 reflags[right--] = ch;
2550 }
2551 reganch >>= 1;
2552 }
2553 if(left != 4) {
2554 reflags[left] = '-';
2555 left = 5;
2556 }
2557
2558 mg->mg_len = re->prelen + 4 + left;
2559 /*
2560 * If /x was used, we have to worry about a regex ending with a
2561 * comment later being embedded within another regex. If so, we don't
2562 * want this regex's "commentization" to leak out to the right part of
2563 * the enclosing regex, we must cap it with a newline.
2564 *
2565 * So, if /x was used, we scan backwards from the end of the regex. If
2566 * we find a '#' before we find a newline, we need to add a newline
2567 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2568 * we don't need to add anything. -jfriedl
2569 */
2570 if (PMf_EXTENDED & re->reganch) {
2571 const char *endptr = re->precomp + re->prelen;
2572 while (endptr >= re->precomp) {
2573 const char c = *(endptr--);
2574 if (c == '\n')
2575 break; /* don't need another */
2576 if (c == '#') {
2577 /* we end while in a comment, so we need a newline */
2578 mg->mg_len++; /* save space for it */
2579 need_newline = 1; /* note to add it */
2580 break;
2581 }
2582 }
2583 }
2584
2585 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2586 mg->mg_ptr[0] = '(';
2587 mg->mg_ptr[1] = '?';
2588 Copy(reflags, mg->mg_ptr+2, left, char);
2589 *(mg->mg_ptr+left+2) = ':';
2590 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2591 if (need_newline)
2592 mg->mg_ptr[mg->mg_len - 2] = '\n';
2593 mg->mg_ptr[mg->mg_len - 1] = ')';
2594 mg->mg_ptr[mg->mg_len] = 0;
2595 }
2596 PL_reginterp_cnt += re->program[0].next_off;
2597
2598 if (re->reganch & ROPT_UTF8)
2599 SvUTF8_on(sv);
2600 else
2601 SvUTF8_off(sv);
2602 if (lp)
2603 *lp = mg->mg_len;
2604 return mg->mg_ptr;
2605}
2606
645c22ef
DM
2607/*
2608=for apidoc sv_2pv_flags
2609
ff276b08 2610Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2611If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2612if necessary.
2613Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2614usually end up here too.
2615
2616=cut
2617*/
2618
8d6d96c1
HS
2619char *
2620Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2621{
97aff369 2622 dVAR;
79072805 2623 register char *s;
79072805 2624
463ee0b2 2625 if (!sv) {
cdb061a3
NC
2626 if (lp)
2627 *lp = 0;
73d840c0 2628 return (char *)"";
463ee0b2 2629 }
8990e307 2630 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2631 if (flags & SV_GMAGIC)
2632 mg_get(sv);
463ee0b2 2633 if (SvPOKp(sv)) {
cdb061a3
NC
2634 if (lp)
2635 *lp = SvCUR(sv);
10516c54
NC
2636 if (flags & SV_MUTABLE_RETURN)
2637 return SvPVX_mutable(sv);
4d84ee25
NC
2638 if (flags & SV_CONST_RETURN)
2639 return (char *)SvPVX_const(sv);
463ee0b2
LW
2640 return SvPVX(sv);
2641 }
75dfc8ec
NC
2642 if (SvIOKp(sv) || SvNOKp(sv)) {
2643 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2644 STRLEN len;
2645
2646 if (SvIOKp(sv)) {
e8ada2d0
NC
2647 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2648 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2649 } else {
e8ada2d0
NC
2650 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2651 len = strlen(tbuf);
75dfc8ec 2652 }
b5b886f0
NC
2653 assert(!SvROK(sv));
2654 {
75dfc8ec
NC
2655 dVAR;
2656
2657#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2658 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2659 tbuf[0] = '0';
2660 tbuf[1] = 0;
75dfc8ec
NC
2661 len = 1;
2662 }
2663#endif
2664 SvUPGRADE(sv, SVt_PV);
2665 if (lp)
2666 *lp = len;
2667 s = SvGROW_mutable(sv, len + 1);
2668 SvCUR_set(sv, len);
2669 SvPOKp_on(sv);
e8ada2d0 2670 return memcpy(s, tbuf, len + 1);
75dfc8ec 2671 }
463ee0b2 2672 }
1c7ff15e
NC
2673 if (SvROK(sv)) {
2674 goto return_rok;
2675 }
2676 assert(SvTYPE(sv) >= SVt_PVMG);
2677 /* This falls through to the report_uninit near the end of the
2678 function. */
2679 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2680 if (SvROK(sv)) {
1c7ff15e 2681 return_rok:
deb46114
NC
2682 if (SvAMAGIC(sv)) {
2683 SV *const tmpstr = AMG_CALLun(sv,string);
2684 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2685 /* Unwrap this: */
2686 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2687 */
2688
2689 char *pv;
2690 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2691 if (flags & SV_CONST_RETURN) {
2692 pv = (char *) SvPVX_const(tmpstr);
2693 } else {
2694 pv = (flags & SV_MUTABLE_RETURN)
2695 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2696 }
2697 if (lp)
2698 *lp = SvCUR(tmpstr);
50adf7d2 2699 } else {
deb46114 2700 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2701 }
deb46114
NC
2702 if (SvUTF8(tmpstr))
2703 SvUTF8_on(sv);
2704 else
2705 SvUTF8_off(sv);
2706 return pv;
50adf7d2 2707 }
deb46114
NC
2708 }
2709 {
75dfc8ec 2710 SV *tsv;
f9277f47 2711 MAGIC *mg;
d8eae41e
NC
2712 const SV *const referent = (SV*)SvRV(sv);
2713
2714 if (!referent) {
396482e1 2715 tsv = sv_2mortal(newSVpvs("NULLREF"));
042dae7a
NC
2716 } else if (SvTYPE(referent) == SVt_PVMG
2717 && ((SvFLAGS(referent) &
2718 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2719 == (SVs_OBJECT|SVs_SMG))
2720 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
c445ea15 2721 return stringify_regexp(sv, mg, lp);
d8eae41e
NC
2722 } else {
2723 const char *const typestr = sv_reftype(referent, 0);
2724
2725 tsv = sv_newmortal();
2726 if (SvOBJECT(referent)) {
2727 const char *const name = HvNAME_get(SvSTASH(referent));
2728 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2729 name ? name : "__ANON__" , typestr,
2730 PTR2UV(referent));
2731 }
2732 else
2733 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2734 PTR2UV(referent));
c080367d 2735 }
042dae7a
NC
2736 if (lp)
2737 *lp = SvCUR(tsv);
2738 return SvPVX(tsv);
463ee0b2 2739 }
79072805 2740 }
0336b60e 2741 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2742 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2743 report_uninit(sv);
cdb061a3
NC
2744 if (lp)
2745 *lp = 0;
73d840c0 2746 return (char *)"";
79072805 2747 }
79072805 2748 }
28e5dec8
JH
2749 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2750 /* I'm assuming that if both IV and NV are equally valid then
2751 converting the IV is going to be more efficient */
e1ec3a88
AL
2752 const U32 isIOK = SvIOK(sv);
2753 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2754 char buf[TYPE_CHARS(UV)];
2755 char *ebuf, *ptr;
2756
2757 if (SvTYPE(sv) < SVt_PVIV)
2758 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2759 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2760 /* inlined from sv_setpvn */
2761 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2762 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2763 SvCUR_set(sv, ebuf - ptr);
2764 s = SvEND(sv);
2765 *s = '\0';
2766 if (isIOK)
2767 SvIOK_on(sv);
2768 else
2769 SvIOKp_on(sv);
2770 if (isUIOK)
2771 SvIsUV_on(sv);
2772 }
2773 else if (SvNOKp(sv)) {
c81271c3 2774 const int olderrno = errno;
79072805
LW
2775 if (SvTYPE(sv) < SVt_PVNV)
2776 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2777 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2778 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2779 /* some Xenix systems wipe out errno here */
79072805 2780#ifdef apollo
463ee0b2 2781 if (SvNVX(sv) == 0.0)
79072805
LW
2782 (void)strcpy(s,"0");
2783 else
2784#endif /*apollo*/
bbce6d69 2785 {
2d4389e4 2786 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2787 }
79072805 2788 errno = olderrno;
a0d0e21e
LW
2789#ifdef FIXNEGATIVEZERO
2790 if (*s == '-' && s[1] == '0' && !s[2])
2791 strcpy(s,"0");
2792#endif
79072805
LW
2793 while (*s) s++;
2794#ifdef hcx
2795 if (s[-1] == '.')
46fc3d4c 2796 *--s = '\0';
79072805
LW
2797#endif
2798 }
79072805 2799 else {
f7877b28 2800 if (isGV_with_GP(sv)) {
180488f8
NC
2801 return glob_2inpuv((GV *)sv, lp, FALSE);
2802 }
2803
041457d9 2804 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2805 report_uninit(sv);
cdb061a3 2806 if (lp)
00b6aa41 2807 *lp = 0;
25da4f38
IZ
2808 if (SvTYPE(sv) < SVt_PV)
2809 /* Typically the caller expects that sv_any is not NULL now. */
2810 sv_upgrade(sv, SVt_PV);
73d840c0 2811 return (char *)"";
79072805 2812 }
cdb061a3 2813 {
823a54a3 2814 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2815 if (lp)
2816 *lp = len;
2817 SvCUR_set(sv, len);
2818 }
79072805 2819 SvPOK_on(sv);
1d7c1841 2820 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2821 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2822 if (flags & SV_CONST_RETURN)
2823 return (char *)SvPVX_const(sv);
10516c54
NC
2824 if (flags & SV_MUTABLE_RETURN)
2825 return SvPVX_mutable(sv);
463ee0b2
LW
2826 return SvPVX(sv);
2827}
2828
645c22ef 2829/*
6050d10e
JP
2830=for apidoc sv_copypv
2831
2832Copies a stringified representation of the source SV into the
2833destination SV. Automatically performs any necessary mg_get and
54f0641b 2834coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2835UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2836sv_2pv[_flags] but operates directly on an SV instead of just the
2837string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2838would lose the UTF-8'ness of the PV.
2839
2840=cut
2841*/
2842
2843void
2844Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2845{
446eaa42 2846 STRLEN len;
53c1dcc0 2847 const char * const s = SvPV_const(ssv,len);
cb50f42d 2848 sv_setpvn(dsv,s,len);
446eaa42 2849 if (SvUTF8(ssv))
cb50f42d 2850 SvUTF8_on(dsv);
446eaa42 2851 else
cb50f42d 2852 SvUTF8_off(dsv);
6050d10e
JP
2853}
2854
2855/*
645c22ef
DM
2856=for apidoc sv_2pvbyte
2857
2858Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2859to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2860side-effect.
2861
2862Usually accessed via the C<SvPVbyte> macro.
2863
2864=cut
2865*/
2866
7340a771
GS
2867char *
2868Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2869{
0875d2fe 2870 sv_utf8_downgrade(sv,0);
97972285 2871 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2872}
2873
645c22ef 2874/*
035cbb0e
RGS
2875=for apidoc sv_2pvutf8
2876
2877Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2878to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2879
2880Usually accessed via the C<SvPVutf8> macro.
2881
2882=cut
2883*/
645c22ef 2884
7340a771
GS
2885char *
2886Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2887{
035cbb0e
RGS
2888 sv_utf8_upgrade(sv);
2889 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2890}
1c846c1f 2891
7ee2227d 2892
645c22ef
DM
2893/*
2894=for apidoc sv_2bool
2895
2896This function is only called on magical items, and is only used by
8cf8f3d1 2897sv_true() or its macro equivalent.
645c22ef
DM
2898
2899=cut
2900*/
2901
463ee0b2 2902bool
864dbfa3 2903Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2904{
97aff369 2905 dVAR;
5b295bef 2906 SvGETMAGIC(sv);
463ee0b2 2907
a0d0e21e
LW
2908 if (!SvOK(sv))
2909 return 0;
2910 if (SvROK(sv)) {
fabdb6c0
AL
2911 if (SvAMAGIC(sv)) {
2912 SV * const tmpsv = AMG_CALLun(sv,bool_);
2913 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2914 return (bool)SvTRUE(tmpsv);
2915 }
2916 return SvRV(sv) != 0;
a0d0e21e 2917 }
463ee0b2 2918 if (SvPOKp(sv)) {
53c1dcc0
AL
2919 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2920 if (Xpvtmp &&
339049b0 2921 (*sv->sv_u.svu_pv > '0' ||
11343788 2922 Xpvtmp->xpv_cur > 1 ||
339049b0 2923 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2924 return 1;
2925 else
2926 return 0;
2927 }
2928 else {
2929 if (SvIOKp(sv))
2930 return SvIVX(sv) != 0;
2931 else {
2932 if (SvNOKp(sv))
2933 return SvNVX(sv) != 0.0;
180488f8 2934 else {
f7877b28 2935 if (isGV_with_GP(sv))
180488f8
NC
2936 return TRUE;
2937 else
2938 return FALSE;
2939 }
463ee0b2
LW
2940 }
2941 }
79072805
LW
2942}
2943
c461cf8f
JH
2944/*
2945=for apidoc sv_utf8_upgrade
2946
78ea37eb 2947Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2948Forces the SV to string form if it is not already.
4411f3b6
NIS
2949Always sets the SvUTF8 flag to avoid future validity checks even
2950if all the bytes have hibit clear.
c461cf8f 2951
13a6c0e0
JH
2952This is not as a general purpose byte encoding to Unicode interface:
2953use the Encode extension for that.
2954
8d6d96c1
HS
2955=for apidoc sv_utf8_upgrade_flags
2956
78ea37eb 2957Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2958Forces the SV to string form if it is not already.
8d6d96c1
HS
2959Always sets the SvUTF8 flag to avoid future validity checks even
2960if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2961will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2962C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2963
13a6c0e0
JH
2964This is not as a general purpose byte encoding to Unicode interface:
2965use the Encode extension for that.
2966
8d6d96c1
HS
2967=cut
2968*/
2969
2970STRLEN
2971Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2972{
97aff369 2973 dVAR;
808c356f
RGS
2974 if (sv == &PL_sv_undef)
2975 return 0;
e0e62c2a
NIS
2976 if (!SvPOK(sv)) {
2977 STRLEN len = 0;
d52b7888
NC
2978 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2979 (void) sv_2pv_flags(sv,&len, flags);
2980 if (SvUTF8(sv))
2981 return len;
2982 } else {
2983 (void) SvPV_force(sv,len);
2984 }
e0e62c2a 2985 }
4411f3b6 2986
f5cee72b 2987 if (SvUTF8(sv)) {
5fec3b1d 2988 return SvCUR(sv);
f5cee72b 2989 }
5fec3b1d 2990
765f542d
NC
2991 if (SvIsCOW(sv)) {
2992 sv_force_normal_flags(sv, 0);
db42d148
NIS
2993 }
2994
88632417 2995 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2996 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2997 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2998 /* This function could be much more efficient if we
2999 * had a FLAG in SVs to signal if there are any hibit
3000 * chars in the PV. Given that there isn't such a flag
3001 * make the loop as fast as possible. */
00b6aa41 3002 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3003 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3004 const U8 *t = s;
c4e7c712
NC
3005
3006 while (t < e) {
53c1dcc0 3007 const U8 ch = *t++;
00b6aa41
AL
3008 /* Check for hi bit */
3009 if (!NATIVE_IS_INVARIANT(ch)) {
3010 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3011 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3012
3013 SvPV_free(sv); /* No longer using what was there before. */
3014 SvPV_set(sv, (char*)recoded);
3015 SvCUR_set(sv, len - 1);
3016 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3017 break;
00b6aa41 3018 }
c4e7c712
NC
3019 }
3020 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3021 SvUTF8_on(sv);
560a288e 3022 }
4411f3b6 3023 return SvCUR(sv);
560a288e
GS
3024}
3025
c461cf8f
JH
3026/*
3027=for apidoc sv_utf8_downgrade
3028
78ea37eb
TS
3029Attempts to convert the PV of an SV from characters to bytes.
3030If the PV contains a character beyond byte, this conversion will fail;
3031in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3032true, croaks.
3033
13a6c0e0
JH
3034This is not as a general purpose Unicode to byte encoding interface:
3035use the Encode extension for that.
3036
c461cf8f
JH
3037=cut
3038*/
3039
560a288e
GS
3040bool
3041Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3042{
97aff369 3043 dVAR;
78ea37eb 3044 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3045 if (SvCUR(sv)) {
03cfe0ae 3046 U8 *s;
652088fc 3047 STRLEN len;
fa301091 3048
765f542d
NC
3049 if (SvIsCOW(sv)) {
3050 sv_force_normal_flags(sv, 0);
3051 }
03cfe0ae
NIS
3052 s = (U8 *) SvPV(sv, len);
3053 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3054 if (fail_ok)
3055 return FALSE;
3056 else {
3057 if (PL_op)
3058 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3059 OP_DESC(PL_op));
fa301091
JH
3060 else
3061 Perl_croak(aTHX_ "Wide character");
3062 }
4b3603a4 3063 }
b162af07 3064 SvCUR_set(sv, len);
67e989fb 3065 }
560a288e 3066 }
ffebcc3e 3067 SvUTF8_off(sv);
560a288e
GS
3068 return TRUE;
3069}
3070
c461cf8f
JH
3071/*
3072=for apidoc sv_utf8_encode
3073
78ea37eb
TS
3074Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3075flag off so that it looks like octets again.
c461cf8f
JH
3076
3077=cut
3078*/
3079
560a288e
GS
3080void
3081Perl_sv_utf8_encode(pTHX_ register SV *sv)
3082{
4411f3b6 3083 (void) sv_utf8_upgrade(sv);
4c94c214
NC
3084 if (SvIsCOW(sv)) {
3085 sv_force_normal_flags(sv, 0);
3086 }
3087 if (SvREADONLY(sv)) {
3088 Perl_croak(aTHX_ PL_no_modify);
3089 }
560a288e
GS
3090 SvUTF8_off(sv);
3091}
3092
4411f3b6
NIS
3093/*
3094=for apidoc sv_utf8_decode
3095
78ea37eb
TS
3096If the PV of the SV is an octet sequence in UTF-8
3097and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3098so that it looks like a character. If the PV contains only single-byte
3099characters, the C<SvUTF8> flag stays being off.
3100Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3101
3102=cut
3103*/
3104
560a288e
GS
3105bool
3106Perl_sv_utf8_decode(pTHX_ register SV *sv)
3107{
78ea37eb 3108 if (SvPOKp(sv)) {
93524f2b
NC
3109 const U8 *c;
3110 const U8 *e;
9cbac4c7 3111
645c22ef
DM
3112 /* The octets may have got themselves encoded - get them back as
3113 * bytes
3114 */
3115 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3116 return FALSE;
3117
3118 /* it is actually just a matter of turning the utf8 flag on, but
3119 * we want to make sure everything inside is valid utf8 first.
3120 */
93524f2b 3121 c = (const U8 *) SvPVX_const(sv);
63cd0674 3122 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3123 return FALSE;
93524f2b 3124 e = (const U8 *) SvEND(sv);
511c2ff0 3125 while (c < e) {
b64e5050 3126 const U8 ch = *c++;
c4d5f83a 3127 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3128 SvUTF8_on(sv);
3129 break;
3130 }
560a288e 3131 }
560a288e
GS
3132 }
3133 return TRUE;
3134}
3135
954c1994
GS
3136/*
3137=for apidoc sv_setsv
3138
645c22ef
DM
3139Copies the contents of the source SV C<ssv> into the destination SV
3140C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3141function if the source SV needs to be reused. Does not handle 'set' magic.
3142Loosely speaking, it performs a copy-by-value, obliterating any previous
3143content of the destination.
3144
3145You probably want to use one of the assortment of wrappers, such as
3146C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3147C<SvSetMagicSV_nosteal>.
3148
8d6d96c1
HS
3149=for apidoc sv_setsv_flags
3150
645c22ef
DM
3151Copies the contents of the source SV C<ssv> into the destination SV
3152C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3153function if the source SV needs to be reused. Does not handle 'set' magic.
3154Loosely speaking, it performs a copy-by-value, obliterating any previous
3155content of the destination.
3156If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3157C<ssv> if appropriate, else not. If the C<flags> parameter has the
3158C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3159and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3160
3161You probably want to use one of the assortment of wrappers, such as
3162C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3163C<SvSetMagicSV_nosteal>.
3164
3165This is the primary function for copying scalars, and most other
3166copy-ish functions and macros use this underneath.
8d6d96c1
HS
3167
3168=cut
3169*/
3170
5d0301b7 3171static void
2eb42952 3172S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3173{
3174 if (dtype != SVt_PVGV) {
3175 const char * const name = GvNAME(sstr);
3176 const STRLEN len = GvNAMELEN(sstr);
3177 /* don't upgrade SVt_PVLV: it can hold a glob */
f7877b28
NC
3178 if (dtype != SVt_PVLV) {
3179 if (dtype >= SVt_PV) {
3180 SvPV_free(dstr);
3181 SvPV_set(dstr, 0);
3182 SvLEN_set(dstr, 0);
3183 SvCUR_set(dstr, 0);
3184 }
5d0301b7 3185 sv_upgrade(dstr, SVt_PVGV);
dedf8e73
NC
3186 (void)SvOK_off(dstr);
3187 SvSCREAM_on(dstr);
f7877b28 3188 }
5d0301b7
NC
3189 GvSTASH(dstr) = GvSTASH(sstr);
3190 if (GvSTASH(dstr))
3191 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3192 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3193 SvFAKE_on(dstr); /* can coerce to non-glob */
3194 }
3195
3196#ifdef GV_UNIQUE_CHECK
3197 if (GvUNIQUE((GV*)dstr)) {
3198 Perl_croak(aTHX_ PL_no_modify);
3199 }
3200#endif
3201
f7877b28
NC
3202 gp_free((GV*)dstr);
3203 SvSCREAM_off(dstr);
5d0301b7 3204 (void)SvOK_off(dstr);
f7877b28 3205 SvSCREAM_on(dstr);
dedf8e73 3206 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3207 GvGP(dstr) = gp_ref(GvGP(sstr));
3208 if (SvTAINTED(sstr))
3209 SvTAINT(dstr);
3210 if (GvIMPORTED(dstr) != GVf_IMPORTED
3211 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3212 {
3213 GvIMPORTED_on(dstr);
3214 }
3215 GvMULTI_on(dstr);
3216 return;
3217}
3218
b8473700 3219static void
2eb42952 3220S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3221 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3222 SV *dref = NULL;
3223 const int intro = GvINTRO(dstr);
2440974c 3224 SV **location;
3386d083 3225 U8 import_flag = 0;
27242d61
NC
3226 const U32 stype = SvTYPE(sref);
3227
b8473700
NC
3228
3229#ifdef GV_UNIQUE_CHECK
3230 if (GvUNIQUE((GV*)dstr)) {
3231 Perl_croak(aTHX_ PL_no_modify);
3232 }
3233#endif
3234
3235 if (intro) {
3236 GvINTRO_off(dstr); /* one-shot flag */
3237 GvLINE(dstr) = CopLINE(PL_curcop);
3238 GvEGV(dstr) = (GV*)dstr;
3239 }
3240 GvMULTI_on(dstr);
27242d61 3241 switch (stype) {
b8473700 3242 case SVt_PVCV:
27242d61
NC
3243 location = (SV **) &GvCV(dstr);
3244 import_flag = GVf_IMPORTED_CV;
3245 goto common;
3246 case SVt_PVHV:
3247 location = (SV **) &GvHV(dstr);
3248 import_flag = GVf_IMPORTED_HV;
3249 goto common;
3250 case SVt_PVAV:
3251 location = (SV **) &GvAV(dstr);
3252 import_flag = GVf_IMPORTED_AV;
3253 goto common;
3254 case SVt_PVIO:
3255 location = (SV **) &GvIOp(dstr);
3256 goto common;
3257 case SVt_PVFM:
3258 location = (SV **) &GvFORM(dstr);
3259 default:
3260 location = &GvSV(dstr);
3261 import_flag = GVf_IMPORTED_SV;
3262 common:
b8473700 3263 if (intro) {
27242d61
NC
3264 if (stype == SVt_PVCV) {
3265 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3266 SvREFCNT_dec(GvCV(dstr));
3267 GvCV(dstr) = NULL;
3268 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3269 PL_sub_generation++;
3270 }
b8473700 3271 }
27242d61 3272 SAVEGENERICSV(*location);
b8473700
NC
3273 }
3274 else
27242d61
NC
3275 dref = *location;
3276 if (stype == SVt_PVCV && *location != sref) {
3277 CV* const cv = (CV*)*location;
b8473700
NC
3278 if (cv) {
3279 if (!GvCVGEN((GV*)dstr) &&
3280 (CvROOT(cv) || CvXSUB(cv)))
3281 {
3282 /* Redefining a sub - warning is mandatory if
3283 it was a const and its value changed. */
3284 if (CvCONST(cv) && CvCONST((CV*)sref)
3285 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
bb263b4e 3286 /*EMPTY*/
b8473700
NC
3287 /* They are 2 constant subroutines generated from
3288 the same constant. This probably means that
3289 they are really the "same" proxy subroutine
3290 instantiated in 2 places. Most likely this is
3291 when a constant is exported twice. Don't warn.
3292 */
3293 }
3294 else if (ckWARN(WARN_REDEFINE)
3295 || (CvCONST(cv)
3296 && (!CvCONST((CV*)sref)
3297 || sv_cmp(cv_const_sv(cv),
3298 cv_const_sv((CV*)sref))))) {
3299 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3300 CvCONST(cv)
3301 ? "Constant subroutine %s::%s redefined"
3302 : "Subroutine %s::%s redefined",
3303 HvNAME_get(GvSTASH((GV*)dstr)),
3304 GvENAME((GV*)dstr));
3305 }
3306 }
3307 if (!intro)
3308 cv_ckproto(cv, (GV*)dstr,
bd61b366 3309 SvPOK(sref) ? SvPVX_const(sref) : NULL);
b8473700 3310 }
b8473700
NC
3311 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3312 GvASSUMECV_on(dstr);
3313 PL_sub_generation++;
3314 }
2440974c 3315 *location = sref;
3386d083
NC
3316 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3317 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3318 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3319 }
3320 break;
3321 }
b37c2d43 3322 SvREFCNT_dec(dref);
b8473700
NC
3323 if (SvTAINTED(sstr))
3324 SvTAINT(dstr);
3325 return;
3326}
3327
8d6d96c1
HS
3328void
3329Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3330{
97aff369 3331 dVAR;
8990e307
LW
3332 register U32 sflags;
3333 register int dtype;
3334 register int stype;
463ee0b2 3335
79072805
LW
3336 if (sstr == dstr)
3337 return;
765f542d 3338 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3339 if (!sstr)
3280af22 3340 sstr = &PL_sv_undef;
8990e307
LW
3341 stype = SvTYPE(sstr);
3342 dtype = SvTYPE(dstr);
79072805 3343
a0d0e21e 3344 SvAMAGIC_off(dstr);
7a5fa8a2 3345 if ( SvVOK(dstr) )
ece467f9
JP
3346 {
3347 /* need to nuke the magic */
3348 mg_free(dstr);
3349 SvRMAGICAL_off(dstr);
3350 }
9e7bc3e8 3351
463ee0b2 3352 /* There's a lot of redundancy below but we're going for speed here */
79072805 3353
8990e307 3354 switch (stype) {
79072805 3355 case SVt_NULL:
aece5585 3356 undef_sstr:
20408e3c
GS
3357 if (dtype != SVt_PVGV) {
3358 (void)SvOK_off(dstr);
3359 return;
3360 }
3361 break;
463ee0b2 3362 case SVt_IV:
aece5585
GA
3363 if (SvIOK(sstr)) {
3364 switch (dtype) {
3365 case SVt_NULL:
8990e307 3366 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3367 break;
3368 case SVt_NV:
aece5585
GA
3369 case SVt_RV:
3370 case SVt_PV:
a0d0e21e 3371 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
3372 break;
3373 }
3374 (void)SvIOK_only(dstr);
45977657 3375 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3376 if (SvIsUV(sstr))
3377 SvIsUV_on(dstr);
37c25af0
NC
3378 /* SvTAINTED can only be true if the SV has taint magic, which in
3379 turn means that the SV type is PVMG (or greater). This is the
3380 case statement for SVt_IV, so this cannot be true (whatever gcov
3381 may say). */
3382 assert(!SvTAINTED(sstr));
aece5585 3383 return;
8990e307 3384 }
aece5585
GA
3385 goto undef_sstr;
3386
463ee0b2 3387 case SVt_NV:
aece5585
GA
3388 if (SvNOK(sstr)) {
3389 switch (dtype) {
3390 case SVt_NULL:
3391 case SVt_IV:
8990e307 3392 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3393 break;
3394 case SVt_RV:
3395 case SVt_PV:
3396 case SVt_PVIV:
a0d0e21e 3397 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
3398 break;
3399 }
9d6ce603 3400 SvNV_set(dstr, SvNVX(sstr));
aece5585 3401 (void)SvNOK_only(dstr);
37c25af0
NC
3402 /* SvTAINTED can only be true if the SV has taint magic, which in
3403 turn means that the SV type is PVMG (or greater). This is the
3404 case statement for SVt_NV, so this cannot be true (whatever gcov
3405 may say). */
3406 assert(!SvTAINTED(sstr));
aece5585 3407 return;
8990e307 3408 }
aece5585
GA
3409 goto undef_sstr;
3410
ed6116ce 3411 case SVt_RV:
8990e307 3412 if (dtype < SVt_RV)
ed6116ce 3413 sv_upgrade(dstr, SVt_RV);
ed6116ce 3414 break;
fc36a67e 3415 case SVt_PVFM:
f8c7b90f 3416#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3417 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3418 if (dtype < SVt_PVIV)
3419 sv_upgrade(dstr, SVt_PVIV);
3420 break;
3421 }
3422 /* Fall through */
3423#endif
3424 case SVt_PV:
8990e307 3425 if (dtype < SVt_PV)
463ee0b2 3426 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3427 break;
3428 case SVt_PVIV:
8990e307 3429 if (dtype < SVt_PVIV)
463ee0b2 3430 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3431 break;
3432 case SVt_PVNV:
8990e307 3433 if (dtype < SVt_PVNV)
463ee0b2 3434 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3435 break;
489f7bfe 3436 default:
a3b680e6
AL
3437 {
3438 const char * const type = sv_reftype(sstr,0);
533c011a 3439 if (PL_op)
a3b680e6 3440 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3441 else
a3b680e6
AL
3442 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3443 }
4633a7c4
LW
3444 break;
3445
79072805 3446 case SVt_PVGV:
8990e307 3447 if (dtype <= SVt_PVGV) {
d4c19fe8 3448 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3449 return;
79072805 3450 }
5f66b61c 3451 /*FALLTHROUGH*/
79072805 3452
489f7bfe
NC
3453 case SVt_PVMG:
3454 case SVt_PVLV:
3455 case SVt_PVBM:
8d6d96c1 3456 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3457 mg_get(sstr);
eb160463 3458 if ((int)SvTYPE(sstr) != stype) {
973f89ab 3459 stype = SvTYPE(sstr);
b8c701c1 3460 if (stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3461 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3462 return;
3463 }
973f89ab
CS
3464 }
3465 }
ded42b9f 3466 if (stype == SVt_PVLV)
862a34c6 3467 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3468 else
862a34c6 3469 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3470 }
3471
ff920335
NC
3472 /* dstr may have been upgraded. */
3473 dtype = SvTYPE(dstr);
8990e307
LW
3474 sflags = SvFLAGS(sstr);
3475
3476 if (sflags & SVf_ROK) {
acaa9288
NC
3477 if (dtype == SVt_PVGV &&
3478 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
3479 sstr = SvRV(sstr);
3480 if (sstr == dstr) {
3481 if (GvIMPORTED(dstr) != GVf_IMPORTED
3482 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3483 {
3484 GvIMPORTED_on(dstr);
3485 }
3486 GvMULTI_on(dstr);
3487 return;
3488 }
d4c19fe8 3489 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3490 return;
3491 }
3492
8990e307 3493 if (dtype >= SVt_PV) {
b8c701c1 3494 if (dtype == SVt_PVGV) {
d4c19fe8 3495 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3496 return;
3497 }
3f7c398e 3498 if (SvPVX_const(dstr)) {
8bd4d4c5 3499 SvPV_free(dstr);
b162af07
SP
3500 SvLEN_set(dstr, 0);
3501 SvCUR_set(dstr, 0);
a0d0e21e 3502 }
8990e307 3503 }
a0d0e21e 3504 (void)SvOK_off(dstr);
b162af07 3505 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
dfd48732
NC
3506 SvFLAGS(dstr) |= sflags & (SVf_ROK|SVf_AMAGIC);
3507 assert(!(sflags & SVp_NOK));
3508 assert(!(sflags & SVp_IOK));
3509 assert(!(sflags & SVf_NOK));
3510 assert(!(sflags & SVf_IOK));
ed6116ce 3511 }
c0c44674
NC
3512 else if (dtype == SVt_PVGV) {
3513 if (!(sflags & SVf_OK)) {
3514 if (ckWARN(WARN_MISC))
3515 Perl_warner(aTHX_ packWARN(WARN_MISC),
3516 "Undefined value assigned to typeglob");
3517 }
3518 else {
3519 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3520 if (dstr != (SV*)gv) {
3521 if (GvGP(dstr))
3522 gp_free((GV*)dstr);
3523 GvGP(dstr) = gp_ref(GvGP(gv));
3524 }
3525 }
3526 }
8990e307 3527 else if (sflags & SVp_POK) {
765f542d 3528 bool isSwipe = 0;
79072805
LW
3529
3530 /*
3531 * Check to see if we can just swipe the string. If so, it's a
3532 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3533 * It might even be a win on short strings if SvPVX_const(dstr)
3534 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3535 */
3536
120fac95
NC
3537 /* Whichever path we take through the next code, we want this true,
3538 and doing it now facilitates the COW check. */
3539 (void)SvPOK_only(dstr);
3540
765f542d 3541 if (
b8f9541a
NC
3542 /* We're not already COW */
3543 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3544#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3545 /* or we are, but dstr isn't a suitable target. */
3546 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3547#endif
3548 )
765f542d 3549 &&
765f542d
NC
3550 !(isSwipe =
3551 (sflags & SVs_TEMP) && /* slated for free anyway? */
3552 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3553 (!(flags & SV_NOSTEAL)) &&
3554 /* and we're allowed to steal temps */
765f542d
NC
3555 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3556 SvLEN(sstr) && /* and really is a string */
645c22ef 3557 /* and won't be needed again, potentially */
765f542d 3558 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3559#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3560 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3561 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3562 && SvTYPE(sstr) >= SVt_PVIV)
3563#endif
3564 ) {
3565 /* Failed the swipe test, and it's not a shared hash key either.
3566 Have to copy the string. */
3567 STRLEN len = SvCUR(sstr);
3568 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3569 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3570 SvCUR_set(dstr, len);
3571 *SvEND(dstr) = '\0';
765f542d 3572 } else {
f8c7b90f 3573 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3574 be true in here. */
765f542d
NC
3575 /* Either it's a shared hash key, or it's suitable for
3576 copy-on-write or we can swipe the string. */
46187eeb 3577 if (DEBUG_C_TEST) {
ed252734 3578 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3579 sv_dump(sstr);
3580 sv_dump(dstr);
46187eeb 3581 }
f8c7b90f 3582#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3583 if (!isSwipe) {
3584 /* I believe I should acquire a global SV mutex if
3585 it's a COW sv (not a shared hash key) to stop
3586 it going un copy-on-write.
3587 If the source SV has gone un copy on write between up there
3588 and down here, then (assert() that) it is of the correct
3589 form to make it copy on write again */
3590 if ((sflags & (SVf_FAKE | SVf_READONLY))
3591 != (SVf_FAKE | SVf_READONLY)) {
3592 SvREADONLY_on(sstr);
3593 SvFAKE_on(sstr);
3594 /* Make the source SV into a loop of 1.
3595 (about to become 2) */
a29f6d03 3596 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3597 }
3598 }
3599#endif
3600 /* Initial code is common. */
94010e71
NC
3601 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3602 SvPV_free(dstr);
79072805 3603 }
765f542d 3604
765f542d
NC
3605 if (!isSwipe) {
3606 /* making another shared SV. */
3607 STRLEN cur = SvCUR(sstr);
3608 STRLEN len = SvLEN(sstr);
f8c7b90f 3609#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3610 if (len) {
b8f9541a 3611 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3612 /* SvIsCOW_normal */
3613 /* splice us in between source and next-after-source. */
a29f6d03
NC
3614 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3615 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3616 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3617 } else
3618#endif
3619 {
765f542d 3620 /* SvIsCOW_shared_hash */
46187eeb
NC
3621 DEBUG_C(PerlIO_printf(Perl_debug_log,
3622 "Copy on write: Sharing hash\n"));
b8f9541a 3623
bdd68bc3 3624 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3625 SvPV_set(dstr,
d1db91c6 3626 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3627 }
87a1ef3d
SP
3628 SvLEN_set(dstr, len);
3629 SvCUR_set(dstr, cur);
765f542d
NC
3630 SvREADONLY_on(dstr);
3631 SvFAKE_on(dstr);
3632 /* Relesase a global SV mutex. */
3633 }
3634 else
765f542d 3635 { /* Passes the swipe test. */
78d1e721 3636 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3637 SvLEN_set(dstr, SvLEN(sstr));
3638 SvCUR_set(dstr, SvCUR(sstr));
3639
3640 SvTEMP_off(dstr);
3641 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3642 SvPV_set(sstr, NULL);
765f542d
NC
3643 SvLEN_set(sstr, 0);
3644 SvCUR_set(sstr, 0);
3645 SvTEMP_off(sstr);
3646 }
3647 }
8990e307 3648 if (sflags & SVp_NOK) {
9d6ce603 3649 SvNV_set(dstr, SvNVX(sstr));
79072805 3650 }
8990e307 3651 if (sflags & SVp_IOK) {
23525414
NC
3652 SvRELEASE_IVX(dstr);
3653 SvIV_set(dstr, SvIVX(sstr));
3654 /* Must do this otherwise some other overloaded use of 0x80000000
3655 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3656 if (sflags & SVf_IVisUV)
25da4f38 3657 SvIsUV_on(dstr);
79072805 3658 }
dd2eae66
NC
3659 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8
3660 |SVf_AMAGIC);
4f2da183
NC
3661 {
3662 const MAGIC * const smg = SvVOK(sstr);
3663 if (smg) {
3664 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3665 smg->mg_ptr, smg->mg_len);
3666 SvRMAGICAL_on(dstr);
3667 }
7a5fa8a2 3668 }
79072805 3669 }
5d581361 3670 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3671 (void)SvOK_off(dstr);
dd2eae66
NC
3672 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK
3673 |SVf_AMAGIC);
5d581361
NC
3674 if (sflags & SVp_IOK) {
3675 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3676 SvIV_set(dstr, SvIVX(sstr));
3677 }
3332b3c1 3678 if (sflags & SVp_NOK) {
9d6ce603 3679 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3680 }
3681 }
79072805 3682 else {
f7877b28 3683 if (isGV_with_GP(sstr)) {
180488f8
NC
3684 /* This stringification rule for globs is spread in 3 places.
3685 This feels bad. FIXME. */
3686 const U32 wasfake = sflags & SVf_FAKE;
3687
3688 /* FAKE globs can get coerced, so need to turn this off
3689 temporarily if it is on. */
3690 SvFAKE_off(sstr);
3691 gv_efullname3(dstr, (GV *)sstr, "*");
3692 SvFLAGS(sstr) |= wasfake;
dd2eae66 3693 SvFLAGS(dstr) |= sflags & SVf_AMAGIC;
180488f8 3694 }
20408e3c
GS
3695 else
3696 (void)SvOK_off(dstr);
a0d0e21e 3697 }
27c9684d
AP
3698 if (SvTAINTED(sstr))
3699 SvTAINT(dstr);
79072805
LW
3700}
3701
954c1994
GS
3702/*
3703=for apidoc sv_setsv_mg
3704
3705Like C<sv_setsv>, but also handles 'set' magic.
3706
3707=cut
3708*/
3709
79072805 3710void
864dbfa3 3711Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3712{
3713 sv_setsv(dstr,sstr);
3714 SvSETMAGIC(dstr);
3715}
3716
f8c7b90f 3717#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3718SV *
3719Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3720{
3721 STRLEN cur = SvCUR(sstr);
3722 STRLEN len = SvLEN(sstr);
3723 register char *new_pv;
3724
3725 if (DEBUG_C_TEST) {
3726 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3727 sstr, dstr);
3728 sv_dump(sstr);
3729 if (dstr)
3730 sv_dump(dstr);
3731 }
3732
3733 if (dstr) {
3734 if (SvTHINKFIRST(dstr))
3735 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3736 else if (SvPVX_const(dstr))
3737 Safefree(SvPVX_const(dstr));
ed252734
NC
3738 }
3739 else
3740 new_SV(dstr);
862a34c6 3741 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3742
3743 assert (SvPOK(sstr));
3744 assert (SvPOKp(sstr));
3745 assert (!SvIOK(sstr));
3746 assert (!SvIOKp(sstr));
3747 assert (!SvNOK(sstr));
3748 assert (!SvNOKp(sstr));
3749
3750 if (SvIsCOW(sstr)) {
3751
3752 if (SvLEN(sstr) == 0) {
3753 /* source is a COW shared hash key. */
ed252734
NC
3754 DEBUG_C(PerlIO_printf(Perl_debug_log,
3755 "Fast copy on write: Sharing hash\n"));
d1db91c6 3756 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3757 goto common_exit;
3758 }
3759 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3760 } else {
3761 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3762 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3763 SvREADONLY_on(sstr);
3764 SvFAKE_on(sstr);
3765 DEBUG_C(PerlIO_printf(Perl_debug_log,
3766 "Fast copy on write: Converting sstr to COW\n"));
3767 SV_COW_NEXT_SV_SET(dstr, sstr);
3768 }
3769 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3770 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3771
3772 common_exit:
3773 SvPV_set(dstr, new_pv);
3774 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3775 if (SvUTF8(sstr))
3776 SvUTF8_on(dstr);
87a1ef3d
SP
3777 SvLEN_set(dstr, len);
3778 SvCUR_set(dstr, cur);
ed252734
NC
3779 if (DEBUG_C_TEST) {
3780 sv_dump(dstr);
3781 }
3782 return dstr;
3783}
3784#endif
3785
954c1994
GS
3786/*
3787=for apidoc sv_setpvn
3788
3789Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3790bytes to be copied. If the C<ptr> argument is NULL the SV will become
3791undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3792
3793=cut
3794*/
3795
ef50df4b 3796void
864dbfa3 3797Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3798{
97aff369 3799 dVAR;
c6f8c383 3800 register char *dptr;
22c522df 3801
765f542d 3802 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3803 if (!ptr) {
a0d0e21e 3804 (void)SvOK_off(sv);
463ee0b2
LW
3805 return;
3806 }
22c522df
JH
3807 else {
3808 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3809 const IV iv = len;
9c5ffd7c
JH
3810 if (iv < 0)
3811 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3812 }
862a34c6 3813 SvUPGRADE(sv, SVt_PV);
c6f8c383 3814
5902b6a9 3815 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3816 Move(ptr,dptr,len,char);
3817 dptr[len] = '\0';
79072805 3818 SvCUR_set(sv, len);
1aa99e6b 3819 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3820 SvTAINT(sv);
79072805
LW
3821}
3822
954c1994
GS
3823/*
3824=for apidoc sv_setpvn_mg
3825
3826Like C<sv_setpvn>, but also handles 'set' magic.
3827
3828=cut
3829*/
3830
79072805 3831void
864dbfa3 3832Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3833{
3834 sv_setpvn(sv,ptr,len);
3835 SvSETMAGIC(sv);
3836}
3837
954c1994
GS
3838/*
3839=for apidoc sv_setpv
3840
3841Copies a string into an SV. The string must be null-terminated. Does not
3842handle 'set' magic. See C<sv_setpv_mg>.
3843
3844=cut
3845*/
3846
ef50df4b 3847void
864dbfa3 3848Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3849{
97aff369 3850 dVAR;
79072805
LW
3851 register STRLEN len;
3852
765f542d 3853 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3854 if (!ptr) {
a0d0e21e 3855 (void)SvOK_off(sv);
463ee0b2
LW
3856 return;
3857 }
79072805 3858 len = strlen(ptr);
862a34c6 3859 SvUPGRADE(sv, SVt_PV);
c6f8c383 3860
79072805 3861 SvGROW(sv, len + 1);
463ee0b2 3862 Move(ptr,SvPVX(sv),len+1,char);
79072805 3863 SvCUR_set(sv, len);
1aa99e6b 3864 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3865 SvTAINT(sv);
3866}
3867
954c1994
GS
3868/*
3869=for apidoc sv_setpv_mg
3870
3871Like C<sv_setpv>, but also handles 'set' magic.
3872
3873=cut
3874*/
3875
463ee0b2 3876void
864dbfa3 3877Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3878{
3879 sv_setpv(sv,ptr);
3880 SvSETMAGIC(sv);
3881}
3882
954c1994
GS
3883/*
3884=for apidoc sv_usepvn
3885
3886Tells an SV to use C<ptr> to find its string value. Normally the string is
1c846c1f 3887stored inside the SV but sv_usepvn allows the SV to use an outside string.
954c1994
GS
3888The C<ptr> should point to memory that was allocated by C<malloc>. The
3889string length, C<len>, must be supplied. This function will realloc the
3890memory pointed to by C<ptr>, so that pointer should not be freed or used by
3891the programmer after giving it to sv_usepvn. Does not handle 'set' magic.
3892See C<sv_usepvn_mg>.
3893
3894=cut
3895*/
3896
ef50df4b 3897void
864dbfa3 3898Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
463ee0b2 3899{
97aff369 3900 dVAR;
1936d2a7 3901 STRLEN allocate;
765f542d 3902 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3903 SvUPGRADE(sv, SVt_PV);
463ee0b2 3904 if (!ptr) {
a0d0e21e 3905 (void)SvOK_off(sv);
463ee0b2
LW
3906 return;
3907 }
3f7c398e 3908 if (SvPVX_const(sv))
8bd4d4c5 3909 SvPV_free(sv);
1936d2a7
NC
3910
3911 allocate = PERL_STRLEN_ROUNDUP(len + 1);
7a9b70e9 3912 ptr = saferealloc (ptr, allocate);
f880fe2f 3913 SvPV_set(sv, ptr);
463ee0b2 3914 SvCUR_set(sv, len);
1936d2a7 3915 SvLEN_set(sv, allocate);
463ee0b2 3916 *SvEND(sv) = '\0';
1aa99e6b 3917 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3918 SvTAINT(sv);
79072805
LW
3919}
3920
954c1994
GS
3921/*
3922=for apidoc sv_usepvn_mg
3923
3924Like C<sv_usepvn>, but also handles 'set' magic.
3925
3926=cut
3927*/
3928
ef50df4b 3929void
864dbfa3 3930Perl_sv_usepvn_mg(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
ef50df4b 3931{
51c1089b 3932 sv_usepvn(sv,ptr,len);
ef50df4b
GS
3933 SvSETMAGIC(sv);
3934}
3935
f8c7b90f 3936#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3937/* Need to do this *after* making the SV normal, as we need the buffer
3938 pointer to remain valid until after we've copied it. If we let go too early,
3939 another thread could invalidate it by unsharing last of the same hash key
3940 (which it can do by means other than releasing copy-on-write Svs)
3941 or by changing the other copy-on-write SVs in the loop. */
3942STATIC void
bdd68bc3 3943S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3944{
3945 if (len) { /* this SV was SvIsCOW_normal(sv) */
3946 /* we need to find the SV pointing to us. */
cf5629ad 3947 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3948
765f542d
NC
3949 if (current == sv) {
3950 /* The SV we point to points back to us (there were only two of us
3951 in the loop.)
3952 Hence other SV is no longer copy on write either. */
3953 SvFAKE_off(after);
3954 SvREADONLY_off(after);
3955 } else {
3956 /* We need to follow the pointers around the loop. */
3957 SV *next;
3958 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3959 assert (next);
3960 current = next;
3961 /* don't loop forever if the structure is bust, and we have
3962 a pointer into a closed loop. */
3963 assert (current != after);
3f7c398e 3964 assert (SvPVX_const(current) == pvx);
765f542d
NC
3965 }
3966 /* Make the SV before us point to the SV after us. */
a29f6d03 3967 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3968 }
3969 } else {
bdd68bc3 3970 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3971 }
3972}
3973
3974int
3975Perl_sv_release_IVX(pTHX_ register SV *sv)
3976{
3977 if (SvIsCOW(sv))
3978 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
3979 SvOOK_off(sv);
3980 return 0;
765f542d
NC
3981}
3982#endif
645c22ef
DM
3983/*
3984=for apidoc sv_force_normal_flags
3985
3986Undo various types of fakery on an SV: if the PV is a shared string, make
3987a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
3988an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
3989we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
3990then a copy-on-write scalar drops its PV buffer (if any) and becomes
3991SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 3992set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
3993C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
3994with flags set to 0.
645c22ef
DM
3995
3996=cut
3997*/
3998
6fc92669 3999void
840a7b70 4000Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4001{
97aff369 4002 dVAR;
f8c7b90f 4003#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4004 if (SvREADONLY(sv)) {
4005 /* At this point I believe I should acquire a global SV mutex. */
4006 if (SvFAKE(sv)) {
b64e5050 4007 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4008 const STRLEN len = SvLEN(sv);
4009 const STRLEN cur = SvCUR(sv);
a28509cc 4010 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4011 if (DEBUG_C_TEST) {
4012 PerlIO_printf(Perl_debug_log,
4013 "Copy on write: Force normal %ld\n",
4014 (long) flags);
e419cbc5 4015 sv_dump(sv);
46187eeb 4016 }
765f542d
NC
4017 SvFAKE_off(sv);
4018 SvREADONLY_off(sv);
9f653bb5 4019 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4020 SvPV_set(sv, NULL);
87a1ef3d 4021 SvLEN_set(sv, 0);
765f542d
NC
4022 if (flags & SV_COW_DROP_PV) {
4023 /* OK, so we don't need to copy our buffer. */
4024 SvPOK_off(sv);
4025 } else {
4026 SvGROW(sv, cur + 1);
4027 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4028 SvCUR_set(sv, cur);
765f542d
NC
4029 *SvEND(sv) = '\0';
4030 }
bdd68bc3 4031 sv_release_COW(sv, pvx, len, next);
46187eeb 4032 if (DEBUG_C_TEST) {
e419cbc5 4033 sv_dump(sv);
46187eeb 4034 }
765f542d 4035 }
923e4eb5 4036 else if (IN_PERL_RUNTIME)
765f542d
NC
4037 Perl_croak(aTHX_ PL_no_modify);
4038 /* At this point I believe that I can drop the global SV mutex. */
4039 }
4040#else
2213622d 4041 if (SvREADONLY(sv)) {
1c846c1f 4042 if (SvFAKE(sv)) {
b64e5050 4043 const char * const pvx = SvPVX_const(sv);
66a1b24b 4044 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4045 SvFAKE_off(sv);
4046 SvREADONLY_off(sv);
bd61b366 4047 SvPV_set(sv, NULL);
66a1b24b 4048 SvLEN_set(sv, 0);
1c846c1f 4049 SvGROW(sv, len + 1);
706aa1c9 4050 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4051 *SvEND(sv) = '\0';
bdd68bc3 4052 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4053 }
923e4eb5 4054 else if (IN_PERL_RUNTIME)
cea2e8a9 4055 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4056 }
765f542d 4057#endif
2213622d 4058 if (SvROK(sv))
840a7b70 4059 sv_unref_flags(sv, flags);
6fc92669
GS
4060 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4061 sv_unglob(sv);
0f15f207 4062}
1c846c1f 4063
645c22ef 4064/*
954c1994
GS
4065=for apidoc sv_chop
4066
1c846c1f 4067Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4068SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4069the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4070string. Uses the "OOK hack".
3f7c398e 4071Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4072refer to the same chunk of data.
954c1994
GS
4073
4074=cut
4075*/
4076
79072805 4077void
f54cb97a 4078Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4079{
4080 register STRLEN delta;
a0d0e21e 4081 if (!ptr || !SvPOKp(sv))
79072805 4082 return;
3f7c398e 4083 delta = ptr - SvPVX_const(sv);
2213622d 4084 SV_CHECK_THINKFIRST(sv);
79072805
LW
4085 if (SvTYPE(sv) < SVt_PVIV)
4086 sv_upgrade(sv,SVt_PVIV);
4087
4088 if (!SvOOK(sv)) {
50483b2c 4089 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4090 const char *pvx = SvPVX_const(sv);
a28509cc 4091 const STRLEN len = SvCUR(sv);
50483b2c 4092 SvGROW(sv, len + 1);
706aa1c9 4093 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4094 *SvEND(sv) = '\0';
4095 }
45977657 4096 SvIV_set(sv, 0);
a4bfb290
AB
4097 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4098 and we do that anyway inside the SvNIOK_off
4099 */
7a5fa8a2 4100 SvFLAGS(sv) |= SVf_OOK;
79072805 4101 }
a4bfb290 4102 SvNIOK_off(sv);
b162af07
SP
4103 SvLEN_set(sv, SvLEN(sv) - delta);
4104 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4105 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4106 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4107}
4108
954c1994
GS
4109/*
4110=for apidoc sv_catpvn
4111
4112Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4113C<len> indicates number of bytes to copy. If the SV has the UTF-8
4114status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4115Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4116
8d6d96c1
HS
4117=for apidoc sv_catpvn_flags
4118
4119Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4120C<len> indicates number of bytes to copy. If the SV has the UTF-8
4121status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4122If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4123appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4124in terms of this function.
4125
4126=cut
4127*/
4128
4129void
4130Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4131{
97aff369 4132 dVAR;
8d6d96c1 4133 STRLEN dlen;
fabdb6c0 4134 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4135
8d6d96c1
HS
4136 SvGROW(dsv, dlen + slen + 1);
4137 if (sstr == dstr)
3f7c398e 4138 sstr = SvPVX_const(dsv);
8d6d96c1 4139 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4140 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4141 *SvEND(dsv) = '\0';
4142 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4143 SvTAINT(dsv);
bddd5118
NC
4144 if (flags & SV_SMAGIC)
4145 SvSETMAGIC(dsv);
79072805
LW
4146}
4147
954c1994 4148/*
954c1994
GS
4149=for apidoc sv_catsv
4150
13e8c8e3
JH
4151Concatenates the string from SV C<ssv> onto the end of the string in
4152SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4153not 'set' magic. See C<sv_catsv_mg>.
954c1994 4154
8d6d96c1
HS
4155=for apidoc sv_catsv_flags
4156
4157Concatenates the string from SV C<ssv> onto the end of the string in
4158SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4159bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4160and C<sv_catsv_nomg> are implemented in terms of this function.
4161
4162=cut */
4163
ef50df4b 4164void
8d6d96c1 4165Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4166{
97aff369 4167 dVAR;
bddd5118 4168 if (ssv) {
00b6aa41
AL
4169 STRLEN slen;
4170 const char *spv = SvPV_const(ssv, slen);
4171 if (spv) {
bddd5118
NC
4172 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4173 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4174 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4175 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4176 dsv->sv_flags doesn't have that bit set.
4fd84b44 4177 Andy Dougherty 12 Oct 2001
bddd5118
NC
4178 */
4179 const I32 sutf8 = DO_UTF8(ssv);
4180 I32 dutf8;
13e8c8e3 4181
bddd5118
NC
4182 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4183 mg_get(dsv);
4184 dutf8 = DO_UTF8(dsv);
8d6d96c1 4185
bddd5118
NC
4186 if (dutf8 != sutf8) {
4187 if (dutf8) {
4188 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4189 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4190
bddd5118
NC
4191 sv_utf8_upgrade(csv);
4192 spv = SvPV_const(csv, slen);
4193 }
4194 else
4195 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4196 }
bddd5118 4197 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4198 }
560a288e 4199 }
bddd5118
NC
4200 if (flags & SV_SMAGIC)
4201 SvSETMAGIC(dsv);
79072805
LW
4202}
4203
954c1994 4204/*
954c1994
GS
4205=for apidoc sv_catpv
4206
4207Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4208If the SV has the UTF-8 status set, then the bytes appended should be
4209valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4210
d5ce4a7c 4211=cut */
954c1994 4212
ef50df4b 4213void
0c981600 4214Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4215{
97aff369 4216 dVAR;
79072805 4217 register STRLEN len;
463ee0b2 4218 STRLEN tlen;
748a9306 4219 char *junk;
79072805 4220
0c981600 4221 if (!ptr)
79072805 4222 return;
748a9306 4223 junk = SvPV_force(sv, tlen);
0c981600 4224 len = strlen(ptr);
463ee0b2 4225 SvGROW(sv, tlen + len + 1);
0c981600 4226 if (ptr == junk)
3f7c398e 4227 ptr = SvPVX_const(sv);
0c981600 4228 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4229 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4230 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4231 SvTAINT(sv);
79072805
LW
4232}
4233
954c1994
GS
4234/*
4235=for apidoc sv_catpv_mg
4236
4237Like C<sv_catpv>, but also handles 'set' magic.
4238
4239=cut
4240*/
4241
ef50df4b 4242void
0c981600 4243Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4244{
0c981600 4245 sv_catpv(sv,ptr);
ef50df4b
GS
4246 SvSETMAGIC(sv);
4247}
4248
645c22ef
DM
4249/*
4250=for apidoc newSV
4251
561b68a9
SH
4252Creates a new SV. A non-zero C<len> parameter indicates the number of
4253bytes of preallocated string space the SV should have. An extra byte for a
4254trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4255space is allocated.) The reference count for the new SV is set to 1.
4256
4257In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4258parameter, I<x>, a debug aid which allowed callers to identify themselves.
4259This aid has been superseded by a new build option, PERL_MEM_LOG (see
4260L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4261modules supporting older perls.
645c22ef
DM
4262
4263=cut
4264*/
4265
79072805 4266SV *
864dbfa3 4267Perl_newSV(pTHX_ STRLEN len)
79072805 4268{
97aff369 4269 dVAR;
79072805 4270 register SV *sv;
1c846c1f 4271
4561caa4 4272 new_SV(sv);
79072805
LW
4273 if (len) {
4274 sv_upgrade(sv, SVt_PV);
4275 SvGROW(sv, len + 1);
4276 }
4277 return sv;
4278}
954c1994 4279/*
92110913 4280=for apidoc sv_magicext
954c1994 4281
68795e93 4282Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4283supplied vtable and returns a pointer to the magic added.
92110913 4284
2d8d5d5a
SH
4285Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4286In particular, you can add magic to SvREADONLY SVs, and add more than
4287one instance of the same 'how'.
645c22ef 4288
2d8d5d5a
SH
4289If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4290stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4291special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4292to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4293
2d8d5d5a 4294(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4295
4296=cut
4297*/
92110913 4298MAGIC *
92e67595 4299Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4300 const char* name, I32 namlen)
79072805 4301{
97aff369 4302 dVAR;
79072805 4303 MAGIC* mg;
68795e93 4304
92110913 4305 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4306 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4307 }
a02a5408 4308 Newxz(mg, 1, MAGIC);
79072805 4309 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4310 SvMAGIC_set(sv, mg);
75f9d97a 4311
05f95b08
SB
4312 /* Sometimes a magic contains a reference loop, where the sv and
4313 object refer to each other. To prevent a reference loop that
4314 would prevent such objects being freed, we look for such loops
4315 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4316
4317 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4318 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4319
4320 */
14befaf4
DM
4321 if (!obj || obj == sv ||
4322 how == PERL_MAGIC_arylen ||
4323 how == PERL_MAGIC_qr ||
8d2f4536 4324 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4325 (SvTYPE(obj) == SVt_PVGV &&
4326 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4327 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4328 GvFORM(obj) == (CV*)sv)))
75f9d97a 4329 {
8990e307 4330 mg->mg_obj = obj;
75f9d97a 4331 }
85e6fe83 4332 else {
b37c2d43 4333 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4334 mg->mg_flags |= MGf_REFCOUNTED;
4335 }
b5ccf5f2
YST
4336
4337 /* Normal self-ties simply pass a null object, and instead of
4338 using mg_obj directly, use the SvTIED_obj macro to produce a
4339 new RV as needed. For glob "self-ties", we are tieing the PVIO
4340 with an RV obj pointing to the glob containing the PVIO. In
4341 this case, to avoid a reference loop, we need to weaken the
4342 reference.
4343 */
4344
4345 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4346 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4347 {
4348 sv_rvweaken(obj);
4349 }
4350
79072805 4351 mg->mg_type = how;
565764a8 4352 mg->mg_len = namlen;
9cbac4c7 4353 if (name) {
92110913 4354 if (namlen > 0)
1edc1566 4355 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4356 else if (namlen == HEf_SVKEY)
b37c2d43 4357 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4358 else
92110913 4359 mg->mg_ptr = (char *) name;
9cbac4c7 4360 }
92110913 4361 mg->mg_virtual = vtable;
68795e93 4362
92110913
NIS
4363 mg_magical(sv);
4364 if (SvGMAGICAL(sv))
4365 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4366 return mg;
4367}
4368
4369/*
4370=for apidoc sv_magic
1c846c1f 4371
92110913
NIS
4372Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4373then adds a new magic item of type C<how> to the head of the magic list.
4374
2d8d5d5a
SH
4375See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4376handling of the C<name> and C<namlen> arguments.
4377
4509d3fb
SB
4378You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4379to add more than one instance of the same 'how'.
4380
92110913
NIS
4381=cut
4382*/
4383
4384void
4385Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4386{
97aff369 4387 dVAR;
92e67595 4388 MGVTBL *vtable;
92110913 4389 MAGIC* mg;
92110913 4390
f8c7b90f 4391#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4392 if (SvIsCOW(sv))
4393 sv_force_normal_flags(sv, 0);
4394#endif
92110913 4395 if (SvREADONLY(sv)) {
d8084ca5
DM
4396 if (
4397 /* its okay to attach magic to shared strings; the subsequent
4398 * upgrade to PVMG will unshare the string */
4399 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4400
4401 && IN_PERL_RUNTIME
92110913
NIS
4402 && how != PERL_MAGIC_regex_global
4403 && how != PERL_MAGIC_bm
4404 && how != PERL_MAGIC_fm
4405 && how != PERL_MAGIC_sv
e6469971 4406 && how != PERL_MAGIC_backref
92110913
NIS
4407 )
4408 {
4409 Perl_croak(aTHX_ PL_no_modify);
4410 }
4411 }
4412 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4413 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4414 /* sv_magic() refuses to add a magic of the same 'how' as an
4415 existing one
92110913 4416 */
2a509ed3 4417 if (how == PERL_MAGIC_taint) {
92110913 4418 mg->mg_len |= 1;
2a509ed3
NC
4419 /* Any scalar which already had taint magic on which someone
4420 (erroneously?) did SvIOK_on() or similar will now be
4421 incorrectly sporting public "OK" flags. */
4422 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4423 }
92110913
NIS
4424 return;
4425 }
4426 }
68795e93 4427
79072805 4428 switch (how) {
14befaf4 4429 case PERL_MAGIC_sv:
92110913 4430 vtable = &PL_vtbl_sv;
79072805 4431 break;
14befaf4 4432 case PERL_MAGIC_overload:
92110913 4433 vtable = &PL_vtbl_amagic;
a0d0e21e 4434 break;
14befaf4 4435 case PERL_MAGIC_overload_elem:
92110913 4436 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4437 break;
14befaf4 4438 case PERL_MAGIC_overload_table:
92110913 4439 vtable = &PL_vtbl_ovrld;
a0d0e21e 4440 break;
14befaf4 4441 case PERL_MAGIC_bm:
92110913 4442 vtable = &PL_vtbl_bm;
79072805 4443 break;
14befaf4 4444 case PERL_MAGIC_regdata:
92110913 4445 vtable = &PL_vtbl_regdata;
6cef1e77 4446 break;
14befaf4 4447 case PERL_MAGIC_regdatum:
92110913 4448 vtable = &PL_vtbl_regdatum;
6cef1e77 4449 break;
14befaf4 4450 case PERL_MAGIC_env:
92110913 4451 vtable = &PL_vtbl_env;
79072805 4452 break;
14befaf4 4453 case PERL_MAGIC_fm:
92110913 4454 vtable = &PL_vtbl_fm;
55497cff 4455 break;
14befaf4 4456 case PERL_MAGIC_envelem:
92110913 4457 vtable = &PL_vtbl_envelem;
79072805 4458 break;
14befaf4 4459 case PERL_MAGIC_regex_global:
92110913 4460 vtable = &PL_vtbl_mglob;
93a17b20 4461 break;
14befaf4 4462 case PERL_MAGIC_isa:
92110913 4463 vtable = &PL_vtbl_isa;
463ee0b2 4464 break;
14befaf4 4465 case PERL_MAGIC_isaelem:
92110913 4466 vtable = &PL_vtbl_isaelem;
463ee0b2 4467 break;
14befaf4 4468 case PERL_MAGIC_nkeys:
92110913 4469 vtable = &PL_vtbl_nkeys;
16660edb 4470 break;
14befaf4 4471 case PERL_MAGIC_dbfile:
aec46f14 4472 vtable = NULL;
93a17b20 4473 break;
14befaf4 4474 case PERL_MAGIC_dbline:
92110913 4475 vtable = &PL_vtbl_dbline;
79072805 4476 break;
36477c24 4477#ifdef USE_LOCALE_COLLATE
14befaf4 4478 case PERL_MAGIC_collxfrm:
92110913 4479 vtable = &PL_vtbl_collxfrm;
bbce6d69 4480 break;
36477c24 4481#endif /* USE_LOCALE_COLLATE */
14befaf4 4482 case PERL_MAGIC_tied:
92110913 4483 vtable = &PL_vtbl_pack;
463ee0b2 4484 break;
14befaf4
DM
4485 case PERL_MAGIC_tiedelem:
4486 case PERL_MAGIC_tiedscalar:
92110913 4487 vtable = &PL_vtbl_packelem;
463ee0b2 4488 break;
14befaf4 4489 case PERL_MAGIC_qr:
92110913 4490 vtable = &PL_vtbl_regexp;
c277df42 4491 break;
b3ca2e83
NC
4492 case PERL_MAGIC_hints:
4493 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4494 case PERL_MAGIC_sig:
92110913 4495 vtable = &PL_vtbl_sig;
79072805 4496 break;
14befaf4 4497 case PERL_MAGIC_sigelem:
92110913 4498 vtable = &PL_vtbl_sigelem;
79072805 4499 break;
14befaf4 4500 case PERL_MAGIC_taint:
92110913 4501 vtable = &PL_vtbl_taint;
463ee0b2 4502 break;
14befaf4 4503 case PERL_MAGIC_uvar:
92110913 4504 vtable = &PL_vtbl_uvar;
79072805 4505 break;
14befaf4 4506 case PERL_MAGIC_vec:
92110913 4507 vtable = &PL_vtbl_vec;
79072805 4508 break;
a3874608 4509 case PERL_MAGIC_arylen_p:
bfcb3514 4510 case PERL_MAGIC_rhash:
8d2f4536 4511 case PERL_MAGIC_symtab:
ece467f9 4512 case PERL_MAGIC_vstring:
aec46f14 4513 vtable = NULL;
ece467f9 4514 break;
7e8c5dac
HS
4515 case PERL_MAGIC_utf8:
4516 vtable = &PL_vtbl_utf8;
4517 break;
14befaf4 4518 case PERL_MAGIC_substr:
92110913 4519 vtable = &PL_vtbl_substr;
79072805 4520 break;
14befaf4 4521 case PERL_MAGIC_defelem:
92110913 4522 vtable = &PL_vtbl_defelem;
5f05dabc 4523 break;
14befaf4 4524 case PERL_MAGIC_arylen:
92110913 4525 vtable = &PL_vtbl_arylen;
79072805 4526 break;
14befaf4 4527 case PERL_MAGIC_pos:
92110913 4528 vtable = &PL_vtbl_pos;
a0d0e21e 4529 break;
14befaf4 4530 case PERL_MAGIC_backref:
92110913 4531 vtable = &PL_vtbl_backref;
810b8aa5 4532 break;
b3ca2e83
NC
4533 case PERL_MAGIC_hintselem:
4534 vtable = &PL_vtbl_hintselem;
4535 break;
14befaf4
DM
4536 case PERL_MAGIC_ext:
4537 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4538 /* Useful for attaching extension internal data to perl vars. */
4539 /* Note that multiple extensions may clash if magical scalars */
4540 /* etc holding private data from one are passed to another. */
aec46f14 4541 vtable = NULL;
a0d0e21e 4542 break;
79072805 4543 default:
14befaf4 4544 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4545 }
68795e93 4546
92110913 4547 /* Rest of work is done else where */
aec46f14 4548 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4549
92110913
NIS
4550 switch (how) {
4551 case PERL_MAGIC_taint:
4552 mg->mg_len = 1;
4553 break;
4554 case PERL_MAGIC_ext:
4555 case PERL_MAGIC_dbfile:
4556 SvRMAGICAL_on(sv);
4557 break;
4558 }
463ee0b2
LW
4559}
4560
c461cf8f
JH
4561/*
4562=for apidoc sv_unmagic
4563
645c22ef 4564Removes all magic of type C<type> from an SV.
c461cf8f
JH
4565
4566=cut
4567*/
4568
463ee0b2 4569int
864dbfa3 4570Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4571{
4572 MAGIC* mg;
4573 MAGIC** mgp;
91bba347 4574 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4575 return 0;
064cf529 4576 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4577 for (mg = *mgp; mg; mg = *mgp) {
4578 if (mg->mg_type == type) {
e1ec3a88 4579 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4580 *mgp = mg->mg_moremagic;
1d7c1841 4581 if (vtbl && vtbl->svt_free)
fc0dc3b3 4582 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4583 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4584 if (mg->mg_len > 0)
1edc1566 4585 Safefree(mg->mg_ptr);
565764a8 4586 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4587 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4588 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4589 Safefree(mg->mg_ptr);
9cbac4c7 4590 }
a0d0e21e
LW
4591 if (mg->mg_flags & MGf_REFCOUNTED)
4592 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4593 Safefree(mg);
4594 }
4595 else
4596 mgp = &mg->mg_moremagic;
79072805 4597 }
91bba347 4598 if (!SvMAGIC(sv)) {
463ee0b2 4599 SvMAGICAL_off(sv);
c268c2a6 4600 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4601 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4602 }
4603
4604 return 0;
79072805
LW
4605}
4606
c461cf8f
JH
4607/*
4608=for apidoc sv_rvweaken
4609
645c22ef
DM
4610Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4611referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4612push a back-reference to this RV onto the array of backreferences
4613associated with that magic.
c461cf8f
JH
4614
4615=cut
4616*/
4617
810b8aa5 4618SV *
864dbfa3 4619Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4620{
4621 SV *tsv;
4622 if (!SvOK(sv)) /* let undefs pass */
4623 return sv;
4624 if (!SvROK(sv))
cea2e8a9 4625 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4626 else if (SvWEAKREF(sv)) {
810b8aa5 4627 if (ckWARN(WARN_MISC))
9014280d 4628 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4629 return sv;
4630 }
4631 tsv = SvRV(sv);
e15faf7d 4632 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4633 SvWEAKREF_on(sv);
1c846c1f 4634 SvREFCNT_dec(tsv);
810b8aa5
GS
4635 return sv;
4636}
4637
645c22ef
DM
4638/* Give tsv backref magic if it hasn't already got it, then push a
4639 * back-reference to sv onto the array associated with the backref magic.
4640 */
4641
e15faf7d
NC
4642void
4643Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4644{
97aff369 4645 dVAR;
810b8aa5 4646 AV *av;
86f55936
NC
4647
4648 if (SvTYPE(tsv) == SVt_PVHV) {
4649 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4650
4651 av = *avp;
4652 if (!av) {
4653 /* There is no AV in the offical place - try a fixup. */
4654 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4655
4656 if (mg) {
4657 /* Aha. They've got it stowed in magic. Bring it back. */
4658 av = (AV*)mg->mg_obj;
4659 /* Stop mg_free decreasing the refernce count. */
4660 mg->mg_obj = NULL;
4661 /* Stop mg_free even calling the destructor, given that
4662 there's no AV to free up. */
4663 mg->mg_virtual = 0;
4664 sv_unmagic(tsv, PERL_MAGIC_backref);
4665 } else {
4666 av = newAV();
4667 AvREAL_off(av);
b37c2d43 4668 SvREFCNT_inc_simple_void(av);
86f55936
NC
4669 }
4670 *avp = av;
4671 }
4672 } else {
4673 const MAGIC *const mg
4674 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4675 if (mg)
4676 av = (AV*)mg->mg_obj;
4677 else {
4678 av = newAV();
4679 AvREAL_off(av);
4680 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4681 /* av now has a refcnt of 2, which avoids it getting freed
4682 * before us during global cleanup. The extra ref is removed
4683 * by magic_killbackrefs() when tsv is being freed */
4684 }
810b8aa5 4685 }
d91d49e8 4686 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4687 av_extend(av, AvFILLp(av)+1);
4688 }
4689 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4690}
4691
645c22ef
DM
4692/* delete a back-reference to ourselves from the backref magic associated
4693 * with the SV we point to.
4694 */
4695
1c846c1f 4696STATIC void
e15faf7d 4697S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4698{
97aff369 4699 dVAR;
86f55936 4700 AV *av = NULL;
810b8aa5
GS
4701 SV **svp;
4702 I32 i;
86f55936
NC
4703
4704 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4705 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4706 /* We mustn't attempt to "fix up" the hash here by moving the
4707 backreference array back to the hv_aux structure, as that is stored
4708 in the main HvARRAY(), and hfreentries assumes that no-one
4709 reallocates HvARRAY() while it is running. */
86f55936
NC
4710 }
4711 if (!av) {
4712 const MAGIC *const mg
4713 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4714 if (mg)
4715 av = (AV *)mg->mg_obj;
4716 }
4717 if (!av) {
e15faf7d
NC
4718 if (PL_in_clean_all)
4719 return;
cea2e8a9 4720 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4721 }
4722
4723 if (SvIS_FREED(av))
4724 return;
4725
810b8aa5 4726 svp = AvARRAY(av);
6a76db8b
NC
4727 /* We shouldn't be in here more than once, but for paranoia reasons lets
4728 not assume this. */
4729 for (i = AvFILLp(av); i >= 0; i--) {
4730 if (svp[i] == sv) {
4731 const SSize_t fill = AvFILLp(av);
4732 if (i != fill) {
4733 /* We weren't the last entry.
4734 An unordered list has this property that you can take the
4735 last element off the end to fill the hole, and it's still
4736 an unordered list :-)
4737 */
4738 svp[i] = svp[fill];
4739 }
a0714e2c 4740 svp[fill] = NULL;
6a76db8b
NC
4741 AvFILLp(av) = fill - 1;
4742 }
4743 }
810b8aa5
GS
4744}
4745
86f55936
NC
4746int
4747Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4748{
4749 SV **svp = AvARRAY(av);
4750
4751 PERL_UNUSED_ARG(sv);
4752
4753 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4754 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4755 if (svp && !SvIS_FREED(av)) {
4756 SV *const *const last = svp + AvFILLp(av);
4757
4758 while (svp <= last) {
4759 if (*svp) {
4760 SV *const referrer = *svp;
4761 if (SvWEAKREF(referrer)) {
4762 /* XXX Should we check that it hasn't changed? */
4763 SvRV_set(referrer, 0);
4764 SvOK_off(referrer);
4765 SvWEAKREF_off(referrer);
4766 } else if (SvTYPE(referrer) == SVt_PVGV ||
4767 SvTYPE(referrer) == SVt_PVLV) {
4768 /* You lookin' at me? */
4769 assert(GvSTASH(referrer));
4770 assert(GvSTASH(referrer) == (HV*)sv);
4771 GvSTASH(referrer) = 0;
4772 } else {
4773 Perl_croak(aTHX_
4774 "panic: magic_killbackrefs (flags=%"UVxf")",
4775 (UV)SvFLAGS(referrer));
4776 }
4777
a0714e2c 4778 *svp = NULL;
86f55936
NC
4779 }
4780 svp++;
4781 }
4782 }
4783 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4784 return 0;
4785}
4786
954c1994
GS
4787/*
4788=for apidoc sv_insert
4789
4790Inserts a string at the specified offset/length within the SV. Similar to
4791the Perl substr() function.
4792
4793=cut
4794*/
4795
79072805 4796void
e1ec3a88 4797Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4798{
97aff369 4799 dVAR;
79072805
LW
4800 register char *big;
4801 register char *mid;
4802 register char *midend;
4803 register char *bigend;
4804 register I32 i;
6ff81951 4805 STRLEN curlen;
1c846c1f 4806
79072805 4807
8990e307 4808 if (!bigstr)
cea2e8a9 4809 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4810 SvPV_force(bigstr, curlen);
60fa28ff 4811 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4812 if (offset + len > curlen) {
4813 SvGROW(bigstr, offset+len+1);
93524f2b 4814 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4815 SvCUR_set(bigstr, offset+len);
4816 }
79072805 4817
69b47968 4818 SvTAINT(bigstr);
79072805
LW
4819 i = littlelen - len;
4820 if (i > 0) { /* string might grow */
a0d0e21e 4821 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4822 mid = big + offset + len;
4823 midend = bigend = big + SvCUR(bigstr);
4824 bigend += i;
4825 *bigend = '\0';
4826 while (midend > mid) /* shove everything down */
4827 *--bigend = *--midend;
4828 Move(little,big+offset,littlelen,char);
b162af07 4829 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4830 SvSETMAGIC(bigstr);
4831 return;
4832 }
4833 else if (i == 0) {
463ee0b2 4834 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4835 SvSETMAGIC(bigstr);
4836 return;
4837 }
4838
463ee0b2 4839 big = SvPVX(bigstr);
79072805
LW
4840 mid = big + offset;
4841 midend = mid + len;
4842 bigend = big + SvCUR(bigstr);
4843
4844 if (midend > bigend)
cea2e8a9 4845 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4846
4847 if (mid - big > bigend - midend) { /* faster to shorten from end */
4848 if (littlelen) {
4849 Move(little, mid, littlelen,char);
4850 mid += littlelen;
4851 }
4852 i = bigend - midend;
4853 if (i > 0) {
4854 Move(midend, mid, i,char);
4855 mid += i;
4856 }
4857 *mid = '\0';
4858 SvCUR_set(bigstr, mid - big);
4859 }
155aba94 4860 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4861 midend -= littlelen;
4862 mid = midend;
4863 sv_chop(bigstr,midend-i);
4864 big += i;
4865 while (i--)
4866 *--midend = *--big;
4867 if (littlelen)
4868 Move(little, mid, littlelen,char);
4869 }
4870 else if (littlelen) {
4871 midend -= littlelen;
4872 sv_chop(bigstr,midend);
4873 Move(little,midend,littlelen,char);
4874 }
4875 else {
4876 sv_chop(bigstr,midend);
4877 }
4878 SvSETMAGIC(bigstr);
4879}
4880
c461cf8f
JH
4881/*
4882=for apidoc sv_replace
4883
4884Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4885The target SV physically takes over ownership of the body of the source SV
4886and inherits its flags; however, the target keeps any magic it owns,
4887and any magic in the source is discarded.
ff276b08 4888Note that this is a rather specialist SV copying operation; most of the
645c22ef 4889time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4890
4891=cut
4892*/
79072805
LW
4893
4894void
864dbfa3 4895Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4896{
97aff369 4897 dVAR;
a3b680e6 4898 const U32 refcnt = SvREFCNT(sv);
765f542d 4899 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4900 if (SvREFCNT(nsv) != 1) {
7437becc 4901 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4902 UVuf " != 1)", (UV) SvREFCNT(nsv));
4903 }
93a17b20 4904 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4905 if (SvMAGICAL(nsv))
4906 mg_free(nsv);
4907 else
4908 sv_upgrade(nsv, SVt_PVMG);
b162af07 4909 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4910 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4911 SvMAGICAL_off(sv);
b162af07 4912 SvMAGIC_set(sv, NULL);
93a17b20 4913 }
79072805
LW
4914 SvREFCNT(sv) = 0;
4915 sv_clear(sv);
477f5d66 4916 assert(!SvREFCNT(sv));
fd0854ff
DM
4917#ifdef DEBUG_LEAKING_SCALARS
4918 sv->sv_flags = nsv->sv_flags;
4919 sv->sv_any = nsv->sv_any;
4920 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4921 sv->sv_u = nsv->sv_u;
fd0854ff 4922#else
79072805 4923 StructCopy(nsv,sv,SV);
fd0854ff 4924#endif
7b2c381c
NC
4925 /* Currently could join these into one piece of pointer arithmetic, but
4926 it would be unclear. */
4927 if(SvTYPE(sv) == SVt_IV)
4928 SvANY(sv)
339049b0 4929 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4930 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4931 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4932 }
4933
fd0854ff 4934
f8c7b90f 4935#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4936 if (SvIsCOW_normal(nsv)) {
4937 /* We need to follow the pointers around the loop to make the
4938 previous SV point to sv, rather than nsv. */
4939 SV *next;
4940 SV *current = nsv;
4941 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4942 assert(next);
4943 current = next;
3f7c398e 4944 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4945 }
4946 /* Make the SV before us point to the SV after us. */
4947 if (DEBUG_C_TEST) {
4948 PerlIO_printf(Perl_debug_log, "previous is\n");
4949 sv_dump(current);
a29f6d03
NC
4950 PerlIO_printf(Perl_debug_log,
4951 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4952 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4953 }
a29f6d03 4954 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4955 }
4956#endif
79072805 4957 SvREFCNT(sv) = refcnt;
1edc1566 4958 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4959 SvREFCNT(nsv) = 0;
463ee0b2 4960 del_SV(nsv);
79072805
LW
4961}
4962
c461cf8f
JH
4963/*
4964=for apidoc sv_clear
4965
645c22ef
DM
4966Clear an SV: call any destructors, free up any memory used by the body,
4967and free the body itself. The SV's head is I<not> freed, although
4968its type is set to all 1's so that it won't inadvertently be assumed
4969to be live during global destruction etc.
4970This function should only be called when REFCNT is zero. Most of the time
4971you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4972instead.
c461cf8f
JH
4973
4974=cut
4975*/
4976
79072805 4977void
864dbfa3 4978Perl_sv_clear(pTHX_ register SV *sv)
79072805 4979{
27da23d5 4980 dVAR;
82bb6deb 4981 const U32 type = SvTYPE(sv);
8edfc514
NC
4982 const struct body_details *const sv_type_details
4983 = bodies_by_type + type;
82bb6deb 4984
79072805
LW
4985 assert(sv);
4986 assert(SvREFCNT(sv) == 0);
4987
d2a0f284
JC
4988 if (type <= SVt_IV) {
4989 /* See the comment in sv.h about the collusion between this early
4990 return and the overloading of the NULL and IV slots in the size
4991 table. */
82bb6deb 4992 return;
d2a0f284 4993 }
82bb6deb 4994
ed6116ce 4995 if (SvOBJECT(sv)) {
3280af22 4996 if (PL_defstash) { /* Still have a symbol table? */
39644a26 4997 dSP;
893645bd 4998 HV* stash;
d460ef45 4999 do {
b464bac0 5000 CV* destructor;
4e8e7886 5001 stash = SvSTASH(sv);
32251b26 5002 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5003 if (destructor) {
1b6737cc 5004 SV* const tmpref = newRV(sv);
5cc433a6 5005 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5006 ENTER;
e788e7d3 5007 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5008 EXTEND(SP, 2);
5009 PUSHMARK(SP);
5cc433a6 5010 PUSHs(tmpref);
4e8e7886 5011 PUTBACK;
44389ee9 5012 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5013
5014
d3acc0f7 5015 POPSTACK;
3095d977 5016 SPAGAIN;
4e8e7886 5017 LEAVE;
5cc433a6
AB
5018 if(SvREFCNT(tmpref) < 2) {
5019 /* tmpref is not kept alive! */
5020 SvREFCNT(sv)--;
b162af07 5021 SvRV_set(tmpref, NULL);
5cc433a6
AB
5022 SvROK_off(tmpref);
5023 }
5024 SvREFCNT_dec(tmpref);
4e8e7886
GS
5025 }
5026 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5027
6f44e0a4
JP
5028
5029 if (SvREFCNT(sv)) {
5030 if (PL_in_clean_objs)
cea2e8a9 5031 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5032 HvNAME_get(stash));
6f44e0a4
JP
5033 /* DESTROY gave object new lease on life */
5034 return;
5035 }
a0d0e21e 5036 }
4e8e7886 5037
a0d0e21e 5038 if (SvOBJECT(sv)) {
4e8e7886 5039 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5040 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5041 if (type != SVt_PVIO)
3280af22 5042 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5043 }
463ee0b2 5044 }
82bb6deb 5045 if (type >= SVt_PVMG) {
e736a858
NC
5046 HV *ourstash;
5047 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5048 (ourstash = OURSTASH(sv))) {
5049 SvREFCNT_dec(ourstash);
5050 } else if (SvMAGIC(sv))
524189f1 5051 mg_free(sv);
00b1698f 5052 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5053 SvREFCNT_dec(SvSTASH(sv));
5054 }
82bb6deb 5055 switch (type) {
8990e307 5056 case SVt_PVIO:
df0bd2f4
GS
5057 if (IoIFP(sv) &&
5058 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5059 IoIFP(sv) != PerlIO_stdout() &&
5060 IoIFP(sv) != PerlIO_stderr())
93578b34 5061 {
f2b5be74 5062 io_close((IO*)sv, FALSE);
93578b34 5063 }
1d7c1841 5064 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5065 PerlDir_close(IoDIRP(sv));
1d7c1841 5066 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5067 Safefree(IoTOP_NAME(sv));
5068 Safefree(IoFMT_NAME(sv));
5069 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5070 goto freescalar;
79072805 5071 case SVt_PVBM:
a0d0e21e 5072 goto freescalar;
79072805 5073 case SVt_PVCV:
748a9306 5074 case SVt_PVFM:
85e6fe83 5075 cv_undef((CV*)sv);
a0d0e21e 5076 goto freescalar;
79072805 5077 case SVt_PVHV:
86f55936 5078 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5079 hv_undef((HV*)sv);
a0d0e21e 5080 break;
79072805 5081 case SVt_PVAV:
85e6fe83 5082 av_undef((AV*)sv);
a0d0e21e 5083 break;
02270b4e 5084 case SVt_PVLV:
dd28f7bb
DM
5085 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5086 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5087 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5088 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5089 }
5090 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5091 SvREFCNT_dec(LvTARG(sv));
02270b4e 5092 goto freescalar;
a0d0e21e 5093 case SVt_PVGV:
1edc1566 5094 gp_free((GV*)sv);
acda4c6a
NC
5095 if (GvNAME_HEK(sv)) {
5096 unshare_hek(GvNAME_HEK(sv));
5097 }
893645bd
NC
5098 /* If we're in a stash, we don't own a reference to it. However it does
5099 have a back reference to us, which needs to be cleared. */
5100 if (GvSTASH(sv))
5101 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5102 case SVt_PVMG:
79072805
LW
5103 case SVt_PVNV:
5104 case SVt_PVIV:
a0d0e21e 5105 freescalar:
5228ca4e
NC
5106 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5107 if (SvOOK(sv)) {
93524f2b 5108 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5109 /* Don't even bother with turning off the OOK flag. */
5110 }
79072805 5111 case SVt_PV:
a0d0e21e 5112 case SVt_RV:
810b8aa5 5113 if (SvROK(sv)) {
b37c2d43 5114 SV * const target = SvRV(sv);
810b8aa5 5115 if (SvWEAKREF(sv))
e15faf7d 5116 sv_del_backref(target, sv);
810b8aa5 5117 else
e15faf7d 5118 SvREFCNT_dec(target);
810b8aa5 5119 }
f8c7b90f 5120#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5121 else if (SvPVX_const(sv)) {
765f542d
NC
5122 if (SvIsCOW(sv)) {
5123 /* I believe I need to grab the global SV mutex here and
5124 then recheck the COW status. */
46187eeb
NC
5125 if (DEBUG_C_TEST) {
5126 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5127 sv_dump(sv);
46187eeb 5128 }
bdd68bc3
NC
5129 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5130 SV_COW_NEXT_SV(sv));
765f542d
NC
5131 /* And drop it here. */
5132 SvFAKE_off(sv);
5133 } else if (SvLEN(sv)) {
3f7c398e 5134 Safefree(SvPVX_const(sv));
765f542d
NC
5135 }
5136 }
5137#else
3f7c398e 5138 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5139 Safefree(SvPVX_mutable(sv));
3f7c398e 5140 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5141 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5142 SvFAKE_off(sv);
5143 }
765f542d 5144#endif
79072805
LW
5145 break;
5146 case SVt_NV:
79072805
LW
5147 break;
5148 }
5149
893645bd
NC
5150 SvFLAGS(sv) &= SVf_BREAK;
5151 SvFLAGS(sv) |= SVTYPEMASK;
5152
8edfc514 5153 if (sv_type_details->arena) {
b9502f15 5154 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5155 &PL_body_roots[type]);
5156 }
d2a0f284 5157 else if (sv_type_details->body_size) {
8edfc514
NC
5158 my_safefree(SvANY(sv));
5159 }
79072805
LW
5160}
5161
645c22ef
DM
5162/*
5163=for apidoc sv_newref
5164
5165Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5166instead.
5167
5168=cut
5169*/
5170
79072805 5171SV *
864dbfa3 5172Perl_sv_newref(pTHX_ SV *sv)
79072805 5173{
96a5add6 5174 PERL_UNUSED_CONTEXT;
463ee0b2 5175 if (sv)
4db098f4 5176 (SvREFCNT(sv))++;
79072805
LW
5177 return sv;
5178}
5179
c461cf8f
JH
5180/*
5181=for apidoc sv_free
5182
645c22ef
DM
5183Decrement an SV's reference count, and if it drops to zero, call
5184C<sv_clear> to invoke destructors and free up any memory used by
5185the body; finally, deallocate the SV's head itself.
5186Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5187
5188=cut
5189*/
5190
79072805 5191void
864dbfa3 5192Perl_sv_free(pTHX_ SV *sv)
79072805 5193{
27da23d5 5194 dVAR;
79072805
LW
5195 if (!sv)
5196 return;
a0d0e21e
LW
5197 if (SvREFCNT(sv) == 0) {
5198 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5199 /* this SV's refcnt has been artificially decremented to
5200 * trigger cleanup */
a0d0e21e 5201 return;
3280af22 5202 if (PL_in_clean_all) /* All is fair */
1edc1566 5203 return;
d689ffdd
JP
5204 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5205 /* make sure SvREFCNT(sv)==0 happens very seldom */
5206 SvREFCNT(sv) = (~(U32)0)/2;
5207 return;
5208 }
41e4abd8 5209 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5210 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5211 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5212 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5213#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5214 Perl_dump_sv_child(aTHX_ sv);
5215#endif
5216 }
79072805
LW
5217 return;
5218 }
4db098f4 5219 if (--(SvREFCNT(sv)) > 0)
8990e307 5220 return;
8c4d3c90
NC
5221 Perl_sv_free2(aTHX_ sv);
5222}
5223
5224void
5225Perl_sv_free2(pTHX_ SV *sv)
5226{
27da23d5 5227 dVAR;
463ee0b2
LW
5228#ifdef DEBUGGING
5229 if (SvTEMP(sv)) {
0453d815 5230 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5231 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5232 "Attempt to free temp prematurely: SV 0x%"UVxf
5233 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5234 return;
79072805 5235 }
463ee0b2 5236#endif
d689ffdd
JP
5237 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5238 /* make sure SvREFCNT(sv)==0 happens very seldom */
5239 SvREFCNT(sv) = (~(U32)0)/2;
5240 return;
5241 }
79072805 5242 sv_clear(sv);
477f5d66
CS
5243 if (! SvREFCNT(sv))
5244 del_SV(sv);
79072805
LW
5245}
5246
954c1994
GS
5247/*
5248=for apidoc sv_len
5249
645c22ef
DM
5250Returns the length of the string in the SV. Handles magic and type
5251coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5252
5253=cut
5254*/
5255
79072805 5256STRLEN
864dbfa3 5257Perl_sv_len(pTHX_ register SV *sv)
79072805 5258{
463ee0b2 5259 STRLEN len;
79072805
LW
5260
5261 if (!sv)
5262 return 0;
5263
8990e307 5264 if (SvGMAGICAL(sv))
565764a8 5265 len = mg_length(sv);
8990e307 5266 else
4d84ee25 5267 (void)SvPV_const(sv, len);
463ee0b2 5268 return len;
79072805
LW
5269}
5270
c461cf8f
JH
5271/*
5272=for apidoc sv_len_utf8
5273
5274Returns the number of characters in the string in an SV, counting wide
1e54db1a 5275UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5276
5277=cut
5278*/
5279
7e8c5dac
HS
5280/*
5281 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5282 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5283 * (Note that the mg_len is not the length of the mg_ptr field.
5284 * This allows the cache to store the character length of the string without
5285 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5286 *
7e8c5dac
HS
5287 */
5288
a0ed51b3 5289STRLEN
864dbfa3 5290Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5291{
a0ed51b3
LW
5292 if (!sv)
5293 return 0;
5294
a0ed51b3 5295 if (SvGMAGICAL(sv))
b76347f2 5296 return mg_length(sv);
a0ed51b3 5297 else
b76347f2 5298 {
26346457 5299 STRLEN len;
e62f0680 5300 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5301
26346457
NC
5302 if (PL_utf8cache) {
5303 STRLEN ulen;
5304 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5305
5306 if (mg && mg->mg_len != -1) {
5307 ulen = mg->mg_len;
5308 if (PL_utf8cache < 0) {
5309 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5310 if (real != ulen) {
5311 /* Need to turn the assertions off otherwise we may
5312 recurse infinitely while printing error messages.
5313 */
5314 SAVEI8(PL_utf8cache);
5315 PL_utf8cache = 0;
5316 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5317 " real %"UVf" for %"SVf,
26346457
NC
5318 (UV) ulen, (UV) real, sv);
5319 }
5320 }
5321 }
5322 else {
5323 ulen = Perl_utf8_length(aTHX_ s, s + len);
5324 if (!SvREADONLY(sv)) {
5325 if (!mg) {
5326 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5327 &PL_vtbl_utf8, 0, 0);
5328 }
cb9e20bb 5329 assert(mg);
26346457 5330 mg->mg_len = ulen;
cb9e20bb 5331 }
cb9e20bb 5332 }
26346457 5333 return ulen;
7e8c5dac 5334 }
26346457 5335 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5336 }
5337}
5338
9564a3bd
NC
5339/* Walk forwards to find the byte corresponding to the passed in UTF-8
5340 offset. */
bdf30dd6
NC
5341static STRLEN
5342S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send,
5343 STRLEN uoffset)
5344{
5345 const U8 *s = start;
5346
d4c19fe8
AL
5347 PERL_UNUSED_CONTEXT;
5348
bdf30dd6
NC
5349 while (s < send && uoffset--)
5350 s += UTF8SKIP(s);
5351 if (s > send) {
5352 /* This is the existing behaviour. Possibly it should be a croak, as
5353 it's actually a bounds error */
5354 s = send;
5355 }
5356 return s - start;
5357}
5358
9564a3bd
NC
5359/* Given the length of the string in both bytes and UTF-8 characters, decide
5360 whether to walk forwards or backwards to find the byte corresponding to
5361 the passed in UTF-8 offset. */
c336ad0b
NC
5362static STRLEN
5363S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send,
5364 STRLEN uoffset, STRLEN uend)
5365{
5366 STRLEN backw = uend - uoffset;
5367 if (uoffset < 2 * backw) {
25a8a4ef 5368 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5369 forward (that's where the 2 * backw comes from).
5370 (The real figure of course depends on the UTF-8 data.) */
5371 return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset);
5372 }
5373
5374 while (backw--) {
5375 send--;
5376 while (UTF8_IS_CONTINUATION(*send))
5377 send--;
5378 }
5379 return send - start;
5380}
5381
9564a3bd
NC
5382/* For the string representation of the given scalar, find the byte
5383 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5384 give another position in the string, *before* the sought offset, which
5385 (which is always true, as 0, 0 is a valid pair of positions), which should
5386 help reduce the amount of linear searching.
5387 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5388 will be used to reduce the amount of linear searching. The cache will be
5389 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5390static STRLEN
5391S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5392 const U8 *const send, STRLEN uoffset,
5393 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5394 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5395 bool found = FALSE;
5396
75c33c12
NC
5397 assert (uoffset >= uoffset0);
5398
c336ad0b 5399 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5400 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5401 if ((*mgp)->mg_ptr) {
5402 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5403 if (cache[0] == uoffset) {
5404 /* An exact match. */
5405 return cache[1];
5406 }
ab455f60
NC
5407 if (cache[2] == uoffset) {
5408 /* An exact match. */
5409 return cache[3];
5410 }
668af93f
NC
5411
5412 if (cache[0] < uoffset) {
d8b2e1f9
NC
5413 /* The cache already knows part of the way. */
5414 if (cache[0] > uoffset0) {
5415 /* The cache knows more than the passed in pair */
5416 uoffset0 = cache[0];
5417 boffset0 = cache[1];
5418 }
5419 if ((*mgp)->mg_len != -1) {
5420 /* And we know the end too. */
5421 boffset = boffset0
5422 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5423 uoffset - uoffset0,
5424 (*mgp)->mg_len - uoffset0);
5425 } else {
5426 boffset = boffset0
5427 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5428 send, uoffset - uoffset0);
5429 }
dd7c5fd3
NC
5430 }
5431 else if (cache[2] < uoffset) {
5432 /* We're between the two cache entries. */
5433 if (cache[2] > uoffset0) {
5434 /* and the cache knows more than the passed in pair */
5435 uoffset0 = cache[2];
5436 boffset0 = cache[3];
5437 }
5438
668af93f
NC
5439 boffset = boffset0
5440 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5441 start + cache[1],
5442 uoffset - uoffset0,
5443 cache[0] - uoffset0);
dd7c5fd3
NC
5444 } else {
5445 boffset = boffset0
5446 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5447 start + cache[3],
5448 uoffset - uoffset0,
5449 cache[2] - uoffset0);
d8b2e1f9 5450 }
668af93f 5451 found = TRUE;
d8b2e1f9
NC
5452 }
5453 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5454 /* If we can take advantage of a passed in offset, do so. */
5455 /* In fact, offset0 is either 0, or less than offset, so don't
5456 need to worry about the other possibility. */
5457 boffset = boffset0
5458 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5459 uoffset - uoffset0,
5460 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5461 found = TRUE;
5462 }
28ccbf94 5463 }
c336ad0b
NC
5464
5465 if (!found || PL_utf8cache < 0) {
75c33c12
NC
5466 const STRLEN real_boffset
5467 = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5468 send, uoffset - uoffset0);
5469
c336ad0b
NC
5470 if (found && PL_utf8cache < 0) {
5471 if (real_boffset != boffset) {
5472 /* Need to turn the assertions off otherwise we may recurse
5473 infinitely while printing error messages. */
5474 SAVEI8(PL_utf8cache);
5475 PL_utf8cache = 0;
5476 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5477 " real %"UVf" for %"SVf,
5478 (UV) boffset, (UV) real_boffset, sv);
5479 }
5480 }
5481 boffset = real_boffset;
28ccbf94 5482 }
0905937d 5483
ab455f60 5484 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5485 return boffset;
5486}
5487
9564a3bd
NC
5488
5489/*
5490=for apidoc sv_pos_u2b
5491
5492Converts the value pointed to by offsetp from a count of UTF-8 chars from
5493the start of the string, to a count of the equivalent number of bytes; if
5494lenp is non-zero, it does the same to lenp, but this time starting from
5495the offset, rather than from the start of the string. Handles magic and
5496type coercion.
5497
5498=cut
5499*/
5500
5501/*
5502 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5503 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5504 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5505 *
5506 */
5507
a0ed51b3 5508void
864dbfa3 5509Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5510{
245d4a47 5511 const U8 *start;
a0ed51b3
LW
5512 STRLEN len;
5513
5514 if (!sv)
5515 return;
5516
245d4a47 5517 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5518 if (len) {
bdf30dd6
NC
5519 STRLEN uoffset = (STRLEN) *offsetp;
5520 const U8 * const send = start + len;
0905937d 5521 MAGIC *mg = NULL;
28ccbf94
NC
5522 STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send,
5523 uoffset, 0, 0);
bdf30dd6
NC
5524
5525 *offsetp = (I32) boffset;
5526
5527 if (lenp) {
28ccbf94
NC
5528 /* Convert the relative offset to absolute. */
5529 STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5530 STRLEN boffset2
5531 = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2,
5532 uoffset, boffset) - boffset;
bdf30dd6 5533
28ccbf94 5534 *lenp = boffset2;
bdf30dd6 5535 }
7e8c5dac
HS
5536 }
5537 else {
5538 *offsetp = 0;
5539 if (lenp)
5540 *lenp = 0;
a0ed51b3 5541 }
e23c8137 5542
a0ed51b3
LW
5543 return;
5544}
5545
9564a3bd
NC
5546/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5547 byte length pairing. The (byte) length of the total SV is passed in too,
5548 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5549 may not have updated SvCUR, so we can't rely on reading it directly.
5550
5551 The proffered utf8/byte length pairing isn't used if the cache already has
5552 two pairs, and swapping either for the proffered pair would increase the
5553 RMS of the intervals between known byte offsets.
5554
5555 The cache itself consists of 4 STRLEN values
5556 0: larger UTF-8 offset
5557 1: corresponding byte offset
5558 2: smaller UTF-8 offset
5559 3: corresponding byte offset
5560
5561 Unused cache pairs have the value 0, 0.
5562 Keeping the cache "backwards" means that the invariant of
5563 cache[0] >= cache[2] is maintained even with empty slots, which means that
5564 the code that uses it doesn't need to worry if only 1 entry has actually
5565 been set to non-zero. It also makes the "position beyond the end of the
5566 cache" logic much simpler, as the first slot is always the one to start
5567 from.
645c22ef 5568*/
ec07b5e0 5569static void
ab455f60
NC
5570S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5571 STRLEN blen)
ec07b5e0
NC
5572{
5573 STRLEN *cache;
5574 if (SvREADONLY(sv))
5575 return;
5576
5577 if (!*mgp) {
5578 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5579 0);
5580 (*mgp)->mg_len = -1;
5581 }
5582 assert(*mgp);
5583
5584 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5585 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5586 (*mgp)->mg_ptr = (char *) cache;
5587 }
5588 assert(cache);
5589
5590 if (PL_utf8cache < 0) {
ef816a78 5591 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5592 const U8 *const end = start + byte;
5593 STRLEN realutf8 = 0;
5594
5595 while (start < end) {
5596 start += UTF8SKIP(start);
5597 realutf8++;
5598 }
5599
5600 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5601 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5602 doesn't? I don't know whether this difference was introduced with
5603 the caching code in 5.8.1. */
ec07b5e0
NC
5604
5605 if (realutf8 != utf8) {
5606 /* Need to turn the assertions off otherwise we may recurse
5607 infinitely while printing error messages. */
5608 SAVEI8(PL_utf8cache);
5609 PL_utf8cache = 0;
5610 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5611 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5612 }
5613 }
ab455f60
NC
5614
5615 /* Cache is held with the later position first, to simplify the code
5616 that deals with unbounded ends. */
5617
5618 ASSERT_UTF8_CACHE(cache);
5619 if (cache[1] == 0) {
5620 /* Cache is totally empty */
5621 cache[0] = utf8;
5622 cache[1] = byte;
5623 } else if (cache[3] == 0) {
5624 if (byte > cache[1]) {
5625 /* New one is larger, so goes first. */
5626 cache[2] = cache[0];
5627 cache[3] = cache[1];
5628 cache[0] = utf8;
5629 cache[1] = byte;
5630 } else {
5631 cache[2] = utf8;
5632 cache[3] = byte;
5633 }
5634 } else {
5635#define THREEWAY_SQUARE(a,b,c,d) \
5636 ((float)((d) - (c))) * ((float)((d) - (c))) \
5637 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5638 + ((float)((b) - (a))) * ((float)((b) - (a)))
5639
5640 /* Cache has 2 slots in use, and we know three potential pairs.
5641 Keep the two that give the lowest RMS distance. Do the
5642 calcualation in bytes simply because we always know the byte
5643 length. squareroot has the same ordering as the positive value,
5644 so don't bother with the actual square root. */
5645 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5646 if (byte > cache[1]) {
5647 /* New position is after the existing pair of pairs. */
5648 const float keep_earlier
5649 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5650 const float keep_later
5651 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5652
5653 if (keep_later < keep_earlier) {
5654 if (keep_later < existing) {
5655 cache[2] = cache[0];
5656 cache[3] = cache[1];
5657 cache[0] = utf8;
5658 cache[1] = byte;
5659 }
5660 }
5661 else {
5662 if (keep_earlier < existing) {
5663 cache[0] = utf8;
5664 cache[1] = byte;
5665 }
5666 }
5667 }
57d7fbf1
NC
5668 else if (byte > cache[3]) {
5669 /* New position is between the existing pair of pairs. */
5670 const float keep_earlier
5671 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5672 const float keep_later
5673 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5674
5675 if (keep_later < keep_earlier) {
5676 if (keep_later < existing) {
5677 cache[2] = utf8;
5678 cache[3] = byte;
5679 }
5680 }
5681 else {
5682 if (keep_earlier < existing) {
5683 cache[0] = utf8;
5684 cache[1] = byte;
5685 }
5686 }
5687 }
5688 else {
5689 /* New position is before the existing pair of pairs. */
5690 const float keep_earlier
5691 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5692 const float keep_later
5693 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5694
5695 if (keep_later < keep_earlier) {
5696 if (keep_later < existing) {
5697 cache[2] = utf8;
5698 cache[3] = byte;
5699 }
5700 }
5701 else {
5702 if (keep_earlier < existing) {
5703 cache[0] = cache[2];
5704 cache[1] = cache[3];
5705 cache[2] = utf8;
5706 cache[3] = byte;
5707 }
5708 }
5709 }
ab455f60 5710 }
0905937d 5711 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5712}
5713
5714/* If we don't know the character offset of the end of a region, our only
5715 option is to walk forwards to the target byte offset. */
5716static STRLEN
5717S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5718{
5719 STRLEN len = 0;
5720 while (s < target) {
5721 STRLEN n = 1;
5722
5723 /* Call utf8n_to_uvchr() to validate the sequence
5724 * (unless a simple non-UTF character) */
5725 if (!UTF8_IS_INVARIANT(*s))
5726 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5727 if (n > 0) {
5728 s += n;
5729 len++;
5730 }
5731 else
5732 break;
5733 }
5734 return len;
5735}
5736
5737/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5738 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5739 backward is half the speed of walking forward. */
ec07b5e0
NC
5740static STRLEN
5741S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5742 STRLEN endu)
5743{
5744 const STRLEN forw = target - s;
5745 STRLEN backw = end - target;
5746
5747 if (forw < 2 * backw) {
5748 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5749 }
5750
5751 while (end > target) {
5752 end--;
5753 while (UTF8_IS_CONTINUATION(*end)) {
5754 end--;
5755 }
5756 endu--;
5757 }
5758 return endu;
5759}
5760
9564a3bd
NC
5761/*
5762=for apidoc sv_pos_b2u
5763
5764Converts the value pointed to by offsetp from a count of bytes from the
5765start of the string, to a count of the equivalent number of UTF-8 chars.
5766Handles magic and type coercion.
5767
5768=cut
5769*/
5770
5771/*
5772 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5773 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5774 * byte offsets.
5775 *
5776 */
a0ed51b3 5777void
7e8c5dac 5778Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5779{
83003860 5780 const U8* s;
ec07b5e0 5781 const STRLEN byte = *offsetp;
7087a21c 5782 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5783 STRLEN blen;
ec07b5e0
NC
5784 MAGIC* mg = NULL;
5785 const U8* send;
a922f900 5786 bool found = FALSE;
a0ed51b3
LW
5787
5788 if (!sv)
5789 return;
5790
ab455f60 5791 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5792
ab455f60 5793 if (blen < byte)
ec07b5e0 5794 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5795
ec07b5e0 5796 send = s + byte;
a67d7df9 5797
ffca234a
NC
5798 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5799 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5800 if (mg->mg_ptr) {
d4c19fe8 5801 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5802 if (cache[1] == byte) {
ec07b5e0
NC
5803 /* An exact match. */
5804 *offsetp = cache[0];
ec07b5e0 5805 return;
7e8c5dac 5806 }
ab455f60
NC
5807 if (cache[3] == byte) {
5808 /* An exact match. */
5809 *offsetp = cache[2];
5810 return;
5811 }
668af93f
NC
5812
5813 if (cache[1] < byte) {
ec07b5e0 5814 /* We already know part of the way. */
b9f984a5
NC
5815 if (mg->mg_len != -1) {
5816 /* Actually, we know the end too. */
5817 len = cache[0]
5818 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5819 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5820 } else {
5821 len = cache[0]
5822 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5823 }
7e8c5dac 5824 }
9f985e4c
NC
5825 else if (cache[3] < byte) {
5826 /* We're between the two cached pairs, so we do the calculation
5827 offset by the byte/utf-8 positions for the earlier pair,
5828 then add the utf-8 characters from the string start to
5829 there. */
5830 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5831 s + cache[1], cache[0] - cache[2])
5832 + cache[2];
5833
5834 }
5835 else { /* cache[3] > byte */
5836 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5837 cache[2]);
7e8c5dac 5838
7e8c5dac 5839 }
ec07b5e0 5840 ASSERT_UTF8_CACHE(cache);
a922f900 5841 found = TRUE;
ffca234a 5842 } else if (mg->mg_len != -1) {
ab455f60 5843 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5844 found = TRUE;
7e8c5dac 5845 }
a0ed51b3 5846 }
a922f900
NC
5847 if (!found || PL_utf8cache < 0) {
5848 const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5849
5850 if (found && PL_utf8cache < 0) {
5851 if (len != real_len) {
5852 /* Need to turn the assertions off otherwise we may recurse
5853 infinitely while printing error messages. */
5854 SAVEI8(PL_utf8cache);
5855 PL_utf8cache = 0;
5856 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5857 " real %"UVf" for %"SVf,
5858 (UV) len, (UV) real_len, sv);
5859 }
5860 }
5861 len = real_len;
ec07b5e0
NC
5862 }
5863 *offsetp = len;
5864
ab455f60 5865 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5866}
5867
954c1994
GS
5868/*
5869=for apidoc sv_eq
5870
5871Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5872identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5873coerce its args to strings if necessary.
954c1994
GS
5874
5875=cut
5876*/
5877
79072805 5878I32
e01b9e88 5879Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5880{
97aff369 5881 dVAR;
e1ec3a88 5882 const char *pv1;
463ee0b2 5883 STRLEN cur1;
e1ec3a88 5884 const char *pv2;
463ee0b2 5885 STRLEN cur2;
e01b9e88 5886 I32 eq = 0;
bd61b366 5887 char *tpv = NULL;
a0714e2c 5888 SV* svrecode = NULL;
79072805 5889
e01b9e88 5890 if (!sv1) {
79072805
LW
5891 pv1 = "";
5892 cur1 = 0;
5893 }
463ee0b2 5894 else
4d84ee25 5895 pv1 = SvPV_const(sv1, cur1);
79072805 5896
e01b9e88
SC
5897 if (!sv2){
5898 pv2 = "";
5899 cur2 = 0;
92d29cee 5900 }
e01b9e88 5901 else
4d84ee25 5902 pv2 = SvPV_const(sv2, cur2);
79072805 5903
cf48d248 5904 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5905 /* Differing utf8ness.
5906 * Do not UTF8size the comparands as a side-effect. */
5907 if (PL_encoding) {
5908 if (SvUTF8(sv1)) {
553e1bcc
AT
5909 svrecode = newSVpvn(pv2, cur2);
5910 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5911 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5912 }
5913 else {
553e1bcc
AT
5914 svrecode = newSVpvn(pv1, cur1);
5915 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5916 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5917 }
5918 /* Now both are in UTF-8. */
0a1bd7ac
DM
5919 if (cur1 != cur2) {
5920 SvREFCNT_dec(svrecode);
799ef3cb 5921 return FALSE;
0a1bd7ac 5922 }
799ef3cb
JH
5923 }
5924 else {
5925 bool is_utf8 = TRUE;
5926
5927 if (SvUTF8(sv1)) {
5928 /* sv1 is the UTF-8 one,
5929 * if is equal it must be downgrade-able */
9d4ba2ae 5930 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5931 &cur1, &is_utf8);
5932 if (pv != pv1)
553e1bcc 5933 pv1 = tpv = pv;
799ef3cb
JH
5934 }
5935 else {
5936 /* sv2 is the UTF-8 one,
5937 * if is equal it must be downgrade-able */
9d4ba2ae 5938 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5939 &cur2, &is_utf8);
5940 if (pv != pv2)
553e1bcc 5941 pv2 = tpv = pv;
799ef3cb
JH
5942 }
5943 if (is_utf8) {
5944 /* Downgrade not possible - cannot be eq */
bf694877 5945 assert (tpv == 0);
799ef3cb
JH
5946 return FALSE;
5947 }
5948 }
cf48d248
JH
5949 }
5950
5951 if (cur1 == cur2)
765f542d 5952 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5953
b37c2d43 5954 SvREFCNT_dec(svrecode);
553e1bcc
AT
5955 if (tpv)
5956 Safefree(tpv);
cf48d248 5957
e01b9e88 5958 return eq;
79072805
LW
5959}
5960
954c1994
GS
5961/*
5962=for apidoc sv_cmp
5963
5964Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5965string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5966C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5967coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5968
5969=cut
5970*/
5971
79072805 5972I32
e01b9e88 5973Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5974{
97aff369 5975 dVAR;
560a288e 5976 STRLEN cur1, cur2;
e1ec3a88 5977 const char *pv1, *pv2;
bd61b366 5978 char *tpv = NULL;
cf48d248 5979 I32 cmp;
a0714e2c 5980 SV *svrecode = NULL;
560a288e 5981
e01b9e88
SC
5982 if (!sv1) {
5983 pv1 = "";
560a288e
GS
5984 cur1 = 0;
5985 }
e01b9e88 5986 else
4d84ee25 5987 pv1 = SvPV_const(sv1, cur1);
560a288e 5988
553e1bcc 5989 if (!sv2) {
e01b9e88 5990 pv2 = "";
560a288e
GS
5991 cur2 = 0;
5992 }
e01b9e88 5993 else
4d84ee25 5994 pv2 = SvPV_const(sv2, cur2);
79072805 5995
cf48d248 5996 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5997 /* Differing utf8ness.
5998 * Do not UTF8size the comparands as a side-effect. */
cf48d248 5999 if (SvUTF8(sv1)) {
799ef3cb 6000 if (PL_encoding) {
553e1bcc
AT
6001 svrecode = newSVpvn(pv2, cur2);
6002 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6003 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6004 }
6005 else {
e1ec3a88 6006 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6007 }
cf48d248
JH
6008 }
6009 else {
799ef3cb 6010 if (PL_encoding) {
553e1bcc
AT
6011 svrecode = newSVpvn(pv1, cur1);
6012 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6013 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6014 }
6015 else {
e1ec3a88 6016 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6017 }
cf48d248
JH
6018 }
6019 }
6020
e01b9e88 6021 if (!cur1) {
cf48d248 6022 cmp = cur2 ? -1 : 0;
e01b9e88 6023 } else if (!cur2) {
cf48d248
JH
6024 cmp = 1;
6025 } else {
e1ec3a88 6026 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6027
6028 if (retval) {
cf48d248 6029 cmp = retval < 0 ? -1 : 1;
e01b9e88 6030 } else if (cur1 == cur2) {
cf48d248
JH
6031 cmp = 0;
6032 } else {
6033 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6034 }
cf48d248 6035 }
16660edb 6036
b37c2d43 6037 SvREFCNT_dec(svrecode);
553e1bcc
AT
6038 if (tpv)
6039 Safefree(tpv);
cf48d248
JH
6040
6041 return cmp;
bbce6d69 6042}
16660edb 6043
c461cf8f
JH
6044/*
6045=for apidoc sv_cmp_locale
6046
645c22ef
DM
6047Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6048'use bytes' aware, handles get magic, and will coerce its args to strings
6049if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6050
6051=cut
6052*/
6053
bbce6d69 6054I32
864dbfa3 6055Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6056{
97aff369 6057 dVAR;
36477c24 6058#ifdef USE_LOCALE_COLLATE
16660edb 6059
bbce6d69 6060 char *pv1, *pv2;
6061 STRLEN len1, len2;
6062 I32 retval;
16660edb 6063
3280af22 6064 if (PL_collation_standard)
bbce6d69 6065 goto raw_compare;
16660edb 6066
bbce6d69 6067 len1 = 0;
8ac85365 6068 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6069 len2 = 0;
8ac85365 6070 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6071
bbce6d69 6072 if (!pv1 || !len1) {
6073 if (pv2 && len2)
6074 return -1;
6075 else
6076 goto raw_compare;
6077 }
6078 else {
6079 if (!pv2 || !len2)
6080 return 1;
6081 }
16660edb 6082
bbce6d69 6083 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6084
bbce6d69 6085 if (retval)
16660edb 6086 return retval < 0 ? -1 : 1;
6087
bbce6d69 6088 /*
6089 * When the result of collation is equality, that doesn't mean
6090 * that there are no differences -- some locales exclude some
6091 * characters from consideration. So to avoid false equalities,
6092 * we use the raw string as a tiebreaker.
6093 */
16660edb 6094
bbce6d69 6095 raw_compare:
5f66b61c 6096 /*FALLTHROUGH*/
16660edb 6097
36477c24 6098#endif /* USE_LOCALE_COLLATE */
16660edb 6099
bbce6d69 6100 return sv_cmp(sv1, sv2);
6101}
79072805 6102
645c22ef 6103
36477c24 6104#ifdef USE_LOCALE_COLLATE
645c22ef 6105
7a4c00b4 6106/*
645c22ef
DM
6107=for apidoc sv_collxfrm
6108
6109Add Collate Transform magic to an SV if it doesn't already have it.
6110
6111Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6112scalar data of the variable, but transformed to such a format that a normal
6113memory comparison can be used to compare the data according to the locale
6114settings.
6115
6116=cut
6117*/
6118
bbce6d69 6119char *
864dbfa3 6120Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6121{
97aff369 6122 dVAR;
7a4c00b4 6123 MAGIC *mg;
16660edb 6124
14befaf4 6125 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6126 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6127 const char *s;
6128 char *xf;
bbce6d69 6129 STRLEN len, xlen;
6130
7a4c00b4 6131 if (mg)
6132 Safefree(mg->mg_ptr);
93524f2b 6133 s = SvPV_const(sv, len);
bbce6d69 6134 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6135 if (SvREADONLY(sv)) {
6136 SAVEFREEPV(xf);
6137 *nxp = xlen;
3280af22 6138 return xf + sizeof(PL_collation_ix);
ff0cee69 6139 }
7a4c00b4 6140 if (! mg) {
d83f0a82
NC
6141#ifdef PERL_OLD_COPY_ON_WRITE
6142 if (SvIsCOW(sv))
6143 sv_force_normal_flags(sv, 0);
6144#endif
6145 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6146 0, 0);
7a4c00b4 6147 assert(mg);
bbce6d69 6148 }
7a4c00b4 6149 mg->mg_ptr = xf;
565764a8 6150 mg->mg_len = xlen;
7a4c00b4 6151 }
6152 else {
ff0cee69 6153 if (mg) {
6154 mg->mg_ptr = NULL;
565764a8 6155 mg->mg_len = -1;
ff0cee69 6156 }
bbce6d69 6157 }
6158 }
7a4c00b4 6159 if (mg && mg->mg_ptr) {
565764a8 6160 *nxp = mg->mg_len;
3280af22 6161 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6162 }
6163 else {
6164 *nxp = 0;
6165 return NULL;
16660edb 6166 }
79072805
LW
6167}
6168
36477c24 6169#endif /* USE_LOCALE_COLLATE */
bbce6d69 6170
c461cf8f
JH
6171/*
6172=for apidoc sv_gets
6173
6174Get a line from the filehandle and store it into the SV, optionally
6175appending to the currently-stored string.
6176
6177=cut
6178*/
6179
79072805 6180char *
864dbfa3 6181Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6182{
97aff369 6183 dVAR;
e1ec3a88 6184 const char *rsptr;
c07a80fd 6185 STRLEN rslen;
6186 register STDCHAR rslast;
6187 register STDCHAR *bp;
6188 register I32 cnt;
9c5ffd7c 6189 I32 i = 0;
8bfdd7d9 6190 I32 rspara = 0;
c07a80fd 6191
bc44a8a2
NC
6192 if (SvTHINKFIRST(sv))
6193 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6194 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6195 from <>.
6196 However, perlbench says it's slower, because the existing swipe code
6197 is faster than copy on write.
6198 Swings and roundabouts. */
862a34c6 6199 SvUPGRADE(sv, SVt_PV);
99491443 6200
ff68c719 6201 SvSCREAM_off(sv);
efd8b2ba
AE
6202
6203 if (append) {
6204 if (PerlIO_isutf8(fp)) {
6205 if (!SvUTF8(sv)) {
6206 sv_utf8_upgrade_nomg(sv);
6207 sv_pos_u2b(sv,&append,0);
6208 }
6209 } else if (SvUTF8(sv)) {
561b68a9 6210 SV * const tsv = newSV(0);
efd8b2ba
AE
6211 sv_gets(tsv, fp, 0);
6212 sv_utf8_upgrade_nomg(tsv);
6213 SvCUR_set(sv,append);
6214 sv_catsv(sv,tsv);
6215 sv_free(tsv);
6216 goto return_string_or_null;
6217 }
6218 }
6219
6220 SvPOK_only(sv);
6221 if (PerlIO_isutf8(fp))
6222 SvUTF8_on(sv);
c07a80fd 6223
923e4eb5 6224 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6225 /* we always read code in line mode */
6226 rsptr = "\n";
6227 rslen = 1;
6228 }
6229 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6230 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6231 of amount we are going to read -- may result in mallocing
6232 more memory than we really need if the layers below reduce
6233 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6234 */
e311fd51 6235 Stat_t st;
e468d35b 6236 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6237 const Off_t offset = PerlIO_tell(fp);
58f1856e 6238 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6239 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6240 }
6241 }
c07a80fd 6242 rsptr = NULL;
6243 rslen = 0;
6244 }
3280af22 6245 else if (RsRECORD(PL_rs)) {
e311fd51 6246 I32 bytesread;
5b2b9c68 6247 char *buffer;
acbd132f 6248 U32 recsize;
5b2b9c68
HM
6249
6250 /* Grab the size of the record we're getting */
acbd132f 6251 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6252 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6253 /* Go yank in */
6254#ifdef VMS
6255 /* VMS wants read instead of fread, because fread doesn't respect */
6256 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6257 /* doing, but we've got no other real choice - except avoid stdio
6258 as implementation - perhaps write a :vms layer ?
6259 */
5b2b9c68
HM
6260 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6261#else
6262 bytesread = PerlIO_read(fp, buffer, recsize);
6263#endif
27e6ca2d
AE
6264 if (bytesread < 0)
6265 bytesread = 0;
e311fd51 6266 SvCUR_set(sv, bytesread += append);
e670df4e 6267 buffer[bytesread] = '\0';
efd8b2ba 6268 goto return_string_or_null;
5b2b9c68 6269 }
3280af22 6270 else if (RsPARA(PL_rs)) {
c07a80fd 6271 rsptr = "\n\n";
6272 rslen = 2;
8bfdd7d9 6273 rspara = 1;
c07a80fd 6274 }
7d59b7e4
NIS
6275 else {
6276 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6277 if (PerlIO_isutf8(fp)) {
6278 rsptr = SvPVutf8(PL_rs, rslen);
6279 }
6280 else {
6281 if (SvUTF8(PL_rs)) {
6282 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6283 Perl_croak(aTHX_ "Wide character in $/");
6284 }
6285 }
93524f2b 6286 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6287 }
6288 }
6289
c07a80fd 6290 rslast = rslen ? rsptr[rslen - 1] : '\0';
6291
8bfdd7d9 6292 if (rspara) { /* have to do this both before and after */
79072805 6293 do { /* to make sure file boundaries work right */
760ac839 6294 if (PerlIO_eof(fp))
a0d0e21e 6295 return 0;
760ac839 6296 i = PerlIO_getc(fp);
79072805 6297 if (i != '\n') {
a0d0e21e
LW
6298 if (i == -1)
6299 return 0;
760ac839 6300 PerlIO_ungetc(fp,i);
79072805
LW
6301 break;
6302 }
6303 } while (i != EOF);
6304 }
c07a80fd 6305
760ac839
LW
6306 /* See if we know enough about I/O mechanism to cheat it ! */
6307
6308 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6309 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6310 enough here - and may even be a macro allowing compile
6311 time optimization.
6312 */
6313
6314 if (PerlIO_fast_gets(fp)) {
6315
6316 /*
6317 * We're going to steal some values from the stdio struct
6318 * and put EVERYTHING in the innermost loop into registers.
6319 */
6320 register STDCHAR *ptr;
6321 STRLEN bpx;
6322 I32 shortbuffered;
6323
16660edb 6324#if defined(VMS) && defined(PERLIO_IS_STDIO)
6325 /* An ungetc()d char is handled separately from the regular
6326 * buffer, so we getc() it back out and stuff it in the buffer.
6327 */
6328 i = PerlIO_getc(fp);
6329 if (i == EOF) return 0;
6330 *(--((*fp)->_ptr)) = (unsigned char) i;
6331 (*fp)->_cnt++;
6332#endif
c07a80fd 6333
c2960299 6334 /* Here is some breathtakingly efficient cheating */
c07a80fd 6335
a20bf0c3 6336 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6337 /* make sure we have the room */
7a5fa8a2 6338 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6339 /* Not room for all of it
7a5fa8a2 6340 if we are looking for a separator and room for some
e468d35b
NIS
6341 */
6342 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6343 /* just process what we have room for */
79072805
LW
6344 shortbuffered = cnt - SvLEN(sv) + append + 1;
6345 cnt -= shortbuffered;
6346 }
6347 else {
6348 shortbuffered = 0;
bbce6d69 6349 /* remember that cnt can be negative */
eb160463 6350 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6351 }
6352 }
7a5fa8a2 6353 else
79072805 6354 shortbuffered = 0;
3f7c398e 6355 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6356 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6357 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6358 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6359 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6360 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6361 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6362 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6363 for (;;) {
6364 screamer:
93a17b20 6365 if (cnt > 0) {
c07a80fd 6366 if (rslen) {
760ac839
LW
6367 while (cnt > 0) { /* this | eat */
6368 cnt--;
c07a80fd 6369 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6370 goto thats_all_folks; /* screams | sed :-) */
6371 }
6372 }
6373 else {
1c846c1f
NIS
6374 Copy(ptr, bp, cnt, char); /* this | eat */
6375 bp += cnt; /* screams | dust */
c07a80fd 6376 ptr += cnt; /* louder | sed :-) */
a5f75d66 6377 cnt = 0;
93a17b20 6378 }
79072805
LW
6379 }
6380
748a9306 6381 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6382 cnt = shortbuffered;
6383 shortbuffered = 0;
3f7c398e 6384 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6385 SvCUR_set(sv, bpx);
6386 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6387 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6388 continue;
6389 }
6390
16660edb 6391 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6392 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6393 PTR2UV(ptr),(long)cnt));
cc00df79 6394 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6395#if 0
16660edb 6396 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6397 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6398 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6399 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6400#endif
1c846c1f 6401 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6402 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6403 another abstraction. */
760ac839 6404 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6405#if 0
16660edb 6406 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6407 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6408 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6409 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6410#endif
a20bf0c3
JH
6411 cnt = PerlIO_get_cnt(fp);
6412 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6413 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6414 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6415
748a9306
LW
6416 if (i == EOF) /* all done for ever? */
6417 goto thats_really_all_folks;
6418
3f7c398e 6419 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6420 SvCUR_set(sv, bpx);
6421 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6422 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6423
eb160463 6424 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6425
c07a80fd 6426 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6427 goto thats_all_folks;
79072805
LW
6428 }
6429
6430thats_all_folks:
3f7c398e 6431 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6432 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6433 goto screamer; /* go back to the fray */
79072805
LW
6434thats_really_all_folks:
6435 if (shortbuffered)
6436 cnt += shortbuffered;
16660edb 6437 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6438 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6439 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6440 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6441 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6442 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6443 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6444 *bp = '\0';
3f7c398e 6445 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6446 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6447 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6448 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6449 }
6450 else
79072805 6451 {
6edd2cd5 6452 /*The big, slow, and stupid way. */
27da23d5 6453#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6454 STDCHAR *buf = NULL;
a02a5408 6455 Newx(buf, 8192, STDCHAR);
6edd2cd5 6456 assert(buf);
4d2c4e07 6457#else
6edd2cd5 6458 STDCHAR buf[8192];
4d2c4e07 6459#endif
79072805 6460
760ac839 6461screamer2:
c07a80fd 6462 if (rslen) {
00b6aa41 6463 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6464 bp = buf;
eb160463 6465 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6466 ; /* keep reading */
6467 cnt = bp - buf;
c07a80fd 6468 }
6469 else {
760ac839 6470 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6471 /* Accomodate broken VAXC compiler, which applies U8 cast to
6472 * both args of ?: operator, causing EOF to change into 255
6473 */
37be0adf 6474 if (cnt > 0)
cbe9e203
JH
6475 i = (U8)buf[cnt - 1];
6476 else
37be0adf 6477 i = EOF;
c07a80fd 6478 }
79072805 6479
cbe9e203
JH
6480 if (cnt < 0)
6481 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6482 if (append)
6483 sv_catpvn(sv, (char *) buf, cnt);
6484 else
6485 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6486
6487 if (i != EOF && /* joy */
6488 (!rslen ||
6489 SvCUR(sv) < rslen ||
3f7c398e 6490 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6491 {
6492 append = -1;
63e4d877
CS
6493 /*
6494 * If we're reading from a TTY and we get a short read,
6495 * indicating that the user hit his EOF character, we need
6496 * to notice it now, because if we try to read from the TTY
6497 * again, the EOF condition will disappear.
6498 *
6499 * The comparison of cnt to sizeof(buf) is an optimization
6500 * that prevents unnecessary calls to feof().
6501 *
6502 * - jik 9/25/96
6503 */
6504 if (!(cnt < sizeof(buf) && PerlIO_eof(fp)))
6505 goto screamer2;
79072805 6506 }
6edd2cd5 6507
27da23d5 6508#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6509 Safefree(buf);
6510#endif
79072805
LW
6511 }
6512
8bfdd7d9 6513 if (rspara) { /* have to do this both before and after */
c07a80fd 6514 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6515 i = PerlIO_getc(fp);
79072805 6516 if (i != '\n') {
760ac839 6517 PerlIO_ungetc(fp,i);
79072805
LW
6518 break;
6519 }
6520 }
6521 }
c07a80fd 6522
efd8b2ba 6523return_string_or_null:
bd61b366 6524 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6525}
6526
954c1994
GS
6527/*
6528=for apidoc sv_inc
6529
645c22ef
DM
6530Auto-increment of the value in the SV, doing string to numeric conversion
6531if necessary. Handles 'get' magic.
954c1994
GS
6532
6533=cut
6534*/
6535
79072805 6536void
864dbfa3 6537Perl_sv_inc(pTHX_ register SV *sv)
79072805 6538{
97aff369 6539 dVAR;
79072805 6540 register char *d;
463ee0b2 6541 int flags;
79072805
LW
6542
6543 if (!sv)
6544 return;
5b295bef 6545 SvGETMAGIC(sv);
ed6116ce 6546 if (SvTHINKFIRST(sv)) {
765f542d
NC
6547 if (SvIsCOW(sv))
6548 sv_force_normal_flags(sv, 0);
0f15f207 6549 if (SvREADONLY(sv)) {
923e4eb5 6550 if (IN_PERL_RUNTIME)
cea2e8a9 6551 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6552 }
a0d0e21e 6553 if (SvROK(sv)) {
b5be31e9 6554 IV i;
9e7bc3e8
JD
6555 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6556 return;
56431972 6557 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6558 sv_unref(sv);
6559 sv_setiv(sv, i);
a0d0e21e 6560 }
ed6116ce 6561 }
8990e307 6562 flags = SvFLAGS(sv);
28e5dec8
JH
6563 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6564 /* It's (privately or publicly) a float, but not tested as an
6565 integer, so test it to see. */
d460ef45 6566 (void) SvIV(sv);
28e5dec8
JH
6567 flags = SvFLAGS(sv);
6568 }
6569 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6570 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6571#ifdef PERL_PRESERVE_IVUV
28e5dec8 6572 oops_its_int:
59d8ce62 6573#endif
25da4f38
IZ
6574 if (SvIsUV(sv)) {
6575 if (SvUVX(sv) == UV_MAX)
a1e868e7 6576 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6577 else
6578 (void)SvIOK_only_UV(sv);
607fa7f2 6579 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6580 } else {
6581 if (SvIVX(sv) == IV_MAX)
28e5dec8 6582 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6583 else {
6584 (void)SvIOK_only(sv);
45977657 6585 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6586 }
55497cff 6587 }
79072805
LW
6588 return;
6589 }
28e5dec8
JH
6590 if (flags & SVp_NOK) {
6591 (void)SvNOK_only(sv);
9d6ce603 6592 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6593 return;
6594 }
6595
3f7c398e 6596 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6597 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6598 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6599 (void)SvIOK_only(sv);
45977657 6600 SvIV_set(sv, 1);
79072805
LW
6601 return;
6602 }
463ee0b2 6603 d = SvPVX(sv);
79072805
LW
6604 while (isALPHA(*d)) d++;
6605 while (isDIGIT(*d)) d++;
6606 if (*d) {
28e5dec8 6607#ifdef PERL_PRESERVE_IVUV
d1be9408 6608 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6609 warnings. Probably ought to make the sv_iv_please() that does
6610 the conversion if possible, and silently. */
504618e9 6611 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6612 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6613 /* Need to try really hard to see if it's an integer.
6614 9.22337203685478e+18 is an integer.
6615 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6616 so $a="9.22337203685478e+18"; $a+0; $a++
6617 needs to be the same as $a="9.22337203685478e+18"; $a++
6618 or we go insane. */
d460ef45 6619
28e5dec8
JH
6620 (void) sv_2iv(sv);
6621 if (SvIOK(sv))
6622 goto oops_its_int;
6623
6624 /* sv_2iv *should* have made this an NV */
6625 if (flags & SVp_NOK) {
6626 (void)SvNOK_only(sv);
9d6ce603 6627 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6628 return;
6629 }
6630 /* I don't think we can get here. Maybe I should assert this
6631 And if we do get here I suspect that sv_setnv will croak. NWC
6632 Fall through. */
6633#if defined(USE_LONG_DOUBLE)
6634 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 6635 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6636#else
1779d84d 6637 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 6638 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6639#endif
6640 }
6641#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6642 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6643 return;
6644 }
6645 d--;
3f7c398e 6646 while (d >= SvPVX_const(sv)) {
79072805
LW
6647 if (isDIGIT(*d)) {
6648 if (++*d <= '9')
6649 return;
6650 *(d--) = '0';
6651 }
6652 else {
9d116dd7
JH
6653#ifdef EBCDIC
6654 /* MKS: The original code here died if letters weren't consecutive.
6655 * at least it didn't have to worry about non-C locales. The
6656 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6657 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6658 * [A-Za-z] are accepted by isALPHA in the C locale.
6659 */
6660 if (*d != 'z' && *d != 'Z') {
6661 do { ++*d; } while (!isALPHA(*d));
6662 return;
6663 }
6664 *(d--) -= 'z' - 'a';
6665#else
79072805
LW
6666 ++*d;
6667 if (isALPHA(*d))
6668 return;
6669 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6670#endif
79072805
LW
6671 }
6672 }
6673 /* oh,oh, the number grew */
6674 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6675 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6676 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6677 *d = d[-1];
6678 if (isDIGIT(d[1]))
6679 *d = '1';
6680 else
6681 *d = d[1];
6682}
6683
954c1994
GS
6684/*
6685=for apidoc sv_dec
6686
645c22ef
DM
6687Auto-decrement of the value in the SV, doing string to numeric conversion
6688if necessary. Handles 'get' magic.
954c1994
GS
6689
6690=cut
6691*/
6692
79072805 6693void
864dbfa3 6694Perl_sv_dec(pTHX_ register SV *sv)
79072805 6695{
97aff369 6696 dVAR;
463ee0b2
LW
6697 int flags;
6698
79072805
LW
6699 if (!sv)
6700 return;
5b295bef 6701 SvGETMAGIC(sv);
ed6116ce 6702 if (SvTHINKFIRST(sv)) {
765f542d
NC
6703 if (SvIsCOW(sv))
6704 sv_force_normal_flags(sv, 0);
0f15f207 6705 if (SvREADONLY(sv)) {
923e4eb5 6706 if (IN_PERL_RUNTIME)
cea2e8a9 6707 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6708 }
a0d0e21e 6709 if (SvROK(sv)) {
b5be31e9 6710 IV i;
9e7bc3e8
JD
6711 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6712 return;
56431972 6713 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6714 sv_unref(sv);
6715 sv_setiv(sv, i);
a0d0e21e 6716 }
ed6116ce 6717 }
28e5dec8
JH
6718 /* Unlike sv_inc we don't have to worry about string-never-numbers
6719 and keeping them magic. But we mustn't warn on punting */
8990e307 6720 flags = SvFLAGS(sv);
28e5dec8
JH
6721 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6722 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6723#ifdef PERL_PRESERVE_IVUV
28e5dec8 6724 oops_its_int:
59d8ce62 6725#endif
25da4f38
IZ
6726 if (SvIsUV(sv)) {
6727 if (SvUVX(sv) == 0) {
6728 (void)SvIOK_only(sv);
45977657 6729 SvIV_set(sv, -1);
25da4f38
IZ
6730 }
6731 else {
6732 (void)SvIOK_only_UV(sv);
f4eee32f 6733 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6734 }
25da4f38
IZ
6735 } else {
6736 if (SvIVX(sv) == IV_MIN)
65202027 6737 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6738 else {
6739 (void)SvIOK_only(sv);
45977657 6740 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6741 }
55497cff 6742 }
6743 return;
6744 }
28e5dec8 6745 if (flags & SVp_NOK) {
9d6ce603 6746 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6747 (void)SvNOK_only(sv);
6748 return;
6749 }
8990e307 6750 if (!(flags & SVp_POK)) {
ef088171
NC
6751 if ((flags & SVTYPEMASK) < SVt_PVIV)
6752 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6753 SvIV_set(sv, -1);
6754 (void)SvIOK_only(sv);
79072805
LW
6755 return;
6756 }
28e5dec8
JH
6757#ifdef PERL_PRESERVE_IVUV
6758 {
504618e9 6759 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6760 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6761 /* Need to try really hard to see if it's an integer.
6762 9.22337203685478e+18 is an integer.
6763 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6764 so $a="9.22337203685478e+18"; $a+0; $a--
6765 needs to be the same as $a="9.22337203685478e+18"; $a--
6766 or we go insane. */
d460ef45 6767
28e5dec8
JH
6768 (void) sv_2iv(sv);
6769 if (SvIOK(sv))
6770 goto oops_its_int;
6771
6772 /* sv_2iv *should* have made this an NV */
6773 if (flags & SVp_NOK) {
6774 (void)SvNOK_only(sv);
9d6ce603 6775 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6776 return;
6777 }
6778 /* I don't think we can get here. Maybe I should assert this
6779 And if we do get here I suspect that sv_setnv will croak. NWC
6780 Fall through. */
6781#if defined(USE_LONG_DOUBLE)
6782 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 6783 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6784#else
1779d84d 6785 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 6786 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6787#endif
6788 }
6789 }
6790#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6791 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6792}
6793
954c1994
GS
6794/*
6795=for apidoc sv_mortalcopy
6796
645c22ef 6797Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6798The new SV is marked as mortal. It will be destroyed "soon", either by an
6799explicit call to FREETMPS, or by an implicit call at places such as
6800statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6801
6802=cut
6803*/
6804
79072805
LW
6805/* Make a string that will exist for the duration of the expression
6806 * evaluation. Actually, it may have to last longer than that, but
6807 * hopefully we won't free it until it has been assigned to a
6808 * permanent location. */
6809
6810SV *
864dbfa3 6811Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6812{
97aff369 6813 dVAR;
463ee0b2 6814 register SV *sv;
b881518d 6815
4561caa4 6816 new_SV(sv);
79072805 6817 sv_setsv(sv,oldstr);
677b06e3
GS
6818 EXTEND_MORTAL(1);
6819 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6820 SvTEMP_on(sv);
6821 return sv;
6822}
6823
954c1994
GS
6824/*
6825=for apidoc sv_newmortal
6826
645c22ef 6827Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6828set to 1. It will be destroyed "soon", either by an explicit call to
6829FREETMPS, or by an implicit call at places such as statement boundaries.
6830See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6831
6832=cut
6833*/
6834
8990e307 6835SV *
864dbfa3 6836Perl_sv_newmortal(pTHX)
8990e307 6837{
97aff369 6838 dVAR;
8990e307
LW
6839 register SV *sv;
6840
4561caa4 6841 new_SV(sv);
8990e307 6842 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6843 EXTEND_MORTAL(1);
6844 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6845 return sv;
6846}
6847
954c1994
GS
6848/*
6849=for apidoc sv_2mortal
6850
d4236ebc
DM
6851Marks an existing SV as mortal. The SV will be destroyed "soon", either
6852by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6853statement boundaries. SvTEMP() is turned on which means that the SV's
6854string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6855and C<sv_mortalcopy>.
954c1994
GS
6856
6857=cut
6858*/
6859
79072805 6860SV *
864dbfa3 6861Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6862{
27da23d5 6863 dVAR;
79072805 6864 if (!sv)
7a5b473e 6865 return NULL;
d689ffdd 6866 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6867 return sv;
677b06e3
GS
6868 EXTEND_MORTAL(1);
6869 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6870 SvTEMP_on(sv);
79072805
LW
6871 return sv;
6872}
6873
954c1994
GS
6874/*
6875=for apidoc newSVpv
6876
6877Creates a new SV and copies a string into it. The reference count for the
6878SV is set to 1. If C<len> is zero, Perl will compute the length using
6879strlen(). For efficiency, consider using C<newSVpvn> instead.
6880
6881=cut
6882*/
6883
79072805 6884SV *
864dbfa3 6885Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6886{
97aff369 6887 dVAR;
463ee0b2 6888 register SV *sv;
79072805 6889
4561caa4 6890 new_SV(sv);
616d8c9c 6891 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6892 return sv;
6893}
6894
954c1994
GS
6895/*
6896=for apidoc newSVpvn
6897
6898Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6899SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6900string. You are responsible for ensuring that the source string is at least
9e09f5f2 6901C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6902
6903=cut
6904*/
6905
9da1e3b5 6906SV *
864dbfa3 6907Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6908{
97aff369 6909 dVAR;
9da1e3b5
MUN
6910 register SV *sv;
6911
6912 new_SV(sv);
9da1e3b5
MUN
6913 sv_setpvn(sv,s,len);
6914 return sv;
6915}
6916
bd08039b
NC
6917
6918/*
926f8064 6919=for apidoc newSVhek
bd08039b
NC
6920
6921Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6922point to the shared string table where possible. Returns a new (undefined)
6923SV if the hek is NULL.
bd08039b
NC
6924
6925=cut
6926*/
6927
6928SV *
c1b02ed8 6929Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6930{
97aff369 6931 dVAR;
5aaec2b4
NC
6932 if (!hek) {
6933 SV *sv;
6934
6935 new_SV(sv);
6936 return sv;
6937 }
6938
bd08039b
NC
6939 if (HEK_LEN(hek) == HEf_SVKEY) {
6940 return newSVsv(*(SV**)HEK_KEY(hek));
6941 } else {
6942 const int flags = HEK_FLAGS(hek);
6943 if (flags & HVhek_WASUTF8) {
6944 /* Trouble :-)
6945 Andreas would like keys he put in as utf8 to come back as utf8
6946 */
6947 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6948 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6949 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6950
6951 SvUTF8_on (sv);
6952 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6953 return sv;
6954 } else if (flags & HVhek_REHASH) {
6955 /* We don't have a pointer to the hv, so we have to replicate the
6956 flag into every HEK. This hv is using custom a hasing
6957 algorithm. Hence we can't return a shared string scalar, as
6958 that would contain the (wrong) hash value, and might get passed
6959 into an hv routine with a regular hash */
6960
b64e5050 6961 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6962 if (HEK_UTF8(hek))
6963 SvUTF8_on (sv);
6964 return sv;
6965 }
6966 /* This will be overwhelminly the most common case. */
6967 return newSVpvn_share(HEK_KEY(hek),
6968 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
6969 HEK_HASH(hek));
6970 }
6971}
6972
1c846c1f
NIS
6973/*
6974=for apidoc newSVpvn_share
6975
3f7c398e 6976Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
6977table. If the string does not already exist in the table, it is created
6978first. Turns on READONLY and FAKE. The string's hash is stored in the UV
6979slot of the SV; if the C<hash> parameter is non-zero, that value is used;
6980otherwise the hash is computed. The idea here is that as the string table
3f7c398e 6981is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 6982hash lookup will avoid string compare.
1c846c1f
NIS
6983
6984=cut
6985*/
6986
6987SV *
c3654f1a 6988Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 6989{
97aff369 6990 dVAR;
1c846c1f 6991 register SV *sv;
c3654f1a
IH
6992 bool is_utf8 = FALSE;
6993 if (len < 0) {
77caf834 6994 STRLEN tmplen = -len;
c3654f1a 6995 is_utf8 = TRUE;
75a54232 6996 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 6997 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
6998 len = tmplen;
6999 }
1c846c1f 7000 if (!hash)
5afd6d42 7001 PERL_HASH(hash, src, len);
1c846c1f 7002 new_SV(sv);
bdd68bc3 7003 sv_upgrade(sv, SVt_PV);
f880fe2f 7004 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7005 SvCUR_set(sv, len);
b162af07 7006 SvLEN_set(sv, 0);
1c846c1f
NIS
7007 SvREADONLY_on(sv);
7008 SvFAKE_on(sv);
7009 SvPOK_on(sv);
c3654f1a
IH
7010 if (is_utf8)
7011 SvUTF8_on(sv);
1c846c1f
NIS
7012 return sv;
7013}
7014
645c22ef 7015
cea2e8a9 7016#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7017
7018/* pTHX_ magic can't cope with varargs, so this is a no-context
7019 * version of the main function, (which may itself be aliased to us).
7020 * Don't access this version directly.
7021 */
7022
46fc3d4c 7023SV *
cea2e8a9 7024Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7025{
cea2e8a9 7026 dTHX;
46fc3d4c 7027 register SV *sv;
7028 va_list args;
46fc3d4c 7029 va_start(args, pat);
c5be433b 7030 sv = vnewSVpvf(pat, &args);
46fc3d4c 7031 va_end(args);
7032 return sv;
7033}
cea2e8a9 7034#endif
46fc3d4c 7035
954c1994
GS
7036/*
7037=for apidoc newSVpvf
7038
645c22ef 7039Creates a new SV and initializes it with the string formatted like
954c1994
GS
7040C<sprintf>.
7041
7042=cut
7043*/
7044
cea2e8a9
GS
7045SV *
7046Perl_newSVpvf(pTHX_ const char* pat, ...)
7047{
7048 register SV *sv;
7049 va_list args;
cea2e8a9 7050 va_start(args, pat);
c5be433b 7051 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7052 va_end(args);
7053 return sv;
7054}
46fc3d4c 7055
645c22ef
DM
7056/* backend for newSVpvf() and newSVpvf_nocontext() */
7057
79072805 7058SV *
c5be433b
GS
7059Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7060{
97aff369 7061 dVAR;
c5be433b
GS
7062 register SV *sv;
7063 new_SV(sv);
4608196e 7064 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7065 return sv;
7066}
7067
954c1994
GS
7068/*
7069=for apidoc newSVnv
7070
7071Creates a new SV and copies a floating point value into it.
7072The reference count for the SV is set to 1.
7073
7074=cut
7075*/
7076
c5be433b 7077SV *
65202027 7078Perl_newSVnv(pTHX_ NV n)
79072805 7079{
97aff369 7080 dVAR;
463ee0b2 7081 register SV *sv;
79072805 7082
4561caa4 7083 new_SV(sv);
79072805
LW
7084 sv_setnv(sv,n);
7085 return sv;
7086}
7087
954c1994
GS
7088/*
7089=for apidoc newSViv
7090
7091Creates a new SV and copies an integer into it. The reference count for the
7092SV is set to 1.
7093
7094=cut
7095*/
7096
79072805 7097SV *
864dbfa3 7098Perl_newSViv(pTHX_ IV i)
79072805 7099{
97aff369 7100 dVAR;
463ee0b2 7101 register SV *sv;
79072805 7102
4561caa4 7103 new_SV(sv);
79072805
LW
7104 sv_setiv(sv,i);
7105 return sv;
7106}
7107
954c1994 7108/*
1a3327fb
JH
7109=for apidoc newSVuv
7110
7111Creates a new SV and copies an unsigned integer into it.
7112The reference count for the SV is set to 1.
7113
7114=cut
7115*/
7116
7117SV *
7118Perl_newSVuv(pTHX_ UV u)
7119{
97aff369 7120 dVAR;
1a3327fb
JH
7121 register SV *sv;
7122
7123 new_SV(sv);
7124 sv_setuv(sv,u);
7125 return sv;
7126}
7127
7128/*
954c1994
GS
7129=for apidoc newRV_noinc
7130
7131Creates an RV wrapper for an SV. The reference count for the original
7132SV is B<not> incremented.
7133
7134=cut
7135*/
7136
2304df62 7137SV *
864dbfa3 7138Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7139{
97aff369 7140 dVAR;
2304df62
AD
7141 register SV *sv;
7142
4561caa4 7143 new_SV(sv);
2304df62 7144 sv_upgrade(sv, SVt_RV);
76e3520e 7145 SvTEMP_off(tmpRef);
b162af07 7146 SvRV_set(sv, tmpRef);
2304df62 7147 SvROK_on(sv);
2304df62
AD
7148 return sv;
7149}
7150
ff276b08 7151/* newRV_inc is the official function name to use now.
645c22ef
DM
7152 * newRV_inc is in fact #defined to newRV in sv.h
7153 */
7154
5f05dabc 7155SV *
7f466ec7 7156Perl_newRV(pTHX_ SV *sv)
5f05dabc 7157{
97aff369 7158 dVAR;
7f466ec7 7159 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7160}
5f05dabc 7161
954c1994
GS
7162/*
7163=for apidoc newSVsv
7164
7165Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7166(Uses C<sv_setsv>).
954c1994
GS
7167
7168=cut
7169*/
7170
79072805 7171SV *
864dbfa3 7172Perl_newSVsv(pTHX_ register SV *old)
79072805 7173{
97aff369 7174 dVAR;
463ee0b2 7175 register SV *sv;
79072805
LW
7176
7177 if (!old)
7a5b473e 7178 return NULL;
8990e307 7179 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7180 if (ckWARN_d(WARN_INTERNAL))
9014280d 7181 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7182 return NULL;
79072805 7183 }
4561caa4 7184 new_SV(sv);
e90aabeb
NC
7185 /* SV_GMAGIC is the default for sv_setv()
7186 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7187 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7188 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7189 return sv;
79072805
LW
7190}
7191
645c22ef
DM
7192/*
7193=for apidoc sv_reset
7194
7195Underlying implementation for the C<reset> Perl function.
7196Note that the perl-level function is vaguely deprecated.
7197
7198=cut
7199*/
7200
79072805 7201void
e1ec3a88 7202Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7203{
27da23d5 7204 dVAR;
4802d5d7 7205 char todo[PERL_UCHAR_MAX+1];
79072805 7206
49d8d3a1
MB
7207 if (!stash)
7208 return;
7209
79072805 7210 if (!*s) { /* reset ?? searches */
aec46f14 7211 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7212 if (mg) {
7213 PMOP *pm = (PMOP *) mg->mg_obj;
7214 while (pm) {
7215 pm->op_pmdynflags &= ~PMdf_USED;
7216 pm = pm->op_pmnext;
7217 }
79072805
LW
7218 }
7219 return;
7220 }
7221
7222 /* reset variables */
7223
7224 if (!HvARRAY(stash))
7225 return;
463ee0b2
LW
7226
7227 Zero(todo, 256, char);
79072805 7228 while (*s) {
b464bac0
AL
7229 I32 max;
7230 I32 i = (unsigned char)*s;
79072805
LW
7231 if (s[1] == '-') {
7232 s += 2;
7233 }
4802d5d7 7234 max = (unsigned char)*s++;
79072805 7235 for ( ; i <= max; i++) {
463ee0b2
LW
7236 todo[i] = 1;
7237 }
a0d0e21e 7238 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7239 HE *entry;
79072805 7240 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7241 entry;
7242 entry = HeNEXT(entry))
7243 {
b464bac0
AL
7244 register GV *gv;
7245 register SV *sv;
7246
1edc1566 7247 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7248 continue;
1edc1566 7249 gv = (GV*)HeVAL(entry);
79072805 7250 sv = GvSV(gv);
e203899d
NC
7251 if (sv) {
7252 if (SvTHINKFIRST(sv)) {
7253 if (!SvREADONLY(sv) && SvROK(sv))
7254 sv_unref(sv);
7255 /* XXX Is this continue a bug? Why should THINKFIRST
7256 exempt us from resetting arrays and hashes? */
7257 continue;
7258 }
7259 SvOK_off(sv);
7260 if (SvTYPE(sv) >= SVt_PV) {
7261 SvCUR_set(sv, 0);
bd61b366 7262 if (SvPVX_const(sv) != NULL)
e203899d
NC
7263 *SvPVX(sv) = '\0';
7264 SvTAINT(sv);
7265 }
79072805
LW
7266 }
7267 if (GvAV(gv)) {
7268 av_clear(GvAV(gv));
7269 }
bfcb3514 7270 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7271#if defined(VMS)
7272 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7273#else /* ! VMS */
463ee0b2 7274 hv_clear(GvHV(gv));
b0269e46
AB
7275# if defined(USE_ENVIRON_ARRAY)
7276 if (gv == PL_envgv)
7277 my_clearenv();
7278# endif /* USE_ENVIRON_ARRAY */
7279#endif /* VMS */
79072805
LW
7280 }
7281 }
7282 }
7283 }
7284}
7285
645c22ef
DM
7286/*
7287=for apidoc sv_2io
7288
7289Using various gambits, try to get an IO from an SV: the IO slot if its a
7290GV; or the recursive result if we're an RV; or the IO slot of the symbol
7291named after the PV if we're a string.
7292
7293=cut
7294*/
7295
46fc3d4c 7296IO*
864dbfa3 7297Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7298{
7299 IO* io;
7300 GV* gv;
7301
7302 switch (SvTYPE(sv)) {
7303 case SVt_PVIO:
7304 io = (IO*)sv;
7305 break;
7306 case SVt_PVGV:
7307 gv = (GV*)sv;
7308 io = GvIO(gv);
7309 if (!io)
cea2e8a9 7310 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7311 break;
7312 default:
7313 if (!SvOK(sv))
cea2e8a9 7314 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7315 if (SvROK(sv))
7316 return sv_2io(SvRV(sv));
f776e3cd 7317 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7318 if (gv)
7319 io = GvIO(gv);
7320 else
7321 io = 0;
7322 if (!io)
35c1215d 7323 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7324 break;
7325 }
7326 return io;
7327}
7328
645c22ef
DM
7329/*
7330=for apidoc sv_2cv
7331
7332Using various gambits, try to get a CV from an SV; in addition, try if
7333possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7334The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7335
7336=cut
7337*/
7338
79072805 7339CV *
864dbfa3 7340Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7341{
27da23d5 7342 dVAR;
a0714e2c 7343 GV *gv = NULL;
601f1833 7344 CV *cv = NULL;
79072805 7345
85dec29a
NC
7346 if (!sv) {
7347 *st = NULL;
7348 *gvp = NULL;
7349 return NULL;
7350 }
79072805 7351 switch (SvTYPE(sv)) {
79072805
LW
7352 case SVt_PVCV:
7353 *st = CvSTASH(sv);
a0714e2c 7354 *gvp = NULL;
79072805
LW
7355 return (CV*)sv;
7356 case SVt_PVHV:
7357 case SVt_PVAV:
ef58ba18 7358 *st = NULL;
a0714e2c 7359 *gvp = NULL;
601f1833 7360 return NULL;
8990e307
LW
7361 case SVt_PVGV:
7362 gv = (GV*)sv;
a0d0e21e 7363 *gvp = gv;
8990e307
LW
7364 *st = GvESTASH(gv);
7365 goto fix_gv;
7366
79072805 7367 default:
5b295bef 7368 SvGETMAGIC(sv);
a0d0e21e 7369 if (SvROK(sv)) {
823a54a3 7370 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7371 tryAMAGICunDEREF(to_cv);
7372
62f274bf
GS
7373 sv = SvRV(sv);
7374 if (SvTYPE(sv) == SVt_PVCV) {
7375 cv = (CV*)sv;
a0714e2c 7376 *gvp = NULL;
62f274bf
GS
7377 *st = CvSTASH(cv);
7378 return cv;
7379 }
7380 else if(isGV(sv))
7381 gv = (GV*)sv;
7382 else
cea2e8a9 7383 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7384 }
62f274bf 7385 else if (isGV(sv))
79072805
LW
7386 gv = (GV*)sv;
7387 else
7a5fd60d 7388 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7389 *gvp = gv;
ef58ba18
NC
7390 if (!gv) {
7391 *st = NULL;
601f1833 7392 return NULL;
ef58ba18 7393 }
e26df76a
NC
7394 /* Some flags to gv_fetchsv mean don't really create the GV */
7395 if (SvTYPE(gv) != SVt_PVGV) {
7396 *st = NULL;
7397 return NULL;
7398 }
79072805 7399 *st = GvESTASH(gv);
8990e307 7400 fix_gv:
8ebc5c01 7401 if (lref && !GvCVu(gv)) {
4633a7c4 7402 SV *tmpsv;
748a9306 7403 ENTER;
561b68a9 7404 tmpsv = newSV(0);
bd61b366 7405 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7406 /* XXX this is probably not what they think they're getting.
7407 * It has the same effect as "sub name;", i.e. just a forward
7408 * declaration! */
774d564b 7409 newSUB(start_subparse(FALSE, 0),
4633a7c4 7410 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7411 NULL, NULL);
748a9306 7412 LEAVE;
8ebc5c01 7413 if (!GvCVu(gv))
35c1215d
NC
7414 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7415 sv);
8990e307 7416 }
8ebc5c01 7417 return GvCVu(gv);
79072805
LW
7418 }
7419}
7420
c461cf8f
JH
7421/*
7422=for apidoc sv_true
7423
7424Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7425Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7426instead use an in-line version.
c461cf8f
JH
7427
7428=cut
7429*/
7430
79072805 7431I32
864dbfa3 7432Perl_sv_true(pTHX_ register SV *sv)
79072805 7433{
8990e307
LW
7434 if (!sv)
7435 return 0;
79072805 7436 if (SvPOK(sv)) {
823a54a3
AL
7437 register const XPV* const tXpv = (XPV*)SvANY(sv);
7438 if (tXpv &&
c2f1de04 7439 (tXpv->xpv_cur > 1 ||
339049b0 7440 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7441 return 1;
7442 else
7443 return 0;
7444 }
7445 else {
7446 if (SvIOK(sv))
463ee0b2 7447 return SvIVX(sv) != 0;
79072805
LW
7448 else {
7449 if (SvNOK(sv))
463ee0b2 7450 return SvNVX(sv) != 0.0;
79072805 7451 else
463ee0b2 7452 return sv_2bool(sv);
79072805
LW
7453 }
7454 }
7455}
79072805 7456
645c22ef 7457/*
c461cf8f
JH
7458=for apidoc sv_pvn_force
7459
7460Get a sensible string out of the SV somehow.
645c22ef
DM
7461A private implementation of the C<SvPV_force> macro for compilers which
7462can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7463
8d6d96c1
HS
7464=for apidoc sv_pvn_force_flags
7465
7466Get a sensible string out of the SV somehow.
7467If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7468appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7469implemented in terms of this function.
645c22ef
DM
7470You normally want to use the various wrapper macros instead: see
7471C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7472
7473=cut
7474*/
7475
7476char *
7477Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7478{
97aff369 7479 dVAR;
6fc92669 7480 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7481 sv_force_normal_flags(sv, 0);
1c846c1f 7482
a0d0e21e 7483 if (SvPOK(sv)) {
13c5b33c
NC
7484 if (lp)
7485 *lp = SvCUR(sv);
a0d0e21e
LW
7486 }
7487 else {
a3b680e6 7488 char *s;
13c5b33c
NC
7489 STRLEN len;
7490
4d84ee25 7491 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7492 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7493 if (PL_op)
7494 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7495 ref, OP_NAME(PL_op));
4d84ee25 7496 else
b64e5050 7497 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7498 }
b64e5050 7499 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7500 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7501 OP_NAME(PL_op));
b64e5050 7502 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7503 if (lp)
7504 *lp = len;
7505
3f7c398e 7506 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7507 if (SvROK(sv))
7508 sv_unref(sv);
862a34c6 7509 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7510 SvGROW(sv, len + 1);
706aa1c9 7511 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7512 SvCUR_set(sv, len);
7513 *SvEND(sv) = '\0';
7514 }
7515 if (!SvPOK(sv)) {
7516 SvPOK_on(sv); /* validate pointer */
7517 SvTAINT(sv);
1d7c1841 7518 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7519 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7520 }
7521 }
4d84ee25 7522 return SvPVX_mutable(sv);
a0d0e21e
LW
7523}
7524
645c22ef 7525/*
645c22ef
DM
7526=for apidoc sv_pvbyten_force
7527
0feed65a 7528The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7529
7530=cut
7531*/
7532
7340a771
GS
7533char *
7534Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7535{
46ec2f14 7536 sv_pvn_force(sv,lp);
ffebcc3e 7537 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7538 *lp = SvCUR(sv);
7539 return SvPVX(sv);
7340a771
GS
7540}
7541
645c22ef 7542/*
c461cf8f
JH
7543=for apidoc sv_pvutf8n_force
7544
0feed65a 7545The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7546
7547=cut
7548*/
7549
7340a771
GS
7550char *
7551Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7552{
46ec2f14 7553 sv_pvn_force(sv,lp);
560a288e 7554 sv_utf8_upgrade(sv);
46ec2f14
TS
7555 *lp = SvCUR(sv);
7556 return SvPVX(sv);
7340a771
GS
7557}
7558
c461cf8f
JH
7559/*
7560=for apidoc sv_reftype
7561
7562Returns a string describing what the SV is a reference to.
7563
7564=cut
7565*/
7566
1cb0ed9b 7567char *
bfed75c6 7568Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7569{
07409e01
NC
7570 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7571 inside return suggests a const propagation bug in g++. */
c86bf373 7572 if (ob && SvOBJECT(sv)) {
1b6737cc 7573 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7574 return name ? name : (char *) "__ANON__";
c86bf373 7575 }
a0d0e21e
LW
7576 else {
7577 switch (SvTYPE(sv)) {
7578 case SVt_NULL:
7579 case SVt_IV:
7580 case SVt_NV:
7581 case SVt_RV:
7582 case SVt_PV:
7583 case SVt_PVIV:
7584 case SVt_PVNV:
7585 case SVt_PVMG:
7586 case SVt_PVBM:
1cb0ed9b 7587 if (SvVOK(sv))
439cb1c4 7588 return "VSTRING";
a0d0e21e
LW
7589 if (SvROK(sv))
7590 return "REF";
7591 else
7592 return "SCALAR";
1cb0ed9b 7593
07409e01 7594 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7595 /* tied lvalues should appear to be
7596 * scalars for backwards compatitbility */
7597 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7598 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7599 case SVt_PVAV: return "ARRAY";
7600 case SVt_PVHV: return "HASH";
7601 case SVt_PVCV: return "CODE";
7602 case SVt_PVGV: return "GLOB";
1d2dff63 7603 case SVt_PVFM: return "FORMAT";
27f9d8f3 7604 case SVt_PVIO: return "IO";
a0d0e21e
LW
7605 default: return "UNKNOWN";
7606 }
7607 }
7608}
7609
954c1994
GS
7610/*
7611=for apidoc sv_isobject
7612
7613Returns a boolean indicating whether the SV is an RV pointing to a blessed
7614object. If the SV is not an RV, or if the object is not blessed, then this
7615will return false.
7616
7617=cut
7618*/
7619
463ee0b2 7620int
864dbfa3 7621Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7622{
68dc0745 7623 if (!sv)
7624 return 0;
5b295bef 7625 SvGETMAGIC(sv);
85e6fe83
LW
7626 if (!SvROK(sv))
7627 return 0;
7628 sv = (SV*)SvRV(sv);
7629 if (!SvOBJECT(sv))
7630 return 0;
7631 return 1;
7632}
7633
954c1994
GS
7634/*
7635=for apidoc sv_isa
7636
7637Returns a boolean indicating whether the SV is blessed into the specified
7638class. This does not check for subtypes; use C<sv_derived_from> to verify
7639an inheritance relationship.
7640
7641=cut
7642*/
7643
85e6fe83 7644int
864dbfa3 7645Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7646{
bfcb3514 7647 const char *hvname;
68dc0745 7648 if (!sv)
7649 return 0;
5b295bef 7650 SvGETMAGIC(sv);
ed6116ce 7651 if (!SvROK(sv))
463ee0b2 7652 return 0;
ed6116ce
LW
7653 sv = (SV*)SvRV(sv);
7654 if (!SvOBJECT(sv))
463ee0b2 7655 return 0;
bfcb3514
NC
7656 hvname = HvNAME_get(SvSTASH(sv));
7657 if (!hvname)
e27ad1f2 7658 return 0;
463ee0b2 7659
bfcb3514 7660 return strEQ(hvname, name);
463ee0b2
LW
7661}
7662
954c1994
GS
7663/*
7664=for apidoc newSVrv
7665
7666Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7667it will be upgraded to one. If C<classname> is non-null then the new SV will
7668be blessed in the specified package. The new SV is returned and its
7669reference count is 1.
7670
7671=cut
7672*/
7673
463ee0b2 7674SV*
864dbfa3 7675Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7676{
97aff369 7677 dVAR;
463ee0b2
LW
7678 SV *sv;
7679
4561caa4 7680 new_SV(sv);
51cf62d8 7681
765f542d 7682 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7683 SvAMAGIC_off(rv);
51cf62d8 7684
0199fce9 7685 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7686 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7687 SvREFCNT(rv) = 0;
7688 sv_clear(rv);
7689 SvFLAGS(rv) = 0;
7690 SvREFCNT(rv) = refcnt;
0199fce9 7691
dc5494d2
NC
7692 sv_upgrade(rv, SVt_RV);
7693 } else if (SvROK(rv)) {
7694 SvREFCNT_dec(SvRV(rv));
7695 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7696 sv_upgrade(rv, SVt_RV);
7697 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7698 SvPV_free(rv);
0199fce9
JD
7699 SvCUR_set(rv, 0);
7700 SvLEN_set(rv, 0);
7701 }
51cf62d8 7702
0c34ef67 7703 SvOK_off(rv);
b162af07 7704 SvRV_set(rv, sv);
ed6116ce 7705 SvROK_on(rv);
463ee0b2 7706
a0d0e21e 7707 if (classname) {
1b6737cc 7708 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7709 (void)sv_bless(rv, stash);
7710 }
7711 return sv;
7712}
7713
954c1994
GS
7714/*
7715=for apidoc sv_setref_pv
7716
7717Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7718argument will be upgraded to an RV. That RV will be modified to point to
7719the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7720into the SV. The C<classname> argument indicates the package for the
bd61b366 7721blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7722will have a reference count of 1, and the RV will be returned.
954c1994
GS
7723
7724Do not use with other Perl types such as HV, AV, SV, CV, because those
7725objects will become corrupted by the pointer copy process.
7726
7727Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7728
7729=cut
7730*/
7731
a0d0e21e 7732SV*
864dbfa3 7733Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7734{
97aff369 7735 dVAR;
189b2af5 7736 if (!pv) {
3280af22 7737 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7738 SvSETMAGIC(rv);
7739 }
a0d0e21e 7740 else
56431972 7741 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7742 return rv;
7743}
7744
954c1994
GS
7745/*
7746=for apidoc sv_setref_iv
7747
7748Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7749argument will be upgraded to an RV. That RV will be modified to point to
7750the new SV. The C<classname> argument indicates the package for the
bd61b366 7751blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7752will have a reference count of 1, and the RV will be returned.
954c1994
GS
7753
7754=cut
7755*/
7756
a0d0e21e 7757SV*
864dbfa3 7758Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7759{
7760 sv_setiv(newSVrv(rv,classname), iv);
7761 return rv;
7762}
7763
954c1994 7764/*
e1c57cef
JH
7765=for apidoc sv_setref_uv
7766
7767Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7768argument will be upgraded to an RV. That RV will be modified to point to
7769the new SV. The C<classname> argument indicates the package for the
bd61b366 7770blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7771will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7772
7773=cut
7774*/
7775
7776SV*
7777Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7778{
7779 sv_setuv(newSVrv(rv,classname), uv);
7780 return rv;
7781}
7782
7783/*
954c1994
GS
7784=for apidoc sv_setref_nv
7785
7786Copies a double into a new SV, optionally blessing the SV. The C<rv>
7787argument will be upgraded to an RV. That RV will be modified to point to
7788the new SV. The C<classname> argument indicates the package for the
bd61b366 7789blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7790will have a reference count of 1, and the RV will be returned.
954c1994
GS
7791
7792=cut
7793*/
7794
a0d0e21e 7795SV*
65202027 7796Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7797{
7798 sv_setnv(newSVrv(rv,classname), nv);
7799 return rv;
7800}
463ee0b2 7801
954c1994
GS
7802/*
7803=for apidoc sv_setref_pvn
7804
7805Copies a string into a new SV, optionally blessing the SV. The length of the
7806string must be specified with C<n>. The C<rv> argument will be upgraded to
7807an RV. That RV will be modified to point to the new SV. The C<classname>
7808argument indicates the package for the blessing. Set C<classname> to
bd61b366 7809C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7810of 1, and the RV will be returned.
954c1994
GS
7811
7812Note that C<sv_setref_pv> copies the pointer while this copies the string.
7813
7814=cut
7815*/
7816
a0d0e21e 7817SV*
1b6737cc 7818Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7819{
7820 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7821 return rv;
7822}
7823
954c1994
GS
7824/*
7825=for apidoc sv_bless
7826
7827Blesses an SV into a specified package. The SV must be an RV. The package
7828must be designated by its stash (see C<gv_stashpv()>). The reference count
7829of the SV is unaffected.
7830
7831=cut
7832*/
7833
a0d0e21e 7834SV*
864dbfa3 7835Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7836{
97aff369 7837 dVAR;
76e3520e 7838 SV *tmpRef;
a0d0e21e 7839 if (!SvROK(sv))
cea2e8a9 7840 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7841 tmpRef = SvRV(sv);
7842 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7843 if (SvREADONLY(tmpRef))
cea2e8a9 7844 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7845 if (SvOBJECT(tmpRef)) {
7846 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7847 --PL_sv_objcount;
76e3520e 7848 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7849 }
a0d0e21e 7850 }
76e3520e
GS
7851 SvOBJECT_on(tmpRef);
7852 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7853 ++PL_sv_objcount;
862a34c6 7854 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7855 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7856
2e3febc6
CS
7857 if (Gv_AMG(stash))
7858 SvAMAGIC_on(sv);
7859 else
7860 SvAMAGIC_off(sv);
a0d0e21e 7861
1edbfb88
AB
7862 if(SvSMAGICAL(tmpRef))
7863 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7864 mg_set(tmpRef);
7865
7866
ecdeb87c 7867
a0d0e21e
LW
7868 return sv;
7869}
7870
645c22ef 7871/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7872 */
7873
76e3520e 7874STATIC void
cea2e8a9 7875S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7876{
97aff369 7877 dVAR;
850fabdf 7878 void *xpvmg;
b37c2d43 7879 SV * const temp = sv_newmortal();
850fabdf 7880
a0d0e21e
LW
7881 assert(SvTYPE(sv) == SVt_PVGV);
7882 SvFAKE_off(sv);
180488f8
NC
7883 gv_efullname3(temp, (GV *) sv, "*");
7884
f7877b28 7885 if (GvGP(sv)) {
1edc1566 7886 gp_free((GV*)sv);
f7877b28 7887 }
e826b3c7 7888 if (GvSTASH(sv)) {
e15faf7d 7889 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7890 GvSTASH(sv) = NULL;
e826b3c7 7891 }
a5f75d66 7892 GvMULTI_off(sv);
acda4c6a
NC
7893 if (GvNAME_HEK(sv)) {
7894 unshare_hek(GvNAME_HEK(sv));
7895 }
dedf8e73 7896 SvSCREAM_off(sv);
850fabdf
GS
7897
7898 /* need to keep SvANY(sv) in the right arena */
7899 xpvmg = new_XPVMG();
7900 StructCopy(SvANY(sv), xpvmg, XPVMG);
7901 del_XPVGV(SvANY(sv));
7902 SvANY(sv) = xpvmg;
7903
a0d0e21e
LW
7904 SvFLAGS(sv) &= ~SVTYPEMASK;
7905 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7906
7907 /* Intentionally not calling any local SET magic, as this isn't so much a
7908 set operation as merely an internal storage change. */
7909 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7910}
7911
954c1994 7912/*
840a7b70 7913=for apidoc sv_unref_flags
954c1994
GS
7914
7915Unsets the RV status of the SV, and decrements the reference count of
7916whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7917as a reversal of C<newSVrv>. The C<cflags> argument can contain
7918C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7919(otherwise the decrementing is conditional on the reference count being
7920different from one or the reference being a readonly SV).
7889fe52 7921See C<SvROK_off>.
954c1994
GS
7922
7923=cut
7924*/
7925
ed6116ce 7926void
e15faf7d 7927Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7928{
b64e5050 7929 SV* const target = SvRV(ref);
810b8aa5 7930
e15faf7d
NC
7931 if (SvWEAKREF(ref)) {
7932 sv_del_backref(target, ref);
7933 SvWEAKREF_off(ref);
7934 SvRV_set(ref, NULL);
810b8aa5
GS
7935 return;
7936 }
e15faf7d
NC
7937 SvRV_set(ref, NULL);
7938 SvROK_off(ref);
7939 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7940 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7941 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7942 SvREFCNT_dec(target);
840a7b70 7943 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7944 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7945}
8990e307 7946
840a7b70 7947/*
645c22ef
DM
7948=for apidoc sv_untaint
7949
7950Untaint an SV. Use C<SvTAINTED_off> instead.
7951=cut
7952*/
7953
bbce6d69 7954void
864dbfa3 7955Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7956{
13f57bf8 7957 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 7958 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 7959 if (mg)
565764a8 7960 mg->mg_len &= ~1;
36477c24 7961 }
bbce6d69 7962}
7963
645c22ef
DM
7964/*
7965=for apidoc sv_tainted
7966
7967Test an SV for taintedness. Use C<SvTAINTED> instead.
7968=cut
7969*/
7970
bbce6d69 7971bool
864dbfa3 7972Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 7973{
13f57bf8 7974 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 7975 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 7976 if (mg && (mg->mg_len & 1) )
36477c24 7977 return TRUE;
7978 }
7979 return FALSE;
bbce6d69 7980}
7981
09540bc3
JH
7982/*
7983=for apidoc sv_setpviv
7984
7985Copies an integer into the given SV, also updating its string value.
7986Does not handle 'set' magic. See C<sv_setpviv_mg>.
7987
7988=cut
7989*/
7990
7991void
7992Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
7993{
7994 char buf[TYPE_CHARS(UV)];
7995 char *ebuf;
b64e5050 7996 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
7997
7998 sv_setpvn(sv, ptr, ebuf - ptr);
7999}
8000
8001/*
8002=for apidoc sv_setpviv_mg
8003
8004Like C<sv_setpviv>, but also handles 'set' magic.
8005
8006=cut
8007*/
8008
8009void
8010Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8011{
df7eb254 8012 sv_setpviv(sv, iv);
09540bc3
JH
8013 SvSETMAGIC(sv);
8014}
8015
cea2e8a9 8016#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8017
8018/* pTHX_ magic can't cope with varargs, so this is a no-context
8019 * version of the main function, (which may itself be aliased to us).
8020 * Don't access this version directly.
8021 */
8022
cea2e8a9
GS
8023void
8024Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8025{
8026 dTHX;
8027 va_list args;
8028 va_start(args, pat);
c5be433b 8029 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8030 va_end(args);
8031}
8032
645c22ef
DM
8033/* pTHX_ magic can't cope with varargs, so this is a no-context
8034 * version of the main function, (which may itself be aliased to us).
8035 * Don't access this version directly.
8036 */
cea2e8a9
GS
8037
8038void
8039Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8040{
8041 dTHX;
8042 va_list args;
8043 va_start(args, pat);
c5be433b 8044 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8045 va_end(args);
cea2e8a9
GS
8046}
8047#endif
8048
954c1994
GS
8049/*
8050=for apidoc sv_setpvf
8051
bffc3d17
SH
8052Works like C<sv_catpvf> but copies the text into the SV instead of
8053appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8054
8055=cut
8056*/
8057
46fc3d4c 8058void
864dbfa3 8059Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8060{
8061 va_list args;
46fc3d4c 8062 va_start(args, pat);
c5be433b 8063 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8064 va_end(args);
8065}
8066
bffc3d17
SH
8067/*
8068=for apidoc sv_vsetpvf
8069
8070Works like C<sv_vcatpvf> but copies the text into the SV instead of
8071appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8072
8073Usually used via its frontend C<sv_setpvf>.
8074
8075=cut
8076*/
645c22ef 8077
c5be433b
GS
8078void
8079Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8080{
4608196e 8081 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8082}
ef50df4b 8083
954c1994
GS
8084/*
8085=for apidoc sv_setpvf_mg
8086
8087Like C<sv_setpvf>, but also handles 'set' magic.
8088
8089=cut
8090*/
8091
ef50df4b 8092void
864dbfa3 8093Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8094{
8095 va_list args;
ef50df4b 8096 va_start(args, pat);
c5be433b 8097 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8098 va_end(args);
c5be433b
GS
8099}
8100
bffc3d17
SH
8101/*
8102=for apidoc sv_vsetpvf_mg
8103
8104Like C<sv_vsetpvf>, but also handles 'set' magic.
8105
8106Usually used via its frontend C<sv_setpvf_mg>.
8107
8108=cut
8109*/
645c22ef 8110
c5be433b
GS
8111void
8112Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8113{
4608196e 8114 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8115 SvSETMAGIC(sv);
8116}
8117
cea2e8a9 8118#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8119
8120/* pTHX_ magic can't cope with varargs, so this is a no-context
8121 * version of the main function, (which may itself be aliased to us).
8122 * Don't access this version directly.
8123 */
8124
cea2e8a9
GS
8125void
8126Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8127{
8128 dTHX;
8129 va_list args;
8130 va_start(args, pat);
c5be433b 8131 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8132 va_end(args);
8133}
8134
645c22ef
DM
8135/* pTHX_ magic can't cope with varargs, so this is a no-context
8136 * version of the main function, (which may itself be aliased to us).
8137 * Don't access this version directly.
8138 */
8139
cea2e8a9
GS
8140void
8141Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8142{
8143 dTHX;
8144 va_list args;
8145 va_start(args, pat);
c5be433b 8146 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8147 va_end(args);
cea2e8a9
GS
8148}
8149#endif
8150
954c1994
GS
8151/*
8152=for apidoc sv_catpvf
8153
d5ce4a7c
GA
8154Processes its arguments like C<sprintf> and appends the formatted
8155output to an SV. If the appended data contains "wide" characters
8156(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8157and characters >255 formatted with %c), the original SV might get
bffc3d17 8158upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8159C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8160valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8161
d5ce4a7c 8162=cut */
954c1994 8163
46fc3d4c 8164void
864dbfa3 8165Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8166{
8167 va_list args;
46fc3d4c 8168 va_start(args, pat);
c5be433b 8169 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8170 va_end(args);
8171}
8172
bffc3d17
SH
8173/*
8174=for apidoc sv_vcatpvf
8175
8176Processes its arguments like C<vsprintf> and appends the formatted output
8177to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8178
8179Usually used via its frontend C<sv_catpvf>.
8180
8181=cut
8182*/
645c22ef 8183
ef50df4b 8184void
c5be433b
GS
8185Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8186{
4608196e 8187 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8188}
8189
954c1994
GS
8190/*
8191=for apidoc sv_catpvf_mg
8192
8193Like C<sv_catpvf>, but also handles 'set' magic.
8194
8195=cut
8196*/
8197
c5be433b 8198void
864dbfa3 8199Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8200{
8201 va_list args;
ef50df4b 8202 va_start(args, pat);
c5be433b 8203 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8204 va_end(args);
c5be433b
GS
8205}
8206
bffc3d17
SH
8207/*
8208=for apidoc sv_vcatpvf_mg
8209
8210Like C<sv_vcatpvf>, but also handles 'set' magic.
8211
8212Usually used via its frontend C<sv_catpvf_mg>.
8213
8214=cut
8215*/
645c22ef 8216
c5be433b
GS
8217void
8218Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8219{
4608196e 8220 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8221 SvSETMAGIC(sv);
8222}
8223
954c1994
GS
8224/*
8225=for apidoc sv_vsetpvfn
8226
bffc3d17 8227Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8228appending it.
8229
bffc3d17 8230Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8231
954c1994
GS
8232=cut
8233*/
8234
46fc3d4c 8235void
7d5ea4e7 8236Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8237{
8238 sv_setpvn(sv, "", 0);
7d5ea4e7 8239 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8240}
8241
2d00ba3b 8242STATIC I32
9dd79c3f 8243S_expect_number(pTHX_ char** pattern)
211dfcf1 8244{
97aff369 8245 dVAR;
211dfcf1
HS
8246 I32 var = 0;
8247 switch (**pattern) {
8248 case '1': case '2': case '3':
8249 case '4': case '5': case '6':
8250 case '7': case '8': case '9':
2fba7546
GA
8251 var = *(*pattern)++ - '0';
8252 while (isDIGIT(**pattern)) {
5f66b61c 8253 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8254 if (tmp < var)
8255 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8256 var = tmp;
8257 }
211dfcf1
HS
8258 }
8259 return var;
8260}
211dfcf1 8261
c445ea15
AL
8262STATIC char *
8263S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8264{
a3b680e6 8265 const int neg = nv < 0;
4151a5fe 8266 UV uv;
4151a5fe
IZ
8267
8268 if (neg)
8269 nv = -nv;
8270 if (nv < UV_MAX) {
b464bac0 8271 char *p = endbuf;
4151a5fe 8272 nv += 0.5;
028f8eaa 8273 uv = (UV)nv;
4151a5fe
IZ
8274 if (uv & 1 && uv == nv)
8275 uv--; /* Round to even */
8276 do {
a3b680e6 8277 const unsigned dig = uv % 10;
4151a5fe
IZ
8278 *--p = '0' + dig;
8279 } while (uv /= 10);
8280 if (neg)
8281 *--p = '-';
8282 *len = endbuf - p;
8283 return p;
8284 }
bd61b366 8285 return NULL;
4151a5fe
IZ
8286}
8287
8288
954c1994
GS
8289/*
8290=for apidoc sv_vcatpvfn
8291
8292Processes its arguments like C<vsprintf> and appends the formatted output
8293to an SV. Uses an array of SVs if the C style variable argument list is
8294missing (NULL). When running with taint checks enabled, indicates via
8295C<maybe_tainted> if results are untrustworthy (often due to the use of
8296locales).
8297
bffc3d17 8298Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8299
954c1994
GS
8300=cut
8301*/
8302
8896765a
RB
8303
8304#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8305 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8306 vec_utf8 = DO_UTF8(vecsv);
8307
1ef29b0e
RGS
8308/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8309
46fc3d4c 8310void
7d5ea4e7 8311Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8312{
97aff369 8313 dVAR;
46fc3d4c 8314 char *p;
8315 char *q;
a3b680e6 8316 const char *patend;
fc36a67e 8317 STRLEN origlen;
46fc3d4c 8318 I32 svix = 0;
27da23d5 8319 static const char nullstr[] = "(null)";
a0714e2c 8320 SV *argsv = NULL;
b464bac0
AL
8321 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8322 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8323 SV *nsv = NULL;
4151a5fe
IZ
8324 /* Times 4: a decimal digit takes more than 3 binary digits.
8325 * NV_DIG: mantissa takes than many decimal digits.
8326 * Plus 32: Playing safe. */
8327 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8328 /* large enough for "%#.#f" --chip */
8329 /* what about long double NVs? --jhi */
db79b45b 8330
53c1dcc0
AL
8331 PERL_UNUSED_ARG(maybe_tainted);
8332
46fc3d4c 8333 /* no matter what, this is a string now */
fc36a67e 8334 (void)SvPV_force(sv, origlen);
46fc3d4c 8335
8896765a 8336 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8337 if (patlen == 0)
8338 return;
0dbb1585 8339 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8340 if (args) {
8341 const char * const s = va_arg(*args, char*);
8342 sv_catpv(sv, s ? s : nullstr);
8343 }
8344 else if (svix < svmax) {
8345 sv_catsv(sv, *svargs);
2d03de9c
AL
8346 }
8347 return;
0dbb1585 8348 }
8896765a
RB
8349 if (args && patlen == 3 && pat[0] == '%' &&
8350 pat[1] == '-' && pat[2] == 'p') {
8351 argsv = va_arg(*args, SV*);
8352 sv_catsv(sv, argsv);
8896765a 8353 return;
46fc3d4c 8354 }
8355
1d917b39 8356#ifndef USE_LONG_DOUBLE
4151a5fe 8357 /* special-case "%.<number>[gf]" */
7af36d83 8358 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8359 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8360 unsigned digits = 0;
8361 const char *pp;
8362
8363 pp = pat + 2;
8364 while (*pp >= '0' && *pp <= '9')
8365 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8366 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8367 NV nv;
8368
7af36d83 8369 if (svix < svmax)
4151a5fe
IZ
8370 nv = SvNV(*svargs);
8371 else
8372 return;
8373 if (*pp == 'g') {
2873255c
NC
8374 /* Add check for digits != 0 because it seems that some
8375 gconverts are buggy in this case, and we don't yet have
8376 a Configure test for this. */
8377 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8378 /* 0, point, slack */
2e59c212 8379 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8380 sv_catpv(sv, ebuf);
8381 if (*ebuf) /* May return an empty string for digits==0 */
8382 return;
8383 }
8384 } else if (!digits) {
8385 STRLEN l;
8386
8387 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8388 sv_catpvn(sv, p, l);
8389 return;
8390 }
8391 }
8392 }
8393 }
1d917b39 8394#endif /* !USE_LONG_DOUBLE */
4151a5fe 8395
2cf2cfc6 8396 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8397 has_utf8 = TRUE;
2cf2cfc6 8398
46fc3d4c 8399 patend = (char*)pat + patlen;
8400 for (p = (char*)pat; p < patend; p = q) {
8401 bool alt = FALSE;
8402 bool left = FALSE;
b22c7a20 8403 bool vectorize = FALSE;
211dfcf1 8404 bool vectorarg = FALSE;
2cf2cfc6 8405 bool vec_utf8 = FALSE;
46fc3d4c 8406 char fill = ' ';
8407 char plus = 0;
8408 char intsize = 0;
8409 STRLEN width = 0;
fc36a67e 8410 STRLEN zeros = 0;
46fc3d4c 8411 bool has_precis = FALSE;
8412 STRLEN precis = 0;
c445ea15 8413 const I32 osvix = svix;
2cf2cfc6 8414 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8415#ifdef HAS_LDBL_SPRINTF_BUG
8416 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8417 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8418 bool fix_ldbl_sprintf_bug = FALSE;
8419#endif
205f51d8 8420
46fc3d4c 8421 char esignbuf[4];
89ebb4a3 8422 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8423 STRLEN esignlen = 0;
8424
bd61b366 8425 const char *eptr = NULL;
fc36a67e 8426 STRLEN elen = 0;
a0714e2c 8427 SV *vecsv = NULL;
4608196e 8428 const U8 *vecstr = NULL;
b22c7a20 8429 STRLEN veclen = 0;
934abaf1 8430 char c = 0;
46fc3d4c 8431 int i;
9c5ffd7c 8432 unsigned base = 0;
8c8eb53c
RB
8433 IV iv = 0;
8434 UV uv = 0;
9e5b023a
JH
8435 /* we need a long double target in case HAS_LONG_DOUBLE but
8436 not USE_LONG_DOUBLE
8437 */
35fff930 8438#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8439 long double nv;
8440#else
65202027 8441 NV nv;
9e5b023a 8442#endif
46fc3d4c 8443 STRLEN have;
8444 STRLEN need;
8445 STRLEN gap;
7af36d83 8446 const char *dotstr = ".";
b22c7a20 8447 STRLEN dotstrlen = 1;
211dfcf1 8448 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8449 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8450 I32 epix = 0; /* explicit precision index */
8451 I32 evix = 0; /* explicit vector index */
eb3fce90 8452 bool asterisk = FALSE;
46fc3d4c 8453
211dfcf1 8454 /* echo everything up to the next format specification */
46fc3d4c 8455 for (q = p; q < patend && *q != '%'; ++q) ;
8456 if (q > p) {
db79b45b
JH
8457 if (has_utf8 && !pat_utf8)
8458 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8459 else
8460 sv_catpvn(sv, p, q - p);
46fc3d4c 8461 p = q;
8462 }
8463 if (q++ >= patend)
8464 break;
8465
211dfcf1
HS
8466/*
8467 We allow format specification elements in this order:
8468 \d+\$ explicit format parameter index
8469 [-+ 0#]+ flags
a472f209 8470 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8471 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8472 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8473 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8474 [hlqLV] size
8896765a
RB
8475 [%bcdefginopsuxDFOUX] format (mandatory)
8476*/
8477
8478 if (args) {
8479/*
8480 As of perl5.9.3, printf format checking is on by default.
8481 Internally, perl uses %p formats to provide an escape to
8482 some extended formatting. This block deals with those
8483 extensions: if it does not match, (char*)q is reset and
8484 the normal format processing code is used.
8485
8486 Currently defined extensions are:
8487 %p include pointer address (standard)
8488 %-p (SVf) include an SV (previously %_)
8489 %-<num>p include an SV with precision <num>
8490 %1p (VDf) include a v-string (as %vd)
8491 %<num>p reserved for future extensions
8492
8493 Robin Barker 2005-07-14
211dfcf1 8494*/
8896765a
RB
8495 char* r = q;
8496 bool sv = FALSE;
8497 STRLEN n = 0;
8498 if (*q == '-')
8499 sv = *q++;
c445ea15 8500 n = expect_number(&q);
8896765a
RB
8501 if (*q++ == 'p') {
8502 if (sv) { /* SVf */
8503 if (n) {
8504 precis = n;
8505 has_precis = TRUE;
8506 }
8507 argsv = va_arg(*args, SV*);
8508 eptr = SvPVx_const(argsv, elen);
8509 if (DO_UTF8(argsv))
8510 is_utf8 = TRUE;
8511 goto string;
8512 }
8513#if vdNUMBER
8514 else if (n == vdNUMBER) { /* VDf */
8515 vectorize = TRUE;
8516 VECTORIZE_ARGS
8517 goto format_vd;
8518 }
8519#endif
8520 else if (n) {
8521 if (ckWARN_d(WARN_INTERNAL))
8522 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8523 "internal %%<num>p might conflict with future printf extensions");
8524 }
8525 }
8526 q = r;
8527 }
8528
c445ea15 8529 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8530 if (*q == '$') {
8531 ++q;
8532 efix = width;
8533 } else {
8534 goto gotwidth;
8535 }
8536 }
8537
fc36a67e 8538 /* FLAGS */
8539
46fc3d4c 8540 while (*q) {
8541 switch (*q) {
8542 case ' ':
8543 case '+':
8544 plus = *q++;
8545 continue;
8546
8547 case '-':
8548 left = TRUE;
8549 q++;
8550 continue;
8551
8552 case '0':
8553 fill = *q++;
8554 continue;
8555
8556 case '#':
8557 alt = TRUE;
8558 q++;
8559 continue;
8560
fc36a67e 8561 default:
8562 break;
8563 }
8564 break;
8565 }
46fc3d4c 8566
211dfcf1 8567 tryasterisk:
eb3fce90 8568 if (*q == '*') {
211dfcf1 8569 q++;
c445ea15 8570 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8571 if (*q++ != '$')
8572 goto unknown;
eb3fce90 8573 asterisk = TRUE;
211dfcf1
HS
8574 }
8575 if (*q == 'v') {
eb3fce90 8576 q++;
211dfcf1
HS
8577 if (vectorize)
8578 goto unknown;
9cbac4c7 8579 if ((vectorarg = asterisk)) {
211dfcf1
HS
8580 evix = ewix;
8581 ewix = 0;
8582 asterisk = FALSE;
8583 }
8584 vectorize = TRUE;
8585 goto tryasterisk;
eb3fce90
JH
8586 }
8587
211dfcf1 8588 if (!asterisk)
858a90f9 8589 {
7a5fa8a2 8590 if( *q == '0' )
f3583277 8591 fill = *q++;
c445ea15 8592 width = expect_number(&q);
858a90f9 8593 }
211dfcf1
HS
8594
8595 if (vectorize) {
8596 if (vectorarg) {
8597 if (args)
8598 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8599 else if (evix) {
8600 vecsv = (evix > 0 && evix <= svmax)
8601 ? svargs[evix-1] : &PL_sv_undef;
8602 } else {
8603 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8604 }
245d4a47 8605 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8606 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8607 bad with tied or overloaded values that return UTF8. */
211dfcf1 8608 if (DO_UTF8(vecsv))
2cf2cfc6 8609 is_utf8 = TRUE;
640283f5
NC
8610 else if (has_utf8) {
8611 vecsv = sv_mortalcopy(vecsv);
8612 sv_utf8_upgrade(vecsv);
8613 dotstr = SvPV_const(vecsv, dotstrlen);
8614 is_utf8 = TRUE;
8615 }
211dfcf1
HS
8616 }
8617 if (args) {
8896765a 8618 VECTORIZE_ARGS
eb3fce90 8619 }
7ad96abb 8620 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8621 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8622 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8623 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8624
8625 /* if this is a version object, we need to convert
8626 * back into v-string notation and then let the
8627 * vectorize happen normally
d7aa5382 8628 */
96b8f7ce
JP
8629 if (sv_derived_from(vecsv, "version")) {
8630 char *version = savesvpv(vecsv);
34ba6322
SP
8631 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8632 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8633 "vector argument not supported with alpha versions");
8634 goto unknown;
8635 }
96b8f7ce
JP
8636 vecsv = sv_newmortal();
8637 /* scan_vstring is expected to be called during
8638 * tokenization, so we need to fake up the end
8639 * of the buffer for it
8640 */
8641 PL_bufend = version + veclen;
8642 scan_vstring(version, vecsv);
8643 vecstr = (U8*)SvPV_const(vecsv, veclen);
8644 vec_utf8 = DO_UTF8(vecsv);
8645 Safefree(version);
d7aa5382 8646 }
211dfcf1
HS
8647 }
8648 else {
8649 vecstr = (U8*)"";
8650 veclen = 0;
8651 }
eb3fce90 8652 }
fc36a67e 8653
eb3fce90 8654 if (asterisk) {
fc36a67e 8655 if (args)
8656 i = va_arg(*args, int);
8657 else
eb3fce90
JH
8658 i = (ewix ? ewix <= svmax : svix < svmax) ?
8659 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8660 left |= (i < 0);
8661 width = (i < 0) ? -i : i;
fc36a67e 8662 }
211dfcf1 8663 gotwidth:
fc36a67e 8664
8665 /* PRECISION */
46fc3d4c 8666
fc36a67e 8667 if (*q == '.') {
8668 q++;
8669 if (*q == '*') {
211dfcf1 8670 q++;
c445ea15 8671 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8672 goto unknown;
8673 /* XXX: todo, support specified precision parameter */
8674 if (epix)
211dfcf1 8675 goto unknown;
46fc3d4c 8676 if (args)
8677 i = va_arg(*args, int);
8678 else
eb3fce90
JH
8679 i = (ewix ? ewix <= svmax : svix < svmax)
8680 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8681 precis = (i < 0) ? 0 : i;
fc36a67e 8682 }
8683 else {
8684 precis = 0;
8685 while (isDIGIT(*q))
8686 precis = precis * 10 + (*q++ - '0');
8687 }
8688 has_precis = TRUE;
8689 }
46fc3d4c 8690
fc36a67e 8691 /* SIZE */
46fc3d4c 8692
fc36a67e 8693 switch (*q) {
c623ac67
GS
8694#ifdef WIN32
8695 case 'I': /* Ix, I32x, and I64x */
8696# ifdef WIN64
8697 if (q[1] == '6' && q[2] == '4') {
8698 q += 3;
8699 intsize = 'q';
8700 break;
8701 }
8702# endif
8703 if (q[1] == '3' && q[2] == '2') {
8704 q += 3;
8705 break;
8706 }
8707# ifdef WIN64
8708 intsize = 'q';
8709# endif
8710 q++;
8711 break;
8712#endif
9e5b023a 8713#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8714 case 'L': /* Ld */
5f66b61c 8715 /*FALLTHROUGH*/
e5c81feb 8716#ifdef HAS_QUAD
6f9bb7fd 8717 case 'q': /* qd */
9e5b023a 8718#endif
6f9bb7fd
GS
8719 intsize = 'q';
8720 q++;
8721 break;
8722#endif
fc36a67e 8723 case 'l':
9e5b023a 8724#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8725 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8726 intsize = 'q';
8727 q += 2;
46fc3d4c 8728 break;
cf2093f6 8729 }
fc36a67e 8730#endif
5f66b61c 8731 /*FALLTHROUGH*/
fc36a67e 8732 case 'h':
5f66b61c 8733 /*FALLTHROUGH*/
fc36a67e 8734 case 'V':
8735 intsize = *q++;
46fc3d4c 8736 break;
8737 }
8738
fc36a67e 8739 /* CONVERSION */
8740
211dfcf1
HS
8741 if (*q == '%') {
8742 eptr = q++;
8743 elen = 1;
26372e71
GA
8744 if (vectorize) {
8745 c = '%';
8746 goto unknown;
8747 }
211dfcf1
HS
8748 goto string;
8749 }
8750
26372e71 8751 if (!vectorize && !args) {
86c51f8b
NC
8752 if (efix) {
8753 const I32 i = efix-1;
8754 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8755 } else {
8756 argsv = (svix >= 0 && svix < svmax)
8757 ? svargs[svix++] : &PL_sv_undef;
8758 }
863811b2 8759 }
211dfcf1 8760
46fc3d4c 8761 switch (c = *q++) {
8762
8763 /* STRINGS */
8764
46fc3d4c 8765 case 'c':
26372e71
GA
8766 if (vectorize)
8767 goto unknown;
8768 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8769 if ((uv > 255 ||
8770 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8771 && !IN_BYTES) {
dfe13c55 8772 eptr = (char*)utf8buf;
9041c2e3 8773 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8774 is_utf8 = TRUE;
7e2040f0
GS
8775 }
8776 else {
8777 c = (char)uv;
8778 eptr = &c;
8779 elen = 1;
a0ed51b3 8780 }
46fc3d4c 8781 goto string;
8782
46fc3d4c 8783 case 's':
26372e71
GA
8784 if (vectorize)
8785 goto unknown;
8786 if (args) {
fc36a67e 8787 eptr = va_arg(*args, char*);
c635e13b 8788 if (eptr)
1d7c1841
GS
8789#ifdef MACOS_TRADITIONAL
8790 /* On MacOS, %#s format is used for Pascal strings */
8791 if (alt)
8792 elen = *eptr++;
8793 else
8794#endif
c635e13b 8795 elen = strlen(eptr);
8796 else {
27da23d5 8797 eptr = (char *)nullstr;
c635e13b 8798 elen = sizeof nullstr - 1;
8799 }
46fc3d4c 8800 }
211dfcf1 8801 else {
4d84ee25 8802 eptr = SvPVx_const(argsv, elen);
7e2040f0 8803 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8804 if (has_precis && precis < elen) {
8805 I32 p = precis;
7e2040f0 8806 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8807 precis = p;
8808 }
8809 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8810 width += elen - sv_len_utf8(argsv);
a0ed51b3 8811 }
2cf2cfc6 8812 is_utf8 = TRUE;
a0ed51b3
LW
8813 }
8814 }
fc36a67e 8815
46fc3d4c 8816 string:
8817 if (has_precis && elen > precis)
8818 elen = precis;
8819 break;
8820
8821 /* INTEGERS */
8822
fc36a67e 8823 case 'p':
be75b157 8824 if (alt || vectorize)
c2e66d9e 8825 goto unknown;
211dfcf1 8826 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8827 base = 16;
8828 goto integer;
8829
46fc3d4c 8830 case 'D':
29fe7a80 8831#ifdef IV_IS_QUAD
22f3ae8c 8832 intsize = 'q';
29fe7a80 8833#else
46fc3d4c 8834 intsize = 'l';
29fe7a80 8835#endif
5f66b61c 8836 /*FALLTHROUGH*/
46fc3d4c 8837 case 'd':
8838 case 'i':
8896765a
RB
8839#if vdNUMBER
8840 format_vd:
8841#endif
b22c7a20 8842 if (vectorize) {
ba210ebe 8843 STRLEN ulen;
211dfcf1
HS
8844 if (!veclen)
8845 continue;
2cf2cfc6
A
8846 if (vec_utf8)
8847 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8848 UTF8_ALLOW_ANYUV);
b22c7a20 8849 else {
e83d50c9 8850 uv = *vecstr;
b22c7a20
GS
8851 ulen = 1;
8852 }
8853 vecstr += ulen;
8854 veclen -= ulen;
e83d50c9
JP
8855 if (plus)
8856 esignbuf[esignlen++] = plus;
b22c7a20
GS
8857 }
8858 else if (args) {
46fc3d4c 8859 switch (intsize) {
8860 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8861 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8862 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8863 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8864#ifdef HAS_QUAD
8865 case 'q': iv = va_arg(*args, Quad_t); break;
8866#endif
46fc3d4c 8867 }
8868 }
8869 else {
b10c0dba 8870 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8871 switch (intsize) {
b10c0dba
MHM
8872 case 'h': iv = (short)tiv; break;
8873 case 'l': iv = (long)tiv; break;
8874 case 'V':
8875 default: iv = tiv; break;
cf2093f6 8876#ifdef HAS_QUAD
b10c0dba 8877 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8878#endif
46fc3d4c 8879 }
8880 }
e83d50c9
JP
8881 if ( !vectorize ) /* we already set uv above */
8882 {
8883 if (iv >= 0) {
8884 uv = iv;
8885 if (plus)
8886 esignbuf[esignlen++] = plus;
8887 }
8888 else {
8889 uv = -iv;
8890 esignbuf[esignlen++] = '-';
8891 }
46fc3d4c 8892 }
8893 base = 10;
8894 goto integer;
8895
fc36a67e 8896 case 'U':
29fe7a80 8897#ifdef IV_IS_QUAD
22f3ae8c 8898 intsize = 'q';
29fe7a80 8899#else
fc36a67e 8900 intsize = 'l';
29fe7a80 8901#endif
5f66b61c 8902 /*FALLTHROUGH*/
fc36a67e 8903 case 'u':
8904 base = 10;
8905 goto uns_integer;
8906
4f19785b
WSI
8907 case 'b':
8908 base = 2;
8909 goto uns_integer;
8910
46fc3d4c 8911 case 'O':
29fe7a80 8912#ifdef IV_IS_QUAD
22f3ae8c 8913 intsize = 'q';
29fe7a80 8914#else
46fc3d4c 8915 intsize = 'l';
29fe7a80 8916#endif
5f66b61c 8917 /*FALLTHROUGH*/
46fc3d4c 8918 case 'o':
8919 base = 8;
8920 goto uns_integer;
8921
8922 case 'X':
46fc3d4c 8923 case 'x':
8924 base = 16;
46fc3d4c 8925
8926 uns_integer:
b22c7a20 8927 if (vectorize) {
ba210ebe 8928 STRLEN ulen;
b22c7a20 8929 vector:
211dfcf1
HS
8930 if (!veclen)
8931 continue;
2cf2cfc6
A
8932 if (vec_utf8)
8933 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8934 UTF8_ALLOW_ANYUV);
b22c7a20 8935 else {
a05b299f 8936 uv = *vecstr;
b22c7a20
GS
8937 ulen = 1;
8938 }
8939 vecstr += ulen;
8940 veclen -= ulen;
8941 }
8942 else if (args) {
46fc3d4c 8943 switch (intsize) {
8944 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8945 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8946 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8947 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8948#ifdef HAS_QUAD
9e3321a5 8949 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8950#endif
46fc3d4c 8951 }
8952 }
8953 else {
b10c0dba 8954 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8955 switch (intsize) {
b10c0dba
MHM
8956 case 'h': uv = (unsigned short)tuv; break;
8957 case 'l': uv = (unsigned long)tuv; break;
8958 case 'V':
8959 default: uv = tuv; break;
cf2093f6 8960#ifdef HAS_QUAD
b10c0dba 8961 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 8962#endif
46fc3d4c 8963 }
8964 }
8965
8966 integer:
4d84ee25
NC
8967 {
8968 char *ptr = ebuf + sizeof ebuf;
8969 switch (base) {
8970 unsigned dig;
8971 case 16:
8972 if (!uv)
8973 alt = FALSE;
8974 p = (char*)((c == 'X')
8975 ? "0123456789ABCDEF" : "0123456789abcdef");
8976 do {
8977 dig = uv & 15;
8978 *--ptr = p[dig];
8979 } while (uv >>= 4);
8980 if (alt) {
8981 esignbuf[esignlen++] = '0';
8982 esignbuf[esignlen++] = c; /* 'x' or 'X' */
8983 }
8984 break;
8985 case 8:
8986 do {
8987 dig = uv & 7;
8988 *--ptr = '0' + dig;
8989 } while (uv >>= 3);
8990 if (alt && *ptr != '0')
8991 *--ptr = '0';
8992 break;
8993 case 2:
ed2b91d2
GA
8994 if (!uv)
8995 alt = FALSE;
4d84ee25
NC
8996 do {
8997 dig = uv & 1;
8998 *--ptr = '0' + dig;
8999 } while (uv >>= 1);
9000 if (alt) {
9001 esignbuf[esignlen++] = '0';
9002 esignbuf[esignlen++] = 'b';
9003 }
9004 break;
9005 default: /* it had better be ten or less */
9006 do {
9007 dig = uv % base;
9008 *--ptr = '0' + dig;
9009 } while (uv /= base);
9010 break;
46fc3d4c 9011 }
4d84ee25
NC
9012 elen = (ebuf + sizeof ebuf) - ptr;
9013 eptr = ptr;
9014 if (has_precis) {
9015 if (precis > elen)
9016 zeros = precis - elen;
9017 else if (precis == 0 && elen == 1 && *eptr == '0')
9018 elen = 0;
eda88b6d 9019 }
c10ed8b9 9020 }
46fc3d4c 9021 break;
9022
9023 /* FLOATING POINT */
9024
fc36a67e 9025 case 'F':
9026 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9027 /*FALLTHROUGH*/
46fc3d4c 9028 case 'e': case 'E':
fc36a67e 9029 case 'f':
46fc3d4c 9030 case 'g': case 'G':
26372e71
GA
9031 if (vectorize)
9032 goto unknown;
46fc3d4c 9033
9034 /* This is evil, but floating point is even more evil */
9035
9e5b023a
JH
9036 /* for SV-style calling, we can only get NV
9037 for C-style calling, we assume %f is double;
9038 for simplicity we allow any of %Lf, %llf, %qf for long double
9039 */
9040 switch (intsize) {
9041 case 'V':
9042#if defined(USE_LONG_DOUBLE)
9043 intsize = 'q';
9044#endif
9045 break;
8a2e3f14 9046/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9047 case 'l':
5f66b61c 9048 /*FALLTHROUGH*/
9e5b023a
JH
9049 default:
9050#if defined(USE_LONG_DOUBLE)
9051 intsize = args ? 0 : 'q';
9052#endif
9053 break;
9054 case 'q':
9055#if defined(HAS_LONG_DOUBLE)
9056 break;
9057#else
5f66b61c 9058 /*FALLTHROUGH*/
9e5b023a
JH
9059#endif
9060 case 'h':
9e5b023a
JH
9061 goto unknown;
9062 }
9063
9064 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9065 nv = (args) ?
35fff930
JH
9066#if LONG_DOUBLESIZE > DOUBLESIZE
9067 intsize == 'q' ?
205f51d8
AS
9068 va_arg(*args, long double) :
9069 va_arg(*args, double)
35fff930 9070#else
205f51d8 9071 va_arg(*args, double)
35fff930 9072#endif
9e5b023a 9073 : SvNVx(argsv);
fc36a67e 9074
9075 need = 0;
9076 if (c != 'e' && c != 'E') {
9077 i = PERL_INT_MIN;
9e5b023a
JH
9078 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9079 will cast our (long double) to (double) */
73b309ea 9080 (void)Perl_frexp(nv, &i);
fc36a67e 9081 if (i == PERL_INT_MIN)
cea2e8a9 9082 Perl_die(aTHX_ "panic: frexp");
c635e13b 9083 if (i > 0)
fc36a67e 9084 need = BIT_DIGITS(i);
9085 }
9086 need += has_precis ? precis : 6; /* known default */
20f6aaab 9087
fc36a67e 9088 if (need < width)
9089 need = width;
9090
20f6aaab
AS
9091#ifdef HAS_LDBL_SPRINTF_BUG
9092 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9093 with sfio - Allen <allens@cpan.org> */
9094
9095# ifdef DBL_MAX
9096# define MY_DBL_MAX DBL_MAX
9097# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9098# if DOUBLESIZE >= 8
9099# define MY_DBL_MAX 1.7976931348623157E+308L
9100# else
9101# define MY_DBL_MAX 3.40282347E+38L
9102# endif
9103# endif
9104
9105# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9106# define MY_DBL_MAX_BUG 1L
20f6aaab 9107# else
205f51d8 9108# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9109# endif
20f6aaab 9110
205f51d8
AS
9111# ifdef DBL_MIN
9112# define MY_DBL_MIN DBL_MIN
9113# else /* XXX guessing! -Allen */
9114# if DOUBLESIZE >= 8
9115# define MY_DBL_MIN 2.2250738585072014E-308L
9116# else
9117# define MY_DBL_MIN 1.17549435E-38L
9118# endif
9119# endif
20f6aaab 9120
205f51d8
AS
9121 if ((intsize == 'q') && (c == 'f') &&
9122 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9123 (need < DBL_DIG)) {
9124 /* it's going to be short enough that
9125 * long double precision is not needed */
9126
9127 if ((nv <= 0L) && (nv >= -0L))
9128 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9129 else {
9130 /* would use Perl_fp_class as a double-check but not
9131 * functional on IRIX - see perl.h comments */
9132
9133 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9134 /* It's within the range that a double can represent */
9135#if defined(DBL_MAX) && !defined(DBL_MIN)
9136 if ((nv >= ((long double)1/DBL_MAX)) ||
9137 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9138#endif
205f51d8 9139 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9140 }
205f51d8
AS
9141 }
9142 if (fix_ldbl_sprintf_bug == TRUE) {
9143 double temp;
9144
9145 intsize = 0;
9146 temp = (double)nv;
9147 nv = (NV)temp;
9148 }
20f6aaab 9149 }
205f51d8
AS
9150
9151# undef MY_DBL_MAX
9152# undef MY_DBL_MAX_BUG
9153# undef MY_DBL_MIN
9154
20f6aaab
AS
9155#endif /* HAS_LDBL_SPRINTF_BUG */
9156
46fc3d4c 9157 need += 20; /* fudge factor */
80252599
GS
9158 if (PL_efloatsize < need) {
9159 Safefree(PL_efloatbuf);
9160 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9161 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9162 PL_efloatbuf[0] = '\0';
46fc3d4c 9163 }
9164
4151a5fe
IZ
9165 if ( !(width || left || plus || alt) && fill != '0'
9166 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9167 /* See earlier comment about buggy Gconvert when digits,
9168 aka precis is 0 */
9169 if ( c == 'g' && precis) {
2e59c212 9170 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9171 /* May return an empty string for digits==0 */
9172 if (*PL_efloatbuf) {
9173 elen = strlen(PL_efloatbuf);
4151a5fe 9174 goto float_converted;
4150c189 9175 }
4151a5fe
IZ
9176 } else if ( c == 'f' && !precis) {
9177 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9178 break;
9179 }
9180 }
4d84ee25
NC
9181 {
9182 char *ptr = ebuf + sizeof ebuf;
9183 *--ptr = '\0';
9184 *--ptr = c;
9185 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9186#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9187 if (intsize == 'q') {
9188 /* Copy the one or more characters in a long double
9189 * format before the 'base' ([efgEFG]) character to
9190 * the format string. */
9191 static char const prifldbl[] = PERL_PRIfldbl;
9192 char const *p = prifldbl + sizeof(prifldbl) - 3;
9193 while (p >= prifldbl) { *--ptr = *p--; }
9194 }
65202027 9195#endif
4d84ee25
NC
9196 if (has_precis) {
9197 base = precis;
9198 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9199 *--ptr = '.';
9200 }
9201 if (width) {
9202 base = width;
9203 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9204 }
9205 if (fill == '0')
9206 *--ptr = fill;
9207 if (left)
9208 *--ptr = '-';
9209 if (plus)
9210 *--ptr = plus;
9211 if (alt)
9212 *--ptr = '#';
9213 *--ptr = '%';
9214
9215 /* No taint. Otherwise we are in the strange situation
9216 * where printf() taints but print($float) doesn't.
9217 * --jhi */
9e5b023a 9218#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9219 elen = ((intsize == 'q')
9220 ? my_sprintf(PL_efloatbuf, ptr, nv)
9221 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9222#else
4150c189 9223 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9224#endif
4d84ee25 9225 }
4151a5fe 9226 float_converted:
80252599 9227 eptr = PL_efloatbuf;
46fc3d4c 9228 break;
9229
fc36a67e 9230 /* SPECIAL */
9231
9232 case 'n':
26372e71
GA
9233 if (vectorize)
9234 goto unknown;
fc36a67e 9235 i = SvCUR(sv) - origlen;
26372e71 9236 if (args) {
c635e13b 9237 switch (intsize) {
9238 case 'h': *(va_arg(*args, short*)) = i; break;
9239 default: *(va_arg(*args, int*)) = i; break;
9240 case 'l': *(va_arg(*args, long*)) = i; break;
9241 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9242#ifdef HAS_QUAD
9243 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9244#endif
c635e13b 9245 }
fc36a67e 9246 }
9dd79c3f 9247 else
211dfcf1 9248 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9249 continue; /* not "break" */
9250
9251 /* UNKNOWN */
9252
46fc3d4c 9253 default:
fc36a67e 9254 unknown:
041457d9
DM
9255 if (!args
9256 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9257 && ckWARN(WARN_PRINTF))
9258 {
c4420975 9259 SV * const msg = sv_newmortal();
35c1215d
NC
9260 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9261 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9262 if (c) {
0f4b6630 9263 if (isPRINT(c))
1c846c1f 9264 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9265 "\"%%%c\"", c & 0xFF);
9266 else
9267 Perl_sv_catpvf(aTHX_ msg,
57def98f 9268 "\"%%\\%03"UVof"\"",
0f4b6630 9269 (UV)c & 0xFF);
0f4b6630 9270 } else
396482e1 9271 sv_catpvs(msg, "end of string");
9014280d 9272 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9273 }
fb73857a 9274
9275 /* output mangled stuff ... */
9276 if (c == '\0')
9277 --q;
46fc3d4c 9278 eptr = p;
9279 elen = q - p;
fb73857a 9280
9281 /* ... right here, because formatting flags should not apply */
9282 SvGROW(sv, SvCUR(sv) + elen + 1);
9283 p = SvEND(sv);
4459522c 9284 Copy(eptr, p, elen, char);
fb73857a 9285 p += elen;
9286 *p = '\0';
3f7c398e 9287 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9288 svix = osvix;
fb73857a 9289 continue; /* not "break" */
46fc3d4c 9290 }
9291
6c94ec8b
HS
9292 /* calculate width before utf8_upgrade changes it */
9293 have = esignlen + zeros + elen;
ed2b91d2
GA
9294 if (have < zeros)
9295 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9296
d2876be5
JH
9297 if (is_utf8 != has_utf8) {
9298 if (is_utf8) {
9299 if (SvCUR(sv))
9300 sv_utf8_upgrade(sv);
9301 }
9302 else {
53c1dcc0 9303 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9304 sv_utf8_upgrade(nsv);
93524f2b 9305 eptr = SvPVX_const(nsv);
d2876be5
JH
9306 elen = SvCUR(nsv);
9307 }
9308 SvGROW(sv, SvCUR(sv) + elen + 1);
9309 p = SvEND(sv);
9310 *p = '\0';
9311 }
6af65485 9312
46fc3d4c 9313 need = (have > width ? have : width);
9314 gap = need - have;
9315
d2641cbd
PC
9316 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9317 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9318 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9319 p = SvEND(sv);
9320 if (esignlen && fill == '0') {
53c1dcc0 9321 int i;
eb160463 9322 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9323 *p++ = esignbuf[i];
9324 }
9325 if (gap && !left) {
9326 memset(p, fill, gap);
9327 p += gap;
9328 }
9329 if (esignlen && fill != '0') {
53c1dcc0 9330 int i;
eb160463 9331 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9332 *p++ = esignbuf[i];
9333 }
fc36a67e 9334 if (zeros) {
53c1dcc0 9335 int i;
fc36a67e 9336 for (i = zeros; i; i--)
9337 *p++ = '0';
9338 }
46fc3d4c 9339 if (elen) {
4459522c 9340 Copy(eptr, p, elen, char);
46fc3d4c 9341 p += elen;
9342 }
9343 if (gap && left) {
9344 memset(p, ' ', gap);
9345 p += gap;
9346 }
b22c7a20
GS
9347 if (vectorize) {
9348 if (veclen) {
4459522c 9349 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9350 p += dotstrlen;
9351 }
9352 else
9353 vectorize = FALSE; /* done iterating over vecstr */
9354 }
2cf2cfc6
A
9355 if (is_utf8)
9356 has_utf8 = TRUE;
9357 if (has_utf8)
7e2040f0 9358 SvUTF8_on(sv);
46fc3d4c 9359 *p = '\0';
3f7c398e 9360 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9361 if (vectorize) {
9362 esignlen = 0;
9363 goto vector;
9364 }
46fc3d4c 9365 }
9366}
51371543 9367
645c22ef
DM
9368/* =========================================================================
9369
9370=head1 Cloning an interpreter
9371
9372All the macros and functions in this section are for the private use of
9373the main function, perl_clone().
9374
9375The foo_dup() functions make an exact copy of an existing foo thinngy.
9376During the course of a cloning, a hash table is used to map old addresses
9377to new addresses. The table is created and manipulated with the
9378ptr_table_* functions.
9379
9380=cut
9381
9382============================================================================*/
9383
9384
1d7c1841
GS
9385#if defined(USE_ITHREADS)
9386
d4c19fe8 9387/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9388#ifndef GpREFCNT_inc
9389# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9390#endif
9391
9392
a41cc44e
NC
9393/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
9394 that currently av_dup and hv_dup are the same as sv_dup. If this changes,
9395 please unmerge ss_dup. */
d2d73c3e 9396#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9397#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9398#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9399#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9400#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9401#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9402#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9403#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9404#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9405#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9406#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9407#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9408#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9409#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9410
d2d73c3e 9411
d2f185dc
AMS
9412/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9413 regcomp.c. AMS 20010712 */
645c22ef 9414
1d7c1841 9415REGEXP *
53c1dcc0 9416Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9417{
27da23d5 9418 dVAR;
d2f185dc
AMS
9419 REGEXP *ret;
9420 int i, len, npar;
9421 struct reg_substr_datum *s;
9422
9423 if (!r)
9424 return (REGEXP *)NULL;
9425
9426 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9427 return ret;
9428
9429 len = r->offsets[0];
9430 npar = r->nparens+1;
9431
a02a5408 9432 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9433 Copy(r->program, ret->program, len+1, regnode);
9434
a02a5408 9435 Newx(ret->startp, npar, I32);
d2f185dc 9436 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9437 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9438 Copy(r->startp, ret->startp, npar, I32);
9439
a02a5408 9440 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9441 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9442 s->min_offset = r->substrs->data[i].min_offset;
9443 s->max_offset = r->substrs->data[i].max_offset;
9444 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9445 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9446 }
9447
70612e96 9448 ret->regstclass = NULL;
d2f185dc
AMS
9449 if (r->data) {
9450 struct reg_data *d;
e1ec3a88 9451 const int count = r->data->count;
53c1dcc0 9452 int i;
d2f185dc 9453
a02a5408 9454 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9455 char, struct reg_data);
a02a5408 9456 Newx(d->what, count, U8);
d2f185dc
AMS
9457
9458 d->count = count;
9459 for (i = 0; i < count; i++) {
9460 d->what[i] = r->data->what[i];
9461 switch (d->what[i]) {
a3621e74
YO
9462 /* legal options are one of: sfpont
9463 see also regcomp.h and pregfree() */
d2f185dc
AMS
9464 case 's':
9465 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9466 break;
9467 case 'p':
9468 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9469 break;
9470 case 'f':
9471 /* This is cheating. */
a02a5408 9472 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9473 StructCopy(r->data->data[i], d->data[i],
9474 struct regnode_charclass_class);
70612e96 9475 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9476 break;
9477 case 'o':
33773810
AMS
9478 /* Compiled op trees are readonly, and can thus be
9479 shared without duplication. */
b34c0dd4 9480 OP_REFCNT_LOCK;
9b978d73 9481 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9482 OP_REFCNT_UNLOCK;
9b978d73 9483 break;
d2f185dc
AMS
9484 case 'n':
9485 d->data[i] = r->data->data[i];
9486 break;
a3621e74
YO
9487 case 't':
9488 d->data[i] = r->data->data[i];
9489 OP_REFCNT_LOCK;
9490 ((reg_trie_data*)d->data[i])->refcount++;
9491 OP_REFCNT_UNLOCK;
9492 break;
9493 default:
9494 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9495 }
9496 }
9497
9498 ret->data = d;
9499 }
9500 else
9501 ret->data = NULL;
9502
a02a5408 9503 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9504 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9505
e01c5899 9506 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9507 ret->refcnt = r->refcnt;
9508 ret->minlen = r->minlen;
9509 ret->prelen = r->prelen;
9510 ret->nparens = r->nparens;
9511 ret->lastparen = r->lastparen;
9512 ret->lastcloseparen = r->lastcloseparen;
9513 ret->reganch = r->reganch;
9514
70612e96
RG
9515 ret->sublen = r->sublen;
9516
9517 if (RX_MATCH_COPIED(ret))
e01c5899 9518 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9519 else
bd61b366 9520 ret->subbeg = NULL;
f8c7b90f 9521#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9522 ret->saved_copy = NULL;
9a26048b 9523#endif
70612e96 9524
d2f185dc
AMS
9525 ptr_table_store(PL_ptr_table, r, ret);
9526 return ret;
1d7c1841
GS
9527}
9528
d2d73c3e 9529/* duplicate a file handle */
645c22ef 9530
1d7c1841 9531PerlIO *
a8fc9800 9532Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9533{
9534 PerlIO *ret;
53c1dcc0
AL
9535
9536 PERL_UNUSED_ARG(type);
73d840c0 9537
1d7c1841
GS
9538 if (!fp)
9539 return (PerlIO*)NULL;
9540
9541 /* look for it in the table first */
9542 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9543 if (ret)
9544 return ret;
9545
9546 /* create anew and remember what it is */
ecdeb87c 9547 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9548 ptr_table_store(PL_ptr_table, fp, ret);
9549 return ret;
9550}
9551
645c22ef
DM
9552/* duplicate a directory handle */
9553
1d7c1841
GS
9554DIR *
9555Perl_dirp_dup(pTHX_ DIR *dp)
9556{
96a5add6 9557 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9558 if (!dp)
9559 return (DIR*)NULL;
9560 /* XXX TODO */
9561 return dp;
9562}
9563
ff276b08 9564/* duplicate a typeglob */
645c22ef 9565
1d7c1841 9566GP *
a8fc9800 9567Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9568{
9569 GP *ret;
b37c2d43 9570
1d7c1841
GS
9571 if (!gp)
9572 return (GP*)NULL;
9573 /* look for it in the table first */
9574 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9575 if (ret)
9576 return ret;
9577
9578 /* create anew and remember what it is */
a02a5408 9579 Newxz(ret, 1, GP);
1d7c1841
GS
9580 ptr_table_store(PL_ptr_table, gp, ret);
9581
9582 /* clone */
9583 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9584 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9585 ret->gp_io = io_dup_inc(gp->gp_io, param);
9586 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9587 ret->gp_av = av_dup_inc(gp->gp_av, param);
9588 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9589 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9590 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9591 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9592 ret->gp_line = gp->gp_line;
9593 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9594 return ret;
9595}
9596
645c22ef
DM
9597/* duplicate a chain of magic */
9598
1d7c1841 9599MAGIC *
a8fc9800 9600Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9601{
cb359b41
JH
9602 MAGIC *mgprev = (MAGIC*)NULL;
9603 MAGIC *mgret;
1d7c1841
GS
9604 if (!mg)
9605 return (MAGIC*)NULL;
9606 /* look for it in the table first */
9607 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9608 if (mgret)
9609 return mgret;
9610
9611 for (; mg; mg = mg->mg_moremagic) {
9612 MAGIC *nmg;
a02a5408 9613 Newxz(nmg, 1, MAGIC);
cb359b41 9614 if (mgprev)
1d7c1841 9615 mgprev->mg_moremagic = nmg;
cb359b41
JH
9616 else
9617 mgret = nmg;
1d7c1841
GS
9618 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9619 nmg->mg_private = mg->mg_private;
9620 nmg->mg_type = mg->mg_type;
9621 nmg->mg_flags = mg->mg_flags;
14befaf4 9622 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9623 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9624 }
05bd4103 9625 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9626 /* The backref AV has its reference count deliberately bumped by
9627 1. */
9628 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9629 }
8d2f4536
NC
9630 else if (mg->mg_type == PERL_MAGIC_symtab) {
9631 nmg->mg_obj = mg->mg_obj;
9632 }
1d7c1841
GS
9633 else {
9634 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9635 ? sv_dup_inc(mg->mg_obj, param)
9636 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9637 }
9638 nmg->mg_len = mg->mg_len;
9639 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9640 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9641 if (mg->mg_len > 0) {
1d7c1841 9642 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9643 if (mg->mg_type == PERL_MAGIC_overload_table &&
9644 AMT_AMAGIC((AMT*)mg->mg_ptr))
9645 {
c445ea15 9646 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9647 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9648 I32 i;
9649 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9650 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9651 }
9652 }
9653 }
9654 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9655 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9656 }
68795e93
NIS
9657 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9658 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9659 }
1d7c1841
GS
9660 mgprev = nmg;
9661 }
9662 return mgret;
9663}
9664
645c22ef
DM
9665/* create a new pointer-mapping table */
9666
1d7c1841
GS
9667PTR_TBL_t *
9668Perl_ptr_table_new(pTHX)
9669{
9670 PTR_TBL_t *tbl;
96a5add6
AL
9671 PERL_UNUSED_CONTEXT;
9672
a02a5408 9673 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9674 tbl->tbl_max = 511;
9675 tbl->tbl_items = 0;
a02a5408 9676 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9677 return tbl;
9678}
9679
7119fd33
NC
9680#define PTR_TABLE_HASH(ptr) \
9681 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9682
93e68bfb
JC
9683/*
9684 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9685 following define) and at call to new_body_inline made below in
9686 Perl_ptr_table_store()
9687 */
9688
9689#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9690
645c22ef
DM
9691/* map an existing pointer using a table */
9692
7bf61b54 9693STATIC PTR_TBL_ENT_t *
b0e6ae5b 9694S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9695 PTR_TBL_ENT_t *tblent;
4373e329 9696 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9697 assert(tbl);
9698 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9699 for (; tblent; tblent = tblent->next) {
9700 if (tblent->oldval == sv)
7bf61b54 9701 return tblent;
1d7c1841 9702 }
d4c19fe8 9703 return NULL;
7bf61b54
NC
9704}
9705
9706void *
9707Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9708{
b0e6ae5b 9709 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9710 PERL_UNUSED_CONTEXT;
d4c19fe8 9711 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9712}
9713
645c22ef
DM
9714/* add a new entry to a pointer-mapping table */
9715
1d7c1841 9716void
44f8325f 9717Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9718{
0c9fdfe0 9719 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9720 PERL_UNUSED_CONTEXT;
1d7c1841 9721
7bf61b54
NC
9722 if (tblent) {
9723 tblent->newval = newsv;
9724 } else {
9725 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9726
d2a0f284
JC
9727 new_body_inline(tblent, PTE_SVSLOT);
9728
7bf61b54
NC
9729 tblent->oldval = oldsv;
9730 tblent->newval = newsv;
9731 tblent->next = tbl->tbl_ary[entry];
9732 tbl->tbl_ary[entry] = tblent;
9733 tbl->tbl_items++;
9734 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9735 ptr_table_split(tbl);
1d7c1841 9736 }
1d7c1841
GS
9737}
9738
645c22ef
DM
9739/* double the hash bucket size of an existing ptr table */
9740
1d7c1841
GS
9741void
9742Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9743{
9744 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9745 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9746 UV newsize = oldsize * 2;
9747 UV i;
96a5add6 9748 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9749
9750 Renew(ary, newsize, PTR_TBL_ENT_t*);
9751 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9752 tbl->tbl_max = --newsize;
9753 tbl->tbl_ary = ary;
9754 for (i=0; i < oldsize; i++, ary++) {
9755 PTR_TBL_ENT_t **curentp, **entp, *ent;
9756 if (!*ary)
9757 continue;
9758 curentp = ary + oldsize;
9759 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9760 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9761 *entp = ent->next;
9762 ent->next = *curentp;
9763 *curentp = ent;
9764 continue;
9765 }
9766 else
9767 entp = &ent->next;
9768 }
9769 }
9770}
9771
645c22ef
DM
9772/* remove all the entries from a ptr table */
9773
a0739874
DM
9774void
9775Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9776{
d5cefff9 9777 if (tbl && tbl->tbl_items) {
c445ea15 9778 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9779 UV riter = tbl->tbl_max;
a0739874 9780
d5cefff9
NC
9781 do {
9782 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9783
d5cefff9 9784 while (entry) {
00b6aa41 9785 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9786 entry = entry->next;
9787 del_pte(oentry);
9788 }
9789 } while (riter--);
a0739874 9790
d5cefff9
NC
9791 tbl->tbl_items = 0;
9792 }
a0739874
DM
9793}
9794
645c22ef
DM
9795/* clear and free a ptr table */
9796
a0739874
DM
9797void
9798Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9799{
9800 if (!tbl) {
9801 return;
9802 }
9803 ptr_table_clear(tbl);
9804 Safefree(tbl->tbl_ary);
9805 Safefree(tbl);
9806}
9807
5bd07a3d 9808
83841fad 9809void
eb86f8b3 9810Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9811{
9812 if (SvROK(sstr)) {
b162af07
SP
9813 SvRV_set(dstr, SvWEAKREF(sstr)
9814 ? sv_dup(SvRV(sstr), param)
9815 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9816
83841fad 9817 }
3f7c398e 9818 else if (SvPVX_const(sstr)) {
83841fad
NIS
9819 /* Has something there */
9820 if (SvLEN(sstr)) {
68795e93 9821 /* Normal PV - clone whole allocated space */
3f7c398e 9822 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9823 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9824 /* Not that normal - actually sstr is copy on write.
9825 But we are a true, independant SV, so: */
9826 SvREADONLY_off(dstr);
9827 SvFAKE_off(dstr);
9828 }
68795e93 9829 }
83841fad
NIS
9830 else {
9831 /* Special case - not normally malloced for some reason */
f7877b28
NC
9832 if (isGV_with_GP(sstr)) {
9833 /* Don't need to do anything here. */
9834 }
9835 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9836 /* A "shared" PV - clone it as "shared" PV */
9837 SvPV_set(dstr,
9838 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9839 param)));
83841fad
NIS
9840 }
9841 else {
9842 /* Some other special case - random pointer */
f880fe2f 9843 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9844 }
83841fad
NIS
9845 }
9846 }
9847 else {
4608196e 9848 /* Copy the NULL */
f880fe2f 9849 if (SvTYPE(dstr) == SVt_RV)
b162af07 9850 SvRV_set(dstr, NULL);
f880fe2f 9851 else
6136c704 9852 SvPV_set(dstr, NULL);
83841fad
NIS
9853 }
9854}
9855
662fb8b2
NC
9856/* duplicate an SV of any type (including AV, HV etc) */
9857
1d7c1841 9858SV *
eb86f8b3 9859Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9860{
27da23d5 9861 dVAR;
1d7c1841
GS
9862 SV *dstr;
9863
9864 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9865 return NULL;
1d7c1841
GS
9866 /* look for it in the table first */
9867 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9868 if (dstr)
9869 return dstr;
9870
0405e91e
AB
9871 if(param->flags & CLONEf_JOIN_IN) {
9872 /** We are joining here so we don't want do clone
9873 something that is bad **/
eb86f8b3
AL
9874 if (SvTYPE(sstr) == SVt_PVHV) {
9875 const char * const hvname = HvNAME_get(sstr);
9876 if (hvname)
9877 /** don't clone stashes if they already exist **/
9878 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9879 }
9880 }
9881
1d7c1841
GS
9882 /* create anew and remember what it is */
9883 new_SV(dstr);
fd0854ff
DM
9884
9885#ifdef DEBUG_LEAKING_SCALARS
9886 dstr->sv_debug_optype = sstr->sv_debug_optype;
9887 dstr->sv_debug_line = sstr->sv_debug_line;
9888 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9889 dstr->sv_debug_cloned = 1;
fd0854ff 9890 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9891#endif
9892
1d7c1841
GS
9893 ptr_table_store(PL_ptr_table, sstr, dstr);
9894
9895 /* clone */
9896 SvFLAGS(dstr) = SvFLAGS(sstr);
9897 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9898 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9899
9900#ifdef DEBUGGING
3f7c398e 9901 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9902 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9903 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9904#endif
9905
9660f481
DM
9906 /* don't clone objects whose class has asked us not to */
9907 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9908 SvFLAGS(dstr) &= ~SVTYPEMASK;
9909 SvOBJECT_off(dstr);
9910 return dstr;
9911 }
9912
1d7c1841
GS
9913 switch (SvTYPE(sstr)) {
9914 case SVt_NULL:
9915 SvANY(dstr) = NULL;
9916 break;
9917 case SVt_IV:
339049b0 9918 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9919 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9920 break;
9921 case SVt_NV:
9922 SvANY(dstr) = new_XNV();
9d6ce603 9923 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9924 break;
9925 case SVt_RV:
339049b0 9926 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9927 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9928 break;
662fb8b2
NC
9929 default:
9930 {
9931 /* These are all the types that need complex bodies allocating. */
662fb8b2 9932 void *new_body;
2bcc16b3
NC
9933 const svtype sv_type = SvTYPE(sstr);
9934 const struct body_details *const sv_type_details
9935 = bodies_by_type + sv_type;
662fb8b2 9936
93e68bfb 9937 switch (sv_type) {
662fb8b2 9938 default:
bb263b4e 9939 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9940 break;
9941
662fb8b2
NC
9942 case SVt_PVGV:
9943 if (GvUNIQUE((GV*)sstr)) {
bb263b4e 9944 /*EMPTY*/; /* Do sharing here, and fall through */
662fb8b2 9945 }
c22188b4
NC
9946 case SVt_PVIO:
9947 case SVt_PVFM:
9948 case SVt_PVHV:
9949 case SVt_PVAV:
93e68bfb 9950 case SVt_PVBM:
662fb8b2 9951 case SVt_PVCV:
662fb8b2 9952 case SVt_PVLV:
662fb8b2 9953 case SVt_PVMG:
662fb8b2 9954 case SVt_PVNV:
662fb8b2 9955 case SVt_PVIV:
662fb8b2 9956 case SVt_PV:
d2a0f284 9957 assert(sv_type_details->body_size);
c22188b4 9958 if (sv_type_details->arena) {
d2a0f284 9959 new_body_inline(new_body, sv_type);
c22188b4 9960 new_body
b9502f15 9961 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9962 } else {
9963 new_body = new_NOARENA(sv_type_details);
9964 }
1d7c1841 9965 }
662fb8b2
NC
9966 assert(new_body);
9967 SvANY(dstr) = new_body;
9968
2bcc16b3 9969#ifndef PURIFY
b9502f15
NC
9970 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
9971 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 9972 sv_type_details->copy, char);
2bcc16b3
NC
9973#else
9974 Copy(((char*)SvANY(sstr)),
9975 ((char*)SvANY(dstr)),
d2a0f284 9976 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 9977#endif
662fb8b2 9978
f7877b28
NC
9979 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
9980 && !isGV_with_GP(dstr))
662fb8b2
NC
9981 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
9982
9983 /* The Copy above means that all the source (unduplicated) pointers
9984 are now in the destination. We can check the flags and the
9985 pointers in either, but it's possible that there's less cache
9986 missing by always going for the destination.
9987 FIXME - instrument and check that assumption */
f32993d6 9988 if (sv_type >= SVt_PVMG) {
e736a858 9989 HV *ourstash;
2892acdb 9990 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
9991 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
9992 } else if (SvMAGIC(dstr))
662fb8b2
NC
9993 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
9994 if (SvSTASH(dstr))
9995 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 9996 }
662fb8b2 9997
f32993d6
NC
9998 /* The cast silences a GCC warning about unhandled types. */
9999 switch ((int)sv_type) {
662fb8b2
NC
10000 case SVt_PV:
10001 break;
10002 case SVt_PVIV:
10003 break;
10004 case SVt_PVNV:
10005 break;
10006 case SVt_PVMG:
10007 break;
10008 case SVt_PVBM:
10009 break;
10010 case SVt_PVLV:
10011 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10012 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10013 LvTARG(dstr) = dstr;
10014 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10015 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10016 else
10017 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10018 break;
10019 case SVt_PVGV:
acda4c6a
NC
10020 if (GvNAME_HEK(dstr))
10021 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 10022
e15faf7d
NC
10023 /* Don't call sv_add_backref here as it's going to be created
10024 as part of the magic cloning of the symbol table. */
f7877b28
NC
10025 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10026 if(isGV_with_GP(sstr)) {
10027 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10028 at the point of this comment. */
10029 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10030 (void)GpREFCNT_inc(GvGP(dstr));
10031 } else
10032 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10033 break;
10034 case SVt_PVIO:
10035 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10036 if (IoOFP(dstr) == IoIFP(sstr))
10037 IoOFP(dstr) = IoIFP(dstr);
10038 else
10039 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10040 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10041 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10042 /* I have no idea why fake dirp (rsfps)
10043 should be treated differently but otherwise
10044 we end up with leaks -- sky*/
10045 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10046 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10047 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10048 } else {
10049 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10050 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10051 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10052 if (IoDIRP(dstr)) {
10053 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10054 } else {
bb263b4e 10055 /*EMPTY*/;
100ce7e1
NC
10056 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10057 }
662fb8b2
NC
10058 }
10059 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10060 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10061 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10062 break;
10063 case SVt_PVAV:
10064 if (AvARRAY((AV*)sstr)) {
10065 SV **dst_ary, **src_ary;
10066 SSize_t items = AvFILLp((AV*)sstr) + 1;
10067
10068 src_ary = AvARRAY((AV*)sstr);
a02a5408 10069 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10070 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10071 SvPV_set(dstr, (char*)dst_ary);
10072 AvALLOC((AV*)dstr) = dst_ary;
10073 if (AvREAL((AV*)sstr)) {
10074 while (items-- > 0)
10075 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10076 }
10077 else {
10078 while (items-- > 0)
10079 *dst_ary++ = sv_dup(*src_ary++, param);
10080 }
10081 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10082 while (items-- > 0) {
10083 *dst_ary++ = &PL_sv_undef;
10084 }
bfcb3514 10085 }
662fb8b2 10086 else {
bd61b366 10087 SvPV_set(dstr, NULL);
662fb8b2 10088 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10089 }
662fb8b2
NC
10090 break;
10091 case SVt_PVHV:
10092 {
cbbf8932 10093 HEK *hvname = NULL;
662fb8b2
NC
10094
10095 if (HvARRAY((HV*)sstr)) {
10096 STRLEN i = 0;
10097 const bool sharekeys = !!HvSHAREKEYS(sstr);
10098 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10099 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10100 char *darray;
a02a5408 10101 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10102 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10103 char);
10104 HvARRAY(dstr) = (HE**)darray;
10105 while (i <= sxhv->xhv_max) {
5c4138a0 10106 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10107 HvARRAY(dstr)[i] = source
10108 ? he_dup(source, sharekeys, param) : 0;
10109 ++i;
10110 }
10111 if (SvOOK(sstr)) {
00b6aa41
AL
10112 struct xpvhv_aux * const saux = HvAUX(sstr);
10113 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10114 /* This flag isn't copied. */
10115 /* SvOOK_on(hv) attacks the IV flags. */
10116 SvFLAGS(dstr) |= SVf_OOK;
10117
10118 hvname = saux->xhv_name;
dd690478
NC
10119 daux->xhv_name
10120 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10121
10122 daux->xhv_riter = saux->xhv_riter;
10123 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10124 ? he_dup(saux->xhv_eiter,
10125 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10126 daux->xhv_backreferences = saux->xhv_backreferences
10127 ? (AV*) SvREFCNT_inc(
10128 sv_dup((SV*)saux->
10129 xhv_backreferences,
10130 param))
10131 : 0;
662fb8b2
NC
10132 }
10133 }
10134 else {
bd61b366 10135 SvPV_set(dstr, NULL);
662fb8b2
NC
10136 }
10137 /* Record stashes for possible cloning in Perl_clone(). */
10138 if(hvname)
10139 av_push(param->stashes, dstr);
10140 }
10141 break;
662fb8b2 10142 case SVt_PVCV:
bb172083
NC
10143 if (!(param->flags & CLONEf_COPY_STACKS)) {
10144 CvDEPTH(dstr) = 0;
10145 }
10146 case SVt_PVFM:
662fb8b2
NC
10147 /* NOTE: not refcounted */
10148 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10149 OP_REFCNT_LOCK;
d04ba589
NC
10150 if (!CvISXSUB(dstr))
10151 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10152 OP_REFCNT_UNLOCK;
cfae286e 10153 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10154 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10155 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10156 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10157 }
10158 /* don't dup if copying back - CvGV isn't refcounted, so the
10159 * duped GV may never be freed. A bit of a hack! DAPM */
10160 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10161 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10162 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10163 CvOUTSIDE(dstr) =
10164 CvWEAKOUTSIDE(sstr)
10165 ? cv_dup( CvOUTSIDE(dstr), param)
10166 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10167 if (!CvISXSUB(dstr))
662fb8b2
NC
10168 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10169 break;
bfcb3514 10170 }
1d7c1841 10171 }
1d7c1841
GS
10172 }
10173
10174 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10175 ++PL_sv_objcount;
10176
10177 return dstr;
d2d73c3e 10178 }
1d7c1841 10179
645c22ef
DM
10180/* duplicate a context */
10181
1d7c1841 10182PERL_CONTEXT *
a8fc9800 10183Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10184{
10185 PERL_CONTEXT *ncxs;
10186
10187 if (!cxs)
10188 return (PERL_CONTEXT*)NULL;
10189
10190 /* look for it in the table first */
10191 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10192 if (ncxs)
10193 return ncxs;
10194
10195 /* create anew and remember what it is */
a02a5408 10196 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10197 ptr_table_store(PL_ptr_table, cxs, ncxs);
10198
10199 while (ix >= 0) {
c445ea15
AL
10200 PERL_CONTEXT * const cx = &cxs[ix];
10201 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10202 ncx->cx_type = cx->cx_type;
10203 if (CxTYPE(cx) == CXt_SUBST) {
10204 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10205 }
10206 else {
10207 ncx->blk_oldsp = cx->blk_oldsp;
10208 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10209 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10210 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10211 ncx->blk_oldpm = cx->blk_oldpm;
10212 ncx->blk_gimme = cx->blk_gimme;
10213 switch (CxTYPE(cx)) {
10214 case CXt_SUB:
10215 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10216 ? cv_dup_inc(cx->blk_sub.cv, param)
10217 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10218 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10219 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10220 : NULL);
d2d73c3e 10221 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10222 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10223 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10224 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10225 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10226 break;
10227 case CXt_EVAL:
10228 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10229 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10230 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10231 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10232 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10233 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10234 break;
10235 case CXt_LOOP:
10236 ncx->blk_loop.label = cx->blk_loop.label;
10237 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10238 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10239 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10240 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10241 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10242 ? cx->blk_loop.iterdata
d2d73c3e 10243 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10244 ncx->blk_loop.oldcomppad
10245 = (PAD*)ptr_table_fetch(PL_ptr_table,
10246 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10247 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10248 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10249 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10250 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10251 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10252 break;
10253 case CXt_FORMAT:
d2d73c3e
AB
10254 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10255 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10256 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10257 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10258 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10259 break;
10260 case CXt_BLOCK:
10261 case CXt_NULL:
10262 break;
10263 }
10264 }
10265 --ix;
10266 }
10267 return ncxs;
10268}
10269
645c22ef
DM
10270/* duplicate a stack info structure */
10271
1d7c1841 10272PERL_SI *
a8fc9800 10273Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10274{
10275 PERL_SI *nsi;
10276
10277 if (!si)
10278 return (PERL_SI*)NULL;
10279
10280 /* look for it in the table first */
10281 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10282 if (nsi)
10283 return nsi;
10284
10285 /* create anew and remember what it is */
a02a5408 10286 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10287 ptr_table_store(PL_ptr_table, si, nsi);
10288
d2d73c3e 10289 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10290 nsi->si_cxix = si->si_cxix;
10291 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10292 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10293 nsi->si_type = si->si_type;
d2d73c3e
AB
10294 nsi->si_prev = si_dup(si->si_prev, param);
10295 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10296 nsi->si_markoff = si->si_markoff;
10297
10298 return nsi;
10299}
10300
10301#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10302#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10303#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10304#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10305#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10306#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10307#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10308#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10309#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10310#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10311#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10312#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10313#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10314#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10315
10316/* XXXXX todo */
10317#define pv_dup_inc(p) SAVEPV(p)
10318#define pv_dup(p) SAVEPV(p)
10319#define svp_dup_inc(p,pp) any_dup(p,pp)
10320
645c22ef
DM
10321/* map any object to the new equivent - either something in the
10322 * ptr table, or something in the interpreter structure
10323 */
10324
1d7c1841 10325void *
53c1dcc0 10326Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10327{
10328 void *ret;
10329
10330 if (!v)
10331 return (void*)NULL;
10332
10333 /* look for it in the table first */
10334 ret = ptr_table_fetch(PL_ptr_table, v);
10335 if (ret)
10336 return ret;
10337
10338 /* see if it is part of the interpreter structure */
10339 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10340 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10341 else {
1d7c1841 10342 ret = v;
05ec9bb3 10343 }
1d7c1841
GS
10344
10345 return ret;
10346}
10347
645c22ef
DM
10348/* duplicate the save stack */
10349
1d7c1841 10350ANY *
a8fc9800 10351Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10352{
53c1dcc0
AL
10353 ANY * const ss = proto_perl->Tsavestack;
10354 const I32 max = proto_perl->Tsavestack_max;
10355 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10356 ANY *nss;
10357 SV *sv;
10358 GV *gv;
10359 AV *av;
10360 HV *hv;
10361 void* ptr;
10362 int intval;
10363 long longval;
10364 GP *gp;
10365 IV iv;
c4e33207 10366 char *c = NULL;
1d7c1841 10367 void (*dptr) (void*);
acfe0abc 10368 void (*dxptr) (pTHX_ void*);
1d7c1841 10369
a02a5408 10370 Newxz(nss, max, ANY);
1d7c1841
GS
10371
10372 while (ix > 0) {
b464bac0 10373 I32 i = POPINT(ss,ix);
1d7c1841
GS
10374 TOPINT(nss,ix) = i;
10375 switch (i) {
10376 case SAVEt_ITEM: /* normal string */
a41cc44e 10377 case SAVEt_SV: /* scalar reference */
1d7c1841 10378 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10379 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10380 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10381 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10382 break;
05ec9bb3
NIS
10383 case SAVEt_SHARED_PVREF: /* char* in shared space */
10384 c = (char*)POPPTR(ss,ix);
10385 TOPPTR(nss,ix) = savesharedpv(c);
10386 ptr = POPPTR(ss,ix);
10387 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10388 break;
1d7c1841
GS
10389 case SAVEt_GENERIC_SVREF: /* generic sv */
10390 case SAVEt_SVREF: /* scalar reference */
10391 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10392 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10393 ptr = POPPTR(ss,ix);
10394 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10395 break;
a41cc44e 10396 case SAVEt_HV: /* hash reference */
1d7c1841 10397 case SAVEt_AV: /* array reference */
337d28f5
NC
10398 sv = POPPTR(ss,ix);
10399 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10400 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10401 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10402 break;
10403 case SAVEt_INT: /* int reference */
10404 ptr = POPPTR(ss,ix);
10405 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10406 intval = (int)POPINT(ss,ix);
10407 TOPINT(nss,ix) = intval;
10408 break;
10409 case SAVEt_LONG: /* long reference */
10410 ptr = POPPTR(ss,ix);
10411 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10412 longval = (long)POPLONG(ss,ix);
10413 TOPLONG(nss,ix) = longval;
10414 break;
10415 case SAVEt_I32: /* I32 reference */
10416 case SAVEt_I16: /* I16 reference */
10417 case SAVEt_I8: /* I8 reference */
88effcc9 10418 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10419 ptr = POPPTR(ss,ix);
10420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10421 i = POPINT(ss,ix);
10422 TOPINT(nss,ix) = i;
10423 break;
10424 case SAVEt_IV: /* IV reference */
10425 ptr = POPPTR(ss,ix);
10426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10427 iv = POPIV(ss,ix);
10428 TOPIV(nss,ix) = iv;
10429 break;
a41cc44e
NC
10430 case SAVEt_HPTR: /* HV* reference */
10431 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10432 case SAVEt_SPTR: /* SV* reference */
10433 ptr = POPPTR(ss,ix);
10434 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10435 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10436 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10437 break;
10438 case SAVEt_VPTR: /* random* reference */
10439 ptr = POPPTR(ss,ix);
10440 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10441 ptr = POPPTR(ss,ix);
10442 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10443 break;
b03d03b0 10444 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10445 case SAVEt_PPTR: /* char* reference */
10446 ptr = POPPTR(ss,ix);
10447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10448 c = (char*)POPPTR(ss,ix);
10449 TOPPTR(nss,ix) = pv_dup(c);
10450 break;
1d7c1841
GS
10451 case SAVEt_NSTAB:
10452 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10453 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10454 break;
10455 case SAVEt_GP: /* scalar reference */
10456 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10457 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10458 (void)GpREFCNT_inc(gp);
10459 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10460 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10461 c = (char*)POPPTR(ss,ix);
10462 TOPPTR(nss,ix) = pv_dup(c);
10463 iv = POPIV(ss,ix);
10464 TOPIV(nss,ix) = iv;
10465 iv = POPIV(ss,ix);
10466 TOPIV(nss,ix) = iv;
10467 break;
10468 case SAVEt_FREESV:
26d9b02f 10469 case SAVEt_MORTALIZESV:
1d7c1841 10470 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10471 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10472 break;
10473 case SAVEt_FREEOP:
10474 ptr = POPPTR(ss,ix);
10475 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10476 /* these are assumed to be refcounted properly */
53c1dcc0 10477 OP *o;
1d7c1841
GS
10478 switch (((OP*)ptr)->op_type) {
10479 case OP_LEAVESUB:
10480 case OP_LEAVESUBLV:
10481 case OP_LEAVEEVAL:
10482 case OP_LEAVE:
10483 case OP_SCOPE:
10484 case OP_LEAVEWRITE:
e977893f
GS
10485 TOPPTR(nss,ix) = ptr;
10486 o = (OP*)ptr;
10487 OpREFCNT_inc(o);
1d7c1841
GS
10488 break;
10489 default:
5f66b61c 10490 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10491 break;
10492 }
10493 }
10494 else
5f66b61c 10495 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10496 break;
10497 case SAVEt_FREEPV:
10498 c = (char*)POPPTR(ss,ix);
10499 TOPPTR(nss,ix) = pv_dup_inc(c);
10500 break;
10501 case SAVEt_CLEARSV:
10502 longval = POPLONG(ss,ix);
10503 TOPLONG(nss,ix) = longval;
10504 break;
10505 case SAVEt_DELETE:
10506 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10507 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10508 c = (char*)POPPTR(ss,ix);
10509 TOPPTR(nss,ix) = pv_dup_inc(c);
10510 i = POPINT(ss,ix);
10511 TOPINT(nss,ix) = i;
10512 break;
10513 case SAVEt_DESTRUCTOR:
10514 ptr = POPPTR(ss,ix);
10515 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10516 dptr = POPDPTR(ss,ix);
8141890a
JH
10517 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10518 any_dup(FPTR2DPTR(void *, dptr),
10519 proto_perl));
1d7c1841
GS
10520 break;
10521 case SAVEt_DESTRUCTOR_X:
10522 ptr = POPPTR(ss,ix);
10523 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10524 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10525 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10526 any_dup(FPTR2DPTR(void *, dxptr),
10527 proto_perl));
1d7c1841
GS
10528 break;
10529 case SAVEt_REGCONTEXT:
10530 case SAVEt_ALLOC:
10531 i = POPINT(ss,ix);
10532 TOPINT(nss,ix) = i;
10533 ix -= i;
10534 break;
10535 case SAVEt_STACK_POS: /* Position on Perl stack */
10536 i = POPINT(ss,ix);
10537 TOPINT(nss,ix) = i;
10538 break;
10539 case SAVEt_AELEM: /* array element */
10540 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10541 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10542 i = POPINT(ss,ix);
10543 TOPINT(nss,ix) = i;
10544 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10545 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10546 break;
10547 case SAVEt_HELEM: /* hash element */
10548 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10549 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10550 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10551 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10552 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10553 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10554 break;
10555 case SAVEt_OP:
10556 ptr = POPPTR(ss,ix);
10557 TOPPTR(nss,ix) = ptr;
10558 break;
10559 case SAVEt_HINTS:
10560 i = POPINT(ss,ix);
10561 TOPINT(nss,ix) = i;
b3ca2e83
NC
10562 ptr = POPPTR(ss,ix);
10563 TOPPTR(nss,ix) = Perl_refcounted_he_dup(aTHX_ ptr, param);
a8f8b6a7
NC
10564 if (i & HINT_LOCALIZE_HH) {
10565 hv = (HV*)POPPTR(ss,ix);
10566 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10567 }
1d7c1841 10568 break;
c4410b1b
GS
10569 case SAVEt_COMPPAD:
10570 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10571 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10572 break;
c3564e5c
GS
10573 case SAVEt_PADSV:
10574 longval = (long)POPLONG(ss,ix);
10575 TOPLONG(nss,ix) = longval;
10576 ptr = POPPTR(ss,ix);
10577 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10578 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10579 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10580 break;
a1bb4754 10581 case SAVEt_BOOL:
38d8b13e 10582 ptr = POPPTR(ss,ix);
b9609c01 10583 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10584 longval = (long)POPBOOL(ss,ix);
b9609c01 10585 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10586 break;
8bd2680e
MHM
10587 case SAVEt_SET_SVFLAGS:
10588 i = POPINT(ss,ix);
10589 TOPINT(nss,ix) = i;
10590 i = POPINT(ss,ix);
10591 TOPINT(nss,ix) = i;
10592 sv = (SV*)POPPTR(ss,ix);
10593 TOPPTR(nss,ix) = sv_dup(sv, param);
10594 break;
5bfb7d0e
NC
10595 case SAVEt_RE_STATE:
10596 {
10597 const struct re_save_state *const old_state
10598 = (struct re_save_state *)
10599 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10600 struct re_save_state *const new_state
10601 = (struct re_save_state *)
10602 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10603
10604 Copy(old_state, new_state, 1, struct re_save_state);
10605 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10606
10607 new_state->re_state_bostr
10608 = pv_dup(old_state->re_state_bostr);
10609 new_state->re_state_reginput
10610 = pv_dup(old_state->re_state_reginput);
10611 new_state->re_state_regbol
10612 = pv_dup(old_state->re_state_regbol);
10613 new_state->re_state_regeol
10614 = pv_dup(old_state->re_state_regeol);
10615 new_state->re_state_regstartp
10616 = any_dup(old_state->re_state_regstartp, proto_perl);
10617 new_state->re_state_regendp
10618 = any_dup(old_state->re_state_regendp, proto_perl);
10619 new_state->re_state_reglastparen
10620 = any_dup(old_state->re_state_reglastparen, proto_perl);
10621 new_state->re_state_reglastcloseparen
10622 = any_dup(old_state->re_state_reglastcloseparen,
10623 proto_perl);
10624 new_state->re_state_regtill
10625 = pv_dup(old_state->re_state_regtill);
10626 /* XXX This just has to be broken. The old save_re_context
10627 code did SAVEGENERICPV(PL_reg_start_tmp);
10628 PL_reg_start_tmp is char **.
10629 Look above to what the dup code does for
10630 SAVEt_GENERIC_PVREF
10631 It can never have worked.
10632 So this is merely a faithful copy of the exiting bug: */
10633 new_state->re_state_reg_start_tmp
10634 = (char **) pv_dup((char *)
10635 old_state->re_state_reg_start_tmp);
10636 /* I assume that it only ever "worked" because no-one called
10637 (pseudo)fork while the regexp engine had re-entered itself.
10638 */
10639 new_state->re_state_reg_call_cc
10640 = any_dup(old_state->re_state_reg_call_cc, proto_perl);
10641 new_state->re_state_reg_re
10642 = any_dup(old_state->re_state_reg_re, proto_perl);
10643 new_state->re_state_reg_ganch
10644 = pv_dup(old_state->re_state_reg_ganch);
10645 new_state->re_state_reg_sv
10646 = sv_dup(old_state->re_state_reg_sv, param);
10647#ifdef PERL_OLD_COPY_ON_WRITE
10648 new_state->re_state_nrs
10649 = sv_dup(old_state->re_state_nrs, param);
10650#endif
10651 new_state->re_state_reg_magic
10652 = any_dup(old_state->re_state_reg_magic, proto_perl);
10653 new_state->re_state_reg_oldcurpm
10654 = any_dup(old_state->re_state_reg_oldcurpm, proto_perl);
10655 new_state->re_state_reg_curpm
10656 = any_dup(old_state->re_state_reg_curpm, proto_perl);
10657 new_state->re_state_reg_oldsaved
10658 = pv_dup(old_state->re_state_reg_oldsaved);
10659 new_state->re_state_reg_poscache
10660 = pv_dup(old_state->re_state_reg_poscache);
10661#ifdef DEBUGGING
10662 new_state->re_state_reg_starttry
10663 = pv_dup(old_state->re_state_reg_starttry);
10664#endif
10665 break;
10666 }
1d7c1841 10667 default:
ca05af4a 10668 Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i);
1d7c1841
GS
10669 }
10670 }
10671
bd81e77b
NC
10672 return nss;
10673}
10674
10675
10676/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10677 * flag to the result. This is done for each stash before cloning starts,
10678 * so we know which stashes want their objects cloned */
10679
10680static void
10681do_mark_cloneable_stash(pTHX_ SV *sv)
10682{
10683 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10684 if (hvname) {
10685 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10686 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10687 if (cloner && GvCV(cloner)) {
10688 dSP;
10689 UV status;
10690
10691 ENTER;
10692 SAVETMPS;
10693 PUSHMARK(SP);
10694 XPUSHs(sv_2mortal(newSVhek(hvname)));
10695 PUTBACK;
10696 call_sv((SV*)GvCV(cloner), G_SCALAR);
10697 SPAGAIN;
10698 status = POPu;
10699 PUTBACK;
10700 FREETMPS;
10701 LEAVE;
10702 if (status)
10703 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10704 }
10705 }
10706}
10707
10708
10709
10710/*
10711=for apidoc perl_clone
10712
10713Create and return a new interpreter by cloning the current one.
10714
10715perl_clone takes these flags as parameters:
10716
10717CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10718without it we only clone the data and zero the stacks,
10719with it we copy the stacks and the new perl interpreter is
10720ready to run at the exact same point as the previous one.
10721The pseudo-fork code uses COPY_STACKS while the
10722threads->new doesn't.
10723
10724CLONEf_KEEP_PTR_TABLE
10725perl_clone keeps a ptr_table with the pointer of the old
10726variable as a key and the new variable as a value,
10727this allows it to check if something has been cloned and not
10728clone it again but rather just use the value and increase the
10729refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10730the ptr_table using the function
10731C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10732reason to keep it around is if you want to dup some of your own
10733variable who are outside the graph perl scans, example of this
10734code is in threads.xs create
10735
10736CLONEf_CLONE_HOST
10737This is a win32 thing, it is ignored on unix, it tells perls
10738win32host code (which is c++) to clone itself, this is needed on
10739win32 if you want to run two threads at the same time,
10740if you just want to do some stuff in a separate perl interpreter
10741and then throw it away and return to the original one,
10742you don't need to do anything.
10743
10744=cut
10745*/
10746
10747/* XXX the above needs expanding by someone who actually understands it ! */
10748EXTERN_C PerlInterpreter *
10749perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10750
10751PerlInterpreter *
10752perl_clone(PerlInterpreter *proto_perl, UV flags)
10753{
10754 dVAR;
10755#ifdef PERL_IMPLICIT_SYS
10756
10757 /* perlhost.h so we need to call into it
10758 to clone the host, CPerlHost should have a c interface, sky */
10759
10760 if (flags & CLONEf_CLONE_HOST) {
10761 return perl_clone_host(proto_perl,flags);
10762 }
10763 return perl_clone_using(proto_perl, flags,
10764 proto_perl->IMem,
10765 proto_perl->IMemShared,
10766 proto_perl->IMemParse,
10767 proto_perl->IEnv,
10768 proto_perl->IStdIO,
10769 proto_perl->ILIO,
10770 proto_perl->IDir,
10771 proto_perl->ISock,
10772 proto_perl->IProc);
10773}
10774
10775PerlInterpreter *
10776perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10777 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10778 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10779 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10780 struct IPerlDir* ipD, struct IPerlSock* ipS,
10781 struct IPerlProc* ipP)
10782{
10783 /* XXX many of the string copies here can be optimized if they're
10784 * constants; they need to be allocated as common memory and just
10785 * their pointers copied. */
10786
10787 IV i;
10788 CLONE_PARAMS clone_params;
5f66b61c 10789 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10790
5f66b61c 10791 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10792 /* for each stash, determine whether its objects should be cloned */
10793 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10794 PERL_SET_THX(my_perl);
10795
10796# ifdef DEBUGGING
7e337ee0 10797 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10798 PL_op = NULL;
10799 PL_curcop = NULL;
bd81e77b
NC
10800 PL_markstack = 0;
10801 PL_scopestack = 0;
10802 PL_savestack = 0;
10803 PL_savestack_ix = 0;
10804 PL_savestack_max = -1;
10805 PL_sig_pending = 0;
10806 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10807# else /* !DEBUGGING */
10808 Zero(my_perl, 1, PerlInterpreter);
10809# endif /* DEBUGGING */
10810
10811 /* host pointers */
10812 PL_Mem = ipM;
10813 PL_MemShared = ipMS;
10814 PL_MemParse = ipMP;
10815 PL_Env = ipE;
10816 PL_StdIO = ipStd;
10817 PL_LIO = ipLIO;
10818 PL_Dir = ipD;
10819 PL_Sock = ipS;
10820 PL_Proc = ipP;
10821#else /* !PERL_IMPLICIT_SYS */
10822 IV i;
10823 CLONE_PARAMS clone_params;
10824 CLONE_PARAMS* param = &clone_params;
5f66b61c 10825 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10826 /* for each stash, determine whether its objects should be cloned */
10827 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10828 PERL_SET_THX(my_perl);
10829
10830# ifdef DEBUGGING
7e337ee0 10831 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10832 PL_op = NULL;
10833 PL_curcop = NULL;
bd81e77b
NC
10834 PL_markstack = 0;
10835 PL_scopestack = 0;
10836 PL_savestack = 0;
10837 PL_savestack_ix = 0;
10838 PL_savestack_max = -1;
10839 PL_sig_pending = 0;
10840 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10841# else /* !DEBUGGING */
10842 Zero(my_perl, 1, PerlInterpreter);
10843# endif /* DEBUGGING */
10844#endif /* PERL_IMPLICIT_SYS */
10845 param->flags = flags;
10846 param->proto_perl = proto_perl;
10847
7cb608b5
NC
10848 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10849
fdda85ca 10850 PL_body_arenas = NULL;
bd81e77b
NC
10851 Zero(&PL_body_roots, 1, PL_body_roots);
10852
10853 PL_nice_chunk = NULL;
10854 PL_nice_chunk_size = 0;
10855 PL_sv_count = 0;
10856 PL_sv_objcount = 0;
a0714e2c
SS
10857 PL_sv_root = NULL;
10858 PL_sv_arenaroot = NULL;
bd81e77b
NC
10859
10860 PL_debug = proto_perl->Idebug;
10861
10862 PL_hash_seed = proto_perl->Ihash_seed;
10863 PL_rehash_seed = proto_perl->Irehash_seed;
10864
10865#ifdef USE_REENTRANT_API
10866 /* XXX: things like -Dm will segfault here in perlio, but doing
10867 * PERL_SET_CONTEXT(proto_perl);
10868 * breaks too many other things
10869 */
10870 Perl_reentrant_init(aTHX);
10871#endif
10872
10873 /* create SV map for pointer relocation */
10874 PL_ptr_table = ptr_table_new();
10875
10876 /* initialize these special pointers as early as possible */
10877 SvANY(&PL_sv_undef) = NULL;
10878 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10879 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10880 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10881
10882 SvANY(&PL_sv_no) = new_XPVNV();
10883 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10884 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10885 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10886 SvPV_set(&PL_sv_no, SAVEPVN(PL_No, 0));
10887 SvCUR_set(&PL_sv_no, 0);
10888 SvLEN_set(&PL_sv_no, 1);
10889 SvIV_set(&PL_sv_no, 0);
10890 SvNV_set(&PL_sv_no, 0);
10891 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10892
10893 SvANY(&PL_sv_yes) = new_XPVNV();
10894 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10895 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10896 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10897 SvPV_set(&PL_sv_yes, SAVEPVN(PL_Yes, 1));
10898 SvCUR_set(&PL_sv_yes, 1);
10899 SvLEN_set(&PL_sv_yes, 2);
10900 SvIV_set(&PL_sv_yes, 1);
10901 SvNV_set(&PL_sv_yes, 1);
10902 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10903
10904 /* create (a non-shared!) shared string table */
10905 PL_strtab = newHV();
10906 HvSHAREKEYS_off(PL_strtab);
10907 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10908 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10909
10910 PL_compiling = proto_perl->Icompiling;
10911
10912 /* These two PVs will be free'd special way so must set them same way op.c does */
10913 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10914 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10915
10916 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10917 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10918
10919 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
10920 if (!specialWARN(PL_compiling.cop_warnings))
10921 PL_compiling.cop_warnings = sv_dup_inc(PL_compiling.cop_warnings, param);
10922 if (!specialCopIO(PL_compiling.cop_io))
10923 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
b3ca2e83 10924 PL_compiling.cop_hints
9b2c10f1 10925 = Perl_refcounted_he_dup(aTHX_ PL_compiling.cop_hints, param);
bd81e77b
NC
10926 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10927
10928 /* pseudo environmental stuff */
10929 PL_origargc = proto_perl->Iorigargc;
10930 PL_origargv = proto_perl->Iorigargv;
10931
10932 param->stashes = newAV(); /* Setup array of objects to call clone on */
10933
10934 /* Set tainting stuff before PerlIO_debug can possibly get called */
10935 PL_tainting = proto_perl->Itainting;
10936 PL_taint_warn = proto_perl->Itaint_warn;
10937
10938#ifdef PERLIO_LAYERS
10939 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10940 PerlIO_clone(aTHX_ proto_perl, param);
10941#endif
10942
10943 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10944 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10945 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10946 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10947 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10948 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10949
10950 /* switches */
10951 PL_minus_c = proto_perl->Iminus_c;
10952 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10953 PL_localpatches = proto_perl->Ilocalpatches;
10954 PL_splitstr = proto_perl->Isplitstr;
10955 PL_preprocess = proto_perl->Ipreprocess;
10956 PL_minus_n = proto_perl->Iminus_n;
10957 PL_minus_p = proto_perl->Iminus_p;
10958 PL_minus_l = proto_perl->Iminus_l;
10959 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10960 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10961 PL_minus_F = proto_perl->Iminus_F;
10962 PL_doswitches = proto_perl->Idoswitches;
10963 PL_dowarn = proto_perl->Idowarn;
10964 PL_doextract = proto_perl->Idoextract;
10965 PL_sawampersand = proto_perl->Isawampersand;
10966 PL_unsafe = proto_perl->Iunsafe;
10967 PL_inplace = SAVEPV(proto_perl->Iinplace);
10968 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10969 PL_perldb = proto_perl->Iperldb;
10970 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10971 PL_exit_flags = proto_perl->Iexit_flags;
10972
10973 /* magical thingies */
10974 /* XXX time(&PL_basetime) when asked for? */
10975 PL_basetime = proto_perl->Ibasetime;
10976 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
10977
10978 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
10979 PL_statusvalue = proto_perl->Istatusvalue;
10980#ifdef VMS
10981 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
10982#else
10983 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
10984#endif
10985 PL_encoding = sv_dup(proto_perl->Iencoding, param);
10986
10987 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
10988 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
10989 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
10990
10991 /* Clone the regex array */
10992 PL_regex_padav = newAV();
10993 {
10994 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 10995 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 10996 IV i;
7f466ec7 10997 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 10998 for(i = 1; i <= len; i++) {
7a5b473e
AL
10999 const SV * const regex = regexen[i];
11000 SV * const sv =
11001 SvREPADTMP(regex)
11002 ? sv_dup_inc(regex, param)
11003 : SvREFCNT_inc(
11004 newSViv(PTR2IV(re_dup(
11005 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11006 ;
11007 av_push(PL_regex_padav, sv);
bd81e77b
NC
11008 }
11009 }
11010 PL_regex_pad = AvARRAY(PL_regex_padav);
11011
11012 /* shortcuts to various I/O objects */
11013 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11014 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11015 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11016 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11017 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11018 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11019
bd81e77b
NC
11020 /* shortcuts to regexp stuff */
11021 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11022
bd81e77b
NC
11023 /* shortcuts to misc objects */
11024 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11025
bd81e77b
NC
11026 /* shortcuts to debugging objects */
11027 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11028 PL_DBline = gv_dup(proto_perl->IDBline, param);
11029 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11030 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11031 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11032 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11033 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11034 PL_lineary = av_dup(proto_perl->Ilineary, param);
11035 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11036
bd81e77b
NC
11037 /* symbol tables */
11038 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11039 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11040 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11041 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11042 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11043
11044 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11045 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11046 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11047 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11048 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11049 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11050
11051 PL_sub_generation = proto_perl->Isub_generation;
11052
11053 /* funky return mechanisms */
11054 PL_forkprocess = proto_perl->Iforkprocess;
11055
11056 /* subprocess state */
11057 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11058
11059 /* internal state */
11060 PL_maxo = proto_perl->Imaxo;
11061 if (proto_perl->Iop_mask)
11062 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11063 else
bd61b366 11064 PL_op_mask = NULL;
bd81e77b
NC
11065 /* PL_asserting = proto_perl->Iasserting; */
11066
11067 /* current interpreter roots */
11068 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
11069 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
11070 PL_main_start = proto_perl->Imain_start;
11071 PL_eval_root = proto_perl->Ieval_root;
11072 PL_eval_start = proto_perl->Ieval_start;
11073
11074 /* runtime control stuff */
11075 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11076 PL_copline = proto_perl->Icopline;
11077
11078 PL_filemode = proto_perl->Ifilemode;
11079 PL_lastfd = proto_perl->Ilastfd;
11080 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11081 PL_Argv = NULL;
bd61b366 11082 PL_Cmd = NULL;
bd81e77b
NC
11083 PL_gensym = proto_perl->Igensym;
11084 PL_preambled = proto_perl->Ipreambled;
11085 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11086 PL_laststatval = proto_perl->Ilaststatval;
11087 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11088 PL_mess_sv = NULL;
bd81e77b
NC
11089
11090 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11091
11092 /* interpreter atexit processing */
11093 PL_exitlistlen = proto_perl->Iexitlistlen;
11094 if (PL_exitlistlen) {
11095 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11096 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11097 }
bd81e77b
NC
11098 else
11099 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11100
11101 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11102 if (PL_my_cxt_size) {
f16dd614
DM
11103 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11104 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11105 }
11106 else
11107 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11108 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11109 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11110 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11111
11112 PL_profiledata = NULL;
11113 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11114 /* PL_rsfp_filters entries have fake IoDIRP() */
11115 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11116
bd81e77b 11117 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11118
bd81e77b 11119 PAD_CLONE_VARS(proto_perl, param);
9660f481 11120
bd81e77b
NC
11121#ifdef HAVE_INTERP_INTERN
11122 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11123#endif
645c22ef 11124
bd81e77b
NC
11125 /* more statics moved here */
11126 PL_generation = proto_perl->Igeneration;
11127 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11128
bd81e77b
NC
11129 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11130 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11131
bd81e77b
NC
11132 PL_uid = proto_perl->Iuid;
11133 PL_euid = proto_perl->Ieuid;
11134 PL_gid = proto_perl->Igid;
11135 PL_egid = proto_perl->Iegid;
11136 PL_nomemok = proto_perl->Inomemok;
11137 PL_an = proto_perl->Ian;
11138 PL_evalseq = proto_perl->Ievalseq;
11139 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11140 PL_origalen = proto_perl->Iorigalen;
11141#ifdef PERL_USES_PL_PIDSTATUS
11142 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11143#endif
11144 PL_osname = SAVEPV(proto_perl->Iosname);
11145 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11146
bd81e77b 11147 PL_runops = proto_perl->Irunops;
6a78b4db 11148
bd81e77b 11149 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11150
bd81e77b
NC
11151#ifdef CSH
11152 PL_cshlen = proto_perl->Icshlen;
11153 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11154#endif
645c22ef 11155
bd81e77b
NC
11156 PL_lex_state = proto_perl->Ilex_state;
11157 PL_lex_defer = proto_perl->Ilex_defer;
11158 PL_lex_expect = proto_perl->Ilex_expect;
11159 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11160 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11161 PL_lex_starts = proto_perl->Ilex_starts;
11162 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11163 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11164 PL_lex_op = proto_perl->Ilex_op;
11165 PL_lex_inpat = proto_perl->Ilex_inpat;
11166 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11167 PL_lex_brackets = proto_perl->Ilex_brackets;
11168 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11169 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11170 PL_lex_casemods = proto_perl->Ilex_casemods;
11171 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11172 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11173
5db06880
NC
11174#ifdef PERL_MAD
11175 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11176 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11177 PL_realtokenstart = proto_perl->Irealtokenstart;
11178 PL_faketokens = proto_perl->Ifaketokens;
11179 PL_thismad = proto_perl->Ithismad;
11180 PL_thistoken = proto_perl->Ithistoken;
11181 PL_thisopen = proto_perl->Ithisopen;
11182 PL_thisstuff = proto_perl->Ithisstuff;
11183 PL_thisclose = proto_perl->Ithisclose;
11184 PL_thiswhite = proto_perl->Ithiswhite;
11185 PL_nextwhite = proto_perl->Inextwhite;
11186 PL_skipwhite = proto_perl->Iskipwhite;
11187 PL_endwhite = proto_perl->Iendwhite;
11188 PL_curforce = proto_perl->Icurforce;
5db06880 11189#else
bd81e77b
NC
11190 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11191 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11192 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11193#endif
c43294b8 11194
bd81e77b
NC
11195 /* XXX This is probably masking the deeper issue of why
11196 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11197 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11198 * (A little debugging with a watchpoint on it may help.)
11199 */
11200 if (SvANY(proto_perl->Ilinestr)) {
11201 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11202 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11203 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11204 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11205 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11206 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11207 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11208 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11209 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11210 }
11211 else {
561b68a9 11212 PL_linestr = newSV(79);
bd81e77b
NC
11213 sv_upgrade(PL_linestr,SVt_PVIV);
11214 sv_setpvn(PL_linestr,"",0);
11215 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11216 }
11217 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11218 PL_pending_ident = proto_perl->Ipending_ident;
11219 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11220
11221 PL_expect = proto_perl->Iexpect;
11222
11223 PL_multi_start = proto_perl->Imulti_start;
11224 PL_multi_end = proto_perl->Imulti_end;
11225 PL_multi_open = proto_perl->Imulti_open;
11226 PL_multi_close = proto_perl->Imulti_close;
11227
11228 PL_error_count = proto_perl->Ierror_count;
11229 PL_subline = proto_perl->Isubline;
11230 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11231
bd81e77b
NC
11232 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11233 if (SvANY(proto_perl->Ilinestr)) {
11234 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11235 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11236 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11237 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11238 PL_last_lop_op = proto_perl->Ilast_lop_op;
11239 }
11240 else {
11241 PL_last_uni = SvPVX(PL_linestr);
11242 PL_last_lop = SvPVX(PL_linestr);
11243 PL_last_lop_op = 0;
11244 }
11245 PL_in_my = proto_perl->Iin_my;
11246 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11247#ifdef FCRYPT
11248 PL_cryptseen = proto_perl->Icryptseen;
11249#endif
1d7c1841 11250
bd81e77b 11251 PL_hints = proto_perl->Ihints;
1d7c1841 11252
bd81e77b 11253 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11254
bd81e77b
NC
11255#ifdef USE_LOCALE_COLLATE
11256 PL_collation_ix = proto_perl->Icollation_ix;
11257 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11258 PL_collation_standard = proto_perl->Icollation_standard;
11259 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11260 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11261#endif /* USE_LOCALE_COLLATE */
1d7c1841 11262
bd81e77b
NC
11263#ifdef USE_LOCALE_NUMERIC
11264 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11265 PL_numeric_standard = proto_perl->Inumeric_standard;
11266 PL_numeric_local = proto_perl->Inumeric_local;
11267 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11268#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11269
bd81e77b
NC
11270 /* utf8 character classes */
11271 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11272 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11273 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11274 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11275 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11276 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11277 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11278 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11279 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11280 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11281 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11282 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11283 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11284 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11285 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11286 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11287 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11288 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11289 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11290 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11291
bd81e77b
NC
11292 /* Did the locale setup indicate UTF-8? */
11293 PL_utf8locale = proto_perl->Iutf8locale;
11294 /* Unicode features (see perlrun/-C) */
11295 PL_unicode = proto_perl->Iunicode;
1d7c1841 11296
bd81e77b
NC
11297 /* Pre-5.8 signals control */
11298 PL_signals = proto_perl->Isignals;
1d7c1841 11299
bd81e77b
NC
11300 /* times() ticks per second */
11301 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11302
bd81e77b
NC
11303 /* Recursion stopper for PerlIO_find_layer */
11304 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11305
bd81e77b
NC
11306 /* sort() routine */
11307 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11308
bd81e77b
NC
11309 /* Not really needed/useful since the reenrant_retint is "volatile",
11310 * but do it for consistency's sake. */
11311 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11312
bd81e77b
NC
11313 /* Hooks to shared SVs and locks. */
11314 PL_sharehook = proto_perl->Isharehook;
11315 PL_lockhook = proto_perl->Ilockhook;
11316 PL_unlockhook = proto_perl->Iunlockhook;
11317 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11318
bd81e77b
NC
11319 PL_runops_std = proto_perl->Irunops_std;
11320 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11321
bd81e77b
NC
11322#ifdef THREADS_HAVE_PIDS
11323 PL_ppid = proto_perl->Ippid;
11324#endif
1d7c1841 11325
bd81e77b 11326 /* swatch cache */
5c284bb0 11327 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11328 PL_last_swash_klen = 0;
11329 PL_last_swash_key[0]= '\0';
11330 PL_last_swash_tmps = (U8*)NULL;
11331 PL_last_swash_slen = 0;
1d7c1841 11332
bd81e77b
NC
11333 PL_glob_index = proto_perl->Iglob_index;
11334 PL_srand_called = proto_perl->Isrand_called;
11335 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11336 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11337
bd81e77b
NC
11338 if (proto_perl->Ipsig_pend) {
11339 Newxz(PL_psig_pend, SIG_SIZE, int);
11340 }
11341 else {
11342 PL_psig_pend = (int*)NULL;
11343 }
05ec9bb3 11344
bd81e77b
NC
11345 if (proto_perl->Ipsig_ptr) {
11346 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11347 Newxz(PL_psig_name, SIG_SIZE, SV*);
11348 for (i = 1; i < SIG_SIZE; i++) {
11349 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11350 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11351 }
11352 }
11353 else {
11354 PL_psig_ptr = (SV**)NULL;
11355 PL_psig_name = (SV**)NULL;
11356 }
05ec9bb3 11357
bd81e77b 11358 /* thrdvar.h stuff */
1d7c1841 11359
bd81e77b
NC
11360 if (flags & CLONEf_COPY_STACKS) {
11361 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11362 PL_tmps_ix = proto_perl->Ttmps_ix;
11363 PL_tmps_max = proto_perl->Ttmps_max;
11364 PL_tmps_floor = proto_perl->Ttmps_floor;
11365 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11366 i = 0;
11367 while (i <= PL_tmps_ix) {
11368 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11369 ++i;
11370 }
d2d73c3e 11371
bd81e77b
NC
11372 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11373 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11374 Newxz(PL_markstack, i, I32);
11375 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11376 - proto_perl->Tmarkstack);
11377 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11378 - proto_perl->Tmarkstack);
11379 Copy(proto_perl->Tmarkstack, PL_markstack,
11380 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11381
bd81e77b
NC
11382 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11383 * NOTE: unlike the others! */
11384 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11385 PL_scopestack_max = proto_perl->Tscopestack_max;
11386 Newxz(PL_scopestack, PL_scopestack_max, I32);
11387 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11388
bd81e77b
NC
11389 /* NOTE: si_dup() looks at PL_markstack */
11390 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11391
bd81e77b
NC
11392 /* PL_curstack = PL_curstackinfo->si_stack; */
11393 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11394 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11395
bd81e77b
NC
11396 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11397 PL_stack_base = AvARRAY(PL_curstack);
11398 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11399 - proto_perl->Tstack_base);
11400 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11401
bd81e77b
NC
11402 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11403 * NOTE: unlike the others! */
11404 PL_savestack_ix = proto_perl->Tsavestack_ix;
11405 PL_savestack_max = proto_perl->Tsavestack_max;
11406 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11407 PL_savestack = ss_dup(proto_perl, param);
11408 }
11409 else {
11410 init_stacks();
11411 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11412
11413 /* although we're not duplicating the tmps stack, we should still
11414 * add entries for any SVs on the tmps stack that got cloned by a
11415 * non-refcount means (eg a temp in @_); otherwise they will be
11416 * orphaned
11417 */
11418 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11419 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11420 proto_perl->Ttmps_stack[i]);
11421 if (nsv && !SvREFCNT(nsv)) {
11422 EXTEND_MORTAL(1);
b37c2d43 11423 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11424 }
11425 }
bd81e77b 11426 }
1d7c1841 11427
bd81e77b
NC
11428 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11429 PL_top_env = &PL_start_env;
1d7c1841 11430
bd81e77b 11431 PL_op = proto_perl->Top;
4a4c6fe3 11432
a0714e2c 11433 PL_Sv = NULL;
bd81e77b
NC
11434 PL_Xpv = (XPV*)NULL;
11435 PL_na = proto_perl->Tna;
1fcf4c12 11436
bd81e77b
NC
11437 PL_statbuf = proto_perl->Tstatbuf;
11438 PL_statcache = proto_perl->Tstatcache;
11439 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11440 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11441#ifdef HAS_TIMES
11442 PL_timesbuf = proto_perl->Ttimesbuf;
11443#endif
1d7c1841 11444
bd81e77b
NC
11445 PL_tainted = proto_perl->Ttainted;
11446 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11447 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11448 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11449 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11450 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11451 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11452 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11453 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11454 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11455
bd81e77b
NC
11456 PL_restartop = proto_perl->Trestartop;
11457 PL_in_eval = proto_perl->Tin_eval;
11458 PL_delaymagic = proto_perl->Tdelaymagic;
11459 PL_dirty = proto_perl->Tdirty;
11460 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11461
bd81e77b 11462 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11463 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11464 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11465 PL_lastgotoprobe = NULL;
bd81e77b 11466 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11467
bd81e77b
NC
11468 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11469 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11470 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11471 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11472 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11473 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11474
bd81e77b 11475 /* regex stuff */
1d7c1841 11476
bd81e77b
NC
11477 PL_screamfirst = NULL;
11478 PL_screamnext = NULL;
11479 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11480 PL_lastscream = NULL;
1d7c1841 11481
bd81e77b 11482 PL_watchaddr = NULL;
bd61b366 11483 PL_watchok = NULL;
1d7c1841 11484
bd81e77b 11485 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11486 PL_colorset = 0; /* reinits PL_colors[] */
11487 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11488
bd81e77b
NC
11489 /* RE engine - function pointers */
11490 PL_regcompp = proto_perl->Tregcompp;
11491 PL_regexecp = proto_perl->Tregexecp;
11492 PL_regint_start = proto_perl->Tregint_start;
11493 PL_regint_string = proto_perl->Tregint_string;
11494 PL_regfree = proto_perl->Tregfree;
46ab3289 11495 Zero(&PL_reg_state, 1, struct re_save_state);
bd81e77b 11496 PL_reginterp_cnt = 0;
5d9a96ca 11497 PL_regmatch_slab = NULL;
1d7c1841 11498
bd81e77b
NC
11499 /* Pluggable optimizer */
11500 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11501
bd81e77b 11502 PL_stashcache = newHV();
1d7c1841 11503
bd81e77b
NC
11504 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11505 ptr_table_free(PL_ptr_table);
11506 PL_ptr_table = NULL;
11507 }
1d7c1841 11508
bd81e77b
NC
11509 /* Call the ->CLONE method, if it exists, for each of the stashes
11510 identified by sv_dup() above.
11511 */
11512 while(av_len(param->stashes) != -1) {
11513 HV* const stash = (HV*) av_shift(param->stashes);
11514 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11515 if (cloner && GvCV(cloner)) {
11516 dSP;
11517 ENTER;
11518 SAVETMPS;
11519 PUSHMARK(SP);
11520 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11521 PUTBACK;
11522 call_sv((SV*)GvCV(cloner), G_DISCARD);
11523 FREETMPS;
11524 LEAVE;
11525 }
1d7c1841 11526 }
1d7c1841 11527
bd81e77b 11528 SvREFCNT_dec(param->stashes);
1d7c1841 11529
bd81e77b
NC
11530 /* orphaned? eg threads->new inside BEGIN or use */
11531 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11532 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11533 SAVEFREESV(PL_compcv);
11534 }
dd2155a4 11535
bd81e77b
NC
11536 return my_perl;
11537}
1d7c1841 11538
bd81e77b 11539#endif /* USE_ITHREADS */
1d7c1841 11540
bd81e77b
NC
11541/*
11542=head1 Unicode Support
1d7c1841 11543
bd81e77b 11544=for apidoc sv_recode_to_utf8
1d7c1841 11545
bd81e77b
NC
11546The encoding is assumed to be an Encode object, on entry the PV
11547of the sv is assumed to be octets in that encoding, and the sv
11548will be converted into Unicode (and UTF-8).
1d7c1841 11549
bd81e77b
NC
11550If the sv already is UTF-8 (or if it is not POK), or if the encoding
11551is not a reference, nothing is done to the sv. If the encoding is not
11552an C<Encode::XS> Encoding object, bad things will happen.
11553(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11554
bd81e77b 11555The PV of the sv is returned.
1d7c1841 11556
bd81e77b 11557=cut */
1d7c1841 11558
bd81e77b
NC
11559char *
11560Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11561{
11562 dVAR;
11563 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11564 SV *uni;
11565 STRLEN len;
11566 const char *s;
11567 dSP;
11568 ENTER;
11569 SAVETMPS;
11570 save_re_context();
11571 PUSHMARK(sp);
11572 EXTEND(SP, 3);
11573 XPUSHs(encoding);
11574 XPUSHs(sv);
11575/*
11576 NI-S 2002/07/09
11577 Passing sv_yes is wrong - it needs to be or'ed set of constants
11578 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11579 remove converted chars from source.
1d7c1841 11580
bd81e77b 11581 Both will default the value - let them.
1d7c1841 11582
bd81e77b
NC
11583 XPUSHs(&PL_sv_yes);
11584*/
11585 PUTBACK;
11586 call_method("decode", G_SCALAR);
11587 SPAGAIN;
11588 uni = POPs;
11589 PUTBACK;
11590 s = SvPV_const(uni, len);
11591 if (s != SvPVX_const(sv)) {
11592 SvGROW(sv, len + 1);
11593 Move(s, SvPVX(sv), len + 1, char);
11594 SvCUR_set(sv, len);
11595 }
11596 FREETMPS;
11597 LEAVE;
11598 SvUTF8_on(sv);
11599 return SvPVX(sv);
389edf32 11600 }
bd81e77b
NC
11601 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11602}
1d7c1841 11603
bd81e77b
NC
11604/*
11605=for apidoc sv_cat_decode
1d7c1841 11606
bd81e77b
NC
11607The encoding is assumed to be an Encode object, the PV of the ssv is
11608assumed to be octets in that encoding and decoding the input starts
11609from the position which (PV + *offset) pointed to. The dsv will be
11610concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11611when the string tstr appears in decoding output or the input ends on
11612the PV of the ssv. The value which the offset points will be modified
11613to the last input position on the ssv.
1d7c1841 11614
bd81e77b 11615Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11616
bd81e77b
NC
11617=cut */
11618
11619bool
11620Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11621 SV *ssv, int *offset, char *tstr, int tlen)
11622{
11623 dVAR;
11624 bool ret = FALSE;
11625 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11626 SV *offsv;
11627 dSP;
11628 ENTER;
11629 SAVETMPS;
11630 save_re_context();
11631 PUSHMARK(sp);
11632 EXTEND(SP, 6);
11633 XPUSHs(encoding);
11634 XPUSHs(dsv);
11635 XPUSHs(ssv);
11636 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11637 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11638 PUTBACK;
11639 call_method("cat_decode", G_SCALAR);
11640 SPAGAIN;
11641 ret = SvTRUE(TOPs);
11642 *offset = SvIV(offsv);
11643 PUTBACK;
11644 FREETMPS;
11645 LEAVE;
389edf32 11646 }
bd81e77b
NC
11647 else
11648 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11649 return ret;
1d7c1841 11650
bd81e77b 11651}
1d7c1841 11652
bd81e77b
NC
11653/* ---------------------------------------------------------------------
11654 *
11655 * support functions for report_uninit()
11656 */
1d7c1841 11657
bd81e77b
NC
11658/* the maxiumum size of array or hash where we will scan looking
11659 * for the undefined element that triggered the warning */
1d7c1841 11660
bd81e77b 11661#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11662
bd81e77b
NC
11663/* Look for an entry in the hash whose value has the same SV as val;
11664 * If so, return a mortal copy of the key. */
1d7c1841 11665
bd81e77b
NC
11666STATIC SV*
11667S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11668{
11669 dVAR;
11670 register HE **array;
11671 I32 i;
6c3182a5 11672
bd81e77b
NC
11673 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11674 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11675 return NULL;
6c3182a5 11676
bd81e77b 11677 array = HvARRAY(hv);
6c3182a5 11678
bd81e77b
NC
11679 for (i=HvMAX(hv); i>0; i--) {
11680 register HE *entry;
11681 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11682 if (HeVAL(entry) != val)
11683 continue;
11684 if ( HeVAL(entry) == &PL_sv_undef ||
11685 HeVAL(entry) == &PL_sv_placeholder)
11686 continue;
11687 if (!HeKEY(entry))
a0714e2c 11688 return NULL;
bd81e77b
NC
11689 if (HeKLEN(entry) == HEf_SVKEY)
11690 return sv_mortalcopy(HeKEY_sv(entry));
11691 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11692 }
11693 }
a0714e2c 11694 return NULL;
bd81e77b 11695}
6c3182a5 11696
bd81e77b
NC
11697/* Look for an entry in the array whose value has the same SV as val;
11698 * If so, return the index, otherwise return -1. */
6c3182a5 11699
bd81e77b
NC
11700STATIC I32
11701S_find_array_subscript(pTHX_ AV *av, SV* val)
11702{
97aff369 11703 dVAR;
bd81e77b
NC
11704 SV** svp;
11705 I32 i;
11706 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11707 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11708 return -1;
57c6e6d2 11709
bd81e77b
NC
11710 svp = AvARRAY(av);
11711 for (i=AvFILLp(av); i>=0; i--) {
11712 if (svp[i] == val && svp[i] != &PL_sv_undef)
11713 return i;
11714 }
11715 return -1;
11716}
15a5279a 11717
bd81e77b
NC
11718/* S_varname(): return the name of a variable, optionally with a subscript.
11719 * If gv is non-zero, use the name of that global, along with gvtype (one
11720 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11721 * targ. Depending on the value of the subscript_type flag, return:
11722 */
bce260cd 11723
bd81e77b
NC
11724#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11725#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11726#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11727#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11728
bd81e77b
NC
11729STATIC SV*
11730S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11731 SV* keyname, I32 aindex, int subscript_type)
11732{
1d7c1841 11733
bd81e77b
NC
11734 SV * const name = sv_newmortal();
11735 if (gv) {
11736 char buffer[2];
11737 buffer[0] = gvtype;
11738 buffer[1] = 0;
1d7c1841 11739
bd81e77b 11740 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11741
bd81e77b 11742 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11743
bd81e77b
NC
11744 if ((unsigned int)SvPVX(name)[1] <= 26) {
11745 buffer[0] = '^';
11746 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11747
bd81e77b
NC
11748 /* Swap the 1 unprintable control character for the 2 byte pretty
11749 version - ie substr($name, 1, 1) = $buffer; */
11750 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11751 }
bd81e77b
NC
11752 }
11753 else {
11754 U32 unused;
11755 CV * const cv = find_runcv(&unused);
11756 SV *sv;
11757 AV *av;
1d7c1841 11758
bd81e77b 11759 if (!cv || !CvPADLIST(cv))
a0714e2c 11760 return NULL;
bd81e77b
NC
11761 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11762 sv = *av_fetch(av, targ, FALSE);
11763 /* SvLEN in a pad name is not to be trusted */
11764 sv_setpv(name, SvPV_nolen_const(sv));
11765 }
1d7c1841 11766
bd81e77b 11767 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11768 SV * const sv = newSV(0);
bd81e77b
NC
11769 *SvPVX(name) = '$';
11770 Perl_sv_catpvf(aTHX_ name, "{%s}",
11771 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11772 SvREFCNT_dec(sv);
11773 }
11774 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11775 *SvPVX(name) = '$';
11776 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11777 }
11778 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11779 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11780
bd81e77b
NC
11781 return name;
11782}
1d7c1841 11783
1d7c1841 11784
bd81e77b
NC
11785/*
11786=for apidoc find_uninit_var
1d7c1841 11787
bd81e77b
NC
11788Find the name of the undefined variable (if any) that caused the operator o
11789to issue a "Use of uninitialized value" warning.
11790If match is true, only return a name if it's value matches uninit_sv.
11791So roughly speaking, if a unary operator (such as OP_COS) generates a
11792warning, then following the direct child of the op may yield an
11793OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11794other hand, with OP_ADD there are two branches to follow, so we only print
11795the variable name if we get an exact match.
1d7c1841 11796
bd81e77b 11797The name is returned as a mortal SV.
1d7c1841 11798
bd81e77b
NC
11799Assumes that PL_op is the op that originally triggered the error, and that
11800PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11801
bd81e77b
NC
11802=cut
11803*/
1d7c1841 11804
bd81e77b
NC
11805STATIC SV *
11806S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11807{
11808 dVAR;
11809 SV *sv;
11810 AV *av;
11811 GV *gv;
11812 OP *o, *o2, *kid;
1d7c1841 11813
bd81e77b
NC
11814 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11815 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11816 return NULL;
1d7c1841 11817
bd81e77b 11818 switch (obase->op_type) {
1d7c1841 11819
bd81e77b
NC
11820 case OP_RV2AV:
11821 case OP_RV2HV:
11822 case OP_PADAV:
11823 case OP_PADHV:
11824 {
11825 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11826 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11827 I32 index = 0;
a0714e2c 11828 SV *keysv = NULL;
bd81e77b 11829 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11830
bd81e77b
NC
11831 if (pad) { /* @lex, %lex */
11832 sv = PAD_SVl(obase->op_targ);
a0714e2c 11833 gv = NULL;
bd81e77b
NC
11834 }
11835 else {
11836 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11837 /* @global, %global */
11838 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11839 if (!gv)
11840 break;
11841 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11842 }
11843 else /* @{expr}, %{expr} */
11844 return find_uninit_var(cUNOPx(obase)->op_first,
11845 uninit_sv, match);
11846 }
1d7c1841 11847
bd81e77b
NC
11848 /* attempt to find a match within the aggregate */
11849 if (hash) {
d4c19fe8 11850 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11851 if (keysv)
11852 subscript_type = FUV_SUBSCRIPT_HASH;
11853 }
11854 else {
11855 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv);
11856 if (index >= 0)
11857 subscript_type = FUV_SUBSCRIPT_ARRAY;
11858 }
1d7c1841 11859
bd81e77b
NC
11860 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11861 break;
1d7c1841 11862
bd81e77b
NC
11863 return varname(gv, hash ? '%' : '@', obase->op_targ,
11864 keysv, index, subscript_type);
11865 }
1d7c1841 11866
bd81e77b
NC
11867 case OP_PADSV:
11868 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11869 break;
a0714e2c
SS
11870 return varname(NULL, '$', obase->op_targ,
11871 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11872
bd81e77b
NC
11873 case OP_GVSV:
11874 gv = cGVOPx_gv(obase);
11875 if (!gv || (match && GvSV(gv) != uninit_sv))
11876 break;
a0714e2c 11877 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11878
bd81e77b
NC
11879 case OP_AELEMFAST:
11880 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11881 if (match) {
11882 SV **svp;
11883 av = (AV*)PAD_SV(obase->op_targ);
11884 if (!av || SvRMAGICAL(av))
11885 break;
11886 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11887 if (!svp || *svp != uninit_sv)
11888 break;
11889 }
a0714e2c
SS
11890 return varname(NULL, '$', obase->op_targ,
11891 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11892 }
11893 else {
11894 gv = cGVOPx_gv(obase);
11895 if (!gv)
11896 break;
11897 if (match) {
11898 SV **svp;
11899 av = GvAV(gv);
11900 if (!av || SvRMAGICAL(av))
11901 break;
11902 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11903 if (!svp || *svp != uninit_sv)
11904 break;
11905 }
11906 return varname(gv, '$', 0,
a0714e2c 11907 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11908 }
11909 break;
1d7c1841 11910
bd81e77b
NC
11911 case OP_EXISTS:
11912 o = cUNOPx(obase)->op_first;
11913 if (!o || o->op_type != OP_NULL ||
11914 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11915 break;
11916 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11917
bd81e77b
NC
11918 case OP_AELEM:
11919 case OP_HELEM:
11920 if (PL_op == obase)
11921 /* $a[uninit_expr] or $h{uninit_expr} */
11922 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11923
a0714e2c 11924 gv = NULL;
bd81e77b
NC
11925 o = cBINOPx(obase)->op_first;
11926 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11927
bd81e77b 11928 /* get the av or hv, and optionally the gv */
a0714e2c 11929 sv = NULL;
bd81e77b
NC
11930 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11931 sv = PAD_SV(o->op_targ);
11932 }
11933 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11934 && cUNOPo->op_first->op_type == OP_GV)
11935 {
11936 gv = cGVOPx_gv(cUNOPo->op_first);
11937 if (!gv)
11938 break;
11939 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11940 }
11941 if (!sv)
11942 break;
11943
11944 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11945 /* index is constant */
11946 if (match) {
11947 if (SvMAGICAL(sv))
11948 break;
11949 if (obase->op_type == OP_HELEM) {
11950 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11951 if (!he || HeVAL(he) != uninit_sv)
11952 break;
11953 }
11954 else {
00b6aa41 11955 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11956 if (!svp || *svp != uninit_sv)
11957 break;
11958 }
11959 }
11960 if (obase->op_type == OP_HELEM)
11961 return varname(gv, '%', o->op_targ,
11962 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11963 else
a0714e2c 11964 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11965 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11966 }
11967 else {
11968 /* index is an expression;
11969 * attempt to find a match within the aggregate */
11970 if (obase->op_type == OP_HELEM) {
d4c19fe8 11971 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11972 if (keysv)
11973 return varname(gv, '%', o->op_targ,
11974 keysv, 0, FUV_SUBSCRIPT_HASH);
11975 }
11976 else {
d4c19fe8 11977 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11978 if (index >= 0)
11979 return varname(gv, '@', o->op_targ,
a0714e2c 11980 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11981 }
11982 if (match)
11983 break;
11984 return varname(gv,
11985 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
11986 ? '@' : '%',
a0714e2c 11987 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 11988 }
bd81e77b 11989 break;
dc507217 11990
bd81e77b
NC
11991 case OP_AASSIGN:
11992 /* only examine RHS */
11993 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 11994
bd81e77b
NC
11995 case OP_OPEN:
11996 o = cUNOPx(obase)->op_first;
11997 if (o->op_type == OP_PUSHMARK)
11998 o = o->op_sibling;
1d7c1841 11999
bd81e77b
NC
12000 if (!o->op_sibling) {
12001 /* one-arg version of open is highly magical */
a0ae6670 12002
bd81e77b
NC
12003 if (o->op_type == OP_GV) { /* open FOO; */
12004 gv = cGVOPx_gv(o);
12005 if (match && GvSV(gv) != uninit_sv)
12006 break;
12007 return varname(gv, '$', 0,
a0714e2c 12008 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12009 }
12010 /* other possibilities not handled are:
12011 * open $x; or open my $x; should return '${*$x}'
12012 * open expr; should return '$'.expr ideally
12013 */
12014 break;
12015 }
12016 goto do_op;
ccfc67b7 12017
bd81e77b
NC
12018 /* ops where $_ may be an implicit arg */
12019 case OP_TRANS:
12020 case OP_SUBST:
12021 case OP_MATCH:
12022 if ( !(obase->op_flags & OPf_STACKED)) {
12023 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12024 ? PAD_SVl(obase->op_targ)
12025 : DEFSV))
12026 {
12027 sv = sv_newmortal();
12028 sv_setpvn(sv, "$_", 2);
12029 return sv;
12030 }
12031 }
12032 goto do_op;
9f4817db 12033
bd81e77b
NC
12034 case OP_PRTF:
12035 case OP_PRINT:
12036 /* skip filehandle as it can't produce 'undef' warning */
12037 o = cUNOPx(obase)->op_first;
12038 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12039 o = o->op_sibling->op_sibling;
12040 goto do_op2;
9f4817db 12041
9f4817db 12042
bd81e77b
NC
12043 case OP_RV2SV:
12044 case OP_CUSTOM:
12045 case OP_ENTERSUB:
12046 match = 1; /* XS or custom code could trigger random warnings */
12047 goto do_op;
9f4817db 12048
bd81e77b
NC
12049 case OP_SCHOMP:
12050 case OP_CHOMP:
12051 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12052 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12053 /*FALLTHROUGH*/
5d170f3a 12054
bd81e77b
NC
12055 default:
12056 do_op:
12057 if (!(obase->op_flags & OPf_KIDS))
12058 break;
12059 o = cUNOPx(obase)->op_first;
12060
12061 do_op2:
12062 if (!o)
12063 break;
f9893866 12064
bd81e77b
NC
12065 /* if all except one arg are constant, or have no side-effects,
12066 * or are optimized away, then it's unambiguous */
5f66b61c 12067 o2 = NULL;
bd81e77b
NC
12068 for (kid=o; kid; kid = kid->op_sibling) {
12069 if (kid &&
12070 ( (kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid)))
12071 || (kid->op_type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12072 || (kid->op_type == OP_PUSHMARK)
12073 )
12074 )
12075 continue;
12076 if (o2) { /* more than one found */
5f66b61c 12077 o2 = NULL;
bd81e77b
NC
12078 break;
12079 }
12080 o2 = kid;
12081 }
12082 if (o2)
12083 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12084
bd81e77b
NC
12085 /* scan all args */
12086 while (o) {
12087 sv = find_uninit_var(o, uninit_sv, 1);
12088 if (sv)
12089 return sv;
12090 o = o->op_sibling;
d0063567 12091 }
bd81e77b 12092 break;
f9893866 12093 }
a0714e2c 12094 return NULL;
9f4817db
JH
12095}
12096
220e2d4e 12097
bd81e77b
NC
12098/*
12099=for apidoc report_uninit
68795e93 12100
bd81e77b 12101Print appropriate "Use of uninitialized variable" warning
220e2d4e 12102
bd81e77b
NC
12103=cut
12104*/
220e2d4e 12105
bd81e77b
NC
12106void
12107Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12108{
97aff369 12109 dVAR;
bd81e77b 12110 if (PL_op) {
a0714e2c 12111 SV* varname = NULL;
bd81e77b
NC
12112 if (uninit_sv) {
12113 varname = find_uninit_var(PL_op, uninit_sv,0);
12114 if (varname)
12115 sv_insert(varname, 0, 0, " ", 1);
12116 }
12117 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12118 varname ? SvPV_nolen_const(varname) : "",
12119 " in ", OP_DESC(PL_op));
220e2d4e 12120 }
a73e8557 12121 else
bd81e77b
NC
12122 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12123 "", "", "");
220e2d4e 12124}
f9893866 12125
241d1a3b
NC
12126/*
12127 * Local variables:
12128 * c-indentation-style: bsd
12129 * c-basic-offset: 4
12130 * indent-tabs-mode: t
12131 * End:
12132 *
37442d52
RGS
12133 * ex: set ts=8 sts=4 sw=4 noet:
12134 */