This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Use both cache entries for Perl_sv_pos_b2u().
[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
193# define POSION_SV_HEAD(sv) Poison(sv, 1, struct STRUCT_SV)
194*/
195# define POSION_SV_HEAD(sv) Poison(&SvANY(sv), 1, void *), \
196 Poison(&SvREFCNT(sv), 1, U32)
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)) \
1092 ? *((void **)(r3wt)) : S_more_bodies(aTHX_ sv_type); \
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) {
1322 int difference
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) {
2eb42952 3448 S_glob_assign_glob(aTHX_ 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) {
2eb42952 3461 S_glob_assign_glob(aTHX_ 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 }
2eb42952 3489 S_glob_assign_glob(aTHX_ dstr, sstr, dtype);
acaa9288
NC
3490 return;
3491 }
3492
8990e307 3493 if (dtype >= SVt_PV) {
b8c701c1 3494 if (dtype == SVt_PVGV) {
2eb42952 3495 S_glob_assign_ref(aTHX_ 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;
14befaf4 4492 case PERL_MAGIC_sig:
92110913 4493 vtable = &PL_vtbl_sig;
79072805 4494 break;
14befaf4 4495 case PERL_MAGIC_sigelem:
92110913 4496 vtable = &PL_vtbl_sigelem;
79072805 4497 break;
14befaf4 4498 case PERL_MAGIC_taint:
92110913 4499 vtable = &PL_vtbl_taint;
463ee0b2 4500 break;
14befaf4 4501 case PERL_MAGIC_uvar:
92110913 4502 vtable = &PL_vtbl_uvar;
79072805 4503 break;
14befaf4 4504 case PERL_MAGIC_vec:
92110913 4505 vtable = &PL_vtbl_vec;
79072805 4506 break;
a3874608 4507 case PERL_MAGIC_arylen_p:
bfcb3514 4508 case PERL_MAGIC_rhash:
8d2f4536 4509 case PERL_MAGIC_symtab:
ece467f9 4510 case PERL_MAGIC_vstring:
aec46f14 4511 vtable = NULL;
ece467f9 4512 break;
7e8c5dac
HS
4513 case PERL_MAGIC_utf8:
4514 vtable = &PL_vtbl_utf8;
4515 break;
14befaf4 4516 case PERL_MAGIC_substr:
92110913 4517 vtable = &PL_vtbl_substr;
79072805 4518 break;
14befaf4 4519 case PERL_MAGIC_defelem:
92110913 4520 vtable = &PL_vtbl_defelem;
5f05dabc 4521 break;
14befaf4 4522 case PERL_MAGIC_arylen:
92110913 4523 vtable = &PL_vtbl_arylen;
79072805 4524 break;
14befaf4 4525 case PERL_MAGIC_pos:
92110913 4526 vtable = &PL_vtbl_pos;
a0d0e21e 4527 break;
14befaf4 4528 case PERL_MAGIC_backref:
92110913 4529 vtable = &PL_vtbl_backref;
810b8aa5 4530 break;
14befaf4
DM
4531 case PERL_MAGIC_ext:
4532 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4533 /* Useful for attaching extension internal data to perl vars. */
4534 /* Note that multiple extensions may clash if magical scalars */
4535 /* etc holding private data from one are passed to another. */
aec46f14 4536 vtable = NULL;
a0d0e21e 4537 break;
79072805 4538 default:
14befaf4 4539 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4540 }
68795e93 4541
92110913 4542 /* Rest of work is done else where */
aec46f14 4543 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4544
92110913
NIS
4545 switch (how) {
4546 case PERL_MAGIC_taint:
4547 mg->mg_len = 1;
4548 break;
4549 case PERL_MAGIC_ext:
4550 case PERL_MAGIC_dbfile:
4551 SvRMAGICAL_on(sv);
4552 break;
4553 }
463ee0b2
LW
4554}
4555
c461cf8f
JH
4556/*
4557=for apidoc sv_unmagic
4558
645c22ef 4559Removes all magic of type C<type> from an SV.
c461cf8f
JH
4560
4561=cut
4562*/
4563
463ee0b2 4564int
864dbfa3 4565Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4566{
4567 MAGIC* mg;
4568 MAGIC** mgp;
91bba347 4569 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4570 return 0;
064cf529 4571 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4572 for (mg = *mgp; mg; mg = *mgp) {
4573 if (mg->mg_type == type) {
e1ec3a88 4574 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4575 *mgp = mg->mg_moremagic;
1d7c1841 4576 if (vtbl && vtbl->svt_free)
fc0dc3b3 4577 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4578 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4579 if (mg->mg_len > 0)
1edc1566 4580 Safefree(mg->mg_ptr);
565764a8 4581 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4582 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4583 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4584 Safefree(mg->mg_ptr);
9cbac4c7 4585 }
a0d0e21e
LW
4586 if (mg->mg_flags & MGf_REFCOUNTED)
4587 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4588 Safefree(mg);
4589 }
4590 else
4591 mgp = &mg->mg_moremagic;
79072805 4592 }
91bba347 4593 if (!SvMAGIC(sv)) {
463ee0b2 4594 SvMAGICAL_off(sv);
c268c2a6 4595 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4596 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4597 }
4598
4599 return 0;
79072805
LW
4600}
4601
c461cf8f
JH
4602/*
4603=for apidoc sv_rvweaken
4604
645c22ef
DM
4605Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4606referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4607push a back-reference to this RV onto the array of backreferences
4608associated with that magic.
c461cf8f
JH
4609
4610=cut
4611*/
4612
810b8aa5 4613SV *
864dbfa3 4614Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4615{
4616 SV *tsv;
4617 if (!SvOK(sv)) /* let undefs pass */
4618 return sv;
4619 if (!SvROK(sv))
cea2e8a9 4620 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4621 else if (SvWEAKREF(sv)) {
810b8aa5 4622 if (ckWARN(WARN_MISC))
9014280d 4623 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4624 return sv;
4625 }
4626 tsv = SvRV(sv);
e15faf7d 4627 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4628 SvWEAKREF_on(sv);
1c846c1f 4629 SvREFCNT_dec(tsv);
810b8aa5
GS
4630 return sv;
4631}
4632
645c22ef
DM
4633/* Give tsv backref magic if it hasn't already got it, then push a
4634 * back-reference to sv onto the array associated with the backref magic.
4635 */
4636
e15faf7d
NC
4637void
4638Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4639{
97aff369 4640 dVAR;
810b8aa5 4641 AV *av;
86f55936
NC
4642
4643 if (SvTYPE(tsv) == SVt_PVHV) {
4644 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4645
4646 av = *avp;
4647 if (!av) {
4648 /* There is no AV in the offical place - try a fixup. */
4649 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4650
4651 if (mg) {
4652 /* Aha. They've got it stowed in magic. Bring it back. */
4653 av = (AV*)mg->mg_obj;
4654 /* Stop mg_free decreasing the refernce count. */
4655 mg->mg_obj = NULL;
4656 /* Stop mg_free even calling the destructor, given that
4657 there's no AV to free up. */
4658 mg->mg_virtual = 0;
4659 sv_unmagic(tsv, PERL_MAGIC_backref);
4660 } else {
4661 av = newAV();
4662 AvREAL_off(av);
b37c2d43 4663 SvREFCNT_inc_simple_void(av);
86f55936
NC
4664 }
4665 *avp = av;
4666 }
4667 } else {
4668 const MAGIC *const mg
4669 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4670 if (mg)
4671 av = (AV*)mg->mg_obj;
4672 else {
4673 av = newAV();
4674 AvREAL_off(av);
4675 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4676 /* av now has a refcnt of 2, which avoids it getting freed
4677 * before us during global cleanup. The extra ref is removed
4678 * by magic_killbackrefs() when tsv is being freed */
4679 }
810b8aa5 4680 }
d91d49e8 4681 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4682 av_extend(av, AvFILLp(av)+1);
4683 }
4684 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4685}
4686
645c22ef
DM
4687/* delete a back-reference to ourselves from the backref magic associated
4688 * with the SV we point to.
4689 */
4690
1c846c1f 4691STATIC void
e15faf7d 4692S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4693{
97aff369 4694 dVAR;
86f55936 4695 AV *av = NULL;
810b8aa5
GS
4696 SV **svp;
4697 I32 i;
86f55936
NC
4698
4699 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4700 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4701 /* We mustn't attempt to "fix up" the hash here by moving the
4702 backreference array back to the hv_aux structure, as that is stored
4703 in the main HvARRAY(), and hfreentries assumes that no-one
4704 reallocates HvARRAY() while it is running. */
86f55936
NC
4705 }
4706 if (!av) {
4707 const MAGIC *const mg
4708 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4709 if (mg)
4710 av = (AV *)mg->mg_obj;
4711 }
4712 if (!av) {
e15faf7d
NC
4713 if (PL_in_clean_all)
4714 return;
cea2e8a9 4715 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4716 }
4717
4718 if (SvIS_FREED(av))
4719 return;
4720
810b8aa5 4721 svp = AvARRAY(av);
6a76db8b
NC
4722 /* We shouldn't be in here more than once, but for paranoia reasons lets
4723 not assume this. */
4724 for (i = AvFILLp(av); i >= 0; i--) {
4725 if (svp[i] == sv) {
4726 const SSize_t fill = AvFILLp(av);
4727 if (i != fill) {
4728 /* We weren't the last entry.
4729 An unordered list has this property that you can take the
4730 last element off the end to fill the hole, and it's still
4731 an unordered list :-)
4732 */
4733 svp[i] = svp[fill];
4734 }
a0714e2c 4735 svp[fill] = NULL;
6a76db8b
NC
4736 AvFILLp(av) = fill - 1;
4737 }
4738 }
810b8aa5
GS
4739}
4740
86f55936
NC
4741int
4742Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4743{
4744 SV **svp = AvARRAY(av);
4745
4746 PERL_UNUSED_ARG(sv);
4747
4748 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4749 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4750 if (svp && !SvIS_FREED(av)) {
4751 SV *const *const last = svp + AvFILLp(av);
4752
4753 while (svp <= last) {
4754 if (*svp) {
4755 SV *const referrer = *svp;
4756 if (SvWEAKREF(referrer)) {
4757 /* XXX Should we check that it hasn't changed? */
4758 SvRV_set(referrer, 0);
4759 SvOK_off(referrer);
4760 SvWEAKREF_off(referrer);
4761 } else if (SvTYPE(referrer) == SVt_PVGV ||
4762 SvTYPE(referrer) == SVt_PVLV) {
4763 /* You lookin' at me? */
4764 assert(GvSTASH(referrer));
4765 assert(GvSTASH(referrer) == (HV*)sv);
4766 GvSTASH(referrer) = 0;
4767 } else {
4768 Perl_croak(aTHX_
4769 "panic: magic_killbackrefs (flags=%"UVxf")",
4770 (UV)SvFLAGS(referrer));
4771 }
4772
a0714e2c 4773 *svp = NULL;
86f55936
NC
4774 }
4775 svp++;
4776 }
4777 }
4778 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4779 return 0;
4780}
4781
954c1994
GS
4782/*
4783=for apidoc sv_insert
4784
4785Inserts a string at the specified offset/length within the SV. Similar to
4786the Perl substr() function.
4787
4788=cut
4789*/
4790
79072805 4791void
e1ec3a88 4792Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4793{
97aff369 4794 dVAR;
79072805
LW
4795 register char *big;
4796 register char *mid;
4797 register char *midend;
4798 register char *bigend;
4799 register I32 i;
6ff81951 4800 STRLEN curlen;
1c846c1f 4801
79072805 4802
8990e307 4803 if (!bigstr)
cea2e8a9 4804 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4805 SvPV_force(bigstr, curlen);
60fa28ff 4806 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4807 if (offset + len > curlen) {
4808 SvGROW(bigstr, offset+len+1);
93524f2b 4809 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4810 SvCUR_set(bigstr, offset+len);
4811 }
79072805 4812
69b47968 4813 SvTAINT(bigstr);
79072805
LW
4814 i = littlelen - len;
4815 if (i > 0) { /* string might grow */
a0d0e21e 4816 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4817 mid = big + offset + len;
4818 midend = bigend = big + SvCUR(bigstr);
4819 bigend += i;
4820 *bigend = '\0';
4821 while (midend > mid) /* shove everything down */
4822 *--bigend = *--midend;
4823 Move(little,big+offset,littlelen,char);
b162af07 4824 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4825 SvSETMAGIC(bigstr);
4826 return;
4827 }
4828 else if (i == 0) {
463ee0b2 4829 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4830 SvSETMAGIC(bigstr);
4831 return;
4832 }
4833
463ee0b2 4834 big = SvPVX(bigstr);
79072805
LW
4835 mid = big + offset;
4836 midend = mid + len;
4837 bigend = big + SvCUR(bigstr);
4838
4839 if (midend > bigend)
cea2e8a9 4840 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4841
4842 if (mid - big > bigend - midend) { /* faster to shorten from end */
4843 if (littlelen) {
4844 Move(little, mid, littlelen,char);
4845 mid += littlelen;
4846 }
4847 i = bigend - midend;
4848 if (i > 0) {
4849 Move(midend, mid, i,char);
4850 mid += i;
4851 }
4852 *mid = '\0';
4853 SvCUR_set(bigstr, mid - big);
4854 }
155aba94 4855 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4856 midend -= littlelen;
4857 mid = midend;
4858 sv_chop(bigstr,midend-i);
4859 big += i;
4860 while (i--)
4861 *--midend = *--big;
4862 if (littlelen)
4863 Move(little, mid, littlelen,char);
4864 }
4865 else if (littlelen) {
4866 midend -= littlelen;
4867 sv_chop(bigstr,midend);
4868 Move(little,midend,littlelen,char);
4869 }
4870 else {
4871 sv_chop(bigstr,midend);
4872 }
4873 SvSETMAGIC(bigstr);
4874}
4875
c461cf8f
JH
4876/*
4877=for apidoc sv_replace
4878
4879Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4880The target SV physically takes over ownership of the body of the source SV
4881and inherits its flags; however, the target keeps any magic it owns,
4882and any magic in the source is discarded.
ff276b08 4883Note that this is a rather specialist SV copying operation; most of the
645c22ef 4884time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4885
4886=cut
4887*/
79072805
LW
4888
4889void
864dbfa3 4890Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4891{
97aff369 4892 dVAR;
a3b680e6 4893 const U32 refcnt = SvREFCNT(sv);
765f542d 4894 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4895 if (SvREFCNT(nsv) != 1) {
7437becc 4896 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4897 UVuf " != 1)", (UV) SvREFCNT(nsv));
4898 }
93a17b20 4899 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4900 if (SvMAGICAL(nsv))
4901 mg_free(nsv);
4902 else
4903 sv_upgrade(nsv, SVt_PVMG);
b162af07 4904 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4905 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4906 SvMAGICAL_off(sv);
b162af07 4907 SvMAGIC_set(sv, NULL);
93a17b20 4908 }
79072805
LW
4909 SvREFCNT(sv) = 0;
4910 sv_clear(sv);
477f5d66 4911 assert(!SvREFCNT(sv));
fd0854ff
DM
4912#ifdef DEBUG_LEAKING_SCALARS
4913 sv->sv_flags = nsv->sv_flags;
4914 sv->sv_any = nsv->sv_any;
4915 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4916 sv->sv_u = nsv->sv_u;
fd0854ff 4917#else
79072805 4918 StructCopy(nsv,sv,SV);
fd0854ff 4919#endif
7b2c381c
NC
4920 /* Currently could join these into one piece of pointer arithmetic, but
4921 it would be unclear. */
4922 if(SvTYPE(sv) == SVt_IV)
4923 SvANY(sv)
339049b0 4924 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4925 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4926 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4927 }
4928
fd0854ff 4929
f8c7b90f 4930#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4931 if (SvIsCOW_normal(nsv)) {
4932 /* We need to follow the pointers around the loop to make the
4933 previous SV point to sv, rather than nsv. */
4934 SV *next;
4935 SV *current = nsv;
4936 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4937 assert(next);
4938 current = next;
3f7c398e 4939 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4940 }
4941 /* Make the SV before us point to the SV after us. */
4942 if (DEBUG_C_TEST) {
4943 PerlIO_printf(Perl_debug_log, "previous is\n");
4944 sv_dump(current);
a29f6d03
NC
4945 PerlIO_printf(Perl_debug_log,
4946 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4947 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4948 }
a29f6d03 4949 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4950 }
4951#endif
79072805 4952 SvREFCNT(sv) = refcnt;
1edc1566 4953 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4954 SvREFCNT(nsv) = 0;
463ee0b2 4955 del_SV(nsv);
79072805
LW
4956}
4957
c461cf8f
JH
4958/*
4959=for apidoc sv_clear
4960
645c22ef
DM
4961Clear an SV: call any destructors, free up any memory used by the body,
4962and free the body itself. The SV's head is I<not> freed, although
4963its type is set to all 1's so that it won't inadvertently be assumed
4964to be live during global destruction etc.
4965This function should only be called when REFCNT is zero. Most of the time
4966you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4967instead.
c461cf8f
JH
4968
4969=cut
4970*/
4971
79072805 4972void
864dbfa3 4973Perl_sv_clear(pTHX_ register SV *sv)
79072805 4974{
27da23d5 4975 dVAR;
82bb6deb 4976 const U32 type = SvTYPE(sv);
8edfc514
NC
4977 const struct body_details *const sv_type_details
4978 = bodies_by_type + type;
82bb6deb 4979
79072805
LW
4980 assert(sv);
4981 assert(SvREFCNT(sv) == 0);
4982
d2a0f284
JC
4983 if (type <= SVt_IV) {
4984 /* See the comment in sv.h about the collusion between this early
4985 return and the overloading of the NULL and IV slots in the size
4986 table. */
82bb6deb 4987 return;
d2a0f284 4988 }
82bb6deb 4989
ed6116ce 4990 if (SvOBJECT(sv)) {
3280af22 4991 if (PL_defstash) { /* Still have a symbol table? */
39644a26 4992 dSP;
893645bd 4993 HV* stash;
d460ef45 4994 do {
b464bac0 4995 CV* destructor;
4e8e7886 4996 stash = SvSTASH(sv);
32251b26 4997 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 4998 if (destructor) {
1b6737cc 4999 SV* const tmpref = newRV(sv);
5cc433a6 5000 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5001 ENTER;
e788e7d3 5002 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5003 EXTEND(SP, 2);
5004 PUSHMARK(SP);
5cc433a6 5005 PUSHs(tmpref);
4e8e7886 5006 PUTBACK;
44389ee9 5007 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5008
5009
d3acc0f7 5010 POPSTACK;
3095d977 5011 SPAGAIN;
4e8e7886 5012 LEAVE;
5cc433a6
AB
5013 if(SvREFCNT(tmpref) < 2) {
5014 /* tmpref is not kept alive! */
5015 SvREFCNT(sv)--;
b162af07 5016 SvRV_set(tmpref, NULL);
5cc433a6
AB
5017 SvROK_off(tmpref);
5018 }
5019 SvREFCNT_dec(tmpref);
4e8e7886
GS
5020 }
5021 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5022
6f44e0a4
JP
5023
5024 if (SvREFCNT(sv)) {
5025 if (PL_in_clean_objs)
cea2e8a9 5026 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5027 HvNAME_get(stash));
6f44e0a4
JP
5028 /* DESTROY gave object new lease on life */
5029 return;
5030 }
a0d0e21e 5031 }
4e8e7886 5032
a0d0e21e 5033 if (SvOBJECT(sv)) {
4e8e7886 5034 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5035 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5036 if (type != SVt_PVIO)
3280af22 5037 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5038 }
463ee0b2 5039 }
82bb6deb 5040 if (type >= SVt_PVMG) {
e736a858
NC
5041 HV *ourstash;
5042 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5043 (ourstash = OURSTASH(sv))) {
5044 SvREFCNT_dec(ourstash);
5045 } else if (SvMAGIC(sv))
524189f1 5046 mg_free(sv);
00b1698f 5047 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5048 SvREFCNT_dec(SvSTASH(sv));
5049 }
82bb6deb 5050 switch (type) {
8990e307 5051 case SVt_PVIO:
df0bd2f4
GS
5052 if (IoIFP(sv) &&
5053 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5054 IoIFP(sv) != PerlIO_stdout() &&
5055 IoIFP(sv) != PerlIO_stderr())
93578b34 5056 {
f2b5be74 5057 io_close((IO*)sv, FALSE);
93578b34 5058 }
1d7c1841 5059 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5060 PerlDir_close(IoDIRP(sv));
1d7c1841 5061 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5062 Safefree(IoTOP_NAME(sv));
5063 Safefree(IoFMT_NAME(sv));
5064 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5065 goto freescalar;
79072805 5066 case SVt_PVBM:
a0d0e21e 5067 goto freescalar;
79072805 5068 case SVt_PVCV:
748a9306 5069 case SVt_PVFM:
85e6fe83 5070 cv_undef((CV*)sv);
a0d0e21e 5071 goto freescalar;
79072805 5072 case SVt_PVHV:
86f55936 5073 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5074 hv_undef((HV*)sv);
a0d0e21e 5075 break;
79072805 5076 case SVt_PVAV:
85e6fe83 5077 av_undef((AV*)sv);
a0d0e21e 5078 break;
02270b4e 5079 case SVt_PVLV:
dd28f7bb
DM
5080 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5081 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5082 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5083 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5084 }
5085 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5086 SvREFCNT_dec(LvTARG(sv));
02270b4e 5087 goto freescalar;
a0d0e21e 5088 case SVt_PVGV:
1edc1566 5089 gp_free((GV*)sv);
acda4c6a
NC
5090 if (GvNAME_HEK(sv)) {
5091 unshare_hek(GvNAME_HEK(sv));
5092 }
893645bd
NC
5093 /* If we're in a stash, we don't own a reference to it. However it does
5094 have a back reference to us, which needs to be cleared. */
5095 if (GvSTASH(sv))
5096 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5097 case SVt_PVMG:
79072805
LW
5098 case SVt_PVNV:
5099 case SVt_PVIV:
a0d0e21e 5100 freescalar:
5228ca4e
NC
5101 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5102 if (SvOOK(sv)) {
93524f2b 5103 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5104 /* Don't even bother with turning off the OOK flag. */
5105 }
79072805 5106 case SVt_PV:
a0d0e21e 5107 case SVt_RV:
810b8aa5 5108 if (SvROK(sv)) {
b37c2d43 5109 SV * const target = SvRV(sv);
810b8aa5 5110 if (SvWEAKREF(sv))
e15faf7d 5111 sv_del_backref(target, sv);
810b8aa5 5112 else
e15faf7d 5113 SvREFCNT_dec(target);
810b8aa5 5114 }
f8c7b90f 5115#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5116 else if (SvPVX_const(sv)) {
765f542d
NC
5117 if (SvIsCOW(sv)) {
5118 /* I believe I need to grab the global SV mutex here and
5119 then recheck the COW status. */
46187eeb
NC
5120 if (DEBUG_C_TEST) {
5121 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5122 sv_dump(sv);
46187eeb 5123 }
bdd68bc3
NC
5124 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5125 SV_COW_NEXT_SV(sv));
765f542d
NC
5126 /* And drop it here. */
5127 SvFAKE_off(sv);
5128 } else if (SvLEN(sv)) {
3f7c398e 5129 Safefree(SvPVX_const(sv));
765f542d
NC
5130 }
5131 }
5132#else
3f7c398e 5133 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5134 Safefree(SvPVX_mutable(sv));
3f7c398e 5135 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5136 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5137 SvFAKE_off(sv);
5138 }
765f542d 5139#endif
79072805
LW
5140 break;
5141 case SVt_NV:
79072805
LW
5142 break;
5143 }
5144
893645bd
NC
5145 SvFLAGS(sv) &= SVf_BREAK;
5146 SvFLAGS(sv) |= SVTYPEMASK;
5147
8edfc514 5148 if (sv_type_details->arena) {
b9502f15 5149 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5150 &PL_body_roots[type]);
5151 }
d2a0f284 5152 else if (sv_type_details->body_size) {
8edfc514
NC
5153 my_safefree(SvANY(sv));
5154 }
79072805
LW
5155}
5156
645c22ef
DM
5157/*
5158=for apidoc sv_newref
5159
5160Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5161instead.
5162
5163=cut
5164*/
5165
79072805 5166SV *
864dbfa3 5167Perl_sv_newref(pTHX_ SV *sv)
79072805 5168{
96a5add6 5169 PERL_UNUSED_CONTEXT;
463ee0b2 5170 if (sv)
4db098f4 5171 (SvREFCNT(sv))++;
79072805
LW
5172 return sv;
5173}
5174
c461cf8f
JH
5175/*
5176=for apidoc sv_free
5177
645c22ef
DM
5178Decrement an SV's reference count, and if it drops to zero, call
5179C<sv_clear> to invoke destructors and free up any memory used by
5180the body; finally, deallocate the SV's head itself.
5181Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5182
5183=cut
5184*/
5185
79072805 5186void
864dbfa3 5187Perl_sv_free(pTHX_ SV *sv)
79072805 5188{
27da23d5 5189 dVAR;
79072805
LW
5190 if (!sv)
5191 return;
a0d0e21e
LW
5192 if (SvREFCNT(sv) == 0) {
5193 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5194 /* this SV's refcnt has been artificially decremented to
5195 * trigger cleanup */
a0d0e21e 5196 return;
3280af22 5197 if (PL_in_clean_all) /* All is fair */
1edc1566 5198 return;
d689ffdd
JP
5199 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5200 /* make sure SvREFCNT(sv)==0 happens very seldom */
5201 SvREFCNT(sv) = (~(U32)0)/2;
5202 return;
5203 }
41e4abd8 5204 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5205 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5206 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5207 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5208#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5209 Perl_dump_sv_child(aTHX_ sv);
5210#endif
5211 }
79072805
LW
5212 return;
5213 }
4db098f4 5214 if (--(SvREFCNT(sv)) > 0)
8990e307 5215 return;
8c4d3c90
NC
5216 Perl_sv_free2(aTHX_ sv);
5217}
5218
5219void
5220Perl_sv_free2(pTHX_ SV *sv)
5221{
27da23d5 5222 dVAR;
463ee0b2
LW
5223#ifdef DEBUGGING
5224 if (SvTEMP(sv)) {
0453d815 5225 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5226 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5227 "Attempt to free temp prematurely: SV 0x%"UVxf
5228 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5229 return;
79072805 5230 }
463ee0b2 5231#endif
d689ffdd
JP
5232 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5233 /* make sure SvREFCNT(sv)==0 happens very seldom */
5234 SvREFCNT(sv) = (~(U32)0)/2;
5235 return;
5236 }
79072805 5237 sv_clear(sv);
477f5d66
CS
5238 if (! SvREFCNT(sv))
5239 del_SV(sv);
79072805
LW
5240}
5241
954c1994
GS
5242/*
5243=for apidoc sv_len
5244
645c22ef
DM
5245Returns the length of the string in the SV. Handles magic and type
5246coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5247
5248=cut
5249*/
5250
79072805 5251STRLEN
864dbfa3 5252Perl_sv_len(pTHX_ register SV *sv)
79072805 5253{
463ee0b2 5254 STRLEN len;
79072805
LW
5255
5256 if (!sv)
5257 return 0;
5258
8990e307 5259 if (SvGMAGICAL(sv))
565764a8 5260 len = mg_length(sv);
8990e307 5261 else
4d84ee25 5262 (void)SvPV_const(sv, len);
463ee0b2 5263 return len;
79072805
LW
5264}
5265
c461cf8f
JH
5266/*
5267=for apidoc sv_len_utf8
5268
5269Returns the number of characters in the string in an SV, counting wide
1e54db1a 5270UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5271
5272=cut
5273*/
5274
7e8c5dac
HS
5275/*
5276 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
5277 * mg_ptr is used, by sv_pos_u2b(), see the comments of S_utf8_mg_pos_init().
5278 * (Note that the mg_len is not the length of the mg_ptr field.)
7a5fa8a2 5279 *
7e8c5dac
HS
5280 */
5281
a0ed51b3 5282STRLEN
864dbfa3 5283Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5284{
a0ed51b3
LW
5285 if (!sv)
5286 return 0;
5287
a0ed51b3 5288 if (SvGMAGICAL(sv))
b76347f2 5289 return mg_length(sv);
a0ed51b3 5290 else
b76347f2 5291 {
26346457 5292 STRLEN len;
e62f0680 5293 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5294
26346457
NC
5295 if (PL_utf8cache) {
5296 STRLEN ulen;
5297 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5298
5299 if (mg && mg->mg_len != -1) {
5300 ulen = mg->mg_len;
5301 if (PL_utf8cache < 0) {
5302 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5303 if (real != ulen) {
5304 /* Need to turn the assertions off otherwise we may
5305 recurse infinitely while printing error messages.
5306 */
5307 SAVEI8(PL_utf8cache);
5308 PL_utf8cache = 0;
5309 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5310 " real %"UVf" for %"SVf,
26346457
NC
5311 (UV) ulen, (UV) real, sv);
5312 }
5313 }
5314 }
5315 else {
5316 ulen = Perl_utf8_length(aTHX_ s, s + len);
5317 if (!SvREADONLY(sv)) {
5318 if (!mg) {
5319 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5320 &PL_vtbl_utf8, 0, 0);
5321 }
cb9e20bb 5322 assert(mg);
26346457 5323 mg->mg_len = ulen;
cb9e20bb 5324 }
cb9e20bb 5325 }
26346457 5326 return ulen;
7e8c5dac 5327 }
26346457 5328 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5329 }
5330}
5331
7e8c5dac 5332/*
645c22ef
DM
5333=for apidoc sv_pos_u2b
5334
1e54db1a 5335Converts the value pointed to by offsetp from a count of UTF-8 chars from
645c22ef
DM
5336the start of the string, to a count of the equivalent number of bytes; if
5337lenp is non-zero, it does the same to lenp, but this time starting from
5338the offset, rather than from the start of the string. Handles magic and
5339type coercion.
5340
5341=cut
5342*/
5343
7e8c5dac
HS
5344/*
5345 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5346 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5347 * byte offsets. See also the comments of S_utf8_mg_pos().
5348 *
5349 */
5350
bdf30dd6
NC
5351static STRLEN
5352S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send,
5353 STRLEN uoffset)
5354{
5355 const U8 *s = start;
5356
5357 while (s < send && uoffset--)
5358 s += UTF8SKIP(s);
5359 if (s > send) {
5360 /* This is the existing behaviour. Possibly it should be a croak, as
5361 it's actually a bounds error */
5362 s = send;
5363 }
5364 return s - start;
5365}
5366
c336ad0b
NC
5367
5368static STRLEN
5369S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send,
5370 STRLEN uoffset, STRLEN uend)
5371{
5372 STRLEN backw = uend - uoffset;
5373 if (uoffset < 2 * backw) {
25a8a4ef 5374 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5375 forward (that's where the 2 * backw comes from).
5376 (The real figure of course depends on the UTF-8 data.) */
5377 return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset);
5378 }
5379
5380 while (backw--) {
5381 send--;
5382 while (UTF8_IS_CONTINUATION(*send))
5383 send--;
5384 }
5385 return send - start;
5386}
5387
28ccbf94
NC
5388static STRLEN
5389S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5390 const U8 *const send, STRLEN uoffset,
5391 STRLEN uoffset0, STRLEN boffset0) {
5392 STRLEN boffset;
c336ad0b
NC
5393 bool found = FALSE;
5394
75c33c12
NC
5395 assert (uoffset >= uoffset0);
5396
c336ad0b 5397 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5398 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5399 if ((*mgp)->mg_ptr) {
5400 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5401 if (cache[0] == uoffset) {
5402 /* An exact match. */
5403 return cache[1];
5404 }
ab455f60
NC
5405 if (cache[2] == uoffset) {
5406 /* An exact match. */
5407 return cache[3];
5408 }
668af93f
NC
5409
5410 if (cache[0] < uoffset) {
d8b2e1f9
NC
5411 /* The cache already knows part of the way. */
5412 if (cache[0] > uoffset0) {
5413 /* The cache knows more than the passed in pair */
5414 uoffset0 = cache[0];
5415 boffset0 = cache[1];
5416 }
5417 if ((*mgp)->mg_len != -1) {
5418 /* And we know the end too. */
5419 boffset = boffset0
5420 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5421 uoffset - uoffset0,
5422 (*mgp)->mg_len - uoffset0);
5423 } else {
5424 boffset = boffset0
5425 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5426 send, uoffset - uoffset0);
5427 }
668af93f
NC
5428 } else {
5429 boffset = boffset0
5430 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5431 start + cache[1],
5432 uoffset - uoffset0,
5433 cache[0] - uoffset0);
d8b2e1f9 5434 }
668af93f 5435 found = TRUE;
d8b2e1f9
NC
5436 }
5437 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5438 /* If we can take advantage of a passed in offset, do so. */
5439 /* In fact, offset0 is either 0, or less than offset, so don't
5440 need to worry about the other possibility. */
5441 boffset = boffset0
5442 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5443 uoffset - uoffset0,
5444 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5445 found = TRUE;
5446 }
28ccbf94 5447 }
c336ad0b
NC
5448
5449 if (!found || PL_utf8cache < 0) {
75c33c12
NC
5450 const STRLEN real_boffset
5451 = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5452 send, uoffset - uoffset0);
5453
c336ad0b
NC
5454 if (found && PL_utf8cache < 0) {
5455 if (real_boffset != boffset) {
5456 /* Need to turn the assertions off otherwise we may recurse
5457 infinitely while printing error messages. */
5458 SAVEI8(PL_utf8cache);
5459 PL_utf8cache = 0;
5460 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5461 " real %"UVf" for %"SVf,
5462 (UV) boffset, (UV) real_boffset, sv);
5463 }
5464 }
5465 boffset = real_boffset;
28ccbf94 5466 }
0905937d 5467
ab455f60 5468 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5469 return boffset;
5470}
5471
a0ed51b3 5472void
864dbfa3 5473Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5474{
245d4a47 5475 const U8 *start;
a0ed51b3
LW
5476 STRLEN len;
5477
5478 if (!sv)
5479 return;
5480
245d4a47 5481 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5482 if (len) {
bdf30dd6
NC
5483 STRLEN uoffset = (STRLEN) *offsetp;
5484 const U8 * const send = start + len;
0905937d 5485 MAGIC *mg = NULL;
28ccbf94
NC
5486 STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send,
5487 uoffset, 0, 0);
bdf30dd6
NC
5488
5489 *offsetp = (I32) boffset;
5490
5491 if (lenp) {
28ccbf94
NC
5492 /* Convert the relative offset to absolute. */
5493 STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5494 STRLEN boffset2
5495 = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2,
5496 uoffset, boffset) - boffset;
bdf30dd6 5497
28ccbf94 5498 *lenp = boffset2;
bdf30dd6 5499 }
7e8c5dac
HS
5500 }
5501 else {
5502 *offsetp = 0;
5503 if (lenp)
5504 *lenp = 0;
a0ed51b3 5505 }
e23c8137 5506
a0ed51b3
LW
5507 return;
5508}
5509
645c22ef
DM
5510/*
5511=for apidoc sv_pos_b2u
5512
5513Converts the value pointed to by offsetp from a count of bytes from the
1e54db1a 5514start of the string, to a count of the equivalent number of UTF-8 chars.
645c22ef
DM
5515Handles magic and type coercion.
5516
5517=cut
5518*/
5519
7e8c5dac
HS
5520/*
5521 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5522 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
25a8a4ef 5523 * byte offsets.
7e8c5dac
HS
5524 *
5525 */
5526
ec07b5e0 5527static void
ab455f60
NC
5528S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5529 STRLEN blen)
ec07b5e0
NC
5530{
5531 STRLEN *cache;
5532 if (SvREADONLY(sv))
5533 return;
5534
5535 if (!*mgp) {
5536 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5537 0);
5538 (*mgp)->mg_len = -1;
5539 }
5540 assert(*mgp);
5541
5542 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5543 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5544 (*mgp)->mg_ptr = (char *) cache;
5545 }
5546 assert(cache);
5547
5548 if (PL_utf8cache < 0) {
ef816a78 5549 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5550 const U8 *const end = start + byte;
5551 STRLEN realutf8 = 0;
5552
5553 while (start < end) {
5554 start += UTF8SKIP(start);
5555 realutf8++;
5556 }
5557
5558 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5559 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5560 doesn't? I don't know whether this difference was introduced with
5561 the caching code in 5.8.1. */
ec07b5e0
NC
5562
5563 if (realutf8 != utf8) {
5564 /* Need to turn the assertions off otherwise we may recurse
5565 infinitely while printing error messages. */
5566 SAVEI8(PL_utf8cache);
5567 PL_utf8cache = 0;
5568 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5569 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5570 }
5571 }
ab455f60
NC
5572
5573 /* Cache is held with the later position first, to simplify the code
5574 that deals with unbounded ends. */
5575
5576 ASSERT_UTF8_CACHE(cache);
5577 if (cache[1] == 0) {
5578 /* Cache is totally empty */
5579 cache[0] = utf8;
5580 cache[1] = byte;
5581 } else if (cache[3] == 0) {
5582 if (byte > cache[1]) {
5583 /* New one is larger, so goes first. */
5584 cache[2] = cache[0];
5585 cache[3] = cache[1];
5586 cache[0] = utf8;
5587 cache[1] = byte;
5588 } else {
5589 cache[2] = utf8;
5590 cache[3] = byte;
5591 }
5592 } else {
5593#define THREEWAY_SQUARE(a,b,c,d) \
5594 ((float)((d) - (c))) * ((float)((d) - (c))) \
5595 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5596 + ((float)((b) - (a))) * ((float)((b) - (a)))
5597
5598 /* Cache has 2 slots in use, and we know three potential pairs.
5599 Keep the two that give the lowest RMS distance. Do the
5600 calcualation in bytes simply because we always know the byte
5601 length. squareroot has the same ordering as the positive value,
5602 so don't bother with the actual square root. */
5603 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5604 if (byte > cache[1]) {
5605 /* New position is after the existing pair of pairs. */
5606 const float keep_earlier
5607 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5608 const float keep_later
5609 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5610
5611 if (keep_later < keep_earlier) {
5612 if (keep_later < existing) {
5613 cache[2] = cache[0];
5614 cache[3] = cache[1];
5615 cache[0] = utf8;
5616 cache[1] = byte;
5617 }
5618 }
5619 else {
5620 if (keep_earlier < existing) {
5621 cache[0] = utf8;
5622 cache[1] = byte;
5623 }
5624 }
5625 }
57d7fbf1
NC
5626 else if (byte > cache[3]) {
5627 /* New position is between the existing pair of pairs. */
5628 const float keep_earlier
5629 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5630 const float keep_later
5631 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5632
5633 if (keep_later < keep_earlier) {
5634 if (keep_later < existing) {
5635 cache[2] = utf8;
5636 cache[3] = byte;
5637 }
5638 }
5639 else {
5640 if (keep_earlier < existing) {
5641 cache[0] = utf8;
5642 cache[1] = byte;
5643 }
5644 }
5645 }
5646 else {
5647 /* New position is before the existing pair of pairs. */
5648 const float keep_earlier
5649 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5650 const float keep_later
5651 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5652
5653 if (keep_later < keep_earlier) {
5654 if (keep_later < existing) {
5655 cache[2] = utf8;
5656 cache[3] = byte;
5657 }
5658 }
5659 else {
5660 if (keep_earlier < existing) {
5661 cache[0] = cache[2];
5662 cache[1] = cache[3];
5663 cache[2] = utf8;
5664 cache[3] = byte;
5665 }
5666 }
5667 }
ab455f60 5668 }
0905937d 5669 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5670}
5671
5672/* If we don't know the character offset of the end of a region, our only
5673 option is to walk forwards to the target byte offset. */
5674static STRLEN
5675S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5676{
5677 STRLEN len = 0;
5678 while (s < target) {
5679 STRLEN n = 1;
5680
5681 /* Call utf8n_to_uvchr() to validate the sequence
5682 * (unless a simple non-UTF character) */
5683 if (!UTF8_IS_INVARIANT(*s))
5684 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5685 if (n > 0) {
5686 s += n;
5687 len++;
5688 }
5689 else
5690 break;
5691 }
5692 return len;
5693}
5694
5695/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5696 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5697 backward is half the speed of walking forward. */
ec07b5e0
NC
5698static STRLEN
5699S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5700 STRLEN endu)
5701{
5702 const STRLEN forw = target - s;
5703 STRLEN backw = end - target;
5704
5705 if (forw < 2 * backw) {
5706 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5707 }
5708
5709 while (end > target) {
5710 end--;
5711 while (UTF8_IS_CONTINUATION(*end)) {
5712 end--;
5713 }
5714 endu--;
5715 }
5716 return endu;
5717}
5718
a0ed51b3 5719void
7e8c5dac 5720Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5721{
83003860 5722 const U8* s;
ec07b5e0 5723 const STRLEN byte = *offsetp;
a0ed51b3 5724 STRLEN len;
ab455f60 5725 STRLEN blen;
ec07b5e0
NC
5726 MAGIC* mg = NULL;
5727 const U8* send;
a0ed51b3
LW
5728
5729 if (!sv)
5730 return;
5731
ab455f60 5732 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5733
ab455f60 5734 if (blen < byte)
ec07b5e0 5735 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5736
ec07b5e0 5737 send = s + byte;
a67d7df9 5738
ffca234a
NC
5739 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5740 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5741 if (mg->mg_ptr) {
ec07b5e0 5742 STRLEN *cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5743 if (cache[1] == byte) {
ec07b5e0
NC
5744 /* An exact match. */
5745 *offsetp = cache[0];
ec07b5e0 5746 return;
7e8c5dac 5747 }
ab455f60
NC
5748 if (cache[3] == byte) {
5749 /* An exact match. */
5750 *offsetp = cache[2];
5751 return;
5752 }
668af93f
NC
5753
5754 if (cache[1] < byte) {
ec07b5e0 5755 /* We already know part of the way. */
b9f984a5
NC
5756 if (mg->mg_len != -1) {
5757 /* Actually, we know the end too. */
5758 len = cache[0]
5759 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5760 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5761 } else {
5762 len = cache[0]
5763 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5764 }
7e8c5dac 5765 }
9f985e4c
NC
5766 else if (cache[3] < byte) {
5767 /* We're between the two cached pairs, so we do the calculation
5768 offset by the byte/utf-8 positions for the earlier pair,
5769 then add the utf-8 characters from the string start to
5770 there. */
5771 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5772 s + cache[1], cache[0] - cache[2])
5773 + cache[2];
5774
5775 }
5776 else { /* cache[3] > byte */
5777 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5778 cache[2]);
7e8c5dac 5779
7e8c5dac 5780 }
ec07b5e0
NC
5781 ASSERT_UTF8_CACHE(cache);
5782 if (PL_utf8cache < 0) {
5783 const STRLEN reallen = S_sv_pos_b2u_forwards(aTHX_ s, send);
5784
5785 if (len != reallen) {
5786 /* Need to turn the assertions off otherwise we may recurse
5787 infinitely while printing error messages. */
5788 SAVEI8(PL_utf8cache);
5789 PL_utf8cache = 0;
5790 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5791 " real %"UVf" for %"SVf,
5792 (UV) len, (UV) reallen, sv);
5793 }
7e8c5dac 5794 }
ffca234a 5795 } else if (mg->mg_len != -1) {
ab455f60 5796 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
ec07b5e0
NC
5797 } else {
5798 len = S_sv_pos_b2u_forwards(aTHX_ s, send);
7e8c5dac 5799 }
a0ed51b3 5800 }
ec07b5e0
NC
5801 else {
5802 len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5803 }
5804 *offsetp = len;
5805
ab455f60 5806 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5807}
5808
954c1994
GS
5809/*
5810=for apidoc sv_eq
5811
5812Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5813identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5814coerce its args to strings if necessary.
954c1994
GS
5815
5816=cut
5817*/
5818
79072805 5819I32
e01b9e88 5820Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5821{
97aff369 5822 dVAR;
e1ec3a88 5823 const char *pv1;
463ee0b2 5824 STRLEN cur1;
e1ec3a88 5825 const char *pv2;
463ee0b2 5826 STRLEN cur2;
e01b9e88 5827 I32 eq = 0;
bd61b366 5828 char *tpv = NULL;
a0714e2c 5829 SV* svrecode = NULL;
79072805 5830
e01b9e88 5831 if (!sv1) {
79072805
LW
5832 pv1 = "";
5833 cur1 = 0;
5834 }
463ee0b2 5835 else
4d84ee25 5836 pv1 = SvPV_const(sv1, cur1);
79072805 5837
e01b9e88
SC
5838 if (!sv2){
5839 pv2 = "";
5840 cur2 = 0;
92d29cee 5841 }
e01b9e88 5842 else
4d84ee25 5843 pv2 = SvPV_const(sv2, cur2);
79072805 5844
cf48d248 5845 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5846 /* Differing utf8ness.
5847 * Do not UTF8size the comparands as a side-effect. */
5848 if (PL_encoding) {
5849 if (SvUTF8(sv1)) {
553e1bcc
AT
5850 svrecode = newSVpvn(pv2, cur2);
5851 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5852 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5853 }
5854 else {
553e1bcc
AT
5855 svrecode = newSVpvn(pv1, cur1);
5856 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5857 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5858 }
5859 /* Now both are in UTF-8. */
0a1bd7ac
DM
5860 if (cur1 != cur2) {
5861 SvREFCNT_dec(svrecode);
799ef3cb 5862 return FALSE;
0a1bd7ac 5863 }
799ef3cb
JH
5864 }
5865 else {
5866 bool is_utf8 = TRUE;
5867
5868 if (SvUTF8(sv1)) {
5869 /* sv1 is the UTF-8 one,
5870 * if is equal it must be downgrade-able */
9d4ba2ae 5871 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5872 &cur1, &is_utf8);
5873 if (pv != pv1)
553e1bcc 5874 pv1 = tpv = pv;
799ef3cb
JH
5875 }
5876 else {
5877 /* sv2 is the UTF-8 one,
5878 * if is equal it must be downgrade-able */
9d4ba2ae 5879 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5880 &cur2, &is_utf8);
5881 if (pv != pv2)
553e1bcc 5882 pv2 = tpv = pv;
799ef3cb
JH
5883 }
5884 if (is_utf8) {
5885 /* Downgrade not possible - cannot be eq */
bf694877 5886 assert (tpv == 0);
799ef3cb
JH
5887 return FALSE;
5888 }
5889 }
cf48d248
JH
5890 }
5891
5892 if (cur1 == cur2)
765f542d 5893 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5894
b37c2d43 5895 SvREFCNT_dec(svrecode);
553e1bcc
AT
5896 if (tpv)
5897 Safefree(tpv);
cf48d248 5898
e01b9e88 5899 return eq;
79072805
LW
5900}
5901
954c1994
GS
5902/*
5903=for apidoc sv_cmp
5904
5905Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5906string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5907C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5908coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5909
5910=cut
5911*/
5912
79072805 5913I32
e01b9e88 5914Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5915{
97aff369 5916 dVAR;
560a288e 5917 STRLEN cur1, cur2;
e1ec3a88 5918 const char *pv1, *pv2;
bd61b366 5919 char *tpv = NULL;
cf48d248 5920 I32 cmp;
a0714e2c 5921 SV *svrecode = NULL;
560a288e 5922
e01b9e88
SC
5923 if (!sv1) {
5924 pv1 = "";
560a288e
GS
5925 cur1 = 0;
5926 }
e01b9e88 5927 else
4d84ee25 5928 pv1 = SvPV_const(sv1, cur1);
560a288e 5929
553e1bcc 5930 if (!sv2) {
e01b9e88 5931 pv2 = "";
560a288e
GS
5932 cur2 = 0;
5933 }
e01b9e88 5934 else
4d84ee25 5935 pv2 = SvPV_const(sv2, cur2);
79072805 5936
cf48d248 5937 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5938 /* Differing utf8ness.
5939 * Do not UTF8size the comparands as a side-effect. */
cf48d248 5940 if (SvUTF8(sv1)) {
799ef3cb 5941 if (PL_encoding) {
553e1bcc
AT
5942 svrecode = newSVpvn(pv2, cur2);
5943 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5944 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5945 }
5946 else {
e1ec3a88 5947 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 5948 }
cf48d248
JH
5949 }
5950 else {
799ef3cb 5951 if (PL_encoding) {
553e1bcc
AT
5952 svrecode = newSVpvn(pv1, cur1);
5953 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5954 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5955 }
5956 else {
e1ec3a88 5957 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 5958 }
cf48d248
JH
5959 }
5960 }
5961
e01b9e88 5962 if (!cur1) {
cf48d248 5963 cmp = cur2 ? -1 : 0;
e01b9e88 5964 } else if (!cur2) {
cf48d248
JH
5965 cmp = 1;
5966 } else {
e1ec3a88 5967 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
5968
5969 if (retval) {
cf48d248 5970 cmp = retval < 0 ? -1 : 1;
e01b9e88 5971 } else if (cur1 == cur2) {
cf48d248
JH
5972 cmp = 0;
5973 } else {
5974 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 5975 }
cf48d248 5976 }
16660edb 5977
b37c2d43 5978 SvREFCNT_dec(svrecode);
553e1bcc
AT
5979 if (tpv)
5980 Safefree(tpv);
cf48d248
JH
5981
5982 return cmp;
bbce6d69 5983}
16660edb 5984
c461cf8f
JH
5985/*
5986=for apidoc sv_cmp_locale
5987
645c22ef
DM
5988Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
5989'use bytes' aware, handles get magic, and will coerce its args to strings
5990if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
5991
5992=cut
5993*/
5994
bbce6d69 5995I32
864dbfa3 5996Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 5997{
97aff369 5998 dVAR;
36477c24 5999#ifdef USE_LOCALE_COLLATE
16660edb 6000
bbce6d69 6001 char *pv1, *pv2;
6002 STRLEN len1, len2;
6003 I32 retval;
16660edb 6004
3280af22 6005 if (PL_collation_standard)
bbce6d69 6006 goto raw_compare;
16660edb 6007
bbce6d69 6008 len1 = 0;
8ac85365 6009 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6010 len2 = 0;
8ac85365 6011 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6012
bbce6d69 6013 if (!pv1 || !len1) {
6014 if (pv2 && len2)
6015 return -1;
6016 else
6017 goto raw_compare;
6018 }
6019 else {
6020 if (!pv2 || !len2)
6021 return 1;
6022 }
16660edb 6023
bbce6d69 6024 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6025
bbce6d69 6026 if (retval)
16660edb 6027 return retval < 0 ? -1 : 1;
6028
bbce6d69 6029 /*
6030 * When the result of collation is equality, that doesn't mean
6031 * that there are no differences -- some locales exclude some
6032 * characters from consideration. So to avoid false equalities,
6033 * we use the raw string as a tiebreaker.
6034 */
16660edb 6035
bbce6d69 6036 raw_compare:
5f66b61c 6037 /*FALLTHROUGH*/
16660edb 6038
36477c24 6039#endif /* USE_LOCALE_COLLATE */
16660edb 6040
bbce6d69 6041 return sv_cmp(sv1, sv2);
6042}
79072805 6043
645c22ef 6044
36477c24 6045#ifdef USE_LOCALE_COLLATE
645c22ef 6046
7a4c00b4 6047/*
645c22ef
DM
6048=for apidoc sv_collxfrm
6049
6050Add Collate Transform magic to an SV if it doesn't already have it.
6051
6052Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6053scalar data of the variable, but transformed to such a format that a normal
6054memory comparison can be used to compare the data according to the locale
6055settings.
6056
6057=cut
6058*/
6059
bbce6d69 6060char *
864dbfa3 6061Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6062{
97aff369 6063 dVAR;
7a4c00b4 6064 MAGIC *mg;
16660edb 6065
14befaf4 6066 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6067 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6068 const char *s;
6069 char *xf;
bbce6d69 6070 STRLEN len, xlen;
6071
7a4c00b4 6072 if (mg)
6073 Safefree(mg->mg_ptr);
93524f2b 6074 s = SvPV_const(sv, len);
bbce6d69 6075 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6076 if (SvREADONLY(sv)) {
6077 SAVEFREEPV(xf);
6078 *nxp = xlen;
3280af22 6079 return xf + sizeof(PL_collation_ix);
ff0cee69 6080 }
7a4c00b4 6081 if (! mg) {
d83f0a82
NC
6082#ifdef PERL_OLD_COPY_ON_WRITE
6083 if (SvIsCOW(sv))
6084 sv_force_normal_flags(sv, 0);
6085#endif
6086 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6087 0, 0);
7a4c00b4 6088 assert(mg);
bbce6d69 6089 }
7a4c00b4 6090 mg->mg_ptr = xf;
565764a8 6091 mg->mg_len = xlen;
7a4c00b4 6092 }
6093 else {
ff0cee69 6094 if (mg) {
6095 mg->mg_ptr = NULL;
565764a8 6096 mg->mg_len = -1;
ff0cee69 6097 }
bbce6d69 6098 }
6099 }
7a4c00b4 6100 if (mg && mg->mg_ptr) {
565764a8 6101 *nxp = mg->mg_len;
3280af22 6102 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6103 }
6104 else {
6105 *nxp = 0;
6106 return NULL;
16660edb 6107 }
79072805
LW
6108}
6109
36477c24 6110#endif /* USE_LOCALE_COLLATE */
bbce6d69 6111
c461cf8f
JH
6112/*
6113=for apidoc sv_gets
6114
6115Get a line from the filehandle and store it into the SV, optionally
6116appending to the currently-stored string.
6117
6118=cut
6119*/
6120
79072805 6121char *
864dbfa3 6122Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6123{
97aff369 6124 dVAR;
e1ec3a88 6125 const char *rsptr;
c07a80fd 6126 STRLEN rslen;
6127 register STDCHAR rslast;
6128 register STDCHAR *bp;
6129 register I32 cnt;
9c5ffd7c 6130 I32 i = 0;
8bfdd7d9 6131 I32 rspara = 0;
e311fd51 6132 I32 recsize;
c07a80fd 6133
bc44a8a2
NC
6134 if (SvTHINKFIRST(sv))
6135 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6136 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6137 from <>.
6138 However, perlbench says it's slower, because the existing swipe code
6139 is faster than copy on write.
6140 Swings and roundabouts. */
862a34c6 6141 SvUPGRADE(sv, SVt_PV);
99491443 6142
ff68c719 6143 SvSCREAM_off(sv);
efd8b2ba
AE
6144
6145 if (append) {
6146 if (PerlIO_isutf8(fp)) {
6147 if (!SvUTF8(sv)) {
6148 sv_utf8_upgrade_nomg(sv);
6149 sv_pos_u2b(sv,&append,0);
6150 }
6151 } else if (SvUTF8(sv)) {
561b68a9 6152 SV * const tsv = newSV(0);
efd8b2ba
AE
6153 sv_gets(tsv, fp, 0);
6154 sv_utf8_upgrade_nomg(tsv);
6155 SvCUR_set(sv,append);
6156 sv_catsv(sv,tsv);
6157 sv_free(tsv);
6158 goto return_string_or_null;
6159 }
6160 }
6161
6162 SvPOK_only(sv);
6163 if (PerlIO_isutf8(fp))
6164 SvUTF8_on(sv);
c07a80fd 6165
923e4eb5 6166 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6167 /* we always read code in line mode */
6168 rsptr = "\n";
6169 rslen = 1;
6170 }
6171 else if (RsSNARF(PL_rs)) {
7a5fa8a2
NIS
6172 /* If it is a regular disk file use size from stat() as estimate
6173 of amount we are going to read - may result in malloc-ing
6174 more memory than we realy need if layers bellow reduce
e468d35b
NIS
6175 size we read (e.g. CRLF or a gzip layer)
6176 */
e311fd51 6177 Stat_t st;
e468d35b 6178 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6179 const Off_t offset = PerlIO_tell(fp);
58f1856e 6180 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6181 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6182 }
6183 }
c07a80fd 6184 rsptr = NULL;
6185 rslen = 0;
6186 }
3280af22 6187 else if (RsRECORD(PL_rs)) {
e311fd51 6188 I32 bytesread;
5b2b9c68
HM
6189 char *buffer;
6190
6191 /* Grab the size of the record we're getting */
3280af22 6192 recsize = SvIV(SvRV(PL_rs));
e311fd51 6193 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6194 /* Go yank in */
6195#ifdef VMS
6196 /* VMS wants read instead of fread, because fread doesn't respect */
6197 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6198 /* doing, but we've got no other real choice - except avoid stdio
6199 as implementation - perhaps write a :vms layer ?
6200 */
5b2b9c68
HM
6201 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6202#else
6203 bytesread = PerlIO_read(fp, buffer, recsize);
6204#endif
27e6ca2d
AE
6205 if (bytesread < 0)
6206 bytesread = 0;
e311fd51 6207 SvCUR_set(sv, bytesread += append);
e670df4e 6208 buffer[bytesread] = '\0';
efd8b2ba 6209 goto return_string_or_null;
5b2b9c68 6210 }
3280af22 6211 else if (RsPARA(PL_rs)) {
c07a80fd 6212 rsptr = "\n\n";
6213 rslen = 2;
8bfdd7d9 6214 rspara = 1;
c07a80fd 6215 }
7d59b7e4
NIS
6216 else {
6217 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6218 if (PerlIO_isutf8(fp)) {
6219 rsptr = SvPVutf8(PL_rs, rslen);
6220 }
6221 else {
6222 if (SvUTF8(PL_rs)) {
6223 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6224 Perl_croak(aTHX_ "Wide character in $/");
6225 }
6226 }
93524f2b 6227 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6228 }
6229 }
6230
c07a80fd 6231 rslast = rslen ? rsptr[rslen - 1] : '\0';
6232
8bfdd7d9 6233 if (rspara) { /* have to do this both before and after */
79072805 6234 do { /* to make sure file boundaries work right */
760ac839 6235 if (PerlIO_eof(fp))
a0d0e21e 6236 return 0;
760ac839 6237 i = PerlIO_getc(fp);
79072805 6238 if (i != '\n') {
a0d0e21e
LW
6239 if (i == -1)
6240 return 0;
760ac839 6241 PerlIO_ungetc(fp,i);
79072805
LW
6242 break;
6243 }
6244 } while (i != EOF);
6245 }
c07a80fd 6246
760ac839
LW
6247 /* See if we know enough about I/O mechanism to cheat it ! */
6248
6249 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6250 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6251 enough here - and may even be a macro allowing compile
6252 time optimization.
6253 */
6254
6255 if (PerlIO_fast_gets(fp)) {
6256
6257 /*
6258 * We're going to steal some values from the stdio struct
6259 * and put EVERYTHING in the innermost loop into registers.
6260 */
6261 register STDCHAR *ptr;
6262 STRLEN bpx;
6263 I32 shortbuffered;
6264
16660edb 6265#if defined(VMS) && defined(PERLIO_IS_STDIO)
6266 /* An ungetc()d char is handled separately from the regular
6267 * buffer, so we getc() it back out and stuff it in the buffer.
6268 */
6269 i = PerlIO_getc(fp);
6270 if (i == EOF) return 0;
6271 *(--((*fp)->_ptr)) = (unsigned char) i;
6272 (*fp)->_cnt++;
6273#endif
c07a80fd 6274
c2960299 6275 /* Here is some breathtakingly efficient cheating */
c07a80fd 6276
a20bf0c3 6277 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6278 /* make sure we have the room */
7a5fa8a2 6279 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6280 /* Not room for all of it
7a5fa8a2 6281 if we are looking for a separator and room for some
e468d35b
NIS
6282 */
6283 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6284 /* just process what we have room for */
79072805
LW
6285 shortbuffered = cnt - SvLEN(sv) + append + 1;
6286 cnt -= shortbuffered;
6287 }
6288 else {
6289 shortbuffered = 0;
bbce6d69 6290 /* remember that cnt can be negative */
eb160463 6291 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6292 }
6293 }
7a5fa8a2 6294 else
79072805 6295 shortbuffered = 0;
3f7c398e 6296 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6297 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6298 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6299 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6300 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6301 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6302 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6303 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6304 for (;;) {
6305 screamer:
93a17b20 6306 if (cnt > 0) {
c07a80fd 6307 if (rslen) {
760ac839
LW
6308 while (cnt > 0) { /* this | eat */
6309 cnt--;
c07a80fd 6310 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6311 goto thats_all_folks; /* screams | sed :-) */
6312 }
6313 }
6314 else {
1c846c1f
NIS
6315 Copy(ptr, bp, cnt, char); /* this | eat */
6316 bp += cnt; /* screams | dust */
c07a80fd 6317 ptr += cnt; /* louder | sed :-) */
a5f75d66 6318 cnt = 0;
93a17b20 6319 }
79072805
LW
6320 }
6321
748a9306 6322 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6323 cnt = shortbuffered;
6324 shortbuffered = 0;
3f7c398e 6325 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6326 SvCUR_set(sv, bpx);
6327 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6328 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6329 continue;
6330 }
6331
16660edb 6332 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6333 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6334 PTR2UV(ptr),(long)cnt));
cc00df79 6335 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6336#if 0
16660edb 6337 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6338 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6339 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6340 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6341#endif
1c846c1f 6342 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6343 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6344 another abstraction. */
760ac839 6345 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6346#if 0
16660edb 6347 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6348 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6349 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6350 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6351#endif
a20bf0c3
JH
6352 cnt = PerlIO_get_cnt(fp);
6353 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6354 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6355 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6356
748a9306
LW
6357 if (i == EOF) /* all done for ever? */
6358 goto thats_really_all_folks;
6359
3f7c398e 6360 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6361 SvCUR_set(sv, bpx);
6362 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6363 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6364
eb160463 6365 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6366
c07a80fd 6367 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6368 goto thats_all_folks;
79072805
LW
6369 }
6370
6371thats_all_folks:
3f7c398e 6372 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6373 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6374 goto screamer; /* go back to the fray */
79072805
LW
6375thats_really_all_folks:
6376 if (shortbuffered)
6377 cnt += shortbuffered;
16660edb 6378 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6379 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6380 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6381 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6382 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6383 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6384 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6385 *bp = '\0';
3f7c398e 6386 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6387 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6388 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6389 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6390 }
6391 else
79072805 6392 {
6edd2cd5 6393 /*The big, slow, and stupid way. */
27da23d5 6394#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6395 STDCHAR *buf = NULL;
a02a5408 6396 Newx(buf, 8192, STDCHAR);
6edd2cd5 6397 assert(buf);
4d2c4e07 6398#else
6edd2cd5 6399 STDCHAR buf[8192];
4d2c4e07 6400#endif
79072805 6401
760ac839 6402screamer2:
c07a80fd 6403 if (rslen) {
00b6aa41 6404 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6405 bp = buf;
eb160463 6406 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6407 ; /* keep reading */
6408 cnt = bp - buf;
c07a80fd 6409 }
6410 else {
760ac839 6411 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6412 /* Accomodate broken VAXC compiler, which applies U8 cast to
6413 * both args of ?: operator, causing EOF to change into 255
6414 */
37be0adf 6415 if (cnt > 0)
cbe9e203
JH
6416 i = (U8)buf[cnt - 1];
6417 else
37be0adf 6418 i = EOF;
c07a80fd 6419 }
79072805 6420
cbe9e203
JH
6421 if (cnt < 0)
6422 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6423 if (append)
6424 sv_catpvn(sv, (char *) buf, cnt);
6425 else
6426 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6427
6428 if (i != EOF && /* joy */
6429 (!rslen ||
6430 SvCUR(sv) < rslen ||
3f7c398e 6431 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6432 {
6433 append = -1;
63e4d877
CS
6434 /*
6435 * If we're reading from a TTY and we get a short read,
6436 * indicating that the user hit his EOF character, we need
6437 * to notice it now, because if we try to read from the TTY
6438 * again, the EOF condition will disappear.
6439 *
6440 * The comparison of cnt to sizeof(buf) is an optimization
6441 * that prevents unnecessary calls to feof().
6442 *
6443 * - jik 9/25/96
6444 */
6445 if (!(cnt < sizeof(buf) && PerlIO_eof(fp)))
6446 goto screamer2;
79072805 6447 }
6edd2cd5 6448
27da23d5 6449#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6450 Safefree(buf);
6451#endif
79072805
LW
6452 }
6453
8bfdd7d9 6454 if (rspara) { /* have to do this both before and after */
c07a80fd 6455 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6456 i = PerlIO_getc(fp);
79072805 6457 if (i != '\n') {
760ac839 6458 PerlIO_ungetc(fp,i);
79072805
LW
6459 break;
6460 }
6461 }
6462 }
c07a80fd 6463
efd8b2ba 6464return_string_or_null:
bd61b366 6465 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6466}
6467
954c1994
GS
6468/*
6469=for apidoc sv_inc
6470
645c22ef
DM
6471Auto-increment of the value in the SV, doing string to numeric conversion
6472if necessary. Handles 'get' magic.
954c1994
GS
6473
6474=cut
6475*/
6476
79072805 6477void
864dbfa3 6478Perl_sv_inc(pTHX_ register SV *sv)
79072805 6479{
97aff369 6480 dVAR;
79072805 6481 register char *d;
463ee0b2 6482 int flags;
79072805
LW
6483
6484 if (!sv)
6485 return;
5b295bef 6486 SvGETMAGIC(sv);
ed6116ce 6487 if (SvTHINKFIRST(sv)) {
765f542d
NC
6488 if (SvIsCOW(sv))
6489 sv_force_normal_flags(sv, 0);
0f15f207 6490 if (SvREADONLY(sv)) {
923e4eb5 6491 if (IN_PERL_RUNTIME)
cea2e8a9 6492 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6493 }
a0d0e21e 6494 if (SvROK(sv)) {
b5be31e9 6495 IV i;
9e7bc3e8
JD
6496 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6497 return;
56431972 6498 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6499 sv_unref(sv);
6500 sv_setiv(sv, i);
a0d0e21e 6501 }
ed6116ce 6502 }
8990e307 6503 flags = SvFLAGS(sv);
28e5dec8
JH
6504 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6505 /* It's (privately or publicly) a float, but not tested as an
6506 integer, so test it to see. */
d460ef45 6507 (void) SvIV(sv);
28e5dec8
JH
6508 flags = SvFLAGS(sv);
6509 }
6510 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6511 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6512#ifdef PERL_PRESERVE_IVUV
28e5dec8 6513 oops_its_int:
59d8ce62 6514#endif
25da4f38
IZ
6515 if (SvIsUV(sv)) {
6516 if (SvUVX(sv) == UV_MAX)
a1e868e7 6517 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6518 else
6519 (void)SvIOK_only_UV(sv);
607fa7f2 6520 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6521 } else {
6522 if (SvIVX(sv) == IV_MAX)
28e5dec8 6523 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6524 else {
6525 (void)SvIOK_only(sv);
45977657 6526 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6527 }
55497cff 6528 }
79072805
LW
6529 return;
6530 }
28e5dec8
JH
6531 if (flags & SVp_NOK) {
6532 (void)SvNOK_only(sv);
9d6ce603 6533 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6534 return;
6535 }
6536
3f7c398e 6537 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6538 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6539 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6540 (void)SvIOK_only(sv);
45977657 6541 SvIV_set(sv, 1);
79072805
LW
6542 return;
6543 }
463ee0b2 6544 d = SvPVX(sv);
79072805
LW
6545 while (isALPHA(*d)) d++;
6546 while (isDIGIT(*d)) d++;
6547 if (*d) {
28e5dec8 6548#ifdef PERL_PRESERVE_IVUV
d1be9408 6549 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6550 warnings. Probably ought to make the sv_iv_please() that does
6551 the conversion if possible, and silently. */
504618e9 6552 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6553 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6554 /* Need to try really hard to see if it's an integer.
6555 9.22337203685478e+18 is an integer.
6556 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6557 so $a="9.22337203685478e+18"; $a+0; $a++
6558 needs to be the same as $a="9.22337203685478e+18"; $a++
6559 or we go insane. */
d460ef45 6560
28e5dec8
JH
6561 (void) sv_2iv(sv);
6562 if (SvIOK(sv))
6563 goto oops_its_int;
6564
6565 /* sv_2iv *should* have made this an NV */
6566 if (flags & SVp_NOK) {
6567 (void)SvNOK_only(sv);
9d6ce603 6568 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6569 return;
6570 }
6571 /* I don't think we can get here. Maybe I should assert this
6572 And if we do get here I suspect that sv_setnv will croak. NWC
6573 Fall through. */
6574#if defined(USE_LONG_DOUBLE)
6575 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 6576 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6577#else
1779d84d 6578 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 6579 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6580#endif
6581 }
6582#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6583 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6584 return;
6585 }
6586 d--;
3f7c398e 6587 while (d >= SvPVX_const(sv)) {
79072805
LW
6588 if (isDIGIT(*d)) {
6589 if (++*d <= '9')
6590 return;
6591 *(d--) = '0';
6592 }
6593 else {
9d116dd7
JH
6594#ifdef EBCDIC
6595 /* MKS: The original code here died if letters weren't consecutive.
6596 * at least it didn't have to worry about non-C locales. The
6597 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6598 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6599 * [A-Za-z] are accepted by isALPHA in the C locale.
6600 */
6601 if (*d != 'z' && *d != 'Z') {
6602 do { ++*d; } while (!isALPHA(*d));
6603 return;
6604 }
6605 *(d--) -= 'z' - 'a';
6606#else
79072805
LW
6607 ++*d;
6608 if (isALPHA(*d))
6609 return;
6610 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6611#endif
79072805
LW
6612 }
6613 }
6614 /* oh,oh, the number grew */
6615 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6616 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6617 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6618 *d = d[-1];
6619 if (isDIGIT(d[1]))
6620 *d = '1';
6621 else
6622 *d = d[1];
6623}
6624
954c1994
GS
6625/*
6626=for apidoc sv_dec
6627
645c22ef
DM
6628Auto-decrement of the value in the SV, doing string to numeric conversion
6629if necessary. Handles 'get' magic.
954c1994
GS
6630
6631=cut
6632*/
6633
79072805 6634void
864dbfa3 6635Perl_sv_dec(pTHX_ register SV *sv)
79072805 6636{
97aff369 6637 dVAR;
463ee0b2
LW
6638 int flags;
6639
79072805
LW
6640 if (!sv)
6641 return;
5b295bef 6642 SvGETMAGIC(sv);
ed6116ce 6643 if (SvTHINKFIRST(sv)) {
765f542d
NC
6644 if (SvIsCOW(sv))
6645 sv_force_normal_flags(sv, 0);
0f15f207 6646 if (SvREADONLY(sv)) {
923e4eb5 6647 if (IN_PERL_RUNTIME)
cea2e8a9 6648 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6649 }
a0d0e21e 6650 if (SvROK(sv)) {
b5be31e9 6651 IV i;
9e7bc3e8
JD
6652 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6653 return;
56431972 6654 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6655 sv_unref(sv);
6656 sv_setiv(sv, i);
a0d0e21e 6657 }
ed6116ce 6658 }
28e5dec8
JH
6659 /* Unlike sv_inc we don't have to worry about string-never-numbers
6660 and keeping them magic. But we mustn't warn on punting */
8990e307 6661 flags = SvFLAGS(sv);
28e5dec8
JH
6662 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6663 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6664#ifdef PERL_PRESERVE_IVUV
28e5dec8 6665 oops_its_int:
59d8ce62 6666#endif
25da4f38
IZ
6667 if (SvIsUV(sv)) {
6668 if (SvUVX(sv) == 0) {
6669 (void)SvIOK_only(sv);
45977657 6670 SvIV_set(sv, -1);
25da4f38
IZ
6671 }
6672 else {
6673 (void)SvIOK_only_UV(sv);
f4eee32f 6674 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6675 }
25da4f38
IZ
6676 } else {
6677 if (SvIVX(sv) == IV_MIN)
65202027 6678 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6679 else {
6680 (void)SvIOK_only(sv);
45977657 6681 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6682 }
55497cff 6683 }
6684 return;
6685 }
28e5dec8 6686 if (flags & SVp_NOK) {
9d6ce603 6687 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6688 (void)SvNOK_only(sv);
6689 return;
6690 }
8990e307 6691 if (!(flags & SVp_POK)) {
ef088171
NC
6692 if ((flags & SVTYPEMASK) < SVt_PVIV)
6693 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6694 SvIV_set(sv, -1);
6695 (void)SvIOK_only(sv);
79072805
LW
6696 return;
6697 }
28e5dec8
JH
6698#ifdef PERL_PRESERVE_IVUV
6699 {
504618e9 6700 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6701 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6702 /* Need to try really hard to see if it's an integer.
6703 9.22337203685478e+18 is an integer.
6704 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6705 so $a="9.22337203685478e+18"; $a+0; $a--
6706 needs to be the same as $a="9.22337203685478e+18"; $a--
6707 or we go insane. */
d460ef45 6708
28e5dec8
JH
6709 (void) sv_2iv(sv);
6710 if (SvIOK(sv))
6711 goto oops_its_int;
6712
6713 /* sv_2iv *should* have made this an NV */
6714 if (flags & SVp_NOK) {
6715 (void)SvNOK_only(sv);
9d6ce603 6716 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6717 return;
6718 }
6719 /* I don't think we can get here. Maybe I should assert this
6720 And if we do get here I suspect that sv_setnv will croak. NWC
6721 Fall through. */
6722#if defined(USE_LONG_DOUBLE)
6723 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 6724 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6725#else
1779d84d 6726 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 6727 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6728#endif
6729 }
6730 }
6731#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6732 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6733}
6734
954c1994
GS
6735/*
6736=for apidoc sv_mortalcopy
6737
645c22ef 6738Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6739The new SV is marked as mortal. It will be destroyed "soon", either by an
6740explicit call to FREETMPS, or by an implicit call at places such as
6741statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6742
6743=cut
6744*/
6745
79072805
LW
6746/* Make a string that will exist for the duration of the expression
6747 * evaluation. Actually, it may have to last longer than that, but
6748 * hopefully we won't free it until it has been assigned to a
6749 * permanent location. */
6750
6751SV *
864dbfa3 6752Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6753{
97aff369 6754 dVAR;
463ee0b2 6755 register SV *sv;
b881518d 6756
4561caa4 6757 new_SV(sv);
79072805 6758 sv_setsv(sv,oldstr);
677b06e3
GS
6759 EXTEND_MORTAL(1);
6760 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6761 SvTEMP_on(sv);
6762 return sv;
6763}
6764
954c1994
GS
6765/*
6766=for apidoc sv_newmortal
6767
645c22ef 6768Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6769set to 1. It will be destroyed "soon", either by an explicit call to
6770FREETMPS, or by an implicit call at places such as statement boundaries.
6771See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6772
6773=cut
6774*/
6775
8990e307 6776SV *
864dbfa3 6777Perl_sv_newmortal(pTHX)
8990e307 6778{
97aff369 6779 dVAR;
8990e307
LW
6780 register SV *sv;
6781
4561caa4 6782 new_SV(sv);
8990e307 6783 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6784 EXTEND_MORTAL(1);
6785 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6786 return sv;
6787}
6788
954c1994
GS
6789/*
6790=for apidoc sv_2mortal
6791
d4236ebc
DM
6792Marks an existing SV as mortal. The SV will be destroyed "soon", either
6793by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6794statement boundaries. SvTEMP() is turned on which means that the SV's
6795string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6796and C<sv_mortalcopy>.
954c1994
GS
6797
6798=cut
6799*/
6800
79072805 6801SV *
864dbfa3 6802Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6803{
27da23d5 6804 dVAR;
79072805 6805 if (!sv)
7a5b473e 6806 return NULL;
d689ffdd 6807 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6808 return sv;
677b06e3
GS
6809 EXTEND_MORTAL(1);
6810 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6811 SvTEMP_on(sv);
79072805
LW
6812 return sv;
6813}
6814
954c1994
GS
6815/*
6816=for apidoc newSVpv
6817
6818Creates a new SV and copies a string into it. The reference count for the
6819SV is set to 1. If C<len> is zero, Perl will compute the length using
6820strlen(). For efficiency, consider using C<newSVpvn> instead.
6821
6822=cut
6823*/
6824
79072805 6825SV *
864dbfa3 6826Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6827{
97aff369 6828 dVAR;
463ee0b2 6829 register SV *sv;
79072805 6830
4561caa4 6831 new_SV(sv);
616d8c9c 6832 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6833 return sv;
6834}
6835
954c1994
GS
6836/*
6837=for apidoc newSVpvn
6838
6839Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6840SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6841string. You are responsible for ensuring that the source string is at least
9e09f5f2 6842C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6843
6844=cut
6845*/
6846
9da1e3b5 6847SV *
864dbfa3 6848Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6849{
97aff369 6850 dVAR;
9da1e3b5
MUN
6851 register SV *sv;
6852
6853 new_SV(sv);
9da1e3b5
MUN
6854 sv_setpvn(sv,s,len);
6855 return sv;
6856}
6857
bd08039b
NC
6858
6859/*
926f8064 6860=for apidoc newSVhek
bd08039b
NC
6861
6862Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6863point to the shared string table where possible. Returns a new (undefined)
6864SV if the hek is NULL.
bd08039b
NC
6865
6866=cut
6867*/
6868
6869SV *
c1b02ed8 6870Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6871{
97aff369 6872 dVAR;
5aaec2b4
NC
6873 if (!hek) {
6874 SV *sv;
6875
6876 new_SV(sv);
6877 return sv;
6878 }
6879
bd08039b
NC
6880 if (HEK_LEN(hek) == HEf_SVKEY) {
6881 return newSVsv(*(SV**)HEK_KEY(hek));
6882 } else {
6883 const int flags = HEK_FLAGS(hek);
6884 if (flags & HVhek_WASUTF8) {
6885 /* Trouble :-)
6886 Andreas would like keys he put in as utf8 to come back as utf8
6887 */
6888 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6889 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6890 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6891
6892 SvUTF8_on (sv);
6893 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6894 return sv;
6895 } else if (flags & HVhek_REHASH) {
6896 /* We don't have a pointer to the hv, so we have to replicate the
6897 flag into every HEK. This hv is using custom a hasing
6898 algorithm. Hence we can't return a shared string scalar, as
6899 that would contain the (wrong) hash value, and might get passed
6900 into an hv routine with a regular hash */
6901
b64e5050 6902 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6903 if (HEK_UTF8(hek))
6904 SvUTF8_on (sv);
6905 return sv;
6906 }
6907 /* This will be overwhelminly the most common case. */
6908 return newSVpvn_share(HEK_KEY(hek),
6909 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
6910 HEK_HASH(hek));
6911 }
6912}
6913
1c846c1f
NIS
6914/*
6915=for apidoc newSVpvn_share
6916
3f7c398e 6917Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
6918table. If the string does not already exist in the table, it is created
6919first. Turns on READONLY and FAKE. The string's hash is stored in the UV
6920slot of the SV; if the C<hash> parameter is non-zero, that value is used;
6921otherwise the hash is computed. The idea here is that as the string table
3f7c398e 6922is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 6923hash lookup will avoid string compare.
1c846c1f
NIS
6924
6925=cut
6926*/
6927
6928SV *
c3654f1a 6929Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 6930{
97aff369 6931 dVAR;
1c846c1f 6932 register SV *sv;
c3654f1a
IH
6933 bool is_utf8 = FALSE;
6934 if (len < 0) {
77caf834 6935 STRLEN tmplen = -len;
c3654f1a 6936 is_utf8 = TRUE;
75a54232 6937 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 6938 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
6939 len = tmplen;
6940 }
1c846c1f 6941 if (!hash)
5afd6d42 6942 PERL_HASH(hash, src, len);
1c846c1f 6943 new_SV(sv);
bdd68bc3 6944 sv_upgrade(sv, SVt_PV);
f880fe2f 6945 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 6946 SvCUR_set(sv, len);
b162af07 6947 SvLEN_set(sv, 0);
1c846c1f
NIS
6948 SvREADONLY_on(sv);
6949 SvFAKE_on(sv);
6950 SvPOK_on(sv);
c3654f1a
IH
6951 if (is_utf8)
6952 SvUTF8_on(sv);
1c846c1f
NIS
6953 return sv;
6954}
6955
645c22ef 6956
cea2e8a9 6957#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
6958
6959/* pTHX_ magic can't cope with varargs, so this is a no-context
6960 * version of the main function, (which may itself be aliased to us).
6961 * Don't access this version directly.
6962 */
6963
46fc3d4c 6964SV *
cea2e8a9 6965Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 6966{
cea2e8a9 6967 dTHX;
46fc3d4c 6968 register SV *sv;
6969 va_list args;
46fc3d4c 6970 va_start(args, pat);
c5be433b 6971 sv = vnewSVpvf(pat, &args);
46fc3d4c 6972 va_end(args);
6973 return sv;
6974}
cea2e8a9 6975#endif
46fc3d4c 6976
954c1994
GS
6977/*
6978=for apidoc newSVpvf
6979
645c22ef 6980Creates a new SV and initializes it with the string formatted like
954c1994
GS
6981C<sprintf>.
6982
6983=cut
6984*/
6985
cea2e8a9
GS
6986SV *
6987Perl_newSVpvf(pTHX_ const char* pat, ...)
6988{
6989 register SV *sv;
6990 va_list args;
cea2e8a9 6991 va_start(args, pat);
c5be433b 6992 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
6993 va_end(args);
6994 return sv;
6995}
46fc3d4c 6996
645c22ef
DM
6997/* backend for newSVpvf() and newSVpvf_nocontext() */
6998
79072805 6999SV *
c5be433b
GS
7000Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7001{
97aff369 7002 dVAR;
c5be433b
GS
7003 register SV *sv;
7004 new_SV(sv);
4608196e 7005 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7006 return sv;
7007}
7008
954c1994
GS
7009/*
7010=for apidoc newSVnv
7011
7012Creates a new SV and copies a floating point value into it.
7013The reference count for the SV is set to 1.
7014
7015=cut
7016*/
7017
c5be433b 7018SV *
65202027 7019Perl_newSVnv(pTHX_ NV n)
79072805 7020{
97aff369 7021 dVAR;
463ee0b2 7022 register SV *sv;
79072805 7023
4561caa4 7024 new_SV(sv);
79072805
LW
7025 sv_setnv(sv,n);
7026 return sv;
7027}
7028
954c1994
GS
7029/*
7030=for apidoc newSViv
7031
7032Creates a new SV and copies an integer into it. The reference count for the
7033SV is set to 1.
7034
7035=cut
7036*/
7037
79072805 7038SV *
864dbfa3 7039Perl_newSViv(pTHX_ IV i)
79072805 7040{
97aff369 7041 dVAR;
463ee0b2 7042 register SV *sv;
79072805 7043
4561caa4 7044 new_SV(sv);
79072805
LW
7045 sv_setiv(sv,i);
7046 return sv;
7047}
7048
954c1994 7049/*
1a3327fb
JH
7050=for apidoc newSVuv
7051
7052Creates a new SV and copies an unsigned integer into it.
7053The reference count for the SV is set to 1.
7054
7055=cut
7056*/
7057
7058SV *
7059Perl_newSVuv(pTHX_ UV u)
7060{
97aff369 7061 dVAR;
1a3327fb
JH
7062 register SV *sv;
7063
7064 new_SV(sv);
7065 sv_setuv(sv,u);
7066 return sv;
7067}
7068
7069/*
954c1994
GS
7070=for apidoc newRV_noinc
7071
7072Creates an RV wrapper for an SV. The reference count for the original
7073SV is B<not> incremented.
7074
7075=cut
7076*/
7077
2304df62 7078SV *
864dbfa3 7079Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7080{
97aff369 7081 dVAR;
2304df62
AD
7082 register SV *sv;
7083
4561caa4 7084 new_SV(sv);
2304df62 7085 sv_upgrade(sv, SVt_RV);
76e3520e 7086 SvTEMP_off(tmpRef);
b162af07 7087 SvRV_set(sv, tmpRef);
2304df62 7088 SvROK_on(sv);
2304df62
AD
7089 return sv;
7090}
7091
ff276b08 7092/* newRV_inc is the official function name to use now.
645c22ef
DM
7093 * newRV_inc is in fact #defined to newRV in sv.h
7094 */
7095
5f05dabc 7096SV *
7f466ec7 7097Perl_newRV(pTHX_ SV *sv)
5f05dabc 7098{
97aff369 7099 dVAR;
7f466ec7 7100 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7101}
5f05dabc 7102
954c1994
GS
7103/*
7104=for apidoc newSVsv
7105
7106Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7107(Uses C<sv_setsv>).
954c1994
GS
7108
7109=cut
7110*/
7111
79072805 7112SV *
864dbfa3 7113Perl_newSVsv(pTHX_ register SV *old)
79072805 7114{
97aff369 7115 dVAR;
463ee0b2 7116 register SV *sv;
79072805
LW
7117
7118 if (!old)
7a5b473e 7119 return NULL;
8990e307 7120 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7121 if (ckWARN_d(WARN_INTERNAL))
9014280d 7122 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7123 return NULL;
79072805 7124 }
4561caa4 7125 new_SV(sv);
e90aabeb
NC
7126 /* SV_GMAGIC is the default for sv_setv()
7127 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7128 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7129 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7130 return sv;
79072805
LW
7131}
7132
645c22ef
DM
7133/*
7134=for apidoc sv_reset
7135
7136Underlying implementation for the C<reset> Perl function.
7137Note that the perl-level function is vaguely deprecated.
7138
7139=cut
7140*/
7141
79072805 7142void
e1ec3a88 7143Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7144{
27da23d5 7145 dVAR;
4802d5d7 7146 char todo[PERL_UCHAR_MAX+1];
79072805 7147
49d8d3a1
MB
7148 if (!stash)
7149 return;
7150
79072805 7151 if (!*s) { /* reset ?? searches */
aec46f14 7152 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7153 if (mg) {
7154 PMOP *pm = (PMOP *) mg->mg_obj;
7155 while (pm) {
7156 pm->op_pmdynflags &= ~PMdf_USED;
7157 pm = pm->op_pmnext;
7158 }
79072805
LW
7159 }
7160 return;
7161 }
7162
7163 /* reset variables */
7164
7165 if (!HvARRAY(stash))
7166 return;
463ee0b2
LW
7167
7168 Zero(todo, 256, char);
79072805 7169 while (*s) {
b464bac0
AL
7170 I32 max;
7171 I32 i = (unsigned char)*s;
79072805
LW
7172 if (s[1] == '-') {
7173 s += 2;
7174 }
4802d5d7 7175 max = (unsigned char)*s++;
79072805 7176 for ( ; i <= max; i++) {
463ee0b2
LW
7177 todo[i] = 1;
7178 }
a0d0e21e 7179 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7180 HE *entry;
79072805 7181 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7182 entry;
7183 entry = HeNEXT(entry))
7184 {
b464bac0
AL
7185 register GV *gv;
7186 register SV *sv;
7187
1edc1566 7188 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7189 continue;
1edc1566 7190 gv = (GV*)HeVAL(entry);
79072805 7191 sv = GvSV(gv);
e203899d
NC
7192 if (sv) {
7193 if (SvTHINKFIRST(sv)) {
7194 if (!SvREADONLY(sv) && SvROK(sv))
7195 sv_unref(sv);
7196 /* XXX Is this continue a bug? Why should THINKFIRST
7197 exempt us from resetting arrays and hashes? */
7198 continue;
7199 }
7200 SvOK_off(sv);
7201 if (SvTYPE(sv) >= SVt_PV) {
7202 SvCUR_set(sv, 0);
bd61b366 7203 if (SvPVX_const(sv) != NULL)
e203899d
NC
7204 *SvPVX(sv) = '\0';
7205 SvTAINT(sv);
7206 }
79072805
LW
7207 }
7208 if (GvAV(gv)) {
7209 av_clear(GvAV(gv));
7210 }
bfcb3514 7211 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7212#if defined(VMS)
7213 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7214#else /* ! VMS */
463ee0b2 7215 hv_clear(GvHV(gv));
b0269e46
AB
7216# if defined(USE_ENVIRON_ARRAY)
7217 if (gv == PL_envgv)
7218 my_clearenv();
7219# endif /* USE_ENVIRON_ARRAY */
7220#endif /* VMS */
79072805
LW
7221 }
7222 }
7223 }
7224 }
7225}
7226
645c22ef
DM
7227/*
7228=for apidoc sv_2io
7229
7230Using various gambits, try to get an IO from an SV: the IO slot if its a
7231GV; or the recursive result if we're an RV; or the IO slot of the symbol
7232named after the PV if we're a string.
7233
7234=cut
7235*/
7236
46fc3d4c 7237IO*
864dbfa3 7238Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7239{
7240 IO* io;
7241 GV* gv;
7242
7243 switch (SvTYPE(sv)) {
7244 case SVt_PVIO:
7245 io = (IO*)sv;
7246 break;
7247 case SVt_PVGV:
7248 gv = (GV*)sv;
7249 io = GvIO(gv);
7250 if (!io)
cea2e8a9 7251 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7252 break;
7253 default:
7254 if (!SvOK(sv))
cea2e8a9 7255 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7256 if (SvROK(sv))
7257 return sv_2io(SvRV(sv));
f776e3cd 7258 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7259 if (gv)
7260 io = GvIO(gv);
7261 else
7262 io = 0;
7263 if (!io)
35c1215d 7264 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7265 break;
7266 }
7267 return io;
7268}
7269
645c22ef
DM
7270/*
7271=for apidoc sv_2cv
7272
7273Using various gambits, try to get a CV from an SV; in addition, try if
7274possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7275The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7276
7277=cut
7278*/
7279
79072805 7280CV *
864dbfa3 7281Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7282{
27da23d5 7283 dVAR;
a0714e2c 7284 GV *gv = NULL;
601f1833 7285 CV *cv = NULL;
79072805 7286
85dec29a
NC
7287 if (!sv) {
7288 *st = NULL;
7289 *gvp = NULL;
7290 return NULL;
7291 }
79072805 7292 switch (SvTYPE(sv)) {
79072805
LW
7293 case SVt_PVCV:
7294 *st = CvSTASH(sv);
a0714e2c 7295 *gvp = NULL;
79072805
LW
7296 return (CV*)sv;
7297 case SVt_PVHV:
7298 case SVt_PVAV:
ef58ba18 7299 *st = NULL;
a0714e2c 7300 *gvp = NULL;
601f1833 7301 return NULL;
8990e307
LW
7302 case SVt_PVGV:
7303 gv = (GV*)sv;
a0d0e21e 7304 *gvp = gv;
8990e307
LW
7305 *st = GvESTASH(gv);
7306 goto fix_gv;
7307
79072805 7308 default:
5b295bef 7309 SvGETMAGIC(sv);
a0d0e21e 7310 if (SvROK(sv)) {
823a54a3 7311 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7312 tryAMAGICunDEREF(to_cv);
7313
62f274bf
GS
7314 sv = SvRV(sv);
7315 if (SvTYPE(sv) == SVt_PVCV) {
7316 cv = (CV*)sv;
a0714e2c 7317 *gvp = NULL;
62f274bf
GS
7318 *st = CvSTASH(cv);
7319 return cv;
7320 }
7321 else if(isGV(sv))
7322 gv = (GV*)sv;
7323 else
cea2e8a9 7324 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7325 }
62f274bf 7326 else if (isGV(sv))
79072805
LW
7327 gv = (GV*)sv;
7328 else
7a5fd60d 7329 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7330 *gvp = gv;
ef58ba18
NC
7331 if (!gv) {
7332 *st = NULL;
601f1833 7333 return NULL;
ef58ba18 7334 }
e26df76a
NC
7335 /* Some flags to gv_fetchsv mean don't really create the GV */
7336 if (SvTYPE(gv) != SVt_PVGV) {
7337 *st = NULL;
7338 return NULL;
7339 }
79072805 7340 *st = GvESTASH(gv);
8990e307 7341 fix_gv:
8ebc5c01 7342 if (lref && !GvCVu(gv)) {
4633a7c4 7343 SV *tmpsv;
748a9306 7344 ENTER;
561b68a9 7345 tmpsv = newSV(0);
bd61b366 7346 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7347 /* XXX this is probably not what they think they're getting.
7348 * It has the same effect as "sub name;", i.e. just a forward
7349 * declaration! */
774d564b 7350 newSUB(start_subparse(FALSE, 0),
4633a7c4 7351 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7352 NULL, NULL);
748a9306 7353 LEAVE;
8ebc5c01 7354 if (!GvCVu(gv))
35c1215d
NC
7355 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7356 sv);
8990e307 7357 }
8ebc5c01 7358 return GvCVu(gv);
79072805
LW
7359 }
7360}
7361
c461cf8f
JH
7362/*
7363=for apidoc sv_true
7364
7365Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7366Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7367instead use an in-line version.
c461cf8f
JH
7368
7369=cut
7370*/
7371
79072805 7372I32
864dbfa3 7373Perl_sv_true(pTHX_ register SV *sv)
79072805 7374{
8990e307
LW
7375 if (!sv)
7376 return 0;
79072805 7377 if (SvPOK(sv)) {
823a54a3
AL
7378 register const XPV* const tXpv = (XPV*)SvANY(sv);
7379 if (tXpv &&
c2f1de04 7380 (tXpv->xpv_cur > 1 ||
339049b0 7381 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7382 return 1;
7383 else
7384 return 0;
7385 }
7386 else {
7387 if (SvIOK(sv))
463ee0b2 7388 return SvIVX(sv) != 0;
79072805
LW
7389 else {
7390 if (SvNOK(sv))
463ee0b2 7391 return SvNVX(sv) != 0.0;
79072805 7392 else
463ee0b2 7393 return sv_2bool(sv);
79072805
LW
7394 }
7395 }
7396}
79072805 7397
645c22ef 7398/*
c461cf8f
JH
7399=for apidoc sv_pvn_force
7400
7401Get a sensible string out of the SV somehow.
645c22ef
DM
7402A private implementation of the C<SvPV_force> macro for compilers which
7403can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7404
8d6d96c1
HS
7405=for apidoc sv_pvn_force_flags
7406
7407Get a sensible string out of the SV somehow.
7408If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7409appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7410implemented in terms of this function.
645c22ef
DM
7411You normally want to use the various wrapper macros instead: see
7412C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7413
7414=cut
7415*/
7416
7417char *
7418Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7419{
97aff369 7420 dVAR;
6fc92669 7421 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7422 sv_force_normal_flags(sv, 0);
1c846c1f 7423
a0d0e21e 7424 if (SvPOK(sv)) {
13c5b33c
NC
7425 if (lp)
7426 *lp = SvCUR(sv);
a0d0e21e
LW
7427 }
7428 else {
a3b680e6 7429 char *s;
13c5b33c
NC
7430 STRLEN len;
7431
4d84ee25 7432 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7433 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7434 if (PL_op)
7435 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7436 ref, OP_NAME(PL_op));
4d84ee25 7437 else
b64e5050 7438 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7439 }
b64e5050 7440 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7441 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7442 OP_NAME(PL_op));
b64e5050 7443 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7444 if (lp)
7445 *lp = len;
7446
3f7c398e 7447 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7448 if (SvROK(sv))
7449 sv_unref(sv);
862a34c6 7450 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7451 SvGROW(sv, len + 1);
706aa1c9 7452 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7453 SvCUR_set(sv, len);
7454 *SvEND(sv) = '\0';
7455 }
7456 if (!SvPOK(sv)) {
7457 SvPOK_on(sv); /* validate pointer */
7458 SvTAINT(sv);
1d7c1841 7459 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7460 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7461 }
7462 }
4d84ee25 7463 return SvPVX_mutable(sv);
a0d0e21e
LW
7464}
7465
645c22ef 7466/*
645c22ef
DM
7467=for apidoc sv_pvbyten_force
7468
0feed65a 7469The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7470
7471=cut
7472*/
7473
7340a771
GS
7474char *
7475Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7476{
46ec2f14 7477 sv_pvn_force(sv,lp);
ffebcc3e 7478 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7479 *lp = SvCUR(sv);
7480 return SvPVX(sv);
7340a771
GS
7481}
7482
645c22ef 7483/*
c461cf8f
JH
7484=for apidoc sv_pvutf8n_force
7485
0feed65a 7486The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7487
7488=cut
7489*/
7490
7340a771
GS
7491char *
7492Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7493{
46ec2f14 7494 sv_pvn_force(sv,lp);
560a288e 7495 sv_utf8_upgrade(sv);
46ec2f14
TS
7496 *lp = SvCUR(sv);
7497 return SvPVX(sv);
7340a771
GS
7498}
7499
c461cf8f
JH
7500/*
7501=for apidoc sv_reftype
7502
7503Returns a string describing what the SV is a reference to.
7504
7505=cut
7506*/
7507
1cb0ed9b 7508char *
bfed75c6 7509Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7510{
07409e01
NC
7511 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7512 inside return suggests a const propagation bug in g++. */
c86bf373 7513 if (ob && SvOBJECT(sv)) {
1b6737cc 7514 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7515 return name ? name : (char *) "__ANON__";
c86bf373 7516 }
a0d0e21e
LW
7517 else {
7518 switch (SvTYPE(sv)) {
7519 case SVt_NULL:
7520 case SVt_IV:
7521 case SVt_NV:
7522 case SVt_RV:
7523 case SVt_PV:
7524 case SVt_PVIV:
7525 case SVt_PVNV:
7526 case SVt_PVMG:
7527 case SVt_PVBM:
1cb0ed9b 7528 if (SvVOK(sv))
439cb1c4 7529 return "VSTRING";
a0d0e21e
LW
7530 if (SvROK(sv))
7531 return "REF";
7532 else
7533 return "SCALAR";
1cb0ed9b 7534
07409e01 7535 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7536 /* tied lvalues should appear to be
7537 * scalars for backwards compatitbility */
7538 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7539 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7540 case SVt_PVAV: return "ARRAY";
7541 case SVt_PVHV: return "HASH";
7542 case SVt_PVCV: return "CODE";
7543 case SVt_PVGV: return "GLOB";
1d2dff63 7544 case SVt_PVFM: return "FORMAT";
27f9d8f3 7545 case SVt_PVIO: return "IO";
a0d0e21e
LW
7546 default: return "UNKNOWN";
7547 }
7548 }
7549}
7550
954c1994
GS
7551/*
7552=for apidoc sv_isobject
7553
7554Returns a boolean indicating whether the SV is an RV pointing to a blessed
7555object. If the SV is not an RV, or if the object is not blessed, then this
7556will return false.
7557
7558=cut
7559*/
7560
463ee0b2 7561int
864dbfa3 7562Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7563{
68dc0745 7564 if (!sv)
7565 return 0;
5b295bef 7566 SvGETMAGIC(sv);
85e6fe83
LW
7567 if (!SvROK(sv))
7568 return 0;
7569 sv = (SV*)SvRV(sv);
7570 if (!SvOBJECT(sv))
7571 return 0;
7572 return 1;
7573}
7574
954c1994
GS
7575/*
7576=for apidoc sv_isa
7577
7578Returns a boolean indicating whether the SV is blessed into the specified
7579class. This does not check for subtypes; use C<sv_derived_from> to verify
7580an inheritance relationship.
7581
7582=cut
7583*/
7584
85e6fe83 7585int
864dbfa3 7586Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7587{
bfcb3514 7588 const char *hvname;
68dc0745 7589 if (!sv)
7590 return 0;
5b295bef 7591 SvGETMAGIC(sv);
ed6116ce 7592 if (!SvROK(sv))
463ee0b2 7593 return 0;
ed6116ce
LW
7594 sv = (SV*)SvRV(sv);
7595 if (!SvOBJECT(sv))
463ee0b2 7596 return 0;
bfcb3514
NC
7597 hvname = HvNAME_get(SvSTASH(sv));
7598 if (!hvname)
e27ad1f2 7599 return 0;
463ee0b2 7600
bfcb3514 7601 return strEQ(hvname, name);
463ee0b2
LW
7602}
7603
954c1994
GS
7604/*
7605=for apidoc newSVrv
7606
7607Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7608it will be upgraded to one. If C<classname> is non-null then the new SV will
7609be blessed in the specified package. The new SV is returned and its
7610reference count is 1.
7611
7612=cut
7613*/
7614
463ee0b2 7615SV*
864dbfa3 7616Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7617{
97aff369 7618 dVAR;
463ee0b2
LW
7619 SV *sv;
7620
4561caa4 7621 new_SV(sv);
51cf62d8 7622
765f542d 7623 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7624 SvAMAGIC_off(rv);
51cf62d8 7625
0199fce9 7626 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7627 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7628 SvREFCNT(rv) = 0;
7629 sv_clear(rv);
7630 SvFLAGS(rv) = 0;
7631 SvREFCNT(rv) = refcnt;
7632 }
7633
51cf62d8 7634 if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7635 sv_upgrade(rv, SVt_RV);
7636 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7637 SvPV_free(rv);
0199fce9
JD
7638 SvCUR_set(rv, 0);
7639 SvLEN_set(rv, 0);
7640 }
51cf62d8 7641
0c34ef67 7642 SvOK_off(rv);
b162af07 7643 SvRV_set(rv, sv);
ed6116ce 7644 SvROK_on(rv);
463ee0b2 7645
a0d0e21e 7646 if (classname) {
1b6737cc 7647 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7648 (void)sv_bless(rv, stash);
7649 }
7650 return sv;
7651}
7652
954c1994
GS
7653/*
7654=for apidoc sv_setref_pv
7655
7656Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7657argument will be upgraded to an RV. That RV will be modified to point to
7658the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7659into the SV. The C<classname> argument indicates the package for the
bd61b366 7660blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7661will have a reference count of 1, and the RV will be returned.
954c1994
GS
7662
7663Do not use with other Perl types such as HV, AV, SV, CV, because those
7664objects will become corrupted by the pointer copy process.
7665
7666Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7667
7668=cut
7669*/
7670
a0d0e21e 7671SV*
864dbfa3 7672Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7673{
97aff369 7674 dVAR;
189b2af5 7675 if (!pv) {
3280af22 7676 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7677 SvSETMAGIC(rv);
7678 }
a0d0e21e 7679 else
56431972 7680 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7681 return rv;
7682}
7683
954c1994
GS
7684/*
7685=for apidoc sv_setref_iv
7686
7687Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7688argument will be upgraded to an RV. That RV will be modified to point to
7689the new SV. The C<classname> argument indicates the package for the
bd61b366 7690blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7691will have a reference count of 1, and the RV will be returned.
954c1994
GS
7692
7693=cut
7694*/
7695
a0d0e21e 7696SV*
864dbfa3 7697Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7698{
7699 sv_setiv(newSVrv(rv,classname), iv);
7700 return rv;
7701}
7702
954c1994 7703/*
e1c57cef
JH
7704=for apidoc sv_setref_uv
7705
7706Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7707argument will be upgraded to an RV. That RV will be modified to point to
7708the new SV. The C<classname> argument indicates the package for the
bd61b366 7709blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7710will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7711
7712=cut
7713*/
7714
7715SV*
7716Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7717{
7718 sv_setuv(newSVrv(rv,classname), uv);
7719 return rv;
7720}
7721
7722/*
954c1994
GS
7723=for apidoc sv_setref_nv
7724
7725Copies a double into a new SV, optionally blessing the SV. The C<rv>
7726argument will be upgraded to an RV. That RV will be modified to point to
7727the new SV. The C<classname> argument indicates the package for the
bd61b366 7728blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7729will have a reference count of 1, and the RV will be returned.
954c1994
GS
7730
7731=cut
7732*/
7733
a0d0e21e 7734SV*
65202027 7735Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7736{
7737 sv_setnv(newSVrv(rv,classname), nv);
7738 return rv;
7739}
463ee0b2 7740
954c1994
GS
7741/*
7742=for apidoc sv_setref_pvn
7743
7744Copies a string into a new SV, optionally blessing the SV. The length of the
7745string must be specified with C<n>. The C<rv> argument will be upgraded to
7746an RV. That RV will be modified to point to the new SV. The C<classname>
7747argument indicates the package for the blessing. Set C<classname> to
bd61b366 7748C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7749of 1, and the RV will be returned.
954c1994
GS
7750
7751Note that C<sv_setref_pv> copies the pointer while this copies the string.
7752
7753=cut
7754*/
7755
a0d0e21e 7756SV*
1b6737cc 7757Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7758{
7759 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7760 return rv;
7761}
7762
954c1994
GS
7763/*
7764=for apidoc sv_bless
7765
7766Blesses an SV into a specified package. The SV must be an RV. The package
7767must be designated by its stash (see C<gv_stashpv()>). The reference count
7768of the SV is unaffected.
7769
7770=cut
7771*/
7772
a0d0e21e 7773SV*
864dbfa3 7774Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7775{
97aff369 7776 dVAR;
76e3520e 7777 SV *tmpRef;
a0d0e21e 7778 if (!SvROK(sv))
cea2e8a9 7779 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7780 tmpRef = SvRV(sv);
7781 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7782 if (SvREADONLY(tmpRef))
cea2e8a9 7783 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7784 if (SvOBJECT(tmpRef)) {
7785 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7786 --PL_sv_objcount;
76e3520e 7787 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7788 }
a0d0e21e 7789 }
76e3520e
GS
7790 SvOBJECT_on(tmpRef);
7791 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7792 ++PL_sv_objcount;
862a34c6 7793 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7794 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7795
2e3febc6
CS
7796 if (Gv_AMG(stash))
7797 SvAMAGIC_on(sv);
7798 else
7799 SvAMAGIC_off(sv);
a0d0e21e 7800
1edbfb88
AB
7801 if(SvSMAGICAL(tmpRef))
7802 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7803 mg_set(tmpRef);
7804
7805
ecdeb87c 7806
a0d0e21e
LW
7807 return sv;
7808}
7809
645c22ef 7810/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7811 */
7812
76e3520e 7813STATIC void
cea2e8a9 7814S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7815{
97aff369 7816 dVAR;
850fabdf 7817 void *xpvmg;
b37c2d43 7818 SV * const temp = sv_newmortal();
850fabdf 7819
a0d0e21e
LW
7820 assert(SvTYPE(sv) == SVt_PVGV);
7821 SvFAKE_off(sv);
180488f8
NC
7822 gv_efullname3(temp, (GV *) sv, "*");
7823
f7877b28 7824 if (GvGP(sv)) {
1edc1566 7825 gp_free((GV*)sv);
f7877b28 7826 }
e826b3c7 7827 if (GvSTASH(sv)) {
e15faf7d 7828 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7829 GvSTASH(sv) = NULL;
e826b3c7 7830 }
a5f75d66 7831 GvMULTI_off(sv);
acda4c6a
NC
7832 if (GvNAME_HEK(sv)) {
7833 unshare_hek(GvNAME_HEK(sv));
7834 }
dedf8e73 7835 SvSCREAM_off(sv);
850fabdf
GS
7836
7837 /* need to keep SvANY(sv) in the right arena */
7838 xpvmg = new_XPVMG();
7839 StructCopy(SvANY(sv), xpvmg, XPVMG);
7840 del_XPVGV(SvANY(sv));
7841 SvANY(sv) = xpvmg;
7842
a0d0e21e
LW
7843 SvFLAGS(sv) &= ~SVTYPEMASK;
7844 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7845
7846 /* Intentionally not calling any local SET magic, as this isn't so much a
7847 set operation as merely an internal storage change. */
7848 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7849}
7850
954c1994 7851/*
840a7b70 7852=for apidoc sv_unref_flags
954c1994
GS
7853
7854Unsets the RV status of the SV, and decrements the reference count of
7855whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7856as a reversal of C<newSVrv>. The C<cflags> argument can contain
7857C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7858(otherwise the decrementing is conditional on the reference count being
7859different from one or the reference being a readonly SV).
7889fe52 7860See C<SvROK_off>.
954c1994
GS
7861
7862=cut
7863*/
7864
ed6116ce 7865void
e15faf7d 7866Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7867{
b64e5050 7868 SV* const target = SvRV(ref);
810b8aa5 7869
e15faf7d
NC
7870 if (SvWEAKREF(ref)) {
7871 sv_del_backref(target, ref);
7872 SvWEAKREF_off(ref);
7873 SvRV_set(ref, NULL);
810b8aa5
GS
7874 return;
7875 }
e15faf7d
NC
7876 SvRV_set(ref, NULL);
7877 SvROK_off(ref);
7878 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7879 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7880 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7881 SvREFCNT_dec(target);
840a7b70 7882 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7883 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7884}
8990e307 7885
840a7b70 7886/*
645c22ef
DM
7887=for apidoc sv_untaint
7888
7889Untaint an SV. Use C<SvTAINTED_off> instead.
7890=cut
7891*/
7892
bbce6d69 7893void
864dbfa3 7894Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7895{
13f57bf8 7896 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 7897 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 7898 if (mg)
565764a8 7899 mg->mg_len &= ~1;
36477c24 7900 }
bbce6d69 7901}
7902
645c22ef
DM
7903/*
7904=for apidoc sv_tainted
7905
7906Test an SV for taintedness. Use C<SvTAINTED> instead.
7907=cut
7908*/
7909
bbce6d69 7910bool
864dbfa3 7911Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 7912{
13f57bf8 7913 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 7914 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 7915 if (mg && (mg->mg_len & 1) )
36477c24 7916 return TRUE;
7917 }
7918 return FALSE;
bbce6d69 7919}
7920
09540bc3
JH
7921/*
7922=for apidoc sv_setpviv
7923
7924Copies an integer into the given SV, also updating its string value.
7925Does not handle 'set' magic. See C<sv_setpviv_mg>.
7926
7927=cut
7928*/
7929
7930void
7931Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
7932{
7933 char buf[TYPE_CHARS(UV)];
7934 char *ebuf;
b64e5050 7935 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
7936
7937 sv_setpvn(sv, ptr, ebuf - ptr);
7938}
7939
7940/*
7941=for apidoc sv_setpviv_mg
7942
7943Like C<sv_setpviv>, but also handles 'set' magic.
7944
7945=cut
7946*/
7947
7948void
7949Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
7950{
df7eb254 7951 sv_setpviv(sv, iv);
09540bc3
JH
7952 SvSETMAGIC(sv);
7953}
7954
cea2e8a9 7955#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7956
7957/* pTHX_ magic can't cope with varargs, so this is a no-context
7958 * version of the main function, (which may itself be aliased to us).
7959 * Don't access this version directly.
7960 */
7961
cea2e8a9
GS
7962void
7963Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
7964{
7965 dTHX;
7966 va_list args;
7967 va_start(args, pat);
c5be433b 7968 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
7969 va_end(args);
7970}
7971
645c22ef
DM
7972/* pTHX_ magic can't cope with varargs, so this is a no-context
7973 * version of the main function, (which may itself be aliased to us).
7974 * Don't access this version directly.
7975 */
cea2e8a9
GS
7976
7977void
7978Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
7979{
7980 dTHX;
7981 va_list args;
7982 va_start(args, pat);
c5be433b 7983 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 7984 va_end(args);
cea2e8a9
GS
7985}
7986#endif
7987
954c1994
GS
7988/*
7989=for apidoc sv_setpvf
7990
bffc3d17
SH
7991Works like C<sv_catpvf> but copies the text into the SV instead of
7992appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
7993
7994=cut
7995*/
7996
46fc3d4c 7997void
864dbfa3 7998Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 7999{
8000 va_list args;
46fc3d4c 8001 va_start(args, pat);
c5be433b 8002 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8003 va_end(args);
8004}
8005
bffc3d17
SH
8006/*
8007=for apidoc sv_vsetpvf
8008
8009Works like C<sv_vcatpvf> but copies the text into the SV instead of
8010appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8011
8012Usually used via its frontend C<sv_setpvf>.
8013
8014=cut
8015*/
645c22ef 8016
c5be433b
GS
8017void
8018Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8019{
4608196e 8020 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8021}
ef50df4b 8022
954c1994
GS
8023/*
8024=for apidoc sv_setpvf_mg
8025
8026Like C<sv_setpvf>, but also handles 'set' magic.
8027
8028=cut
8029*/
8030
ef50df4b 8031void
864dbfa3 8032Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8033{
8034 va_list args;
ef50df4b 8035 va_start(args, pat);
c5be433b 8036 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8037 va_end(args);
c5be433b
GS
8038}
8039
bffc3d17
SH
8040/*
8041=for apidoc sv_vsetpvf_mg
8042
8043Like C<sv_vsetpvf>, but also handles 'set' magic.
8044
8045Usually used via its frontend C<sv_setpvf_mg>.
8046
8047=cut
8048*/
645c22ef 8049
c5be433b
GS
8050void
8051Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8052{
4608196e 8053 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8054 SvSETMAGIC(sv);
8055}
8056
cea2e8a9 8057#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8058
8059/* pTHX_ magic can't cope with varargs, so this is a no-context
8060 * version of the main function, (which may itself be aliased to us).
8061 * Don't access this version directly.
8062 */
8063
cea2e8a9
GS
8064void
8065Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8066{
8067 dTHX;
8068 va_list args;
8069 va_start(args, pat);
c5be433b 8070 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8071 va_end(args);
8072}
8073
645c22ef
DM
8074/* pTHX_ magic can't cope with varargs, so this is a no-context
8075 * version of the main function, (which may itself be aliased to us).
8076 * Don't access this version directly.
8077 */
8078
cea2e8a9
GS
8079void
8080Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8081{
8082 dTHX;
8083 va_list args;
8084 va_start(args, pat);
c5be433b 8085 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8086 va_end(args);
cea2e8a9
GS
8087}
8088#endif
8089
954c1994
GS
8090/*
8091=for apidoc sv_catpvf
8092
d5ce4a7c
GA
8093Processes its arguments like C<sprintf> and appends the formatted
8094output to an SV. If the appended data contains "wide" characters
8095(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8096and characters >255 formatted with %c), the original SV might get
bffc3d17 8097upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8098C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8099valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8100
d5ce4a7c 8101=cut */
954c1994 8102
46fc3d4c 8103void
864dbfa3 8104Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8105{
8106 va_list args;
46fc3d4c 8107 va_start(args, pat);
c5be433b 8108 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8109 va_end(args);
8110}
8111
bffc3d17
SH
8112/*
8113=for apidoc sv_vcatpvf
8114
8115Processes its arguments like C<vsprintf> and appends the formatted output
8116to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8117
8118Usually used via its frontend C<sv_catpvf>.
8119
8120=cut
8121*/
645c22ef 8122
ef50df4b 8123void
c5be433b
GS
8124Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8125{
4608196e 8126 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8127}
8128
954c1994
GS
8129/*
8130=for apidoc sv_catpvf_mg
8131
8132Like C<sv_catpvf>, but also handles 'set' magic.
8133
8134=cut
8135*/
8136
c5be433b 8137void
864dbfa3 8138Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8139{
8140 va_list args;
ef50df4b 8141 va_start(args, pat);
c5be433b 8142 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8143 va_end(args);
c5be433b
GS
8144}
8145
bffc3d17
SH
8146/*
8147=for apidoc sv_vcatpvf_mg
8148
8149Like C<sv_vcatpvf>, but also handles 'set' magic.
8150
8151Usually used via its frontend C<sv_catpvf_mg>.
8152
8153=cut
8154*/
645c22ef 8155
c5be433b
GS
8156void
8157Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8158{
4608196e 8159 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8160 SvSETMAGIC(sv);
8161}
8162
954c1994
GS
8163/*
8164=for apidoc sv_vsetpvfn
8165
bffc3d17 8166Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8167appending it.
8168
bffc3d17 8169Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8170
954c1994
GS
8171=cut
8172*/
8173
46fc3d4c 8174void
7d5ea4e7 8175Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8176{
8177 sv_setpvn(sv, "", 0);
7d5ea4e7 8178 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8179}
8180
2d00ba3b 8181STATIC I32
9dd79c3f 8182S_expect_number(pTHX_ char** pattern)
211dfcf1 8183{
97aff369 8184 dVAR;
211dfcf1
HS
8185 I32 var = 0;
8186 switch (**pattern) {
8187 case '1': case '2': case '3':
8188 case '4': case '5': case '6':
8189 case '7': case '8': case '9':
2fba7546
GA
8190 var = *(*pattern)++ - '0';
8191 while (isDIGIT(**pattern)) {
5f66b61c 8192 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8193 if (tmp < var)
8194 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8195 var = tmp;
8196 }
211dfcf1
HS
8197 }
8198 return var;
8199}
211dfcf1 8200
c445ea15
AL
8201STATIC char *
8202S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8203{
a3b680e6 8204 const int neg = nv < 0;
4151a5fe 8205 UV uv;
4151a5fe
IZ
8206
8207 if (neg)
8208 nv = -nv;
8209 if (nv < UV_MAX) {
b464bac0 8210 char *p = endbuf;
4151a5fe 8211 nv += 0.5;
028f8eaa 8212 uv = (UV)nv;
4151a5fe
IZ
8213 if (uv & 1 && uv == nv)
8214 uv--; /* Round to even */
8215 do {
a3b680e6 8216 const unsigned dig = uv % 10;
4151a5fe
IZ
8217 *--p = '0' + dig;
8218 } while (uv /= 10);
8219 if (neg)
8220 *--p = '-';
8221 *len = endbuf - p;
8222 return p;
8223 }
bd61b366 8224 return NULL;
4151a5fe
IZ
8225}
8226
8227
954c1994
GS
8228/*
8229=for apidoc sv_vcatpvfn
8230
8231Processes its arguments like C<vsprintf> and appends the formatted output
8232to an SV. Uses an array of SVs if the C style variable argument list is
8233missing (NULL). When running with taint checks enabled, indicates via
8234C<maybe_tainted> if results are untrustworthy (often due to the use of
8235locales).
8236
bffc3d17 8237Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8238
954c1994
GS
8239=cut
8240*/
8241
8896765a
RB
8242
8243#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8244 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8245 vec_utf8 = DO_UTF8(vecsv);
8246
1ef29b0e
RGS
8247/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8248
46fc3d4c 8249void
7d5ea4e7 8250Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8251{
97aff369 8252 dVAR;
46fc3d4c 8253 char *p;
8254 char *q;
a3b680e6 8255 const char *patend;
fc36a67e 8256 STRLEN origlen;
46fc3d4c 8257 I32 svix = 0;
27da23d5 8258 static const char nullstr[] = "(null)";
a0714e2c 8259 SV *argsv = NULL;
b464bac0
AL
8260 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8261 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8262 SV *nsv = NULL;
4151a5fe
IZ
8263 /* Times 4: a decimal digit takes more than 3 binary digits.
8264 * NV_DIG: mantissa takes than many decimal digits.
8265 * Plus 32: Playing safe. */
8266 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8267 /* large enough for "%#.#f" --chip */
8268 /* what about long double NVs? --jhi */
db79b45b 8269
53c1dcc0
AL
8270 PERL_UNUSED_ARG(maybe_tainted);
8271
46fc3d4c 8272 /* no matter what, this is a string now */
fc36a67e 8273 (void)SvPV_force(sv, origlen);
46fc3d4c 8274
8896765a 8275 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8276 if (patlen == 0)
8277 return;
0dbb1585 8278 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8279 if (args) {
8280 const char * const s = va_arg(*args, char*);
8281 sv_catpv(sv, s ? s : nullstr);
8282 }
8283 else if (svix < svmax) {
8284 sv_catsv(sv, *svargs);
2d03de9c
AL
8285 }
8286 return;
0dbb1585 8287 }
8896765a
RB
8288 if (args && patlen == 3 && pat[0] == '%' &&
8289 pat[1] == '-' && pat[2] == 'p') {
8290 argsv = va_arg(*args, SV*);
8291 sv_catsv(sv, argsv);
8896765a 8292 return;
46fc3d4c 8293 }
8294
1d917b39 8295#ifndef USE_LONG_DOUBLE
4151a5fe 8296 /* special-case "%.<number>[gf]" */
7af36d83 8297 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8298 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8299 unsigned digits = 0;
8300 const char *pp;
8301
8302 pp = pat + 2;
8303 while (*pp >= '0' && *pp <= '9')
8304 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8305 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8306 NV nv;
8307
7af36d83 8308 if (svix < svmax)
4151a5fe
IZ
8309 nv = SvNV(*svargs);
8310 else
8311 return;
8312 if (*pp == 'g') {
2873255c
NC
8313 /* Add check for digits != 0 because it seems that some
8314 gconverts are buggy in this case, and we don't yet have
8315 a Configure test for this. */
8316 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8317 /* 0, point, slack */
2e59c212 8318 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8319 sv_catpv(sv, ebuf);
8320 if (*ebuf) /* May return an empty string for digits==0 */
8321 return;
8322 }
8323 } else if (!digits) {
8324 STRLEN l;
8325
8326 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8327 sv_catpvn(sv, p, l);
8328 return;
8329 }
8330 }
8331 }
8332 }
1d917b39 8333#endif /* !USE_LONG_DOUBLE */
4151a5fe 8334
2cf2cfc6 8335 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8336 has_utf8 = TRUE;
2cf2cfc6 8337
46fc3d4c 8338 patend = (char*)pat + patlen;
8339 for (p = (char*)pat; p < patend; p = q) {
8340 bool alt = FALSE;
8341 bool left = FALSE;
b22c7a20 8342 bool vectorize = FALSE;
211dfcf1 8343 bool vectorarg = FALSE;
2cf2cfc6 8344 bool vec_utf8 = FALSE;
46fc3d4c 8345 char fill = ' ';
8346 char plus = 0;
8347 char intsize = 0;
8348 STRLEN width = 0;
fc36a67e 8349 STRLEN zeros = 0;
46fc3d4c 8350 bool has_precis = FALSE;
8351 STRLEN precis = 0;
c445ea15 8352 const I32 osvix = svix;
2cf2cfc6 8353 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8354#ifdef HAS_LDBL_SPRINTF_BUG
8355 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8356 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8357 bool fix_ldbl_sprintf_bug = FALSE;
8358#endif
205f51d8 8359
46fc3d4c 8360 char esignbuf[4];
89ebb4a3 8361 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8362 STRLEN esignlen = 0;
8363
bd61b366 8364 const char *eptr = NULL;
fc36a67e 8365 STRLEN elen = 0;
a0714e2c 8366 SV *vecsv = NULL;
4608196e 8367 const U8 *vecstr = NULL;
b22c7a20 8368 STRLEN veclen = 0;
934abaf1 8369 char c = 0;
46fc3d4c 8370 int i;
9c5ffd7c 8371 unsigned base = 0;
8c8eb53c
RB
8372 IV iv = 0;
8373 UV uv = 0;
9e5b023a
JH
8374 /* we need a long double target in case HAS_LONG_DOUBLE but
8375 not USE_LONG_DOUBLE
8376 */
35fff930 8377#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8378 long double nv;
8379#else
65202027 8380 NV nv;
9e5b023a 8381#endif
46fc3d4c 8382 STRLEN have;
8383 STRLEN need;
8384 STRLEN gap;
7af36d83 8385 const char *dotstr = ".";
b22c7a20 8386 STRLEN dotstrlen = 1;
211dfcf1 8387 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8388 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8389 I32 epix = 0; /* explicit precision index */
8390 I32 evix = 0; /* explicit vector index */
eb3fce90 8391 bool asterisk = FALSE;
46fc3d4c 8392
211dfcf1 8393 /* echo everything up to the next format specification */
46fc3d4c 8394 for (q = p; q < patend && *q != '%'; ++q) ;
8395 if (q > p) {
db79b45b
JH
8396 if (has_utf8 && !pat_utf8)
8397 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8398 else
8399 sv_catpvn(sv, p, q - p);
46fc3d4c 8400 p = q;
8401 }
8402 if (q++ >= patend)
8403 break;
8404
211dfcf1
HS
8405/*
8406 We allow format specification elements in this order:
8407 \d+\$ explicit format parameter index
8408 [-+ 0#]+ flags
a472f209 8409 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8410 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8411 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8412 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8413 [hlqLV] size
8896765a
RB
8414 [%bcdefginopsuxDFOUX] format (mandatory)
8415*/
8416
8417 if (args) {
8418/*
8419 As of perl5.9.3, printf format checking is on by default.
8420 Internally, perl uses %p formats to provide an escape to
8421 some extended formatting. This block deals with those
8422 extensions: if it does not match, (char*)q is reset and
8423 the normal format processing code is used.
8424
8425 Currently defined extensions are:
8426 %p include pointer address (standard)
8427 %-p (SVf) include an SV (previously %_)
8428 %-<num>p include an SV with precision <num>
8429 %1p (VDf) include a v-string (as %vd)
8430 %<num>p reserved for future extensions
8431
8432 Robin Barker 2005-07-14
211dfcf1 8433*/
8896765a
RB
8434 char* r = q;
8435 bool sv = FALSE;
8436 STRLEN n = 0;
8437 if (*q == '-')
8438 sv = *q++;
c445ea15 8439 n = expect_number(&q);
8896765a
RB
8440 if (*q++ == 'p') {
8441 if (sv) { /* SVf */
8442 if (n) {
8443 precis = n;
8444 has_precis = TRUE;
8445 }
8446 argsv = va_arg(*args, SV*);
8447 eptr = SvPVx_const(argsv, elen);
8448 if (DO_UTF8(argsv))
8449 is_utf8 = TRUE;
8450 goto string;
8451 }
8452#if vdNUMBER
8453 else if (n == vdNUMBER) { /* VDf */
8454 vectorize = TRUE;
8455 VECTORIZE_ARGS
8456 goto format_vd;
8457 }
8458#endif
8459 else if (n) {
8460 if (ckWARN_d(WARN_INTERNAL))
8461 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8462 "internal %%<num>p might conflict with future printf extensions");
8463 }
8464 }
8465 q = r;
8466 }
8467
c445ea15 8468 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8469 if (*q == '$') {
8470 ++q;
8471 efix = width;
8472 } else {
8473 goto gotwidth;
8474 }
8475 }
8476
fc36a67e 8477 /* FLAGS */
8478
46fc3d4c 8479 while (*q) {
8480 switch (*q) {
8481 case ' ':
8482 case '+':
8483 plus = *q++;
8484 continue;
8485
8486 case '-':
8487 left = TRUE;
8488 q++;
8489 continue;
8490
8491 case '0':
8492 fill = *q++;
8493 continue;
8494
8495 case '#':
8496 alt = TRUE;
8497 q++;
8498 continue;
8499
fc36a67e 8500 default:
8501 break;
8502 }
8503 break;
8504 }
46fc3d4c 8505
211dfcf1 8506 tryasterisk:
eb3fce90 8507 if (*q == '*') {
211dfcf1 8508 q++;
c445ea15 8509 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8510 if (*q++ != '$')
8511 goto unknown;
eb3fce90 8512 asterisk = TRUE;
211dfcf1
HS
8513 }
8514 if (*q == 'v') {
eb3fce90 8515 q++;
211dfcf1
HS
8516 if (vectorize)
8517 goto unknown;
9cbac4c7 8518 if ((vectorarg = asterisk)) {
211dfcf1
HS
8519 evix = ewix;
8520 ewix = 0;
8521 asterisk = FALSE;
8522 }
8523 vectorize = TRUE;
8524 goto tryasterisk;
eb3fce90
JH
8525 }
8526
211dfcf1 8527 if (!asterisk)
858a90f9 8528 {
7a5fa8a2 8529 if( *q == '0' )
f3583277 8530 fill = *q++;
c445ea15 8531 width = expect_number(&q);
858a90f9 8532 }
211dfcf1
HS
8533
8534 if (vectorize) {
8535 if (vectorarg) {
8536 if (args)
8537 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8538 else if (evix) {
8539 vecsv = (evix > 0 && evix <= svmax)
8540 ? svargs[evix-1] : &PL_sv_undef;
8541 } else {
8542 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8543 }
245d4a47 8544 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8545 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8546 bad with tied or overloaded values that return UTF8. */
211dfcf1 8547 if (DO_UTF8(vecsv))
2cf2cfc6 8548 is_utf8 = TRUE;
640283f5
NC
8549 else if (has_utf8) {
8550 vecsv = sv_mortalcopy(vecsv);
8551 sv_utf8_upgrade(vecsv);
8552 dotstr = SvPV_const(vecsv, dotstrlen);
8553 is_utf8 = TRUE;
8554 }
211dfcf1
HS
8555 }
8556 if (args) {
8896765a 8557 VECTORIZE_ARGS
eb3fce90 8558 }
7ad96abb 8559 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8560 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8561 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8562 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8563
8564 /* if this is a version object, we need to convert
8565 * back into v-string notation and then let the
8566 * vectorize happen normally
d7aa5382 8567 */
96b8f7ce
JP
8568 if (sv_derived_from(vecsv, "version")) {
8569 char *version = savesvpv(vecsv);
34ba6322
SP
8570 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8571 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8572 "vector argument not supported with alpha versions");
8573 goto unknown;
8574 }
96b8f7ce
JP
8575 vecsv = sv_newmortal();
8576 /* scan_vstring is expected to be called during
8577 * tokenization, so we need to fake up the end
8578 * of the buffer for it
8579 */
8580 PL_bufend = version + veclen;
8581 scan_vstring(version, vecsv);
8582 vecstr = (U8*)SvPV_const(vecsv, veclen);
8583 vec_utf8 = DO_UTF8(vecsv);
8584 Safefree(version);
d7aa5382 8585 }
211dfcf1
HS
8586 }
8587 else {
8588 vecstr = (U8*)"";
8589 veclen = 0;
8590 }
eb3fce90 8591 }
fc36a67e 8592
eb3fce90 8593 if (asterisk) {
fc36a67e 8594 if (args)
8595 i = va_arg(*args, int);
8596 else
eb3fce90
JH
8597 i = (ewix ? ewix <= svmax : svix < svmax) ?
8598 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8599 left |= (i < 0);
8600 width = (i < 0) ? -i : i;
fc36a67e 8601 }
211dfcf1 8602 gotwidth:
fc36a67e 8603
8604 /* PRECISION */
46fc3d4c 8605
fc36a67e 8606 if (*q == '.') {
8607 q++;
8608 if (*q == '*') {
211dfcf1 8609 q++;
c445ea15 8610 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8611 goto unknown;
8612 /* XXX: todo, support specified precision parameter */
8613 if (epix)
211dfcf1 8614 goto unknown;
46fc3d4c 8615 if (args)
8616 i = va_arg(*args, int);
8617 else
eb3fce90
JH
8618 i = (ewix ? ewix <= svmax : svix < svmax)
8619 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8620 precis = (i < 0) ? 0 : i;
fc36a67e 8621 }
8622 else {
8623 precis = 0;
8624 while (isDIGIT(*q))
8625 precis = precis * 10 + (*q++ - '0');
8626 }
8627 has_precis = TRUE;
8628 }
46fc3d4c 8629
fc36a67e 8630 /* SIZE */
46fc3d4c 8631
fc36a67e 8632 switch (*q) {
c623ac67
GS
8633#ifdef WIN32
8634 case 'I': /* Ix, I32x, and I64x */
8635# ifdef WIN64
8636 if (q[1] == '6' && q[2] == '4') {
8637 q += 3;
8638 intsize = 'q';
8639 break;
8640 }
8641# endif
8642 if (q[1] == '3' && q[2] == '2') {
8643 q += 3;
8644 break;
8645 }
8646# ifdef WIN64
8647 intsize = 'q';
8648# endif
8649 q++;
8650 break;
8651#endif
9e5b023a 8652#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8653 case 'L': /* Ld */
5f66b61c 8654 /*FALLTHROUGH*/
e5c81feb 8655#ifdef HAS_QUAD
6f9bb7fd 8656 case 'q': /* qd */
9e5b023a 8657#endif
6f9bb7fd
GS
8658 intsize = 'q';
8659 q++;
8660 break;
8661#endif
fc36a67e 8662 case 'l':
9e5b023a 8663#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8664 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8665 intsize = 'q';
8666 q += 2;
46fc3d4c 8667 break;
cf2093f6 8668 }
fc36a67e 8669#endif
5f66b61c 8670 /*FALLTHROUGH*/
fc36a67e 8671 case 'h':
5f66b61c 8672 /*FALLTHROUGH*/
fc36a67e 8673 case 'V':
8674 intsize = *q++;
46fc3d4c 8675 break;
8676 }
8677
fc36a67e 8678 /* CONVERSION */
8679
211dfcf1
HS
8680 if (*q == '%') {
8681 eptr = q++;
8682 elen = 1;
26372e71
GA
8683 if (vectorize) {
8684 c = '%';
8685 goto unknown;
8686 }
211dfcf1
HS
8687 goto string;
8688 }
8689
26372e71 8690 if (!vectorize && !args) {
86c51f8b
NC
8691 if (efix) {
8692 const I32 i = efix-1;
8693 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8694 } else {
8695 argsv = (svix >= 0 && svix < svmax)
8696 ? svargs[svix++] : &PL_sv_undef;
8697 }
863811b2 8698 }
211dfcf1 8699
46fc3d4c 8700 switch (c = *q++) {
8701
8702 /* STRINGS */
8703
46fc3d4c 8704 case 'c':
26372e71
GA
8705 if (vectorize)
8706 goto unknown;
8707 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8708 if ((uv > 255 ||
8709 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8710 && !IN_BYTES) {
dfe13c55 8711 eptr = (char*)utf8buf;
9041c2e3 8712 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8713 is_utf8 = TRUE;
7e2040f0
GS
8714 }
8715 else {
8716 c = (char)uv;
8717 eptr = &c;
8718 elen = 1;
a0ed51b3 8719 }
46fc3d4c 8720 goto string;
8721
46fc3d4c 8722 case 's':
26372e71
GA
8723 if (vectorize)
8724 goto unknown;
8725 if (args) {
fc36a67e 8726 eptr = va_arg(*args, char*);
c635e13b 8727 if (eptr)
1d7c1841
GS
8728#ifdef MACOS_TRADITIONAL
8729 /* On MacOS, %#s format is used for Pascal strings */
8730 if (alt)
8731 elen = *eptr++;
8732 else
8733#endif
c635e13b 8734 elen = strlen(eptr);
8735 else {
27da23d5 8736 eptr = (char *)nullstr;
c635e13b 8737 elen = sizeof nullstr - 1;
8738 }
46fc3d4c 8739 }
211dfcf1 8740 else {
4d84ee25 8741 eptr = SvPVx_const(argsv, elen);
7e2040f0 8742 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8743 if (has_precis && precis < elen) {
8744 I32 p = precis;
7e2040f0 8745 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8746 precis = p;
8747 }
8748 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8749 width += elen - sv_len_utf8(argsv);
a0ed51b3 8750 }
2cf2cfc6 8751 is_utf8 = TRUE;
a0ed51b3
LW
8752 }
8753 }
fc36a67e 8754
46fc3d4c 8755 string:
8756 if (has_precis && elen > precis)
8757 elen = precis;
8758 break;
8759
8760 /* INTEGERS */
8761
fc36a67e 8762 case 'p':
be75b157 8763 if (alt || vectorize)
c2e66d9e 8764 goto unknown;
211dfcf1 8765 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8766 base = 16;
8767 goto integer;
8768
46fc3d4c 8769 case 'D':
29fe7a80 8770#ifdef IV_IS_QUAD
22f3ae8c 8771 intsize = 'q';
29fe7a80 8772#else
46fc3d4c 8773 intsize = 'l';
29fe7a80 8774#endif
5f66b61c 8775 /*FALLTHROUGH*/
46fc3d4c 8776 case 'd':
8777 case 'i':
8896765a
RB
8778#if vdNUMBER
8779 format_vd:
8780#endif
b22c7a20 8781 if (vectorize) {
ba210ebe 8782 STRLEN ulen;
211dfcf1
HS
8783 if (!veclen)
8784 continue;
2cf2cfc6
A
8785 if (vec_utf8)
8786 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8787 UTF8_ALLOW_ANYUV);
b22c7a20 8788 else {
e83d50c9 8789 uv = *vecstr;
b22c7a20
GS
8790 ulen = 1;
8791 }
8792 vecstr += ulen;
8793 veclen -= ulen;
e83d50c9
JP
8794 if (plus)
8795 esignbuf[esignlen++] = plus;
b22c7a20
GS
8796 }
8797 else if (args) {
46fc3d4c 8798 switch (intsize) {
8799 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8800 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8801 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8802 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8803#ifdef HAS_QUAD
8804 case 'q': iv = va_arg(*args, Quad_t); break;
8805#endif
46fc3d4c 8806 }
8807 }
8808 else {
b10c0dba 8809 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8810 switch (intsize) {
b10c0dba
MHM
8811 case 'h': iv = (short)tiv; break;
8812 case 'l': iv = (long)tiv; break;
8813 case 'V':
8814 default: iv = tiv; break;
cf2093f6 8815#ifdef HAS_QUAD
b10c0dba 8816 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8817#endif
46fc3d4c 8818 }
8819 }
e83d50c9
JP
8820 if ( !vectorize ) /* we already set uv above */
8821 {
8822 if (iv >= 0) {
8823 uv = iv;
8824 if (plus)
8825 esignbuf[esignlen++] = plus;
8826 }
8827 else {
8828 uv = -iv;
8829 esignbuf[esignlen++] = '-';
8830 }
46fc3d4c 8831 }
8832 base = 10;
8833 goto integer;
8834
fc36a67e 8835 case 'U':
29fe7a80 8836#ifdef IV_IS_QUAD
22f3ae8c 8837 intsize = 'q';
29fe7a80 8838#else
fc36a67e 8839 intsize = 'l';
29fe7a80 8840#endif
5f66b61c 8841 /*FALLTHROUGH*/
fc36a67e 8842 case 'u':
8843 base = 10;
8844 goto uns_integer;
8845
4f19785b
WSI
8846 case 'b':
8847 base = 2;
8848 goto uns_integer;
8849
46fc3d4c 8850 case 'O':
29fe7a80 8851#ifdef IV_IS_QUAD
22f3ae8c 8852 intsize = 'q';
29fe7a80 8853#else
46fc3d4c 8854 intsize = 'l';
29fe7a80 8855#endif
5f66b61c 8856 /*FALLTHROUGH*/
46fc3d4c 8857 case 'o':
8858 base = 8;
8859 goto uns_integer;
8860
8861 case 'X':
46fc3d4c 8862 case 'x':
8863 base = 16;
46fc3d4c 8864
8865 uns_integer:
b22c7a20 8866 if (vectorize) {
ba210ebe 8867 STRLEN ulen;
b22c7a20 8868 vector:
211dfcf1
HS
8869 if (!veclen)
8870 continue;
2cf2cfc6
A
8871 if (vec_utf8)
8872 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8873 UTF8_ALLOW_ANYUV);
b22c7a20 8874 else {
a05b299f 8875 uv = *vecstr;
b22c7a20
GS
8876 ulen = 1;
8877 }
8878 vecstr += ulen;
8879 veclen -= ulen;
8880 }
8881 else if (args) {
46fc3d4c 8882 switch (intsize) {
8883 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8884 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8885 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8886 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8887#ifdef HAS_QUAD
9e3321a5 8888 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8889#endif
46fc3d4c 8890 }
8891 }
8892 else {
b10c0dba 8893 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8894 switch (intsize) {
b10c0dba
MHM
8895 case 'h': uv = (unsigned short)tuv; break;
8896 case 'l': uv = (unsigned long)tuv; break;
8897 case 'V':
8898 default: uv = tuv; break;
cf2093f6 8899#ifdef HAS_QUAD
b10c0dba 8900 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 8901#endif
46fc3d4c 8902 }
8903 }
8904
8905 integer:
4d84ee25
NC
8906 {
8907 char *ptr = ebuf + sizeof ebuf;
8908 switch (base) {
8909 unsigned dig;
8910 case 16:
8911 if (!uv)
8912 alt = FALSE;
8913 p = (char*)((c == 'X')
8914 ? "0123456789ABCDEF" : "0123456789abcdef");
8915 do {
8916 dig = uv & 15;
8917 *--ptr = p[dig];
8918 } while (uv >>= 4);
8919 if (alt) {
8920 esignbuf[esignlen++] = '0';
8921 esignbuf[esignlen++] = c; /* 'x' or 'X' */
8922 }
8923 break;
8924 case 8:
8925 do {
8926 dig = uv & 7;
8927 *--ptr = '0' + dig;
8928 } while (uv >>= 3);
8929 if (alt && *ptr != '0')
8930 *--ptr = '0';
8931 break;
8932 case 2:
ed2b91d2
GA
8933 if (!uv)
8934 alt = FALSE;
4d84ee25
NC
8935 do {
8936 dig = uv & 1;
8937 *--ptr = '0' + dig;
8938 } while (uv >>= 1);
8939 if (alt) {
8940 esignbuf[esignlen++] = '0';
8941 esignbuf[esignlen++] = 'b';
8942 }
8943 break;
8944 default: /* it had better be ten or less */
8945 do {
8946 dig = uv % base;
8947 *--ptr = '0' + dig;
8948 } while (uv /= base);
8949 break;
46fc3d4c 8950 }
4d84ee25
NC
8951 elen = (ebuf + sizeof ebuf) - ptr;
8952 eptr = ptr;
8953 if (has_precis) {
8954 if (precis > elen)
8955 zeros = precis - elen;
8956 else if (precis == 0 && elen == 1 && *eptr == '0')
8957 elen = 0;
eda88b6d 8958 }
c10ed8b9 8959 }
46fc3d4c 8960 break;
8961
8962 /* FLOATING POINT */
8963
fc36a67e 8964 case 'F':
8965 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 8966 /*FALLTHROUGH*/
46fc3d4c 8967 case 'e': case 'E':
fc36a67e 8968 case 'f':
46fc3d4c 8969 case 'g': case 'G':
26372e71
GA
8970 if (vectorize)
8971 goto unknown;
46fc3d4c 8972
8973 /* This is evil, but floating point is even more evil */
8974
9e5b023a
JH
8975 /* for SV-style calling, we can only get NV
8976 for C-style calling, we assume %f is double;
8977 for simplicity we allow any of %Lf, %llf, %qf for long double
8978 */
8979 switch (intsize) {
8980 case 'V':
8981#if defined(USE_LONG_DOUBLE)
8982 intsize = 'q';
8983#endif
8984 break;
8a2e3f14 8985/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 8986 case 'l':
5f66b61c 8987 /*FALLTHROUGH*/
9e5b023a
JH
8988 default:
8989#if defined(USE_LONG_DOUBLE)
8990 intsize = args ? 0 : 'q';
8991#endif
8992 break;
8993 case 'q':
8994#if defined(HAS_LONG_DOUBLE)
8995 break;
8996#else
5f66b61c 8997 /*FALLTHROUGH*/
9e5b023a
JH
8998#endif
8999 case 'h':
9e5b023a
JH
9000 goto unknown;
9001 }
9002
9003 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9004 nv = (args) ?
35fff930
JH
9005#if LONG_DOUBLESIZE > DOUBLESIZE
9006 intsize == 'q' ?
205f51d8
AS
9007 va_arg(*args, long double) :
9008 va_arg(*args, double)
35fff930 9009#else
205f51d8 9010 va_arg(*args, double)
35fff930 9011#endif
9e5b023a 9012 : SvNVx(argsv);
fc36a67e 9013
9014 need = 0;
9015 if (c != 'e' && c != 'E') {
9016 i = PERL_INT_MIN;
9e5b023a
JH
9017 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9018 will cast our (long double) to (double) */
73b309ea 9019 (void)Perl_frexp(nv, &i);
fc36a67e 9020 if (i == PERL_INT_MIN)
cea2e8a9 9021 Perl_die(aTHX_ "panic: frexp");
c635e13b 9022 if (i > 0)
fc36a67e 9023 need = BIT_DIGITS(i);
9024 }
9025 need += has_precis ? precis : 6; /* known default */
20f6aaab 9026
fc36a67e 9027 if (need < width)
9028 need = width;
9029
20f6aaab
AS
9030#ifdef HAS_LDBL_SPRINTF_BUG
9031 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9032 with sfio - Allen <allens@cpan.org> */
9033
9034# ifdef DBL_MAX
9035# define MY_DBL_MAX DBL_MAX
9036# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9037# if DOUBLESIZE >= 8
9038# define MY_DBL_MAX 1.7976931348623157E+308L
9039# else
9040# define MY_DBL_MAX 3.40282347E+38L
9041# endif
9042# endif
9043
9044# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9045# define MY_DBL_MAX_BUG 1L
20f6aaab 9046# else
205f51d8 9047# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9048# endif
20f6aaab 9049
205f51d8
AS
9050# ifdef DBL_MIN
9051# define MY_DBL_MIN DBL_MIN
9052# else /* XXX guessing! -Allen */
9053# if DOUBLESIZE >= 8
9054# define MY_DBL_MIN 2.2250738585072014E-308L
9055# else
9056# define MY_DBL_MIN 1.17549435E-38L
9057# endif
9058# endif
20f6aaab 9059
205f51d8
AS
9060 if ((intsize == 'q') && (c == 'f') &&
9061 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9062 (need < DBL_DIG)) {
9063 /* it's going to be short enough that
9064 * long double precision is not needed */
9065
9066 if ((nv <= 0L) && (nv >= -0L))
9067 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9068 else {
9069 /* would use Perl_fp_class as a double-check but not
9070 * functional on IRIX - see perl.h comments */
9071
9072 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9073 /* It's within the range that a double can represent */
9074#if defined(DBL_MAX) && !defined(DBL_MIN)
9075 if ((nv >= ((long double)1/DBL_MAX)) ||
9076 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9077#endif
205f51d8 9078 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9079 }
205f51d8
AS
9080 }
9081 if (fix_ldbl_sprintf_bug == TRUE) {
9082 double temp;
9083
9084 intsize = 0;
9085 temp = (double)nv;
9086 nv = (NV)temp;
9087 }
20f6aaab 9088 }
205f51d8
AS
9089
9090# undef MY_DBL_MAX
9091# undef MY_DBL_MAX_BUG
9092# undef MY_DBL_MIN
9093
20f6aaab
AS
9094#endif /* HAS_LDBL_SPRINTF_BUG */
9095
46fc3d4c 9096 need += 20; /* fudge factor */
80252599
GS
9097 if (PL_efloatsize < need) {
9098 Safefree(PL_efloatbuf);
9099 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9100 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9101 PL_efloatbuf[0] = '\0';
46fc3d4c 9102 }
9103
4151a5fe
IZ
9104 if ( !(width || left || plus || alt) && fill != '0'
9105 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9106 /* See earlier comment about buggy Gconvert when digits,
9107 aka precis is 0 */
9108 if ( c == 'g' && precis) {
2e59c212 9109 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9110 /* May return an empty string for digits==0 */
9111 if (*PL_efloatbuf) {
9112 elen = strlen(PL_efloatbuf);
4151a5fe 9113 goto float_converted;
4150c189 9114 }
4151a5fe
IZ
9115 } else if ( c == 'f' && !precis) {
9116 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9117 break;
9118 }
9119 }
4d84ee25
NC
9120 {
9121 char *ptr = ebuf + sizeof ebuf;
9122 *--ptr = '\0';
9123 *--ptr = c;
9124 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9125#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9126 if (intsize == 'q') {
9127 /* Copy the one or more characters in a long double
9128 * format before the 'base' ([efgEFG]) character to
9129 * the format string. */
9130 static char const prifldbl[] = PERL_PRIfldbl;
9131 char const *p = prifldbl + sizeof(prifldbl) - 3;
9132 while (p >= prifldbl) { *--ptr = *p--; }
9133 }
65202027 9134#endif
4d84ee25
NC
9135 if (has_precis) {
9136 base = precis;
9137 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9138 *--ptr = '.';
9139 }
9140 if (width) {
9141 base = width;
9142 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9143 }
9144 if (fill == '0')
9145 *--ptr = fill;
9146 if (left)
9147 *--ptr = '-';
9148 if (plus)
9149 *--ptr = plus;
9150 if (alt)
9151 *--ptr = '#';
9152 *--ptr = '%';
9153
9154 /* No taint. Otherwise we are in the strange situation
9155 * where printf() taints but print($float) doesn't.
9156 * --jhi */
9e5b023a 9157#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9158 elen = ((intsize == 'q')
9159 ? my_sprintf(PL_efloatbuf, ptr, nv)
9160 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9161#else
4150c189 9162 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9163#endif
4d84ee25 9164 }
4151a5fe 9165 float_converted:
80252599 9166 eptr = PL_efloatbuf;
46fc3d4c 9167 break;
9168
fc36a67e 9169 /* SPECIAL */
9170
9171 case 'n':
26372e71
GA
9172 if (vectorize)
9173 goto unknown;
fc36a67e 9174 i = SvCUR(sv) - origlen;
26372e71 9175 if (args) {
c635e13b 9176 switch (intsize) {
9177 case 'h': *(va_arg(*args, short*)) = i; break;
9178 default: *(va_arg(*args, int*)) = i; break;
9179 case 'l': *(va_arg(*args, long*)) = i; break;
9180 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9181#ifdef HAS_QUAD
9182 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9183#endif
c635e13b 9184 }
fc36a67e 9185 }
9dd79c3f 9186 else
211dfcf1 9187 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9188 continue; /* not "break" */
9189
9190 /* UNKNOWN */
9191
46fc3d4c 9192 default:
fc36a67e 9193 unknown:
041457d9
DM
9194 if (!args
9195 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9196 && ckWARN(WARN_PRINTF))
9197 {
c4420975 9198 SV * const msg = sv_newmortal();
35c1215d
NC
9199 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9200 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9201 if (c) {
0f4b6630 9202 if (isPRINT(c))
1c846c1f 9203 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9204 "\"%%%c\"", c & 0xFF);
9205 else
9206 Perl_sv_catpvf(aTHX_ msg,
57def98f 9207 "\"%%\\%03"UVof"\"",
0f4b6630 9208 (UV)c & 0xFF);
0f4b6630 9209 } else
396482e1 9210 sv_catpvs(msg, "end of string");
9014280d 9211 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9212 }
fb73857a 9213
9214 /* output mangled stuff ... */
9215 if (c == '\0')
9216 --q;
46fc3d4c 9217 eptr = p;
9218 elen = q - p;
fb73857a 9219
9220 /* ... right here, because formatting flags should not apply */
9221 SvGROW(sv, SvCUR(sv) + elen + 1);
9222 p = SvEND(sv);
4459522c 9223 Copy(eptr, p, elen, char);
fb73857a 9224 p += elen;
9225 *p = '\0';
3f7c398e 9226 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9227 svix = osvix;
fb73857a 9228 continue; /* not "break" */
46fc3d4c 9229 }
9230
6c94ec8b
HS
9231 /* calculate width before utf8_upgrade changes it */
9232 have = esignlen + zeros + elen;
ed2b91d2
GA
9233 if (have < zeros)
9234 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9235
d2876be5
JH
9236 if (is_utf8 != has_utf8) {
9237 if (is_utf8) {
9238 if (SvCUR(sv))
9239 sv_utf8_upgrade(sv);
9240 }
9241 else {
53c1dcc0 9242 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9243 sv_utf8_upgrade(nsv);
93524f2b 9244 eptr = SvPVX_const(nsv);
d2876be5
JH
9245 elen = SvCUR(nsv);
9246 }
9247 SvGROW(sv, SvCUR(sv) + elen + 1);
9248 p = SvEND(sv);
9249 *p = '\0';
9250 }
6af65485 9251
46fc3d4c 9252 need = (have > width ? have : width);
9253 gap = need - have;
9254
d2641cbd
PC
9255 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9256 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9257 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9258 p = SvEND(sv);
9259 if (esignlen && fill == '0') {
53c1dcc0 9260 int i;
eb160463 9261 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9262 *p++ = esignbuf[i];
9263 }
9264 if (gap && !left) {
9265 memset(p, fill, gap);
9266 p += gap;
9267 }
9268 if (esignlen && fill != '0') {
53c1dcc0 9269 int i;
eb160463 9270 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9271 *p++ = esignbuf[i];
9272 }
fc36a67e 9273 if (zeros) {
53c1dcc0 9274 int i;
fc36a67e 9275 for (i = zeros; i; i--)
9276 *p++ = '0';
9277 }
46fc3d4c 9278 if (elen) {
4459522c 9279 Copy(eptr, p, elen, char);
46fc3d4c 9280 p += elen;
9281 }
9282 if (gap && left) {
9283 memset(p, ' ', gap);
9284 p += gap;
9285 }
b22c7a20
GS
9286 if (vectorize) {
9287 if (veclen) {
4459522c 9288 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9289 p += dotstrlen;
9290 }
9291 else
9292 vectorize = FALSE; /* done iterating over vecstr */
9293 }
2cf2cfc6
A
9294 if (is_utf8)
9295 has_utf8 = TRUE;
9296 if (has_utf8)
7e2040f0 9297 SvUTF8_on(sv);
46fc3d4c 9298 *p = '\0';
3f7c398e 9299 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9300 if (vectorize) {
9301 esignlen = 0;
9302 goto vector;
9303 }
46fc3d4c 9304 }
9305}
51371543 9306
645c22ef
DM
9307/* =========================================================================
9308
9309=head1 Cloning an interpreter
9310
9311All the macros and functions in this section are for the private use of
9312the main function, perl_clone().
9313
9314The foo_dup() functions make an exact copy of an existing foo thinngy.
9315During the course of a cloning, a hash table is used to map old addresses
9316to new addresses. The table is created and manipulated with the
9317ptr_table_* functions.
9318
9319=cut
9320
9321============================================================================*/
9322
9323
1d7c1841
GS
9324#if defined(USE_ITHREADS)
9325
1d7c1841
GS
9326#ifndef GpREFCNT_inc
9327# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9328#endif
9329
9330
d2d73c3e 9331#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9332#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9333#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9334#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9335#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9336#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9337#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9338#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9339#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9340#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9341#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9342#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9343#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9344#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9345
d2d73c3e 9346
d2f185dc
AMS
9347/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9348 regcomp.c. AMS 20010712 */
645c22ef 9349
1d7c1841 9350REGEXP *
53c1dcc0 9351Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9352{
27da23d5 9353 dVAR;
d2f185dc
AMS
9354 REGEXP *ret;
9355 int i, len, npar;
9356 struct reg_substr_datum *s;
9357
9358 if (!r)
9359 return (REGEXP *)NULL;
9360
9361 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9362 return ret;
9363
9364 len = r->offsets[0];
9365 npar = r->nparens+1;
9366
a02a5408 9367 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9368 Copy(r->program, ret->program, len+1, regnode);
9369
a02a5408 9370 Newx(ret->startp, npar, I32);
d2f185dc 9371 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9372 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9373 Copy(r->startp, ret->startp, npar, I32);
9374
a02a5408 9375 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9376 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9377 s->min_offset = r->substrs->data[i].min_offset;
9378 s->max_offset = r->substrs->data[i].max_offset;
9379 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9380 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9381 }
9382
70612e96 9383 ret->regstclass = NULL;
d2f185dc
AMS
9384 if (r->data) {
9385 struct reg_data *d;
e1ec3a88 9386 const int count = r->data->count;
53c1dcc0 9387 int i;
d2f185dc 9388
a02a5408 9389 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9390 char, struct reg_data);
a02a5408 9391 Newx(d->what, count, U8);
d2f185dc
AMS
9392
9393 d->count = count;
9394 for (i = 0; i < count; i++) {
9395 d->what[i] = r->data->what[i];
9396 switch (d->what[i]) {
a3621e74
YO
9397 /* legal options are one of: sfpont
9398 see also regcomp.h and pregfree() */
d2f185dc
AMS
9399 case 's':
9400 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9401 break;
9402 case 'p':
9403 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9404 break;
9405 case 'f':
9406 /* This is cheating. */
a02a5408 9407 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9408 StructCopy(r->data->data[i], d->data[i],
9409 struct regnode_charclass_class);
70612e96 9410 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9411 break;
9412 case 'o':
33773810
AMS
9413 /* Compiled op trees are readonly, and can thus be
9414 shared without duplication. */
b34c0dd4 9415 OP_REFCNT_LOCK;
9b978d73 9416 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9417 OP_REFCNT_UNLOCK;
9b978d73 9418 break;
d2f185dc
AMS
9419 case 'n':
9420 d->data[i] = r->data->data[i];
9421 break;
a3621e74
YO
9422 case 't':
9423 d->data[i] = r->data->data[i];
9424 OP_REFCNT_LOCK;
9425 ((reg_trie_data*)d->data[i])->refcount++;
9426 OP_REFCNT_UNLOCK;
9427 break;
9428 default:
9429 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9430 }
9431 }
9432
9433 ret->data = d;
9434 }
9435 else
9436 ret->data = NULL;
9437
a02a5408 9438 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9439 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9440
e01c5899 9441 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9442 ret->refcnt = r->refcnt;
9443 ret->minlen = r->minlen;
9444 ret->prelen = r->prelen;
9445 ret->nparens = r->nparens;
9446 ret->lastparen = r->lastparen;
9447 ret->lastcloseparen = r->lastcloseparen;
9448 ret->reganch = r->reganch;
9449
70612e96
RG
9450 ret->sublen = r->sublen;
9451
9452 if (RX_MATCH_COPIED(ret))
e01c5899 9453 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9454 else
bd61b366 9455 ret->subbeg = NULL;
f8c7b90f 9456#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9457 ret->saved_copy = NULL;
9a26048b 9458#endif
70612e96 9459
d2f185dc
AMS
9460 ptr_table_store(PL_ptr_table, r, ret);
9461 return ret;
1d7c1841
GS
9462}
9463
d2d73c3e 9464/* duplicate a file handle */
645c22ef 9465
1d7c1841 9466PerlIO *
a8fc9800 9467Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9468{
9469 PerlIO *ret;
53c1dcc0
AL
9470
9471 PERL_UNUSED_ARG(type);
73d840c0 9472
1d7c1841
GS
9473 if (!fp)
9474 return (PerlIO*)NULL;
9475
9476 /* look for it in the table first */
9477 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9478 if (ret)
9479 return ret;
9480
9481 /* create anew and remember what it is */
ecdeb87c 9482 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9483 ptr_table_store(PL_ptr_table, fp, ret);
9484 return ret;
9485}
9486
645c22ef
DM
9487/* duplicate a directory handle */
9488
1d7c1841
GS
9489DIR *
9490Perl_dirp_dup(pTHX_ DIR *dp)
9491{
96a5add6 9492 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9493 if (!dp)
9494 return (DIR*)NULL;
9495 /* XXX TODO */
9496 return dp;
9497}
9498
ff276b08 9499/* duplicate a typeglob */
645c22ef 9500
1d7c1841 9501GP *
a8fc9800 9502Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9503{
9504 GP *ret;
b37c2d43 9505
1d7c1841
GS
9506 if (!gp)
9507 return (GP*)NULL;
9508 /* look for it in the table first */
9509 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9510 if (ret)
9511 return ret;
9512
9513 /* create anew and remember what it is */
a02a5408 9514 Newxz(ret, 1, GP);
1d7c1841
GS
9515 ptr_table_store(PL_ptr_table, gp, ret);
9516
9517 /* clone */
9518 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9519 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9520 ret->gp_io = io_dup_inc(gp->gp_io, param);
9521 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9522 ret->gp_av = av_dup_inc(gp->gp_av, param);
9523 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9524 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9525 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9526 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9527 ret->gp_line = gp->gp_line;
9528 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9529 return ret;
9530}
9531
645c22ef
DM
9532/* duplicate a chain of magic */
9533
1d7c1841 9534MAGIC *
a8fc9800 9535Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9536{
cb359b41
JH
9537 MAGIC *mgprev = (MAGIC*)NULL;
9538 MAGIC *mgret;
1d7c1841
GS
9539 if (!mg)
9540 return (MAGIC*)NULL;
9541 /* look for it in the table first */
9542 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9543 if (mgret)
9544 return mgret;
9545
9546 for (; mg; mg = mg->mg_moremagic) {
9547 MAGIC *nmg;
a02a5408 9548 Newxz(nmg, 1, MAGIC);
cb359b41 9549 if (mgprev)
1d7c1841 9550 mgprev->mg_moremagic = nmg;
cb359b41
JH
9551 else
9552 mgret = nmg;
1d7c1841
GS
9553 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9554 nmg->mg_private = mg->mg_private;
9555 nmg->mg_type = mg->mg_type;
9556 nmg->mg_flags = mg->mg_flags;
14befaf4 9557 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9558 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9559 }
05bd4103 9560 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9561 /* The backref AV has its reference count deliberately bumped by
9562 1. */
9563 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9564 }
8d2f4536
NC
9565 else if (mg->mg_type == PERL_MAGIC_symtab) {
9566 nmg->mg_obj = mg->mg_obj;
9567 }
1d7c1841
GS
9568 else {
9569 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9570 ? sv_dup_inc(mg->mg_obj, param)
9571 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9572 }
9573 nmg->mg_len = mg->mg_len;
9574 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9575 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9576 if (mg->mg_len > 0) {
1d7c1841 9577 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9578 if (mg->mg_type == PERL_MAGIC_overload_table &&
9579 AMT_AMAGIC((AMT*)mg->mg_ptr))
9580 {
c445ea15 9581 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9582 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9583 I32 i;
9584 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9585 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9586 }
9587 }
9588 }
9589 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9590 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9591 }
68795e93
NIS
9592 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9593 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9594 }
1d7c1841
GS
9595 mgprev = nmg;
9596 }
9597 return mgret;
9598}
9599
645c22ef
DM
9600/* create a new pointer-mapping table */
9601
1d7c1841
GS
9602PTR_TBL_t *
9603Perl_ptr_table_new(pTHX)
9604{
9605 PTR_TBL_t *tbl;
96a5add6
AL
9606 PERL_UNUSED_CONTEXT;
9607
a02a5408 9608 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9609 tbl->tbl_max = 511;
9610 tbl->tbl_items = 0;
a02a5408 9611 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9612 return tbl;
9613}
9614
7119fd33
NC
9615#define PTR_TABLE_HASH(ptr) \
9616 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9617
93e68bfb
JC
9618/*
9619 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9620 following define) and at call to new_body_inline made below in
9621 Perl_ptr_table_store()
9622 */
9623
9624#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9625
645c22ef
DM
9626/* map an existing pointer using a table */
9627
7bf61b54 9628STATIC PTR_TBL_ENT_t *
b0e6ae5b 9629S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9630 PTR_TBL_ENT_t *tblent;
4373e329 9631 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9632 assert(tbl);
9633 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9634 for (; tblent; tblent = tblent->next) {
9635 if (tblent->oldval == sv)
7bf61b54 9636 return tblent;
1d7c1841 9637 }
7bf61b54
NC
9638 return 0;
9639}
9640
9641void *
9642Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9643{
b0e6ae5b 9644 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9645 PERL_UNUSED_CONTEXT;
7bf61b54 9646 return tblent ? tblent->newval : (void *) 0;
1d7c1841
GS
9647}
9648
645c22ef
DM
9649/* add a new entry to a pointer-mapping table */
9650
1d7c1841 9651void
44f8325f 9652Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9653{
0c9fdfe0 9654 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9655 PERL_UNUSED_CONTEXT;
1d7c1841 9656
7bf61b54
NC
9657 if (tblent) {
9658 tblent->newval = newsv;
9659 } else {
9660 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9661
d2a0f284
JC
9662 new_body_inline(tblent, PTE_SVSLOT);
9663
7bf61b54
NC
9664 tblent->oldval = oldsv;
9665 tblent->newval = newsv;
9666 tblent->next = tbl->tbl_ary[entry];
9667 tbl->tbl_ary[entry] = tblent;
9668 tbl->tbl_items++;
9669 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9670 ptr_table_split(tbl);
1d7c1841 9671 }
1d7c1841
GS
9672}
9673
645c22ef
DM
9674/* double the hash bucket size of an existing ptr table */
9675
1d7c1841
GS
9676void
9677Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9678{
9679 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9680 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9681 UV newsize = oldsize * 2;
9682 UV i;
96a5add6 9683 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9684
9685 Renew(ary, newsize, PTR_TBL_ENT_t*);
9686 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9687 tbl->tbl_max = --newsize;
9688 tbl->tbl_ary = ary;
9689 for (i=0; i < oldsize; i++, ary++) {
9690 PTR_TBL_ENT_t **curentp, **entp, *ent;
9691 if (!*ary)
9692 continue;
9693 curentp = ary + oldsize;
9694 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9695 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9696 *entp = ent->next;
9697 ent->next = *curentp;
9698 *curentp = ent;
9699 continue;
9700 }
9701 else
9702 entp = &ent->next;
9703 }
9704 }
9705}
9706
645c22ef
DM
9707/* remove all the entries from a ptr table */
9708
a0739874
DM
9709void
9710Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9711{
d5cefff9 9712 if (tbl && tbl->tbl_items) {
c445ea15 9713 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9714 UV riter = tbl->tbl_max;
a0739874 9715
d5cefff9
NC
9716 do {
9717 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9718
d5cefff9 9719 while (entry) {
00b6aa41 9720 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9721 entry = entry->next;
9722 del_pte(oentry);
9723 }
9724 } while (riter--);
a0739874 9725
d5cefff9
NC
9726 tbl->tbl_items = 0;
9727 }
a0739874
DM
9728}
9729
645c22ef
DM
9730/* clear and free a ptr table */
9731
a0739874
DM
9732void
9733Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9734{
9735 if (!tbl) {
9736 return;
9737 }
9738 ptr_table_clear(tbl);
9739 Safefree(tbl->tbl_ary);
9740 Safefree(tbl);
9741}
9742
5bd07a3d 9743
83841fad 9744void
eb86f8b3 9745Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9746{
9747 if (SvROK(sstr)) {
b162af07
SP
9748 SvRV_set(dstr, SvWEAKREF(sstr)
9749 ? sv_dup(SvRV(sstr), param)
9750 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9751
83841fad 9752 }
3f7c398e 9753 else if (SvPVX_const(sstr)) {
83841fad
NIS
9754 /* Has something there */
9755 if (SvLEN(sstr)) {
68795e93 9756 /* Normal PV - clone whole allocated space */
3f7c398e 9757 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9758 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9759 /* Not that normal - actually sstr is copy on write.
9760 But we are a true, independant SV, so: */
9761 SvREADONLY_off(dstr);
9762 SvFAKE_off(dstr);
9763 }
68795e93 9764 }
83841fad
NIS
9765 else {
9766 /* Special case - not normally malloced for some reason */
f7877b28
NC
9767 if (isGV_with_GP(sstr)) {
9768 /* Don't need to do anything here. */
9769 }
9770 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9771 /* A "shared" PV - clone it as "shared" PV */
9772 SvPV_set(dstr,
9773 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9774 param)));
83841fad
NIS
9775 }
9776 else {
9777 /* Some other special case - random pointer */
f880fe2f 9778 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9779 }
83841fad
NIS
9780 }
9781 }
9782 else {
4608196e 9783 /* Copy the NULL */
f880fe2f 9784 if (SvTYPE(dstr) == SVt_RV)
b162af07 9785 SvRV_set(dstr, NULL);
f880fe2f 9786 else
6136c704 9787 SvPV_set(dstr, NULL);
83841fad
NIS
9788 }
9789}
9790
662fb8b2
NC
9791/* duplicate an SV of any type (including AV, HV etc) */
9792
1d7c1841 9793SV *
eb86f8b3 9794Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9795{
27da23d5 9796 dVAR;
1d7c1841
GS
9797 SV *dstr;
9798
9799 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9800 return NULL;
1d7c1841
GS
9801 /* look for it in the table first */
9802 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9803 if (dstr)
9804 return dstr;
9805
0405e91e
AB
9806 if(param->flags & CLONEf_JOIN_IN) {
9807 /** We are joining here so we don't want do clone
9808 something that is bad **/
eb86f8b3
AL
9809 if (SvTYPE(sstr) == SVt_PVHV) {
9810 const char * const hvname = HvNAME_get(sstr);
9811 if (hvname)
9812 /** don't clone stashes if they already exist **/
9813 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9814 }
9815 }
9816
1d7c1841
GS
9817 /* create anew and remember what it is */
9818 new_SV(dstr);
fd0854ff
DM
9819
9820#ifdef DEBUG_LEAKING_SCALARS
9821 dstr->sv_debug_optype = sstr->sv_debug_optype;
9822 dstr->sv_debug_line = sstr->sv_debug_line;
9823 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9824 dstr->sv_debug_cloned = 1;
fd0854ff 9825 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9826#endif
9827
1d7c1841
GS
9828 ptr_table_store(PL_ptr_table, sstr, dstr);
9829
9830 /* clone */
9831 SvFLAGS(dstr) = SvFLAGS(sstr);
9832 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9833 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9834
9835#ifdef DEBUGGING
3f7c398e 9836 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9837 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9838 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9839#endif
9840
9660f481
DM
9841 /* don't clone objects whose class has asked us not to */
9842 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9843 SvFLAGS(dstr) &= ~SVTYPEMASK;
9844 SvOBJECT_off(dstr);
9845 return dstr;
9846 }
9847
1d7c1841
GS
9848 switch (SvTYPE(sstr)) {
9849 case SVt_NULL:
9850 SvANY(dstr) = NULL;
9851 break;
9852 case SVt_IV:
339049b0 9853 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9854 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9855 break;
9856 case SVt_NV:
9857 SvANY(dstr) = new_XNV();
9d6ce603 9858 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9859 break;
9860 case SVt_RV:
339049b0 9861 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9862 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9863 break;
662fb8b2
NC
9864 default:
9865 {
9866 /* These are all the types that need complex bodies allocating. */
662fb8b2 9867 void *new_body;
2bcc16b3
NC
9868 const svtype sv_type = SvTYPE(sstr);
9869 const struct body_details *const sv_type_details
9870 = bodies_by_type + sv_type;
662fb8b2 9871
93e68bfb 9872 switch (sv_type) {
662fb8b2 9873 default:
bb263b4e 9874 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9875 break;
9876
662fb8b2
NC
9877 case SVt_PVGV:
9878 if (GvUNIQUE((GV*)sstr)) {
bb263b4e 9879 /*EMPTY*/; /* Do sharing here, and fall through */
662fb8b2 9880 }
c22188b4
NC
9881 case SVt_PVIO:
9882 case SVt_PVFM:
9883 case SVt_PVHV:
9884 case SVt_PVAV:
93e68bfb 9885 case SVt_PVBM:
662fb8b2 9886 case SVt_PVCV:
662fb8b2 9887 case SVt_PVLV:
662fb8b2 9888 case SVt_PVMG:
662fb8b2 9889 case SVt_PVNV:
662fb8b2 9890 case SVt_PVIV:
662fb8b2 9891 case SVt_PV:
d2a0f284 9892 assert(sv_type_details->body_size);
c22188b4 9893 if (sv_type_details->arena) {
d2a0f284 9894 new_body_inline(new_body, sv_type);
c22188b4 9895 new_body
b9502f15 9896 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9897 } else {
9898 new_body = new_NOARENA(sv_type_details);
9899 }
1d7c1841 9900 }
662fb8b2
NC
9901 assert(new_body);
9902 SvANY(dstr) = new_body;
9903
2bcc16b3 9904#ifndef PURIFY
b9502f15
NC
9905 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
9906 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 9907 sv_type_details->copy, char);
2bcc16b3
NC
9908#else
9909 Copy(((char*)SvANY(sstr)),
9910 ((char*)SvANY(dstr)),
d2a0f284 9911 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 9912#endif
662fb8b2 9913
f7877b28
NC
9914 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
9915 && !isGV_with_GP(dstr))
662fb8b2
NC
9916 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
9917
9918 /* The Copy above means that all the source (unduplicated) pointers
9919 are now in the destination. We can check the flags and the
9920 pointers in either, but it's possible that there's less cache
9921 missing by always going for the destination.
9922 FIXME - instrument and check that assumption */
f32993d6 9923 if (sv_type >= SVt_PVMG) {
e736a858 9924 HV *ourstash;
2892acdb 9925 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
9926 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
9927 } else if (SvMAGIC(dstr))
662fb8b2
NC
9928 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
9929 if (SvSTASH(dstr))
9930 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 9931 }
662fb8b2 9932
f32993d6
NC
9933 /* The cast silences a GCC warning about unhandled types. */
9934 switch ((int)sv_type) {
662fb8b2
NC
9935 case SVt_PV:
9936 break;
9937 case SVt_PVIV:
9938 break;
9939 case SVt_PVNV:
9940 break;
9941 case SVt_PVMG:
9942 break;
9943 case SVt_PVBM:
9944 break;
9945 case SVt_PVLV:
9946 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
9947 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
9948 LvTARG(dstr) = dstr;
9949 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
9950 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
9951 else
9952 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
9953 break;
9954 case SVt_PVGV:
acda4c6a
NC
9955 if (GvNAME_HEK(dstr))
9956 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 9957
e15faf7d
NC
9958 /* Don't call sv_add_backref here as it's going to be created
9959 as part of the magic cloning of the symbol table. */
f7877b28
NC
9960 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
9961 if(isGV_with_GP(sstr)) {
9962 /* Danger Will Robinson - GvGP(dstr) isn't initialised
9963 at the point of this comment. */
9964 GvGP(dstr) = gp_dup(GvGP(sstr), param);
9965 (void)GpREFCNT_inc(GvGP(dstr));
9966 } else
9967 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
9968 break;
9969 case SVt_PVIO:
9970 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
9971 if (IoOFP(dstr) == IoIFP(sstr))
9972 IoOFP(dstr) = IoIFP(dstr);
9973 else
9974 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
9975 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
9976 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
9977 /* I have no idea why fake dirp (rsfps)
9978 should be treated differently but otherwise
9979 we end up with leaks -- sky*/
9980 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
9981 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
9982 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
9983 } else {
9984 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
9985 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
9986 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
9987 if (IoDIRP(dstr)) {
9988 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
9989 } else {
bb263b4e 9990 /*EMPTY*/;
100ce7e1
NC
9991 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
9992 }
662fb8b2
NC
9993 }
9994 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
9995 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
9996 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
9997 break;
9998 case SVt_PVAV:
9999 if (AvARRAY((AV*)sstr)) {
10000 SV **dst_ary, **src_ary;
10001 SSize_t items = AvFILLp((AV*)sstr) + 1;
10002
10003 src_ary = AvARRAY((AV*)sstr);
a02a5408 10004 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10005 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10006 SvPV_set(dstr, (char*)dst_ary);
10007 AvALLOC((AV*)dstr) = dst_ary;
10008 if (AvREAL((AV*)sstr)) {
10009 while (items-- > 0)
10010 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10011 }
10012 else {
10013 while (items-- > 0)
10014 *dst_ary++ = sv_dup(*src_ary++, param);
10015 }
10016 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10017 while (items-- > 0) {
10018 *dst_ary++ = &PL_sv_undef;
10019 }
bfcb3514 10020 }
662fb8b2 10021 else {
bd61b366 10022 SvPV_set(dstr, NULL);
662fb8b2 10023 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10024 }
662fb8b2
NC
10025 break;
10026 case SVt_PVHV:
10027 {
cbbf8932 10028 HEK *hvname = NULL;
662fb8b2
NC
10029
10030 if (HvARRAY((HV*)sstr)) {
10031 STRLEN i = 0;
10032 const bool sharekeys = !!HvSHAREKEYS(sstr);
10033 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10034 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10035 char *darray;
a02a5408 10036 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10037 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10038 char);
10039 HvARRAY(dstr) = (HE**)darray;
10040 while (i <= sxhv->xhv_max) {
5c4138a0 10041 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10042 HvARRAY(dstr)[i] = source
10043 ? he_dup(source, sharekeys, param) : 0;
10044 ++i;
10045 }
10046 if (SvOOK(sstr)) {
00b6aa41
AL
10047 struct xpvhv_aux * const saux = HvAUX(sstr);
10048 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10049 /* This flag isn't copied. */
10050 /* SvOOK_on(hv) attacks the IV flags. */
10051 SvFLAGS(dstr) |= SVf_OOK;
10052
10053 hvname = saux->xhv_name;
dd690478
NC
10054 daux->xhv_name
10055 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10056
10057 daux->xhv_riter = saux->xhv_riter;
10058 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10059 ? he_dup(saux->xhv_eiter,
10060 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10061 daux->xhv_backreferences = saux->xhv_backreferences
10062 ? (AV*) SvREFCNT_inc(
10063 sv_dup((SV*)saux->
10064 xhv_backreferences,
10065 param))
10066 : 0;
662fb8b2
NC
10067 }
10068 }
10069 else {
bd61b366 10070 SvPV_set(dstr, NULL);
662fb8b2
NC
10071 }
10072 /* Record stashes for possible cloning in Perl_clone(). */
10073 if(hvname)
10074 av_push(param->stashes, dstr);
10075 }
10076 break;
662fb8b2 10077 case SVt_PVCV:
bb172083
NC
10078 if (!(param->flags & CLONEf_COPY_STACKS)) {
10079 CvDEPTH(dstr) = 0;
10080 }
10081 case SVt_PVFM:
662fb8b2
NC
10082 /* NOTE: not refcounted */
10083 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10084 OP_REFCNT_LOCK;
d04ba589
NC
10085 if (!CvISXSUB(dstr))
10086 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10087 OP_REFCNT_UNLOCK;
cfae286e 10088 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10089 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10090 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10091 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10092 }
10093 /* don't dup if copying back - CvGV isn't refcounted, so the
10094 * duped GV may never be freed. A bit of a hack! DAPM */
10095 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10096 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10097 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10098 CvOUTSIDE(dstr) =
10099 CvWEAKOUTSIDE(sstr)
10100 ? cv_dup( CvOUTSIDE(dstr), param)
10101 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10102 if (!CvISXSUB(dstr))
662fb8b2
NC
10103 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10104 break;
bfcb3514 10105 }
1d7c1841 10106 }
1d7c1841
GS
10107 }
10108
10109 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10110 ++PL_sv_objcount;
10111
10112 return dstr;
d2d73c3e 10113 }
1d7c1841 10114
645c22ef
DM
10115/* duplicate a context */
10116
1d7c1841 10117PERL_CONTEXT *
a8fc9800 10118Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10119{
10120 PERL_CONTEXT *ncxs;
10121
10122 if (!cxs)
10123 return (PERL_CONTEXT*)NULL;
10124
10125 /* look for it in the table first */
10126 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10127 if (ncxs)
10128 return ncxs;
10129
10130 /* create anew and remember what it is */
a02a5408 10131 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10132 ptr_table_store(PL_ptr_table, cxs, ncxs);
10133
10134 while (ix >= 0) {
c445ea15
AL
10135 PERL_CONTEXT * const cx = &cxs[ix];
10136 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10137 ncx->cx_type = cx->cx_type;
10138 if (CxTYPE(cx) == CXt_SUBST) {
10139 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10140 }
10141 else {
10142 ncx->blk_oldsp = cx->blk_oldsp;
10143 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10144 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10145 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10146 ncx->blk_oldpm = cx->blk_oldpm;
10147 ncx->blk_gimme = cx->blk_gimme;
10148 switch (CxTYPE(cx)) {
10149 case CXt_SUB:
10150 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10151 ? cv_dup_inc(cx->blk_sub.cv, param)
10152 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10153 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10154 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10155 : NULL);
d2d73c3e 10156 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10157 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10158 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10159 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10160 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10161 break;
10162 case CXt_EVAL:
10163 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10164 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10165 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10166 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10167 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10168 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10169 break;
10170 case CXt_LOOP:
10171 ncx->blk_loop.label = cx->blk_loop.label;
10172 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10173 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10174 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10175 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10176 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10177 ? cx->blk_loop.iterdata
d2d73c3e 10178 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10179 ncx->blk_loop.oldcomppad
10180 = (PAD*)ptr_table_fetch(PL_ptr_table,
10181 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10182 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10183 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10184 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10185 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10186 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10187 break;
10188 case CXt_FORMAT:
d2d73c3e
AB
10189 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10190 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10191 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10192 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10193 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10194 break;
10195 case CXt_BLOCK:
10196 case CXt_NULL:
10197 break;
10198 }
10199 }
10200 --ix;
10201 }
10202 return ncxs;
10203}
10204
645c22ef
DM
10205/* duplicate a stack info structure */
10206
1d7c1841 10207PERL_SI *
a8fc9800 10208Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10209{
10210 PERL_SI *nsi;
10211
10212 if (!si)
10213 return (PERL_SI*)NULL;
10214
10215 /* look for it in the table first */
10216 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10217 if (nsi)
10218 return nsi;
10219
10220 /* create anew and remember what it is */
a02a5408 10221 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10222 ptr_table_store(PL_ptr_table, si, nsi);
10223
d2d73c3e 10224 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10225 nsi->si_cxix = si->si_cxix;
10226 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10227 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10228 nsi->si_type = si->si_type;
d2d73c3e
AB
10229 nsi->si_prev = si_dup(si->si_prev, param);
10230 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10231 nsi->si_markoff = si->si_markoff;
10232
10233 return nsi;
10234}
10235
10236#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10237#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10238#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10239#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10240#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10241#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10242#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10243#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10244#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10245#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10246#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10247#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10248#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10249#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10250
10251/* XXXXX todo */
10252#define pv_dup_inc(p) SAVEPV(p)
10253#define pv_dup(p) SAVEPV(p)
10254#define svp_dup_inc(p,pp) any_dup(p,pp)
10255
645c22ef
DM
10256/* map any object to the new equivent - either something in the
10257 * ptr table, or something in the interpreter structure
10258 */
10259
1d7c1841 10260void *
53c1dcc0 10261Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10262{
10263 void *ret;
10264
10265 if (!v)
10266 return (void*)NULL;
10267
10268 /* look for it in the table first */
10269 ret = ptr_table_fetch(PL_ptr_table, v);
10270 if (ret)
10271 return ret;
10272
10273 /* see if it is part of the interpreter structure */
10274 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10275 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10276 else {
1d7c1841 10277 ret = v;
05ec9bb3 10278 }
1d7c1841
GS
10279
10280 return ret;
10281}
10282
645c22ef
DM
10283/* duplicate the save stack */
10284
1d7c1841 10285ANY *
a8fc9800 10286Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10287{
53c1dcc0
AL
10288 ANY * const ss = proto_perl->Tsavestack;
10289 const I32 max = proto_perl->Tsavestack_max;
10290 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10291 ANY *nss;
10292 SV *sv;
10293 GV *gv;
10294 AV *av;
10295 HV *hv;
10296 void* ptr;
10297 int intval;
10298 long longval;
10299 GP *gp;
10300 IV iv;
c4e33207 10301 char *c = NULL;
1d7c1841 10302 void (*dptr) (void*);
acfe0abc 10303 void (*dxptr) (pTHX_ void*);
1d7c1841 10304
a02a5408 10305 Newxz(nss, max, ANY);
1d7c1841
GS
10306
10307 while (ix > 0) {
b464bac0 10308 I32 i = POPINT(ss,ix);
1d7c1841
GS
10309 TOPINT(nss,ix) = i;
10310 switch (i) {
10311 case SAVEt_ITEM: /* normal string */
10312 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10313 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10314 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10315 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10316 break;
10317 case SAVEt_SV: /* scalar reference */
10318 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10319 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10320 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10321 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841 10322 break;
f4dd75d9
GS
10323 case SAVEt_GENERIC_PVREF: /* generic char* */
10324 c = (char*)POPPTR(ss,ix);
10325 TOPPTR(nss,ix) = pv_dup(c);
10326 ptr = POPPTR(ss,ix);
10327 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10328 break;
05ec9bb3
NIS
10329 case SAVEt_SHARED_PVREF: /* char* in shared space */
10330 c = (char*)POPPTR(ss,ix);
10331 TOPPTR(nss,ix) = savesharedpv(c);
10332 ptr = POPPTR(ss,ix);
10333 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10334 break;
1d7c1841
GS
10335 case SAVEt_GENERIC_SVREF: /* generic sv */
10336 case SAVEt_SVREF: /* scalar reference */
10337 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10338 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10339 ptr = POPPTR(ss,ix);
10340 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10341 break;
10342 case SAVEt_AV: /* array reference */
10343 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10344 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841 10345 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10346 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10347 break;
10348 case SAVEt_HV: /* hash reference */
10349 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10350 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841 10351 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10352 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10353 break;
10354 case SAVEt_INT: /* int reference */
10355 ptr = POPPTR(ss,ix);
10356 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10357 intval = (int)POPINT(ss,ix);
10358 TOPINT(nss,ix) = intval;
10359 break;
10360 case SAVEt_LONG: /* long reference */
10361 ptr = POPPTR(ss,ix);
10362 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10363 longval = (long)POPLONG(ss,ix);
10364 TOPLONG(nss,ix) = longval;
10365 break;
10366 case SAVEt_I32: /* I32 reference */
10367 case SAVEt_I16: /* I16 reference */
10368 case SAVEt_I8: /* I8 reference */
10369 ptr = POPPTR(ss,ix);
10370 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10371 i = POPINT(ss,ix);
10372 TOPINT(nss,ix) = i;
10373 break;
10374 case SAVEt_IV: /* IV reference */
10375 ptr = POPPTR(ss,ix);
10376 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10377 iv = POPIV(ss,ix);
10378 TOPIV(nss,ix) = iv;
10379 break;
10380 case SAVEt_SPTR: /* SV* reference */
10381 ptr = POPPTR(ss,ix);
10382 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10383 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10384 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10385 break;
10386 case SAVEt_VPTR: /* random* reference */
10387 ptr = POPPTR(ss,ix);
10388 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10389 ptr = POPPTR(ss,ix);
10390 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10391 break;
10392 case SAVEt_PPTR: /* char* reference */
10393 ptr = POPPTR(ss,ix);
10394 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10395 c = (char*)POPPTR(ss,ix);
10396 TOPPTR(nss,ix) = pv_dup(c);
10397 break;
10398 case SAVEt_HPTR: /* HV* reference */
10399 ptr = POPPTR(ss,ix);
10400 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10401 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10402 TOPPTR(nss,ix) = hv_dup(hv, param);
1d7c1841
GS
10403 break;
10404 case SAVEt_APTR: /* AV* reference */
10405 ptr = POPPTR(ss,ix);
10406 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10407 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10408 TOPPTR(nss,ix) = av_dup(av, param);
1d7c1841
GS
10409 break;
10410 case SAVEt_NSTAB:
10411 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10412 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10413 break;
10414 case SAVEt_GP: /* scalar reference */
10415 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10416 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10417 (void)GpREFCNT_inc(gp);
10418 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10419 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10420 c = (char*)POPPTR(ss,ix);
10421 TOPPTR(nss,ix) = pv_dup(c);
10422 iv = POPIV(ss,ix);
10423 TOPIV(nss,ix) = iv;
10424 iv = POPIV(ss,ix);
10425 TOPIV(nss,ix) = iv;
10426 break;
10427 case SAVEt_FREESV:
26d9b02f 10428 case SAVEt_MORTALIZESV:
1d7c1841 10429 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10430 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10431 break;
10432 case SAVEt_FREEOP:
10433 ptr = POPPTR(ss,ix);
10434 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10435 /* these are assumed to be refcounted properly */
53c1dcc0 10436 OP *o;
1d7c1841
GS
10437 switch (((OP*)ptr)->op_type) {
10438 case OP_LEAVESUB:
10439 case OP_LEAVESUBLV:
10440 case OP_LEAVEEVAL:
10441 case OP_LEAVE:
10442 case OP_SCOPE:
10443 case OP_LEAVEWRITE:
e977893f
GS
10444 TOPPTR(nss,ix) = ptr;
10445 o = (OP*)ptr;
10446 OpREFCNT_inc(o);
1d7c1841
GS
10447 break;
10448 default:
5f66b61c 10449 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10450 break;
10451 }
10452 }
10453 else
5f66b61c 10454 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10455 break;
10456 case SAVEt_FREEPV:
10457 c = (char*)POPPTR(ss,ix);
10458 TOPPTR(nss,ix) = pv_dup_inc(c);
10459 break;
10460 case SAVEt_CLEARSV:
10461 longval = POPLONG(ss,ix);
10462 TOPLONG(nss,ix) = longval;
10463 break;
10464 case SAVEt_DELETE:
10465 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10466 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10467 c = (char*)POPPTR(ss,ix);
10468 TOPPTR(nss,ix) = pv_dup_inc(c);
10469 i = POPINT(ss,ix);
10470 TOPINT(nss,ix) = i;
10471 break;
10472 case SAVEt_DESTRUCTOR:
10473 ptr = POPPTR(ss,ix);
10474 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10475 dptr = POPDPTR(ss,ix);
8141890a
JH
10476 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10477 any_dup(FPTR2DPTR(void *, dptr),
10478 proto_perl));
1d7c1841
GS
10479 break;
10480 case SAVEt_DESTRUCTOR_X:
10481 ptr = POPPTR(ss,ix);
10482 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10483 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10484 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10485 any_dup(FPTR2DPTR(void *, dxptr),
10486 proto_perl));
1d7c1841
GS
10487 break;
10488 case SAVEt_REGCONTEXT:
10489 case SAVEt_ALLOC:
10490 i = POPINT(ss,ix);
10491 TOPINT(nss,ix) = i;
10492 ix -= i;
10493 break;
10494 case SAVEt_STACK_POS: /* Position on Perl stack */
10495 i = POPINT(ss,ix);
10496 TOPINT(nss,ix) = i;
10497 break;
10498 case SAVEt_AELEM: /* array element */
10499 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10500 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10501 i = POPINT(ss,ix);
10502 TOPINT(nss,ix) = i;
10503 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10504 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10505 break;
10506 case SAVEt_HELEM: /* hash element */
10507 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10508 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10509 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10510 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10511 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10512 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10513 break;
10514 case SAVEt_OP:
10515 ptr = POPPTR(ss,ix);
10516 TOPPTR(nss,ix) = ptr;
10517 break;
10518 case SAVEt_HINTS:
10519 i = POPINT(ss,ix);
10520 TOPINT(nss,ix) = i;
10521 break;
c4410b1b
GS
10522 case SAVEt_COMPPAD:
10523 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10524 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10525 break;
c3564e5c
GS
10526 case SAVEt_PADSV:
10527 longval = (long)POPLONG(ss,ix);
10528 TOPLONG(nss,ix) = longval;
10529 ptr = POPPTR(ss,ix);
10530 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10531 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10532 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10533 break;
a1bb4754 10534 case SAVEt_BOOL:
38d8b13e 10535 ptr = POPPTR(ss,ix);
b9609c01 10536 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10537 longval = (long)POPBOOL(ss,ix);
b9609c01 10538 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10539 break;
8bd2680e
MHM
10540 case SAVEt_SET_SVFLAGS:
10541 i = POPINT(ss,ix);
10542 TOPINT(nss,ix) = i;
10543 i = POPINT(ss,ix);
10544 TOPINT(nss,ix) = i;
10545 sv = (SV*)POPPTR(ss,ix);
10546 TOPPTR(nss,ix) = sv_dup(sv, param);
10547 break;
1d7c1841
GS
10548 default:
10549 Perl_croak(aTHX_ "panic: ss_dup inconsistency");
10550 }
10551 }
10552
bd81e77b
NC
10553 return nss;
10554}
10555
10556
10557/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10558 * flag to the result. This is done for each stash before cloning starts,
10559 * so we know which stashes want their objects cloned */
10560
10561static void
10562do_mark_cloneable_stash(pTHX_ SV *sv)
10563{
10564 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10565 if (hvname) {
10566 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10567 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10568 if (cloner && GvCV(cloner)) {
10569 dSP;
10570 UV status;
10571
10572 ENTER;
10573 SAVETMPS;
10574 PUSHMARK(SP);
10575 XPUSHs(sv_2mortal(newSVhek(hvname)));
10576 PUTBACK;
10577 call_sv((SV*)GvCV(cloner), G_SCALAR);
10578 SPAGAIN;
10579 status = POPu;
10580 PUTBACK;
10581 FREETMPS;
10582 LEAVE;
10583 if (status)
10584 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10585 }
10586 }
10587}
10588
10589
10590
10591/*
10592=for apidoc perl_clone
10593
10594Create and return a new interpreter by cloning the current one.
10595
10596perl_clone takes these flags as parameters:
10597
10598CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10599without it we only clone the data and zero the stacks,
10600with it we copy the stacks and the new perl interpreter is
10601ready to run at the exact same point as the previous one.
10602The pseudo-fork code uses COPY_STACKS while the
10603threads->new doesn't.
10604
10605CLONEf_KEEP_PTR_TABLE
10606perl_clone keeps a ptr_table with the pointer of the old
10607variable as a key and the new variable as a value,
10608this allows it to check if something has been cloned and not
10609clone it again but rather just use the value and increase the
10610refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10611the ptr_table using the function
10612C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10613reason to keep it around is if you want to dup some of your own
10614variable who are outside the graph perl scans, example of this
10615code is in threads.xs create
10616
10617CLONEf_CLONE_HOST
10618This is a win32 thing, it is ignored on unix, it tells perls
10619win32host code (which is c++) to clone itself, this is needed on
10620win32 if you want to run two threads at the same time,
10621if you just want to do some stuff in a separate perl interpreter
10622and then throw it away and return to the original one,
10623you don't need to do anything.
10624
10625=cut
10626*/
10627
10628/* XXX the above needs expanding by someone who actually understands it ! */
10629EXTERN_C PerlInterpreter *
10630perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10631
10632PerlInterpreter *
10633perl_clone(PerlInterpreter *proto_perl, UV flags)
10634{
10635 dVAR;
10636#ifdef PERL_IMPLICIT_SYS
10637
10638 /* perlhost.h so we need to call into it
10639 to clone the host, CPerlHost should have a c interface, sky */
10640
10641 if (flags & CLONEf_CLONE_HOST) {
10642 return perl_clone_host(proto_perl,flags);
10643 }
10644 return perl_clone_using(proto_perl, flags,
10645 proto_perl->IMem,
10646 proto_perl->IMemShared,
10647 proto_perl->IMemParse,
10648 proto_perl->IEnv,
10649 proto_perl->IStdIO,
10650 proto_perl->ILIO,
10651 proto_perl->IDir,
10652 proto_perl->ISock,
10653 proto_perl->IProc);
10654}
10655
10656PerlInterpreter *
10657perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10658 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10659 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10660 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10661 struct IPerlDir* ipD, struct IPerlSock* ipS,
10662 struct IPerlProc* ipP)
10663{
10664 /* XXX many of the string copies here can be optimized if they're
10665 * constants; they need to be allocated as common memory and just
10666 * their pointers copied. */
10667
10668 IV i;
10669 CLONE_PARAMS clone_params;
5f66b61c 10670 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10671
5f66b61c 10672 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10673 /* for each stash, determine whether its objects should be cloned */
10674 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10675 PERL_SET_THX(my_perl);
10676
10677# ifdef DEBUGGING
10678 Poison(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10679 PL_op = NULL;
10680 PL_curcop = NULL;
bd81e77b
NC
10681 PL_markstack = 0;
10682 PL_scopestack = 0;
10683 PL_savestack = 0;
10684 PL_savestack_ix = 0;
10685 PL_savestack_max = -1;
10686 PL_sig_pending = 0;
10687 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10688# else /* !DEBUGGING */
10689 Zero(my_perl, 1, PerlInterpreter);
10690# endif /* DEBUGGING */
10691
10692 /* host pointers */
10693 PL_Mem = ipM;
10694 PL_MemShared = ipMS;
10695 PL_MemParse = ipMP;
10696 PL_Env = ipE;
10697 PL_StdIO = ipStd;
10698 PL_LIO = ipLIO;
10699 PL_Dir = ipD;
10700 PL_Sock = ipS;
10701 PL_Proc = ipP;
10702#else /* !PERL_IMPLICIT_SYS */
10703 IV i;
10704 CLONE_PARAMS clone_params;
10705 CLONE_PARAMS* param = &clone_params;
5f66b61c 10706 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10707 /* for each stash, determine whether its objects should be cloned */
10708 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10709 PERL_SET_THX(my_perl);
10710
10711# ifdef DEBUGGING
10712 Poison(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10713 PL_op = NULL;
10714 PL_curcop = NULL;
bd81e77b
NC
10715 PL_markstack = 0;
10716 PL_scopestack = 0;
10717 PL_savestack = 0;
10718 PL_savestack_ix = 0;
10719 PL_savestack_max = -1;
10720 PL_sig_pending = 0;
10721 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10722# else /* !DEBUGGING */
10723 Zero(my_perl, 1, PerlInterpreter);
10724# endif /* DEBUGGING */
10725#endif /* PERL_IMPLICIT_SYS */
10726 param->flags = flags;
10727 param->proto_perl = proto_perl;
10728
7cb608b5
NC
10729 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10730
fdda85ca 10731 PL_body_arenas = NULL;
bd81e77b
NC
10732 Zero(&PL_body_roots, 1, PL_body_roots);
10733
10734 PL_nice_chunk = NULL;
10735 PL_nice_chunk_size = 0;
10736 PL_sv_count = 0;
10737 PL_sv_objcount = 0;
a0714e2c
SS
10738 PL_sv_root = NULL;
10739 PL_sv_arenaroot = NULL;
bd81e77b
NC
10740
10741 PL_debug = proto_perl->Idebug;
10742
10743 PL_hash_seed = proto_perl->Ihash_seed;
10744 PL_rehash_seed = proto_perl->Irehash_seed;
10745
10746#ifdef USE_REENTRANT_API
10747 /* XXX: things like -Dm will segfault here in perlio, but doing
10748 * PERL_SET_CONTEXT(proto_perl);
10749 * breaks too many other things
10750 */
10751 Perl_reentrant_init(aTHX);
10752#endif
10753
10754 /* create SV map for pointer relocation */
10755 PL_ptr_table = ptr_table_new();
10756
10757 /* initialize these special pointers as early as possible */
10758 SvANY(&PL_sv_undef) = NULL;
10759 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10760 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10761 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10762
10763 SvANY(&PL_sv_no) = new_XPVNV();
10764 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10765 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10766 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10767 SvPV_set(&PL_sv_no, SAVEPVN(PL_No, 0));
10768 SvCUR_set(&PL_sv_no, 0);
10769 SvLEN_set(&PL_sv_no, 1);
10770 SvIV_set(&PL_sv_no, 0);
10771 SvNV_set(&PL_sv_no, 0);
10772 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10773
10774 SvANY(&PL_sv_yes) = new_XPVNV();
10775 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10776 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10777 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10778 SvPV_set(&PL_sv_yes, SAVEPVN(PL_Yes, 1));
10779 SvCUR_set(&PL_sv_yes, 1);
10780 SvLEN_set(&PL_sv_yes, 2);
10781 SvIV_set(&PL_sv_yes, 1);
10782 SvNV_set(&PL_sv_yes, 1);
10783 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10784
10785 /* create (a non-shared!) shared string table */
10786 PL_strtab = newHV();
10787 HvSHAREKEYS_off(PL_strtab);
10788 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10789 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10790
10791 PL_compiling = proto_perl->Icompiling;
10792
10793 /* These two PVs will be free'd special way so must set them same way op.c does */
10794 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10795 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10796
10797 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10798 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10799
10800 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
10801 if (!specialWARN(PL_compiling.cop_warnings))
10802 PL_compiling.cop_warnings = sv_dup_inc(PL_compiling.cop_warnings, param);
10803 if (!specialCopIO(PL_compiling.cop_io))
10804 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
10805 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10806
10807 /* pseudo environmental stuff */
10808 PL_origargc = proto_perl->Iorigargc;
10809 PL_origargv = proto_perl->Iorigargv;
10810
10811 param->stashes = newAV(); /* Setup array of objects to call clone on */
10812
10813 /* Set tainting stuff before PerlIO_debug can possibly get called */
10814 PL_tainting = proto_perl->Itainting;
10815 PL_taint_warn = proto_perl->Itaint_warn;
10816
10817#ifdef PERLIO_LAYERS
10818 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10819 PerlIO_clone(aTHX_ proto_perl, param);
10820#endif
10821
10822 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10823 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10824 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10825 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10826 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10827 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10828
10829 /* switches */
10830 PL_minus_c = proto_perl->Iminus_c;
10831 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10832 PL_localpatches = proto_perl->Ilocalpatches;
10833 PL_splitstr = proto_perl->Isplitstr;
10834 PL_preprocess = proto_perl->Ipreprocess;
10835 PL_minus_n = proto_perl->Iminus_n;
10836 PL_minus_p = proto_perl->Iminus_p;
10837 PL_minus_l = proto_perl->Iminus_l;
10838 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10839 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10840 PL_minus_F = proto_perl->Iminus_F;
10841 PL_doswitches = proto_perl->Idoswitches;
10842 PL_dowarn = proto_perl->Idowarn;
10843 PL_doextract = proto_perl->Idoextract;
10844 PL_sawampersand = proto_perl->Isawampersand;
10845 PL_unsafe = proto_perl->Iunsafe;
10846 PL_inplace = SAVEPV(proto_perl->Iinplace);
10847 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10848 PL_perldb = proto_perl->Iperldb;
10849 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10850 PL_exit_flags = proto_perl->Iexit_flags;
10851
10852 /* magical thingies */
10853 /* XXX time(&PL_basetime) when asked for? */
10854 PL_basetime = proto_perl->Ibasetime;
10855 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
10856
10857 PL_maxsysfd = proto_perl->Imaxsysfd;
10858 PL_multiline = proto_perl->Imultiline;
10859 PL_statusvalue = proto_perl->Istatusvalue;
10860#ifdef VMS
10861 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
10862#else
10863 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
10864#endif
10865 PL_encoding = sv_dup(proto_perl->Iencoding, param);
10866
10867 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
10868 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
10869 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
10870
10871 /* Clone the regex array */
10872 PL_regex_padav = newAV();
10873 {
10874 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 10875 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 10876 IV i;
7f466ec7 10877 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 10878 for(i = 1; i <= len; i++) {
7a5b473e
AL
10879 const SV * const regex = regexen[i];
10880 SV * const sv =
10881 SvREPADTMP(regex)
10882 ? sv_dup_inc(regex, param)
10883 : SvREFCNT_inc(
10884 newSViv(PTR2IV(re_dup(
10885 INT2PTR(REGEXP *, SvIVX(regex)), param))))
10886 ;
10887 av_push(PL_regex_padav, sv);
bd81e77b
NC
10888 }
10889 }
10890 PL_regex_pad = AvARRAY(PL_regex_padav);
10891
10892 /* shortcuts to various I/O objects */
10893 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
10894 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
10895 PL_defgv = gv_dup(proto_perl->Idefgv, param);
10896 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
10897 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
10898 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 10899
bd81e77b
NC
10900 /* shortcuts to regexp stuff */
10901 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 10902
bd81e77b
NC
10903 /* shortcuts to misc objects */
10904 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 10905
bd81e77b
NC
10906 /* shortcuts to debugging objects */
10907 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
10908 PL_DBline = gv_dup(proto_perl->IDBline, param);
10909 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
10910 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
10911 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
10912 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
10913 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
10914 PL_lineary = av_dup(proto_perl->Ilineary, param);
10915 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 10916
bd81e77b
NC
10917 /* symbol tables */
10918 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
10919 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
10920 PL_debstash = hv_dup(proto_perl->Idebstash, param);
10921 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
10922 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
10923
10924 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
10925 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
10926 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
10927 PL_endav = av_dup_inc(proto_perl->Iendav, param);
10928 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
10929 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
10930
10931 PL_sub_generation = proto_perl->Isub_generation;
10932
10933 /* funky return mechanisms */
10934 PL_forkprocess = proto_perl->Iforkprocess;
10935
10936 /* subprocess state */
10937 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
10938
10939 /* internal state */
10940 PL_maxo = proto_perl->Imaxo;
10941 if (proto_perl->Iop_mask)
10942 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
10943 else
bd61b366 10944 PL_op_mask = NULL;
bd81e77b
NC
10945 /* PL_asserting = proto_perl->Iasserting; */
10946
10947 /* current interpreter roots */
10948 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
10949 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
10950 PL_main_start = proto_perl->Imain_start;
10951 PL_eval_root = proto_perl->Ieval_root;
10952 PL_eval_start = proto_perl->Ieval_start;
10953
10954 /* runtime control stuff */
10955 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
10956 PL_copline = proto_perl->Icopline;
10957
10958 PL_filemode = proto_perl->Ifilemode;
10959 PL_lastfd = proto_perl->Ilastfd;
10960 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
10961 PL_Argv = NULL;
bd61b366 10962 PL_Cmd = NULL;
bd81e77b
NC
10963 PL_gensym = proto_perl->Igensym;
10964 PL_preambled = proto_perl->Ipreambled;
10965 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
10966 PL_laststatval = proto_perl->Ilaststatval;
10967 PL_laststype = proto_perl->Ilaststype;
a0714e2c 10968 PL_mess_sv = NULL;
bd81e77b
NC
10969
10970 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
10971
10972 /* interpreter atexit processing */
10973 PL_exitlistlen = proto_perl->Iexitlistlen;
10974 if (PL_exitlistlen) {
10975 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
10976 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 10977 }
bd81e77b
NC
10978 else
10979 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
10980
10981 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 10982 if (PL_my_cxt_size) {
f16dd614
DM
10983 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
10984 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
10985 }
10986 else
10987 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
10988 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
10989 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
10990 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
10991
10992 PL_profiledata = NULL;
10993 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
10994 /* PL_rsfp_filters entries have fake IoDIRP() */
10995 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 10996
bd81e77b 10997 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 10998
bd81e77b 10999 PAD_CLONE_VARS(proto_perl, param);
9660f481 11000
bd81e77b
NC
11001#ifdef HAVE_INTERP_INTERN
11002 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11003#endif
645c22ef 11004
bd81e77b
NC
11005 /* more statics moved here */
11006 PL_generation = proto_perl->Igeneration;
11007 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11008
bd81e77b
NC
11009 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11010 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11011
bd81e77b
NC
11012 PL_uid = proto_perl->Iuid;
11013 PL_euid = proto_perl->Ieuid;
11014 PL_gid = proto_perl->Igid;
11015 PL_egid = proto_perl->Iegid;
11016 PL_nomemok = proto_perl->Inomemok;
11017 PL_an = proto_perl->Ian;
11018 PL_evalseq = proto_perl->Ievalseq;
11019 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11020 PL_origalen = proto_perl->Iorigalen;
11021#ifdef PERL_USES_PL_PIDSTATUS
11022 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11023#endif
11024 PL_osname = SAVEPV(proto_perl->Iosname);
11025 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11026
bd81e77b 11027 PL_runops = proto_perl->Irunops;
6a78b4db 11028
bd81e77b 11029 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11030
bd81e77b
NC
11031#ifdef CSH
11032 PL_cshlen = proto_perl->Icshlen;
11033 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11034#endif
645c22ef 11035
bd81e77b
NC
11036 PL_lex_state = proto_perl->Ilex_state;
11037 PL_lex_defer = proto_perl->Ilex_defer;
11038 PL_lex_expect = proto_perl->Ilex_expect;
11039 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11040 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11041 PL_lex_starts = proto_perl->Ilex_starts;
11042 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11043 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11044 PL_lex_op = proto_perl->Ilex_op;
11045 PL_lex_inpat = proto_perl->Ilex_inpat;
11046 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11047 PL_lex_brackets = proto_perl->Ilex_brackets;
11048 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11049 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11050 PL_lex_casemods = proto_perl->Ilex_casemods;
11051 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11052 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11053
5db06880
NC
11054#ifdef PERL_MAD
11055 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11056 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11057 PL_realtokenstart = proto_perl->Irealtokenstart;
11058 PL_faketokens = proto_perl->Ifaketokens;
11059 PL_thismad = proto_perl->Ithismad;
11060 PL_thistoken = proto_perl->Ithistoken;
11061 PL_thisopen = proto_perl->Ithisopen;
11062 PL_thisstuff = proto_perl->Ithisstuff;
11063 PL_thisclose = proto_perl->Ithisclose;
11064 PL_thiswhite = proto_perl->Ithiswhite;
11065 PL_nextwhite = proto_perl->Inextwhite;
11066 PL_skipwhite = proto_perl->Iskipwhite;
11067 PL_endwhite = proto_perl->Iendwhite;
11068 PL_curforce = proto_perl->Icurforce;
5db06880 11069#else
bd81e77b
NC
11070 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11071 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11072 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11073#endif
c43294b8 11074
bd81e77b
NC
11075 /* XXX This is probably masking the deeper issue of why
11076 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11077 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11078 * (A little debugging with a watchpoint on it may help.)
11079 */
11080 if (SvANY(proto_perl->Ilinestr)) {
11081 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11082 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11083 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11084 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11085 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11086 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11087 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11088 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11089 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11090 }
11091 else {
561b68a9 11092 PL_linestr = newSV(79);
bd81e77b
NC
11093 sv_upgrade(PL_linestr,SVt_PVIV);
11094 sv_setpvn(PL_linestr,"",0);
11095 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11096 }
11097 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11098 PL_pending_ident = proto_perl->Ipending_ident;
11099 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11100
11101 PL_expect = proto_perl->Iexpect;
11102
11103 PL_multi_start = proto_perl->Imulti_start;
11104 PL_multi_end = proto_perl->Imulti_end;
11105 PL_multi_open = proto_perl->Imulti_open;
11106 PL_multi_close = proto_perl->Imulti_close;
11107
11108 PL_error_count = proto_perl->Ierror_count;
11109 PL_subline = proto_perl->Isubline;
11110 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11111
bd81e77b
NC
11112 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11113 if (SvANY(proto_perl->Ilinestr)) {
11114 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11115 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11116 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11117 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11118 PL_last_lop_op = proto_perl->Ilast_lop_op;
11119 }
11120 else {
11121 PL_last_uni = SvPVX(PL_linestr);
11122 PL_last_lop = SvPVX(PL_linestr);
11123 PL_last_lop_op = 0;
11124 }
11125 PL_in_my = proto_perl->Iin_my;
11126 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11127#ifdef FCRYPT
11128 PL_cryptseen = proto_perl->Icryptseen;
11129#endif
1d7c1841 11130
bd81e77b 11131 PL_hints = proto_perl->Ihints;
1d7c1841 11132
bd81e77b 11133 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11134
bd81e77b
NC
11135#ifdef USE_LOCALE_COLLATE
11136 PL_collation_ix = proto_perl->Icollation_ix;
11137 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11138 PL_collation_standard = proto_perl->Icollation_standard;
11139 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11140 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11141#endif /* USE_LOCALE_COLLATE */
1d7c1841 11142
bd81e77b
NC
11143#ifdef USE_LOCALE_NUMERIC
11144 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11145 PL_numeric_standard = proto_perl->Inumeric_standard;
11146 PL_numeric_local = proto_perl->Inumeric_local;
11147 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11148#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11149
bd81e77b
NC
11150 /* utf8 character classes */
11151 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11152 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11153 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11154 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11155 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11156 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11157 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11158 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11159 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11160 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11161 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11162 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11163 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11164 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11165 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11166 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11167 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11168 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11169 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11170 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11171
bd81e77b
NC
11172 /* Did the locale setup indicate UTF-8? */
11173 PL_utf8locale = proto_perl->Iutf8locale;
11174 /* Unicode features (see perlrun/-C) */
11175 PL_unicode = proto_perl->Iunicode;
1d7c1841 11176
bd81e77b
NC
11177 /* Pre-5.8 signals control */
11178 PL_signals = proto_perl->Isignals;
1d7c1841 11179
bd81e77b
NC
11180 /* times() ticks per second */
11181 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11182
bd81e77b
NC
11183 /* Recursion stopper for PerlIO_find_layer */
11184 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11185
bd81e77b
NC
11186 /* sort() routine */
11187 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11188
bd81e77b
NC
11189 /* Not really needed/useful since the reenrant_retint is "volatile",
11190 * but do it for consistency's sake. */
11191 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11192
bd81e77b
NC
11193 /* Hooks to shared SVs and locks. */
11194 PL_sharehook = proto_perl->Isharehook;
11195 PL_lockhook = proto_perl->Ilockhook;
11196 PL_unlockhook = proto_perl->Iunlockhook;
11197 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11198
bd81e77b
NC
11199 PL_runops_std = proto_perl->Irunops_std;
11200 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11201
bd81e77b
NC
11202#ifdef THREADS_HAVE_PIDS
11203 PL_ppid = proto_perl->Ippid;
11204#endif
1d7c1841 11205
bd81e77b 11206 /* swatch cache */
5c284bb0 11207 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11208 PL_last_swash_klen = 0;
11209 PL_last_swash_key[0]= '\0';
11210 PL_last_swash_tmps = (U8*)NULL;
11211 PL_last_swash_slen = 0;
1d7c1841 11212
bd81e77b
NC
11213 PL_glob_index = proto_perl->Iglob_index;
11214 PL_srand_called = proto_perl->Isrand_called;
11215 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11216 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11217
bd81e77b
NC
11218 if (proto_perl->Ipsig_pend) {
11219 Newxz(PL_psig_pend, SIG_SIZE, int);
11220 }
11221 else {
11222 PL_psig_pend = (int*)NULL;
11223 }
05ec9bb3 11224
bd81e77b
NC
11225 if (proto_perl->Ipsig_ptr) {
11226 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11227 Newxz(PL_psig_name, SIG_SIZE, SV*);
11228 for (i = 1; i < SIG_SIZE; i++) {
11229 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11230 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11231 }
11232 }
11233 else {
11234 PL_psig_ptr = (SV**)NULL;
11235 PL_psig_name = (SV**)NULL;
11236 }
05ec9bb3 11237
bd81e77b 11238 /* thrdvar.h stuff */
1d7c1841 11239
bd81e77b
NC
11240 if (flags & CLONEf_COPY_STACKS) {
11241 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11242 PL_tmps_ix = proto_perl->Ttmps_ix;
11243 PL_tmps_max = proto_perl->Ttmps_max;
11244 PL_tmps_floor = proto_perl->Ttmps_floor;
11245 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11246 i = 0;
11247 while (i <= PL_tmps_ix) {
11248 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11249 ++i;
11250 }
d2d73c3e 11251
bd81e77b
NC
11252 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11253 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11254 Newxz(PL_markstack, i, I32);
11255 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11256 - proto_perl->Tmarkstack);
11257 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11258 - proto_perl->Tmarkstack);
11259 Copy(proto_perl->Tmarkstack, PL_markstack,
11260 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11261
bd81e77b
NC
11262 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11263 * NOTE: unlike the others! */
11264 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11265 PL_scopestack_max = proto_perl->Tscopestack_max;
11266 Newxz(PL_scopestack, PL_scopestack_max, I32);
11267 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11268
bd81e77b
NC
11269 /* NOTE: si_dup() looks at PL_markstack */
11270 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11271
bd81e77b
NC
11272 /* PL_curstack = PL_curstackinfo->si_stack; */
11273 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11274 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11275
bd81e77b
NC
11276 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11277 PL_stack_base = AvARRAY(PL_curstack);
11278 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11279 - proto_perl->Tstack_base);
11280 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11281
bd81e77b
NC
11282 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11283 * NOTE: unlike the others! */
11284 PL_savestack_ix = proto_perl->Tsavestack_ix;
11285 PL_savestack_max = proto_perl->Tsavestack_max;
11286 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11287 PL_savestack = ss_dup(proto_perl, param);
11288 }
11289 else {
11290 init_stacks();
11291 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11292
11293 /* although we're not duplicating the tmps stack, we should still
11294 * add entries for any SVs on the tmps stack that got cloned by a
11295 * non-refcount means (eg a temp in @_); otherwise they will be
11296 * orphaned
11297 */
11298 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11299 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11300 proto_perl->Ttmps_stack[i]);
11301 if (nsv && !SvREFCNT(nsv)) {
11302 EXTEND_MORTAL(1);
b37c2d43 11303 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11304 }
11305 }
bd81e77b 11306 }
1d7c1841 11307
bd81e77b
NC
11308 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11309 PL_top_env = &PL_start_env;
1d7c1841 11310
bd81e77b 11311 PL_op = proto_perl->Top;
4a4c6fe3 11312
a0714e2c 11313 PL_Sv = NULL;
bd81e77b
NC
11314 PL_Xpv = (XPV*)NULL;
11315 PL_na = proto_perl->Tna;
1fcf4c12 11316
bd81e77b
NC
11317 PL_statbuf = proto_perl->Tstatbuf;
11318 PL_statcache = proto_perl->Tstatcache;
11319 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11320 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11321#ifdef HAS_TIMES
11322 PL_timesbuf = proto_perl->Ttimesbuf;
11323#endif
1d7c1841 11324
bd81e77b
NC
11325 PL_tainted = proto_perl->Ttainted;
11326 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11327 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11328 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11329 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11330 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11331 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11332 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11333 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11334 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11335
bd81e77b
NC
11336 PL_restartop = proto_perl->Trestartop;
11337 PL_in_eval = proto_perl->Tin_eval;
11338 PL_delaymagic = proto_perl->Tdelaymagic;
11339 PL_dirty = proto_perl->Tdirty;
11340 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11341
bd81e77b 11342 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11343 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11344 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11345 PL_lastgotoprobe = NULL;
bd81e77b 11346 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11347
bd81e77b
NC
11348 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11349 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11350 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11351 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11352 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11353 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11354
bd81e77b 11355 /* regex stuff */
1d7c1841 11356
bd81e77b
NC
11357 PL_screamfirst = NULL;
11358 PL_screamnext = NULL;
11359 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11360 PL_lastscream = NULL;
1d7c1841 11361
bd81e77b 11362 PL_watchaddr = NULL;
bd61b366 11363 PL_watchok = NULL;
1d7c1841 11364
bd81e77b 11365 PL_regdummy = proto_perl->Tregdummy;
bd61b366 11366 PL_regprecomp = NULL;
bd81e77b
NC
11367 PL_regnpar = 0;
11368 PL_regsize = 0;
11369 PL_colorset = 0; /* reinits PL_colors[] */
11370 /*PL_colors[6] = {0,0,0,0,0,0};*/
bd61b366
SS
11371 PL_reginput = NULL;
11372 PL_regbol = NULL;
11373 PL_regeol = NULL;
bd81e77b
NC
11374 PL_regstartp = (I32*)NULL;
11375 PL_regendp = (I32*)NULL;
11376 PL_reglastparen = (U32*)NULL;
11377 PL_reglastcloseparen = (U32*)NULL;
bd61b366 11378 PL_regtill = NULL;
bd81e77b
NC
11379 PL_reg_start_tmp = (char**)NULL;
11380 PL_reg_start_tmpl = 0;
11381 PL_regdata = (struct reg_data*)NULL;
bd61b366 11382 PL_bostr = NULL;
bd81e77b
NC
11383 PL_reg_flags = 0;
11384 PL_reg_eval_set = 0;
11385 PL_regnarrate = 0;
11386 PL_regprogram = (regnode*)NULL;
11387 PL_regindent = 0;
11388 PL_regcc = (CURCUR*)NULL;
11389 PL_reg_call_cc = (struct re_cc_state*)NULL;
11390 PL_reg_re = (regexp*)NULL;
bd61b366 11391 PL_reg_ganch = NULL;
a0714e2c 11392 PL_reg_sv = NULL;
bd81e77b
NC
11393 PL_reg_match_utf8 = FALSE;
11394 PL_reg_magic = (MAGIC*)NULL;
11395 PL_reg_oldpos = 0;
11396 PL_reg_oldcurpm = (PMOP*)NULL;
11397 PL_reg_curpm = (PMOP*)NULL;
bd61b366 11398 PL_reg_oldsaved = NULL;
bd81e77b
NC
11399 PL_reg_oldsavedlen = 0;
11400#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 11401 PL_nrs = NULL;
bd81e77b
NC
11402#endif
11403 PL_reg_maxiter = 0;
11404 PL_reg_leftiter = 0;
bd61b366 11405 PL_reg_poscache = NULL;
bd81e77b 11406 PL_reg_poscache_size= 0;
1d7c1841 11407
bd81e77b
NC
11408 /* RE engine - function pointers */
11409 PL_regcompp = proto_perl->Tregcompp;
11410 PL_regexecp = proto_perl->Tregexecp;
11411 PL_regint_start = proto_perl->Tregint_start;
11412 PL_regint_string = proto_perl->Tregint_string;
11413 PL_regfree = proto_perl->Tregfree;
1d7c1841 11414
bd81e77b
NC
11415 PL_reginterp_cnt = 0;
11416 PL_reg_starttry = 0;
1d7c1841 11417
bd81e77b
NC
11418 /* Pluggable optimizer */
11419 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11420
bd81e77b 11421 PL_stashcache = newHV();
1d7c1841 11422
bd81e77b
NC
11423 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11424 ptr_table_free(PL_ptr_table);
11425 PL_ptr_table = NULL;
11426 }
1d7c1841 11427
bd81e77b
NC
11428 /* Call the ->CLONE method, if it exists, for each of the stashes
11429 identified by sv_dup() above.
11430 */
11431 while(av_len(param->stashes) != -1) {
11432 HV* const stash = (HV*) av_shift(param->stashes);
11433 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11434 if (cloner && GvCV(cloner)) {
11435 dSP;
11436 ENTER;
11437 SAVETMPS;
11438 PUSHMARK(SP);
11439 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11440 PUTBACK;
11441 call_sv((SV*)GvCV(cloner), G_DISCARD);
11442 FREETMPS;
11443 LEAVE;
11444 }
1d7c1841 11445 }
1d7c1841 11446
bd81e77b 11447 SvREFCNT_dec(param->stashes);
1d7c1841 11448
bd81e77b
NC
11449 /* orphaned? eg threads->new inside BEGIN or use */
11450 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11451 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11452 SAVEFREESV(PL_compcv);
11453 }
dd2155a4 11454
bd81e77b
NC
11455 return my_perl;
11456}
1d7c1841 11457
bd81e77b 11458#endif /* USE_ITHREADS */
1d7c1841 11459
bd81e77b
NC
11460/*
11461=head1 Unicode Support
1d7c1841 11462
bd81e77b 11463=for apidoc sv_recode_to_utf8
1d7c1841 11464
bd81e77b
NC
11465The encoding is assumed to be an Encode object, on entry the PV
11466of the sv is assumed to be octets in that encoding, and the sv
11467will be converted into Unicode (and UTF-8).
1d7c1841 11468
bd81e77b
NC
11469If the sv already is UTF-8 (or if it is not POK), or if the encoding
11470is not a reference, nothing is done to the sv. If the encoding is not
11471an C<Encode::XS> Encoding object, bad things will happen.
11472(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11473
bd81e77b 11474The PV of the sv is returned.
1d7c1841 11475
bd81e77b 11476=cut */
1d7c1841 11477
bd81e77b
NC
11478char *
11479Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11480{
11481 dVAR;
11482 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11483 SV *uni;
11484 STRLEN len;
11485 const char *s;
11486 dSP;
11487 ENTER;
11488 SAVETMPS;
11489 save_re_context();
11490 PUSHMARK(sp);
11491 EXTEND(SP, 3);
11492 XPUSHs(encoding);
11493 XPUSHs(sv);
11494/*
11495 NI-S 2002/07/09
11496 Passing sv_yes is wrong - it needs to be or'ed set of constants
11497 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11498 remove converted chars from source.
1d7c1841 11499
bd81e77b 11500 Both will default the value - let them.
1d7c1841 11501
bd81e77b
NC
11502 XPUSHs(&PL_sv_yes);
11503*/
11504 PUTBACK;
11505 call_method("decode", G_SCALAR);
11506 SPAGAIN;
11507 uni = POPs;
11508 PUTBACK;
11509 s = SvPV_const(uni, len);
11510 if (s != SvPVX_const(sv)) {
11511 SvGROW(sv, len + 1);
11512 Move(s, SvPVX(sv), len + 1, char);
11513 SvCUR_set(sv, len);
11514 }
11515 FREETMPS;
11516 LEAVE;
11517 SvUTF8_on(sv);
11518 return SvPVX(sv);
389edf32 11519 }
bd81e77b
NC
11520 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11521}
1d7c1841 11522
bd81e77b
NC
11523/*
11524=for apidoc sv_cat_decode
1d7c1841 11525
bd81e77b
NC
11526The encoding is assumed to be an Encode object, the PV of the ssv is
11527assumed to be octets in that encoding and decoding the input starts
11528from the position which (PV + *offset) pointed to. The dsv will be
11529concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11530when the string tstr appears in decoding output or the input ends on
11531the PV of the ssv. The value which the offset points will be modified
11532to the last input position on the ssv.
1d7c1841 11533
bd81e77b 11534Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11535
bd81e77b
NC
11536=cut */
11537
11538bool
11539Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11540 SV *ssv, int *offset, char *tstr, int tlen)
11541{
11542 dVAR;
11543 bool ret = FALSE;
11544 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11545 SV *offsv;
11546 dSP;
11547 ENTER;
11548 SAVETMPS;
11549 save_re_context();
11550 PUSHMARK(sp);
11551 EXTEND(SP, 6);
11552 XPUSHs(encoding);
11553 XPUSHs(dsv);
11554 XPUSHs(ssv);
11555 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11556 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11557 PUTBACK;
11558 call_method("cat_decode", G_SCALAR);
11559 SPAGAIN;
11560 ret = SvTRUE(TOPs);
11561 *offset = SvIV(offsv);
11562 PUTBACK;
11563 FREETMPS;
11564 LEAVE;
389edf32 11565 }
bd81e77b
NC
11566 else
11567 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11568 return ret;
1d7c1841 11569
bd81e77b 11570}
1d7c1841 11571
bd81e77b
NC
11572/* ---------------------------------------------------------------------
11573 *
11574 * support functions for report_uninit()
11575 */
1d7c1841 11576
bd81e77b
NC
11577/* the maxiumum size of array or hash where we will scan looking
11578 * for the undefined element that triggered the warning */
1d7c1841 11579
bd81e77b 11580#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11581
bd81e77b
NC
11582/* Look for an entry in the hash whose value has the same SV as val;
11583 * If so, return a mortal copy of the key. */
1d7c1841 11584
bd81e77b
NC
11585STATIC SV*
11586S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11587{
11588 dVAR;
11589 register HE **array;
11590 I32 i;
6c3182a5 11591
bd81e77b
NC
11592 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11593 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11594 return NULL;
6c3182a5 11595
bd81e77b 11596 array = HvARRAY(hv);
6c3182a5 11597
bd81e77b
NC
11598 for (i=HvMAX(hv); i>0; i--) {
11599 register HE *entry;
11600 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11601 if (HeVAL(entry) != val)
11602 continue;
11603 if ( HeVAL(entry) == &PL_sv_undef ||
11604 HeVAL(entry) == &PL_sv_placeholder)
11605 continue;
11606 if (!HeKEY(entry))
a0714e2c 11607 return NULL;
bd81e77b
NC
11608 if (HeKLEN(entry) == HEf_SVKEY)
11609 return sv_mortalcopy(HeKEY_sv(entry));
11610 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11611 }
11612 }
a0714e2c 11613 return NULL;
bd81e77b 11614}
6c3182a5 11615
bd81e77b
NC
11616/* Look for an entry in the array whose value has the same SV as val;
11617 * If so, return the index, otherwise return -1. */
6c3182a5 11618
bd81e77b
NC
11619STATIC I32
11620S_find_array_subscript(pTHX_ AV *av, SV* val)
11621{
97aff369 11622 dVAR;
bd81e77b
NC
11623 SV** svp;
11624 I32 i;
11625 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11626 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11627 return -1;
57c6e6d2 11628
bd81e77b
NC
11629 svp = AvARRAY(av);
11630 for (i=AvFILLp(av); i>=0; i--) {
11631 if (svp[i] == val && svp[i] != &PL_sv_undef)
11632 return i;
11633 }
11634 return -1;
11635}
15a5279a 11636
bd81e77b
NC
11637/* S_varname(): return the name of a variable, optionally with a subscript.
11638 * If gv is non-zero, use the name of that global, along with gvtype (one
11639 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11640 * targ. Depending on the value of the subscript_type flag, return:
11641 */
bce260cd 11642
bd81e77b
NC
11643#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11644#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11645#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11646#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11647
bd81e77b
NC
11648STATIC SV*
11649S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11650 SV* keyname, I32 aindex, int subscript_type)
11651{
1d7c1841 11652
bd81e77b
NC
11653 SV * const name = sv_newmortal();
11654 if (gv) {
11655 char buffer[2];
11656 buffer[0] = gvtype;
11657 buffer[1] = 0;
1d7c1841 11658
bd81e77b 11659 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11660
bd81e77b 11661 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11662
bd81e77b
NC
11663 if ((unsigned int)SvPVX(name)[1] <= 26) {
11664 buffer[0] = '^';
11665 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11666
bd81e77b
NC
11667 /* Swap the 1 unprintable control character for the 2 byte pretty
11668 version - ie substr($name, 1, 1) = $buffer; */
11669 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11670 }
bd81e77b
NC
11671 }
11672 else {
11673 U32 unused;
11674 CV * const cv = find_runcv(&unused);
11675 SV *sv;
11676 AV *av;
1d7c1841 11677
bd81e77b 11678 if (!cv || !CvPADLIST(cv))
a0714e2c 11679 return NULL;
bd81e77b
NC
11680 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11681 sv = *av_fetch(av, targ, FALSE);
11682 /* SvLEN in a pad name is not to be trusted */
11683 sv_setpv(name, SvPV_nolen_const(sv));
11684 }
1d7c1841 11685
bd81e77b 11686 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11687 SV * const sv = newSV(0);
bd81e77b
NC
11688 *SvPVX(name) = '$';
11689 Perl_sv_catpvf(aTHX_ name, "{%s}",
11690 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11691 SvREFCNT_dec(sv);
11692 }
11693 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11694 *SvPVX(name) = '$';
11695 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11696 }
11697 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11698 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11699
bd81e77b
NC
11700 return name;
11701}
1d7c1841 11702
1d7c1841 11703
bd81e77b
NC
11704/*
11705=for apidoc find_uninit_var
1d7c1841 11706
bd81e77b
NC
11707Find the name of the undefined variable (if any) that caused the operator o
11708to issue a "Use of uninitialized value" warning.
11709If match is true, only return a name if it's value matches uninit_sv.
11710So roughly speaking, if a unary operator (such as OP_COS) generates a
11711warning, then following the direct child of the op may yield an
11712OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11713other hand, with OP_ADD there are two branches to follow, so we only print
11714the variable name if we get an exact match.
1d7c1841 11715
bd81e77b 11716The name is returned as a mortal SV.
1d7c1841 11717
bd81e77b
NC
11718Assumes that PL_op is the op that originally triggered the error, and that
11719PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11720
bd81e77b
NC
11721=cut
11722*/
1d7c1841 11723
bd81e77b
NC
11724STATIC SV *
11725S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11726{
11727 dVAR;
11728 SV *sv;
11729 AV *av;
11730 GV *gv;
11731 OP *o, *o2, *kid;
1d7c1841 11732
bd81e77b
NC
11733 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11734 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11735 return NULL;
1d7c1841 11736
bd81e77b 11737 switch (obase->op_type) {
1d7c1841 11738
bd81e77b
NC
11739 case OP_RV2AV:
11740 case OP_RV2HV:
11741 case OP_PADAV:
11742 case OP_PADHV:
11743 {
11744 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11745 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11746 I32 index = 0;
a0714e2c 11747 SV *keysv = NULL;
bd81e77b 11748 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11749
bd81e77b
NC
11750 if (pad) { /* @lex, %lex */
11751 sv = PAD_SVl(obase->op_targ);
a0714e2c 11752 gv = NULL;
bd81e77b
NC
11753 }
11754 else {
11755 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11756 /* @global, %global */
11757 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11758 if (!gv)
11759 break;
11760 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11761 }
11762 else /* @{expr}, %{expr} */
11763 return find_uninit_var(cUNOPx(obase)->op_first,
11764 uninit_sv, match);
11765 }
1d7c1841 11766
bd81e77b
NC
11767 /* attempt to find a match within the aggregate */
11768 if (hash) {
11769 keysv = S_find_hash_subscript(aTHX_ (HV*)sv, uninit_sv);
11770 if (keysv)
11771 subscript_type = FUV_SUBSCRIPT_HASH;
11772 }
11773 else {
11774 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv);
11775 if (index >= 0)
11776 subscript_type = FUV_SUBSCRIPT_ARRAY;
11777 }
1d7c1841 11778
bd81e77b
NC
11779 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11780 break;
1d7c1841 11781
bd81e77b
NC
11782 return varname(gv, hash ? '%' : '@', obase->op_targ,
11783 keysv, index, subscript_type);
11784 }
1d7c1841 11785
bd81e77b
NC
11786 case OP_PADSV:
11787 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11788 break;
a0714e2c
SS
11789 return varname(NULL, '$', obase->op_targ,
11790 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11791
bd81e77b
NC
11792 case OP_GVSV:
11793 gv = cGVOPx_gv(obase);
11794 if (!gv || (match && GvSV(gv) != uninit_sv))
11795 break;
a0714e2c 11796 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11797
bd81e77b
NC
11798 case OP_AELEMFAST:
11799 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11800 if (match) {
11801 SV **svp;
11802 av = (AV*)PAD_SV(obase->op_targ);
11803 if (!av || SvRMAGICAL(av))
11804 break;
11805 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11806 if (!svp || *svp != uninit_sv)
11807 break;
11808 }
a0714e2c
SS
11809 return varname(NULL, '$', obase->op_targ,
11810 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11811 }
11812 else {
11813 gv = cGVOPx_gv(obase);
11814 if (!gv)
11815 break;
11816 if (match) {
11817 SV **svp;
11818 av = GvAV(gv);
11819 if (!av || SvRMAGICAL(av))
11820 break;
11821 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11822 if (!svp || *svp != uninit_sv)
11823 break;
11824 }
11825 return varname(gv, '$', 0,
a0714e2c 11826 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11827 }
11828 break;
1d7c1841 11829
bd81e77b
NC
11830 case OP_EXISTS:
11831 o = cUNOPx(obase)->op_first;
11832 if (!o || o->op_type != OP_NULL ||
11833 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11834 break;
11835 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11836
bd81e77b
NC
11837 case OP_AELEM:
11838 case OP_HELEM:
11839 if (PL_op == obase)
11840 /* $a[uninit_expr] or $h{uninit_expr} */
11841 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11842
a0714e2c 11843 gv = NULL;
bd81e77b
NC
11844 o = cBINOPx(obase)->op_first;
11845 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11846
bd81e77b 11847 /* get the av or hv, and optionally the gv */
a0714e2c 11848 sv = NULL;
bd81e77b
NC
11849 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11850 sv = PAD_SV(o->op_targ);
11851 }
11852 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11853 && cUNOPo->op_first->op_type == OP_GV)
11854 {
11855 gv = cGVOPx_gv(cUNOPo->op_first);
11856 if (!gv)
11857 break;
11858 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11859 }
11860 if (!sv)
11861 break;
11862
11863 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11864 /* index is constant */
11865 if (match) {
11866 if (SvMAGICAL(sv))
11867 break;
11868 if (obase->op_type == OP_HELEM) {
11869 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11870 if (!he || HeVAL(he) != uninit_sv)
11871 break;
11872 }
11873 else {
00b6aa41 11874 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11875 if (!svp || *svp != uninit_sv)
11876 break;
11877 }
11878 }
11879 if (obase->op_type == OP_HELEM)
11880 return varname(gv, '%', o->op_targ,
11881 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11882 else
a0714e2c 11883 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11884 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11885 }
11886 else {
11887 /* index is an expression;
11888 * attempt to find a match within the aggregate */
11889 if (obase->op_type == OP_HELEM) {
11890 SV * const keysv = S_find_hash_subscript(aTHX_ (HV*)sv, uninit_sv);
11891 if (keysv)
11892 return varname(gv, '%', o->op_targ,
11893 keysv, 0, FUV_SUBSCRIPT_HASH);
11894 }
11895 else {
11896 const I32 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv);
11897 if (index >= 0)
11898 return varname(gv, '@', o->op_targ,
a0714e2c 11899 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11900 }
11901 if (match)
11902 break;
11903 return varname(gv,
11904 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
11905 ? '@' : '%',
a0714e2c 11906 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 11907 }
bd81e77b 11908 break;
dc507217 11909
bd81e77b
NC
11910 case OP_AASSIGN:
11911 /* only examine RHS */
11912 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 11913
bd81e77b
NC
11914 case OP_OPEN:
11915 o = cUNOPx(obase)->op_first;
11916 if (o->op_type == OP_PUSHMARK)
11917 o = o->op_sibling;
1d7c1841 11918
bd81e77b
NC
11919 if (!o->op_sibling) {
11920 /* one-arg version of open is highly magical */
a0ae6670 11921
bd81e77b
NC
11922 if (o->op_type == OP_GV) { /* open FOO; */
11923 gv = cGVOPx_gv(o);
11924 if (match && GvSV(gv) != uninit_sv)
11925 break;
11926 return varname(gv, '$', 0,
a0714e2c 11927 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
11928 }
11929 /* other possibilities not handled are:
11930 * open $x; or open my $x; should return '${*$x}'
11931 * open expr; should return '$'.expr ideally
11932 */
11933 break;
11934 }
11935 goto do_op;
ccfc67b7 11936
bd81e77b
NC
11937 /* ops where $_ may be an implicit arg */
11938 case OP_TRANS:
11939 case OP_SUBST:
11940 case OP_MATCH:
11941 if ( !(obase->op_flags & OPf_STACKED)) {
11942 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
11943 ? PAD_SVl(obase->op_targ)
11944 : DEFSV))
11945 {
11946 sv = sv_newmortal();
11947 sv_setpvn(sv, "$_", 2);
11948 return sv;
11949 }
11950 }
11951 goto do_op;
9f4817db 11952
bd81e77b
NC
11953 case OP_PRTF:
11954 case OP_PRINT:
11955 /* skip filehandle as it can't produce 'undef' warning */
11956 o = cUNOPx(obase)->op_first;
11957 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
11958 o = o->op_sibling->op_sibling;
11959 goto do_op2;
9f4817db 11960
9f4817db 11961
bd81e77b
NC
11962 case OP_RV2SV:
11963 case OP_CUSTOM:
11964 case OP_ENTERSUB:
11965 match = 1; /* XS or custom code could trigger random warnings */
11966 goto do_op;
9f4817db 11967
bd81e77b
NC
11968 case OP_SCHOMP:
11969 case OP_CHOMP:
11970 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 11971 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 11972 /*FALLTHROUGH*/
5d170f3a 11973
bd81e77b
NC
11974 default:
11975 do_op:
11976 if (!(obase->op_flags & OPf_KIDS))
11977 break;
11978 o = cUNOPx(obase)->op_first;
11979
11980 do_op2:
11981 if (!o)
11982 break;
f9893866 11983
bd81e77b
NC
11984 /* if all except one arg are constant, or have no side-effects,
11985 * or are optimized away, then it's unambiguous */
5f66b61c 11986 o2 = NULL;
bd81e77b
NC
11987 for (kid=o; kid; kid = kid->op_sibling) {
11988 if (kid &&
11989 ( (kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid)))
11990 || (kid->op_type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
11991 || (kid->op_type == OP_PUSHMARK)
11992 )
11993 )
11994 continue;
11995 if (o2) { /* more than one found */
5f66b61c 11996 o2 = NULL;
bd81e77b
NC
11997 break;
11998 }
11999 o2 = kid;
12000 }
12001 if (o2)
12002 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12003
bd81e77b
NC
12004 /* scan all args */
12005 while (o) {
12006 sv = find_uninit_var(o, uninit_sv, 1);
12007 if (sv)
12008 return sv;
12009 o = o->op_sibling;
d0063567 12010 }
bd81e77b 12011 break;
f9893866 12012 }
a0714e2c 12013 return NULL;
9f4817db
JH
12014}
12015
220e2d4e 12016
bd81e77b
NC
12017/*
12018=for apidoc report_uninit
68795e93 12019
bd81e77b 12020Print appropriate "Use of uninitialized variable" warning
220e2d4e 12021
bd81e77b
NC
12022=cut
12023*/
220e2d4e 12024
bd81e77b
NC
12025void
12026Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12027{
97aff369 12028 dVAR;
bd81e77b 12029 if (PL_op) {
a0714e2c 12030 SV* varname = NULL;
bd81e77b
NC
12031 if (uninit_sv) {
12032 varname = find_uninit_var(PL_op, uninit_sv,0);
12033 if (varname)
12034 sv_insert(varname, 0, 0, " ", 1);
12035 }
12036 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12037 varname ? SvPV_nolen_const(varname) : "",
12038 " in ", OP_DESC(PL_op));
220e2d4e 12039 }
a73e8557 12040 else
bd81e77b
NC
12041 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12042 "", "", "");
220e2d4e 12043}
f9893866 12044
241d1a3b
NC
12045/*
12046 * Local variables:
12047 * c-indentation-style: bsd
12048 * c-basic-offset: 4
12049 * indent-tabs-mode: t
12050 * End:
12051 *
37442d52
RGS
12052 * ex: set ts=8 sts=4 sw=4 noet:
12053 */