This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
opening $0 doesn't work if you change directory. So use test.pl
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
7272f7c1 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137 32#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 33/* if adding more checks watch out for the following tests:
e23c8137
JH
34 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
35 * lib/utf8.t lib/Unicode/Collate/t/index.t
36 * --jhi
37 */
6f207bd3 38# define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
39 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
40 assert((cache)[2] <= (cache)[3]); \
41 assert((cache)[3] <= (cache)[1]);} \
42 } STMT_END
e23c8137 43#else
6f207bd3 44# define ASSERT_UTF8_CACHE(cache) NOOP
e23c8137
JH
45#endif
46
f8c7b90f 47#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 48#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 49#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 50/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 51 on-write. */
765f542d 52#endif
645c22ef
DM
53
54/* ============================================================================
55
56=head1 Allocation and deallocation of SVs.
57
d2a0f284
JC
58An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
59sv, av, hv...) contains type and reference count information, and for
60many types, a pointer to the body (struct xrv, xpv, xpviv...), which
61contains fields specific to each type. Some types store all they need
62in the head, so don't have a body.
63
64In all but the most memory-paranoid configuations (ex: PURIFY), heads
65and bodies are allocated out of arenas, which by default are
66approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
67Sv-bodies are allocated by their sv-type, guaranteeing size
68consistency needed to allocate safely from arrays.
69
d2a0f284
JC
70For SV-heads, the first slot in each arena is reserved, and holds a
71link to the next arena, some flags, and a note of the number of slots.
72Snaked through each arena chain is a linked list of free items; when
73this becomes empty, an extra arena is allocated and divided up into N
74items which are threaded into the free list.
75
76SV-bodies are similar, but they use arena-sets by default, which
77separate the link and info from the arena itself, and reclaim the 1st
78slot in the arena. SV-bodies are further described later.
645c22ef
DM
79
80The following global variables are associated with arenas:
81
82 PL_sv_arenaroot pointer to list of SV arenas
83 PL_sv_root pointer to list of free SV structures
84
d2a0f284
JC
85 PL_body_arenas head of linked-list of body arenas
86 PL_body_roots[] array of pointers to list of free bodies of svtype
87 arrays are indexed by the svtype needed
93e68bfb 88
d2a0f284
JC
89A few special SV heads are not allocated from an arena, but are
90instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
91The size of arenas can be changed from the default by setting
92PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
93
94The SV arena serves the secondary purpose of allowing still-live SVs
95to be located and destroyed during final cleanup.
96
97At the lowest level, the macros new_SV() and del_SV() grab and free
98an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
99to return the SV to the free list with error checking.) new_SV() calls
100more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
101SVs in the free list have their SvTYPE field set to all ones.
102
ff276b08 103At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 104perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 105start of the interpreter.
645c22ef 106
645c22ef
DM
107The function visit() scans the SV arenas list, and calls a specified
108function for each SV it finds which is still live - ie which has an SvTYPE
109other than all 1's, and a non-zero SvREFCNT. visit() is used by the
110following functions (specified as [function that calls visit()] / [function
111called by visit() for each SV]):
112
113 sv_report_used() / do_report_used()
f2524eef 114 dump all remaining SVs (debugging aid)
645c22ef
DM
115
116 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
117 Attempt to free all objects pointed to by RVs,
118 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
119 try to do the same for all objects indirectly
120 referenced by typeglobs too. Called once from
121 perl_destruct(), prior to calling sv_clean_all()
122 below.
123
124 sv_clean_all() / do_clean_all()
125 SvREFCNT_dec(sv) each remaining SV, possibly
126 triggering an sv_free(). It also sets the
127 SVf_BREAK flag on the SV to indicate that the
128 refcnt has been artificially lowered, and thus
129 stopping sv_free() from giving spurious warnings
130 about SVs which unexpectedly have a refcnt
131 of zero. called repeatedly from perl_destruct()
132 until there are no SVs left.
133
93e68bfb 134=head2 Arena allocator API Summary
645c22ef
DM
135
136Private API to rest of sv.c
137
138 new_SV(), del_SV(),
139
140 new_XIV(), del_XIV(),
141 new_XNV(), del_XNV(),
142 etc
143
144Public API:
145
8cf8f3d1 146 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 147
645c22ef
DM
148=cut
149
150============================================================================ */
151
4561caa4
CS
152/*
153 * "A time to plant, and a time to uproot what was planted..."
154 */
155
77354fb4
NC
156void
157Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
158{
97aff369 159 dVAR;
77354fb4
NC
160 void *new_chunk;
161 U32 new_chunk_size;
77354fb4
NC
162 new_chunk = (void *)(chunk);
163 new_chunk_size = (chunk_size);
164 if (new_chunk_size > PL_nice_chunk_size) {
165 Safefree(PL_nice_chunk);
166 PL_nice_chunk = (char *) new_chunk;
167 PL_nice_chunk_size = new_chunk_size;
168 } else {
169 Safefree(chunk);
170 }
77354fb4 171}
cac9b346 172
fd0854ff 173#ifdef DEBUG_LEAKING_SCALARS
22162ca8 174# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
175#else
176# define FREE_SV_DEBUG_FILE(sv)
177#endif
178
48614a46
NC
179#ifdef PERL_POISON
180# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
181/* Whilst I'd love to do this, it seems that things like to check on
182 unreferenced scalars
7e337ee0 183# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 184*/
7e337ee0
JH
185# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
186 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
187#else
188# define SvARENA_CHAIN(sv) SvANY(sv)
189# define POSION_SV_HEAD(sv)
190#endif
191
053fc874
GS
192#define plant_SV(p) \
193 STMT_START { \
fd0854ff 194 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
195 POSION_SV_HEAD(p); \
196 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
197 SvFLAGS(p) = SVTYPEMASK; \
198 PL_sv_root = (p); \
199 --PL_sv_count; \
200 } STMT_END
a0d0e21e 201
053fc874
GS
202#define uproot_SV(p) \
203 STMT_START { \
204 (p) = PL_sv_root; \
bb7bbd9c 205 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
206 ++PL_sv_count; \
207 } STMT_END
208
645c22ef 209
cac9b346
NC
210/* make some more SVs by adding another arena */
211
cac9b346
NC
212STATIC SV*
213S_more_sv(pTHX)
214{
97aff369 215 dVAR;
cac9b346
NC
216 SV* sv;
217
218 if (PL_nice_chunk) {
219 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 220 PL_nice_chunk = NULL;
cac9b346
NC
221 PL_nice_chunk_size = 0;
222 }
223 else {
224 char *chunk; /* must use New here to match call to */
d2a0f284 225 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 226 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
227 }
228 uproot_SV(sv);
229 return sv;
230}
231
645c22ef
DM
232/* new_SV(): return a new, empty SV head */
233
eba0f806
DM
234#ifdef DEBUG_LEAKING_SCALARS
235/* provide a real function for a debugger to play with */
236STATIC SV*
237S_new_SV(pTHX)
238{
239 SV* sv;
240
eba0f806
DM
241 if (PL_sv_root)
242 uproot_SV(sv);
243 else
cac9b346 244 sv = S_more_sv(aTHX);
eba0f806
DM
245 SvANY(sv) = 0;
246 SvREFCNT(sv) = 1;
247 SvFLAGS(sv) = 0;
fd0854ff
DM
248 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
249 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
250 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
251 sv->sv_debug_inpad = 0;
252 sv->sv_debug_cloned = 0;
fd0854ff 253 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 254
eba0f806
DM
255 return sv;
256}
257# define new_SV(p) (p)=S_new_SV(aTHX)
258
259#else
260# define new_SV(p) \
053fc874 261 STMT_START { \
053fc874
GS
262 if (PL_sv_root) \
263 uproot_SV(p); \
264 else \
cac9b346 265 (p) = S_more_sv(aTHX); \
053fc874
GS
266 SvANY(p) = 0; \
267 SvREFCNT(p) = 1; \
268 SvFLAGS(p) = 0; \
269 } STMT_END
eba0f806 270#endif
463ee0b2 271
645c22ef
DM
272
273/* del_SV(): return an empty SV head to the free list */
274
a0d0e21e 275#ifdef DEBUGGING
4561caa4 276
053fc874
GS
277#define del_SV(p) \
278 STMT_START { \
aea4f609 279 if (DEBUG_D_TEST) \
053fc874
GS
280 del_sv(p); \
281 else \
282 plant_SV(p); \
053fc874 283 } STMT_END
a0d0e21e 284
76e3520e 285STATIC void
cea2e8a9 286S_del_sv(pTHX_ SV *p)
463ee0b2 287{
97aff369 288 dVAR;
aea4f609 289 if (DEBUG_D_TEST) {
4633a7c4 290 SV* sva;
a3b680e6 291 bool ok = 0;
3280af22 292 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
293 const SV * const sv = sva + 1;
294 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 295 if (p >= sv && p < svend) {
a0d0e21e 296 ok = 1;
c0ff570e
NC
297 break;
298 }
a0d0e21e
LW
299 }
300 if (!ok) {
0453d815 301 if (ckWARN_d(WARN_INTERNAL))
9014280d 302 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
303 "Attempt to free non-arena SV: 0x%"UVxf
304 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
305 return;
306 }
307 }
4561caa4 308 plant_SV(p);
463ee0b2 309}
a0d0e21e 310
4561caa4
CS
311#else /* ! DEBUGGING */
312
313#define del_SV(p) plant_SV(p)
314
315#endif /* DEBUGGING */
463ee0b2 316
645c22ef
DM
317
318/*
ccfc67b7
JH
319=head1 SV Manipulation Functions
320
645c22ef
DM
321=for apidoc sv_add_arena
322
323Given a chunk of memory, link it to the head of the list of arenas,
324and split it into a list of free SVs.
325
326=cut
327*/
328
4633a7c4 329void
864dbfa3 330Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 331{
97aff369 332 dVAR;
0bd48802 333 SV* const sva = (SV*)ptr;
463ee0b2
LW
334 register SV* sv;
335 register SV* svend;
4633a7c4
LW
336
337 /* The first SV in an arena isn't an SV. */
3280af22 338 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
339 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
340 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
341
3280af22
NIS
342 PL_sv_arenaroot = sva;
343 PL_sv_root = sva + 1;
4633a7c4
LW
344
345 svend = &sva[SvREFCNT(sva) - 1];
346 sv = sva + 1;
463ee0b2 347 while (sv < svend) {
48614a46 348 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 349#ifdef DEBUGGING
978b032e 350 SvREFCNT(sv) = 0;
03e36789
NC
351#endif
352 /* Must always set typemask because it's awlays checked in on cleanup
353 when the arenas are walked looking for objects. */
8990e307 354 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
355 sv++;
356 }
48614a46 357 SvARENA_CHAIN(sv) = 0;
03e36789
NC
358#ifdef DEBUGGING
359 SvREFCNT(sv) = 0;
360#endif
4633a7c4
LW
361 SvFLAGS(sv) = SVTYPEMASK;
362}
363
055972dc
DM
364/* visit(): call the named function for each non-free SV in the arenas
365 * whose flags field matches the flags/mask args. */
645c22ef 366
5226ed68 367STATIC I32
055972dc 368S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 369{
97aff369 370 dVAR;
4633a7c4 371 SV* sva;
5226ed68 372 I32 visited = 0;
8990e307 373
3280af22 374 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 375 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 376 register SV* sv;
4561caa4 377 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
378 if (SvTYPE(sv) != SVTYPEMASK
379 && (sv->sv_flags & mask) == flags
380 && SvREFCNT(sv))
381 {
acfe0abc 382 (FCALL)(aTHX_ sv);
5226ed68
JH
383 ++visited;
384 }
8990e307
LW
385 }
386 }
5226ed68 387 return visited;
8990e307
LW
388}
389
758a08c3
JH
390#ifdef DEBUGGING
391
645c22ef
DM
392/* called by sv_report_used() for each live SV */
393
394static void
acfe0abc 395do_report_used(pTHX_ SV *sv)
645c22ef
DM
396{
397 if (SvTYPE(sv) != SVTYPEMASK) {
398 PerlIO_printf(Perl_debug_log, "****\n");
399 sv_dump(sv);
400 }
401}
758a08c3 402#endif
645c22ef
DM
403
404/*
405=for apidoc sv_report_used
406
407Dump the contents of all SVs not yet freed. (Debugging aid).
408
409=cut
410*/
411
8990e307 412void
864dbfa3 413Perl_sv_report_used(pTHX)
4561caa4 414{
ff270d3a 415#ifdef DEBUGGING
055972dc 416 visit(do_report_used, 0, 0);
96a5add6
AL
417#else
418 PERL_UNUSED_CONTEXT;
ff270d3a 419#endif
4561caa4
CS
420}
421
645c22ef
DM
422/* called by sv_clean_objs() for each live SV */
423
424static void
e15faf7d 425do_clean_objs(pTHX_ SV *ref)
645c22ef 426{
97aff369 427 dVAR;
ea724faa
NC
428 assert (SvROK(ref));
429 {
823a54a3
AL
430 SV * const target = SvRV(ref);
431 if (SvOBJECT(target)) {
432 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
433 if (SvWEAKREF(ref)) {
434 sv_del_backref(target, ref);
435 SvWEAKREF_off(ref);
436 SvRV_set(ref, NULL);
437 } else {
438 SvROK_off(ref);
439 SvRV_set(ref, NULL);
440 SvREFCNT_dec(target);
441 }
645c22ef
DM
442 }
443 }
444
445 /* XXX Might want to check arrays, etc. */
446}
447
448/* called by sv_clean_objs() for each live SV */
449
450#ifndef DISABLE_DESTRUCTOR_KLUDGE
451static void
acfe0abc 452do_clean_named_objs(pTHX_ SV *sv)
645c22ef 453{
97aff369 454 dVAR;
ea724faa 455 assert(SvTYPE(sv) == SVt_PVGV);
d011219a
NC
456 assert(isGV_with_GP(sv));
457 if (GvGP(sv)) {
c69033f2
NC
458 if ((
459#ifdef PERL_DONT_CREATE_GVSV
460 GvSV(sv) &&
461#endif
462 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
463 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
464 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
465 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
466 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
467 {
468 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 469 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
470 SvREFCNT_dec(sv);
471 }
472 }
473}
474#endif
475
476/*
477=for apidoc sv_clean_objs
478
479Attempt to destroy all objects not yet freed
480
481=cut
482*/
483
4561caa4 484void
864dbfa3 485Perl_sv_clean_objs(pTHX)
4561caa4 486{
97aff369 487 dVAR;
3280af22 488 PL_in_clean_objs = TRUE;
055972dc 489 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 490#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 491 /* some barnacles may yet remain, clinging to typeglobs */
d011219a 492 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
4561caa4 493#endif
3280af22 494 PL_in_clean_objs = FALSE;
4561caa4
CS
495}
496
645c22ef
DM
497/* called by sv_clean_all() for each live SV */
498
499static void
acfe0abc 500do_clean_all(pTHX_ SV *sv)
645c22ef 501{
97aff369 502 dVAR;
645c22ef
DM
503 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
504 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b 505 if (PL_comppad == (AV*)sv) {
7d49f689 506 PL_comppad = NULL;
4608196e 507 PL_curpad = NULL;
0e705b3b 508 }
645c22ef
DM
509 SvREFCNT_dec(sv);
510}
511
512/*
513=for apidoc sv_clean_all
514
515Decrement the refcnt of each remaining SV, possibly triggering a
516cleanup. This function may have to be called multiple times to free
ff276b08 517SVs which are in complex self-referential hierarchies.
645c22ef
DM
518
519=cut
520*/
521
5226ed68 522I32
864dbfa3 523Perl_sv_clean_all(pTHX)
8990e307 524{
97aff369 525 dVAR;
5226ed68 526 I32 cleaned;
3280af22 527 PL_in_clean_all = TRUE;
055972dc 528 cleaned = visit(do_clean_all, 0,0);
3280af22 529 PL_in_clean_all = FALSE;
5226ed68 530 return cleaned;
8990e307 531}
463ee0b2 532
5e258f8c
JC
533/*
534 ARENASETS: a meta-arena implementation which separates arena-info
535 into struct arena_set, which contains an array of struct
536 arena_descs, each holding info for a single arena. By separating
537 the meta-info from the arena, we recover the 1st slot, formerly
538 borrowed for list management. The arena_set is about the size of an
39244528 539 arena, avoiding the needless malloc overhead of a naive linked-list.
5e258f8c
JC
540
541 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
542 memory in the last arena-set (1/2 on average). In trade, we get
543 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284
JC
544 smaller types). The recovery of the wasted space allows use of
545 small arenas for large, rare body types,
5e258f8c 546*/
5e258f8c 547struct arena_desc {
398c677b
NC
548 char *arena; /* the raw storage, allocated aligned */
549 size_t size; /* its size ~4k typ */
0a848332 550 U32 misc; /* type, and in future other things. */
5e258f8c
JC
551};
552
e6148039
NC
553struct arena_set;
554
555/* Get the maximum number of elements in set[] such that struct arena_set
556 will fit within PERL_ARENA_SIZE, which is probabably just under 4K, and
557 therefore likely to be 1 aligned memory page. */
558
559#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
560 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
561
562struct arena_set {
563 struct arena_set* next;
0a848332
NC
564 unsigned int set_size; /* ie ARENAS_PER_SET */
565 unsigned int curr; /* index of next available arena-desc */
5e258f8c
JC
566 struct arena_desc set[ARENAS_PER_SET];
567};
568
645c22ef
DM
569/*
570=for apidoc sv_free_arenas
571
572Deallocate the memory used by all arenas. Note that all the individual SV
573heads and bodies within the arenas must already have been freed.
574
575=cut
576*/
4633a7c4 577void
864dbfa3 578Perl_sv_free_arenas(pTHX)
4633a7c4 579{
97aff369 580 dVAR;
4633a7c4
LW
581 SV* sva;
582 SV* svanext;
0a848332 583 unsigned int i;
4633a7c4
LW
584
585 /* Free arenas here, but be careful about fake ones. (We assume
586 contiguity of the fake ones with the corresponding real ones.) */
587
3280af22 588 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
589 svanext = (SV*) SvANY(sva);
590 while (svanext && SvFAKE(svanext))
591 svanext = (SV*) SvANY(svanext);
592
593 if (!SvFAKE(sva))
1df70142 594 Safefree(sva);
4633a7c4 595 }
93e68bfb 596
5e258f8c 597 {
0a848332
NC
598 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
599
600 while (aroot) {
601 struct arena_set *current = aroot;
602 i = aroot->curr;
603 while (i--) {
5e258f8c
JC
604 assert(aroot->set[i].arena);
605 Safefree(aroot->set[i].arena);
606 }
0a848332
NC
607 aroot = aroot->next;
608 Safefree(current);
5e258f8c
JC
609 }
610 }
dc8220bf 611 PL_body_arenas = 0;
fdda85ca 612
0a848332
NC
613 i = PERL_ARENA_ROOTS_SIZE;
614 while (i--)
93e68bfb 615 PL_body_roots[i] = 0;
93e68bfb 616
43c5f42d 617 Safefree(PL_nice_chunk);
bd61b366 618 PL_nice_chunk = NULL;
3280af22
NIS
619 PL_nice_chunk_size = 0;
620 PL_sv_arenaroot = 0;
621 PL_sv_root = 0;
4633a7c4
LW
622}
623
bd81e77b
NC
624/*
625 Here are mid-level routines that manage the allocation of bodies out
626 of the various arenas. There are 5 kinds of arenas:
29489e7c 627
bd81e77b
NC
628 1. SV-head arenas, which are discussed and handled above
629 2. regular body arenas
630 3. arenas for reduced-size bodies
631 4. Hash-Entry arenas
632 5. pte arenas (thread related)
29489e7c 633
bd81e77b
NC
634 Arena types 2 & 3 are chained by body-type off an array of
635 arena-root pointers, which is indexed by svtype. Some of the
636 larger/less used body types are malloced singly, since a large
637 unused block of them is wasteful. Also, several svtypes dont have
638 bodies; the data fits into the sv-head itself. The arena-root
639 pointer thus has a few unused root-pointers (which may be hijacked
640 later for arena types 4,5)
29489e7c 641
bd81e77b
NC
642 3 differs from 2 as an optimization; some body types have several
643 unused fields in the front of the structure (which are kept in-place
644 for consistency). These bodies can be allocated in smaller chunks,
645 because the leading fields arent accessed. Pointers to such bodies
646 are decremented to point at the unused 'ghost' memory, knowing that
647 the pointers are used with offsets to the real memory.
29489e7c 648
bd81e77b
NC
649 HE, HEK arenas are managed separately, with separate code, but may
650 be merge-able later..
651
652 PTE arenas are not sv-bodies, but they share these mid-level
653 mechanics, so are considered here. The new mid-level mechanics rely
654 on the sv_type of the body being allocated, so we just reserve one
655 of the unused body-slots for PTEs, then use it in those (2) PTE
656 contexts below (line ~10k)
657*/
658
bd26d9a3 659/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
660 TBD: export properly for hv.c: S_more_he().
661*/
662void*
0a848332 663Perl_get_arena(pTHX_ size_t arena_size, U32 misc)
5e258f8c 664{
7a89be66 665 dVAR;
5e258f8c 666 struct arena_desc* adesc;
39244528 667 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
0a848332 668 unsigned int curr;
5e258f8c 669
476a1e16
JC
670 /* shouldnt need this
671 if (!arena_size) arena_size = PERL_ARENA_SIZE;
672 */
5e258f8c
JC
673
674 /* may need new arena-set to hold new arena */
39244528
NC
675 if (!aroot || aroot->curr >= aroot->set_size) {
676 struct arena_set *newroot;
5e258f8c
JC
677 Newxz(newroot, 1, struct arena_set);
678 newroot->set_size = ARENAS_PER_SET;
39244528
NC
679 newroot->next = aroot;
680 aroot = newroot;
681 PL_body_arenas = (void *) newroot;
52944de8 682 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
5e258f8c
JC
683 }
684
685 /* ok, now have arena-set with at least 1 empty/available arena-desc */
39244528
NC
686 curr = aroot->curr++;
687 adesc = &(aroot->set[curr]);
5e258f8c
JC
688 assert(!adesc->arena);
689
89086707 690 Newx(adesc->arena, arena_size, char);
5e258f8c 691 adesc->size = arena_size;
0a848332 692 adesc->misc = misc;
d2a0f284 693 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
6c9570dc 694 curr, (void*)adesc->arena, arena_size));
5e258f8c
JC
695
696 return adesc->arena;
5e258f8c
JC
697}
698
53c1dcc0 699
bd81e77b 700/* return a thing to the free list */
29489e7c 701
bd81e77b
NC
702#define del_body(thing, root) \
703 STMT_START { \
00b6aa41 704 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
705 *thing_copy = *root; \
706 *root = (void*)thing_copy; \
bd81e77b 707 } STMT_END
29489e7c 708
bd81e77b 709/*
d2a0f284
JC
710
711=head1 SV-Body Allocation
712
713Allocation of SV-bodies is similar to SV-heads, differing as follows;
714the allocation mechanism is used for many body types, so is somewhat
715more complicated, it uses arena-sets, and has no need for still-live
716SV detection.
717
718At the outermost level, (new|del)_X*V macros return bodies of the
719appropriate type. These macros call either (new|del)_body_type or
720(new|del)_body_allocated macro pairs, depending on specifics of the
721type. Most body types use the former pair, the latter pair is used to
722allocate body types with "ghost fields".
723
724"ghost fields" are fields that are unused in certain types, and
725consequently dont need to actually exist. They are declared because
726they're part of a "base type", which allows use of functions as
727methods. The simplest examples are AVs and HVs, 2 aggregate types
728which don't use the fields which support SCALAR semantics.
729
730For these types, the arenas are carved up into *_allocated size
731chunks, we thus avoid wasted memory for those unaccessed members.
732When bodies are allocated, we adjust the pointer back in memory by the
733size of the bit not allocated, so it's as if we allocated the full
734structure. (But things will all go boom if you write to the part that
735is "not there", because you'll be overwriting the last members of the
736preceding structure in memory.)
737
738We calculate the correction using the STRUCT_OFFSET macro. For
739example, if xpv_allocated is the same structure as XPV then the two
740OFFSETs sum to zero, and the pointer is unchanged. If the allocated
741structure is smaller (no initial NV actually allocated) then the net
742effect is to subtract the size of the NV from the pointer, to return a
743new pointer as if an initial NV were actually allocated.
744
745This is the same trick as was used for NV and IV bodies. Ironically it
746doesn't need to be used for NV bodies any more, because NV is now at
747the start of the structure. IV bodies don't need it either, because
748they are no longer allocated.
749
750In turn, the new_body_* allocators call S_new_body(), which invokes
751new_body_inline macro, which takes a lock, and takes a body off the
752linked list at PL_body_roots[sv_type], calling S_more_bodies() if
753necessary to refresh an empty list. Then the lock is released, and
754the body is returned.
755
756S_more_bodies calls get_arena(), and carves it up into an array of N
757bodies, which it strings into a linked list. It looks up arena-size
758and body-size from the body_details table described below, thus
759supporting the multiple body-types.
760
761If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
762the (new|del)_X*V macros are mapped directly to malloc/free.
763
764*/
765
766/*
767
768For each sv-type, struct body_details bodies_by_type[] carries
769parameters which control these aspects of SV handling:
770
771Arena_size determines whether arenas are used for this body type, and if
772so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
773zero, forcing individual mallocs and frees.
774
775Body_size determines how big a body is, and therefore how many fit into
776each arena. Offset carries the body-pointer adjustment needed for
777*_allocated body types, and is used in *_allocated macros.
778
779But its main purpose is to parameterize info needed in
780Perl_sv_upgrade(). The info here dramatically simplifies the function
781vs the implementation in 5.8.7, making it table-driven. All fields
782are used for this, except for arena_size.
783
784For the sv-types that have no bodies, arenas are not used, so those
785PL_body_roots[sv_type] are unused, and can be overloaded. In
786something of a special case, SVt_NULL is borrowed for HE arenas;
787PL_body_roots[SVt_NULL] is filled by S_more_he, but the
788bodies_by_type[SVt_NULL] slot is not used, as the table is not
789available in hv.c,
790
791PTEs also use arenas, but are never seen in Perl_sv_upgrade.
792Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
793they can just use the same allocation semantics. At first, PTEs were
794also overloaded to a non-body sv-type, but this yielded hard-to-find
795malloc bugs, so was simplified by claiming a new slot. This choice
796has no consequence at this time.
797
29489e7c
DM
798*/
799
bd81e77b 800struct body_details {
0fb58b32 801 U8 body_size; /* Size to allocate */
10666ae3 802 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 803 U8 offset;
10666ae3
NC
804 unsigned int type : 4; /* We have space for a sanity check. */
805 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
806 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
807 unsigned int arena : 1; /* Allocated from an arena */
808 size_t arena_size; /* Size of arena to allocate */
bd81e77b 809};
29489e7c 810
bd81e77b
NC
811#define HADNV FALSE
812#define NONV TRUE
29489e7c 813
d2a0f284 814
bd81e77b
NC
815#ifdef PURIFY
816/* With -DPURFIY we allocate everything directly, and don't use arenas.
817 This seems a rather elegant way to simplify some of the code below. */
818#define HASARENA FALSE
819#else
820#define HASARENA TRUE
821#endif
822#define NOARENA FALSE
29489e7c 823
d2a0f284
JC
824/* Size the arenas to exactly fit a given number of bodies. A count
825 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
826 simplifying the default. If count > 0, the arena is sized to fit
827 only that many bodies, allowing arenas to be used for large, rare
828 bodies (XPVFM, XPVIO) without undue waste. The arena size is
829 limited by PERL_ARENA_SIZE, so we can safely oversize the
830 declarations.
831 */
95db5f15
MB
832#define FIT_ARENA0(body_size) \
833 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
834#define FIT_ARENAn(count,body_size) \
835 ( count * body_size <= PERL_ARENA_SIZE) \
836 ? count * body_size \
837 : FIT_ARENA0 (body_size)
838#define FIT_ARENA(count,body_size) \
839 count \
840 ? FIT_ARENAn (count, body_size) \
841 : FIT_ARENA0 (body_size)
d2a0f284 842
bd81e77b 843/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 844
bd81e77b
NC
845typedef struct {
846 STRLEN xpv_cur;
847 STRLEN xpv_len;
848} xpv_allocated;
29489e7c 849
bd81e77b 850to make its members accessible via a pointer to (say)
29489e7c 851
bd81e77b
NC
852struct xpv {
853 NV xnv_nv;
854 STRLEN xpv_cur;
855 STRLEN xpv_len;
856};
29489e7c 857
bd81e77b 858*/
29489e7c 859
bd81e77b
NC
860#define relative_STRUCT_OFFSET(longer, shorter, member) \
861 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 862
bd81e77b
NC
863/* Calculate the length to copy. Specifically work out the length less any
864 final padding the compiler needed to add. See the comment in sv_upgrade
865 for why copying the padding proved to be a bug. */
29489e7c 866
bd81e77b
NC
867#define copy_length(type, last_member) \
868 STRUCT_OFFSET(type, last_member) \
869 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 870
bd81e77b 871static const struct body_details bodies_by_type[] = {
10666ae3
NC
872 { sizeof(HE), 0, 0, SVt_NULL,
873 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284 874
1cb9cd50
NC
875 /* The bind placeholder pretends to be an RV for now.
876 Also it's marked as "can't upgrade" top stop anyone using it before it's
877 implemented. */
878 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
879
d2a0f284
JC
880 /* IVs are in the head, so the allocation size is 0.
881 However, the slot is overloaded for PTEs. */
882 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
883 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 884 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
885 NOARENA /* IVS don't need an arena */,
886 /* But PTEs need to know the size of their arena */
887 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
888 },
889
bd81e77b 890 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 891 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
892 FIT_ARENA(0, sizeof(NV)) },
893
894 /* RVs are in the head now. */
10666ae3 895 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 896
bd81e77b 897 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
898 { sizeof(xpv_allocated),
899 copy_length(XPV, xpv_len)
900 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
901 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 902 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 903
bd81e77b 904 /* 12 */
d2a0f284
JC
905 { sizeof(xpviv_allocated),
906 copy_length(XPVIV, xiv_u)
907 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
908 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 909 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 910
bd81e77b 911 /* 20 */
10666ae3 912 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
913 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
914
bd81e77b 915 /* 28 */
10666ae3 916 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
917 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
918
bd81e77b 919 /* 48 */
10666ae3 920 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
921 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
922
bd81e77b 923 /* 64 */
10666ae3 924 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
925 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
926
927 { sizeof(xpvav_allocated),
928 copy_length(XPVAV, xmg_stash)
929 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
930 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 931 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
932
933 { sizeof(xpvhv_allocated),
934 copy_length(XPVHV, xmg_stash)
935 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
936 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 937 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 938
c84c4652 939 /* 56 */
4115f141 940 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 941 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 942 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 943
4115f141 944 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 945 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 946 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
947
948 /* XPVIO is 84 bytes, fits 48x */
10666ae3 949 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 950 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 951};
29489e7c 952
d2a0f284
JC
953#define new_body_type(sv_type) \
954 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 955
bd81e77b
NC
956#define del_body_type(p, sv_type) \
957 del_body(p, &PL_body_roots[sv_type])
29489e7c 958
29489e7c 959
bd81e77b 960#define new_body_allocated(sv_type) \
d2a0f284 961 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 962 - bodies_by_type[sv_type].offset)
29489e7c 963
bd81e77b
NC
964#define del_body_allocated(p, sv_type) \
965 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 966
29489e7c 967
bd81e77b
NC
968#define my_safemalloc(s) (void*)safemalloc(s)
969#define my_safecalloc(s) (void*)safecalloc(s, 1)
970#define my_safefree(p) safefree((char*)p)
29489e7c 971
bd81e77b 972#ifdef PURIFY
29489e7c 973
bd81e77b
NC
974#define new_XNV() my_safemalloc(sizeof(XPVNV))
975#define del_XNV(p) my_safefree(p)
29489e7c 976
bd81e77b
NC
977#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
978#define del_XPVNV(p) my_safefree(p)
29489e7c 979
bd81e77b
NC
980#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
981#define del_XPVAV(p) my_safefree(p)
29489e7c 982
bd81e77b
NC
983#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
984#define del_XPVHV(p) my_safefree(p)
29489e7c 985
bd81e77b
NC
986#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
987#define del_XPVMG(p) my_safefree(p)
29489e7c 988
bd81e77b
NC
989#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
990#define del_XPVGV(p) my_safefree(p)
29489e7c 991
bd81e77b 992#else /* !PURIFY */
29489e7c 993
bd81e77b
NC
994#define new_XNV() new_body_type(SVt_NV)
995#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 996
bd81e77b
NC
997#define new_XPVNV() new_body_type(SVt_PVNV)
998#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 999
bd81e77b
NC
1000#define new_XPVAV() new_body_allocated(SVt_PVAV)
1001#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1002
bd81e77b
NC
1003#define new_XPVHV() new_body_allocated(SVt_PVHV)
1004#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1005
bd81e77b
NC
1006#define new_XPVMG() new_body_type(SVt_PVMG)
1007#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1008
bd81e77b
NC
1009#define new_XPVGV() new_body_type(SVt_PVGV)
1010#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1011
bd81e77b 1012#endif /* PURIFY */
93e68bfb 1013
bd81e77b 1014/* no arena for you! */
93e68bfb 1015
bd81e77b 1016#define new_NOARENA(details) \
d2a0f284 1017 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1018#define new_NOARENAZ(details) \
d2a0f284
JC
1019 my_safecalloc((details)->body_size + (details)->offset)
1020
1021STATIC void *
1022S_more_bodies (pTHX_ svtype sv_type)
1023{
1024 dVAR;
1025 void ** const root = &PL_body_roots[sv_type];
96a5add6 1026 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1027 const size_t body_size = bdp->body_size;
1028 char *start;
1029 const char *end;
0b2d3faa 1030#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
23e9d66c
NC
1031 static bool done_sanity_check;
1032
0b2d3faa
JH
1033 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1034 * variables like done_sanity_check. */
10666ae3 1035 if (!done_sanity_check) {
ea471437 1036 unsigned int i = SVt_LAST;
10666ae3
NC
1037
1038 done_sanity_check = TRUE;
1039
1040 while (i--)
1041 assert (bodies_by_type[i].type == i);
1042 }
1043#endif
1044
23e9d66c
NC
1045 assert(bdp->arena_size);
1046
0a848332 1047 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size, sv_type);
d2a0f284
JC
1048
1049 end = start + bdp->arena_size - body_size;
1050
d2a0f284
JC
1051 /* computed count doesnt reflect the 1st slot reservation */
1052 DEBUG_m(PerlIO_printf(Perl_debug_log,
1053 "arena %p end %p arena-size %d type %d size %d ct %d\n",
6c9570dc 1054 (void*)start, (void*)end,
0e84aef4
JH
1055 (int)bdp->arena_size, sv_type, (int)body_size,
1056 (int)bdp->arena_size / (int)body_size));
d2a0f284
JC
1057
1058 *root = (void *)start;
1059
1060 while (start < end) {
1061 char * const next = start + body_size;
1062 *(void**) start = (void *)next;
1063 start = next;
1064 }
1065 *(void **)start = 0;
1066
1067 return *root;
1068}
1069
1070/* grab a new thing from the free list, allocating more if necessary.
1071 The inline version is used for speed in hot routines, and the
1072 function using it serves the rest (unless PURIFY).
1073*/
1074#define new_body_inline(xpv, sv_type) \
1075 STMT_START { \
1076 void ** const r3wt = &PL_body_roots[sv_type]; \
11b79775
DD
1077 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1078 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284 1079 *(r3wt) = *(void**)(xpv); \
d2a0f284
JC
1080 } STMT_END
1081
1082#ifndef PURIFY
1083
1084STATIC void *
1085S_new_body(pTHX_ svtype sv_type)
1086{
1087 dVAR;
1088 void *xpv;
1089 new_body_inline(xpv, sv_type);
1090 return xpv;
1091}
1092
1093#endif
93e68bfb 1094
bd81e77b
NC
1095/*
1096=for apidoc sv_upgrade
93e68bfb 1097
bd81e77b
NC
1098Upgrade an SV to a more complex form. Generally adds a new body type to the
1099SV, then copies across as much information as possible from the old body.
1100You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1101
bd81e77b 1102=cut
93e68bfb 1103*/
93e68bfb 1104
bd81e77b 1105void
42d0e0b7 1106Perl_sv_upgrade(pTHX_ register SV *sv, svtype new_type)
cac9b346 1107{
97aff369 1108 dVAR;
bd81e77b
NC
1109 void* old_body;
1110 void* new_body;
42d0e0b7 1111 const svtype old_type = SvTYPE(sv);
d2a0f284 1112 const struct body_details *new_type_details;
bd81e77b
NC
1113 const struct body_details *const old_type_details
1114 = bodies_by_type + old_type;
cac9b346 1115
bd81e77b
NC
1116 if (new_type != SVt_PV && SvIsCOW(sv)) {
1117 sv_force_normal_flags(sv, 0);
1118 }
cac9b346 1119
bd81e77b
NC
1120 if (old_type == new_type)
1121 return;
cac9b346 1122
bd81e77b
NC
1123 if (old_type > new_type)
1124 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1125 (int)old_type, (int)new_type);
cac9b346 1126
cac9b346 1127
bd81e77b 1128 old_body = SvANY(sv);
de042e1d 1129
bd81e77b
NC
1130 /* Copying structures onto other structures that have been neatly zeroed
1131 has a subtle gotcha. Consider XPVMG
cac9b346 1132
bd81e77b
NC
1133 +------+------+------+------+------+-------+-------+
1134 | NV | CUR | LEN | IV | MAGIC | STASH |
1135 +------+------+------+------+------+-------+-------+
1136 0 4 8 12 16 20 24 28
645c22ef 1137
bd81e77b
NC
1138 where NVs are aligned to 8 bytes, so that sizeof that structure is
1139 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1140
bd81e77b
NC
1141 +------+------+------+------+------+-------+-------+------+
1142 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1143 +------+------+------+------+------+-------+-------+------+
1144 0 4 8 12 16 20 24 28 32
08742458 1145
bd81e77b 1146 so what happens if you allocate memory for this structure:
30f9da9e 1147
bd81e77b
NC
1148 +------+------+------+------+------+-------+-------+------+------+...
1149 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1150 +------+------+------+------+------+-------+-------+------+------+...
1151 0 4 8 12 16 20 24 28 32 36
bfc44f79 1152
bd81e77b
NC
1153 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1154 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1155 started out as zero once, but it's quite possible that it isn't. So now,
1156 rather than a nicely zeroed GP, you have it pointing somewhere random.
1157 Bugs ensue.
bfc44f79 1158
bd81e77b
NC
1159 (In fact, GP ends up pointing at a previous GP structure, because the
1160 principle cause of the padding in XPVMG getting garbage is a copy of
6c9e42f7
NC
1161 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1162 this happens to be moot because XPVGV has been re-ordered, with GP
1163 no longer after STASH)
30f9da9e 1164
bd81e77b
NC
1165 So we are careful and work out the size of used parts of all the
1166 structures. */
bfc44f79 1167
bd81e77b
NC
1168 switch (old_type) {
1169 case SVt_NULL:
1170 break;
1171 case SVt_IV:
1172 if (new_type < SVt_PVIV) {
1173 new_type = (new_type == SVt_NV)
1174 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1175 }
1176 break;
1177 case SVt_NV:
1178 if (new_type < SVt_PVNV) {
1179 new_type = SVt_PVNV;
bd81e77b
NC
1180 }
1181 break;
1182 case SVt_RV:
1183 break;
1184 case SVt_PV:
1185 assert(new_type > SVt_PV);
1186 assert(SVt_IV < SVt_PV);
1187 assert(SVt_NV < SVt_PV);
1188 break;
1189 case SVt_PVIV:
1190 break;
1191 case SVt_PVNV:
1192 break;
1193 case SVt_PVMG:
1194 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1195 there's no way that it can be safely upgraded, because perl.c
1196 expects to Safefree(SvANY(PL_mess_sv)) */
1197 assert(sv != PL_mess_sv);
1198 /* This flag bit is used to mean other things in other scalar types.
1199 Given that it only has meaning inside the pad, it shouldn't be set
1200 on anything that can get upgraded. */
00b1698f 1201 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1202 break;
1203 default:
1204 if (old_type_details->cant_upgrade)
c81225bc
NC
1205 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1206 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1207 }
2fa1109b 1208 new_type_details = bodies_by_type + new_type;
645c22ef 1209
bd81e77b
NC
1210 SvFLAGS(sv) &= ~SVTYPEMASK;
1211 SvFLAGS(sv) |= new_type;
932e9ff9 1212
ab4416c0
NC
1213 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1214 the return statements above will have triggered. */
1215 assert (new_type != SVt_NULL);
bd81e77b 1216 switch (new_type) {
bd81e77b
NC
1217 case SVt_IV:
1218 assert(old_type == SVt_NULL);
1219 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1220 SvIV_set(sv, 0);
1221 return;
1222 case SVt_NV:
1223 assert(old_type == SVt_NULL);
1224 SvANY(sv) = new_XNV();
1225 SvNV_set(sv, 0);
1226 return;
1227 case SVt_RV:
1228 assert(old_type == SVt_NULL);
1229 SvANY(sv) = &sv->sv_u.svu_rv;
1230 SvRV_set(sv, 0);
1231 return;
1232 case SVt_PVHV:
bd81e77b 1233 case SVt_PVAV:
d2a0f284 1234 assert(new_type_details->body_size);
c1ae03ae
NC
1235
1236#ifndef PURIFY
1237 assert(new_type_details->arena);
d2a0f284 1238 assert(new_type_details->arena_size);
c1ae03ae 1239 /* This points to the start of the allocated area. */
d2a0f284
JC
1240 new_body_inline(new_body, new_type);
1241 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1242 new_body = ((char *)new_body) - new_type_details->offset;
1243#else
1244 /* We always allocated the full length item with PURIFY. To do this
1245 we fake things so that arena is false for all 16 types.. */
1246 new_body = new_NOARENAZ(new_type_details);
1247#endif
1248 SvANY(sv) = new_body;
1249 if (new_type == SVt_PVAV) {
1250 AvMAX(sv) = -1;
1251 AvFILLp(sv) = -1;
1252 AvREAL_only(sv);
1253 }
aeb18a1e 1254
bd81e77b
NC
1255 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1256 The target created by newSVrv also is, and it can have magic.
1257 However, it never has SvPVX set.
1258 */
1259 if (old_type >= SVt_RV) {
1260 assert(SvPVX_const(sv) == 0);
1261 }
aeb18a1e 1262
bd81e77b 1263 if (old_type >= SVt_PVMG) {
e736a858 1264 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1265 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1266 } else {
1267 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1268 }
1269 break;
93e68bfb 1270
93e68bfb 1271
bd81e77b
NC
1272 case SVt_PVIV:
1273 /* XXX Is this still needed? Was it ever needed? Surely as there is
1274 no route from NV to PVIV, NOK can never be true */
1275 assert(!SvNOKp(sv));
1276 assert(!SvNOK(sv));
1277 case SVt_PVIO:
1278 case SVt_PVFM:
bd81e77b
NC
1279 case SVt_PVGV:
1280 case SVt_PVCV:
1281 case SVt_PVLV:
1282 case SVt_PVMG:
1283 case SVt_PVNV:
1284 case SVt_PV:
93e68bfb 1285
d2a0f284 1286 assert(new_type_details->body_size);
bd81e77b
NC
1287 /* We always allocated the full length item with PURIFY. To do this
1288 we fake things so that arena is false for all 16 types.. */
1289 if(new_type_details->arena) {
1290 /* This points to the start of the allocated area. */
d2a0f284
JC
1291 new_body_inline(new_body, new_type);
1292 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1293 new_body = ((char *)new_body) - new_type_details->offset;
1294 } else {
1295 new_body = new_NOARENAZ(new_type_details);
1296 }
1297 SvANY(sv) = new_body;
5e2fc214 1298
bd81e77b 1299 if (old_type_details->copy) {
f9ba3d20
NC
1300 /* There is now the potential for an upgrade from something without
1301 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1302 int offset = old_type_details->offset;
1303 int length = old_type_details->copy;
1304
1305 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1306 const int difference
f9ba3d20
NC
1307 = new_type_details->offset - old_type_details->offset;
1308 offset += difference;
1309 length -= difference;
1310 }
1311 assert (length >= 0);
1312
1313 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1314 char);
bd81e77b
NC
1315 }
1316
1317#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1318 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1319 * correct 0.0 for us. Otherwise, if the old body didn't have an
1320 * NV slot, but the new one does, then we need to initialise the
1321 * freshly created NV slot with whatever the correct bit pattern is
1322 * for 0.0 */
1323 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1324 SvNV_set(sv, 0);
82048762 1325#endif
5e2fc214 1326
bd81e77b 1327 if (new_type == SVt_PVIO)
f2524eef 1328 IoPAGE_LEN(sv) = 60;
bd81e77b 1329 if (old_type < SVt_RV)
6136c704 1330 SvPV_set(sv, NULL);
bd81e77b
NC
1331 break;
1332 default:
afd78fd5
JH
1333 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1334 (unsigned long)new_type);
bd81e77b 1335 }
73171d91 1336
d2a0f284
JC
1337 if (old_type_details->arena) {
1338 /* If there was an old body, then we need to free it.
1339 Note that there is an assumption that all bodies of types that
1340 can be upgraded came from arenas. Only the more complex non-
1341 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1342#ifdef PURIFY
1343 my_safefree(old_body);
1344#else
1345 del_body((void*)((char*)old_body + old_type_details->offset),
1346 &PL_body_roots[old_type]);
1347#endif
1348 }
1349}
73171d91 1350
bd81e77b
NC
1351/*
1352=for apidoc sv_backoff
73171d91 1353
bd81e77b
NC
1354Remove any string offset. You should normally use the C<SvOOK_off> macro
1355wrapper instead.
73171d91 1356
bd81e77b 1357=cut
73171d91
NC
1358*/
1359
bd81e77b
NC
1360int
1361Perl_sv_backoff(pTHX_ register SV *sv)
1362{
96a5add6 1363 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1364 assert(SvOOK(sv));
1365 assert(SvTYPE(sv) != SVt_PVHV);
1366 assert(SvTYPE(sv) != SVt_PVAV);
1367 if (SvIVX(sv)) {
1368 const char * const s = SvPVX_const(sv);
1369 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1370 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1371 SvIV_set(sv, 0);
1372 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1373 }
1374 SvFLAGS(sv) &= ~SVf_OOK;
1375 return 0;
1376}
73171d91 1377
bd81e77b
NC
1378/*
1379=for apidoc sv_grow
73171d91 1380
bd81e77b
NC
1381Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1382upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1383Use the C<SvGROW> wrapper instead.
93e68bfb 1384
bd81e77b
NC
1385=cut
1386*/
93e68bfb 1387
bd81e77b
NC
1388char *
1389Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1390{
1391 register char *s;
93e68bfb 1392
5db06880
NC
1393 if (PL_madskills && newlen >= 0x100000) {
1394 PerlIO_printf(Perl_debug_log,
1395 "Allocation too large: %"UVxf"\n", (UV)newlen);
1396 }
bd81e77b
NC
1397#ifdef HAS_64K_LIMIT
1398 if (newlen >= 0x10000) {
1399 PerlIO_printf(Perl_debug_log,
1400 "Allocation too large: %"UVxf"\n", (UV)newlen);
1401 my_exit(1);
1402 }
1403#endif /* HAS_64K_LIMIT */
1404 if (SvROK(sv))
1405 sv_unref(sv);
1406 if (SvTYPE(sv) < SVt_PV) {
1407 sv_upgrade(sv, SVt_PV);
1408 s = SvPVX_mutable(sv);
1409 }
1410 else if (SvOOK(sv)) { /* pv is offset? */
1411 sv_backoff(sv);
1412 s = SvPVX_mutable(sv);
1413 if (newlen > SvLEN(sv))
1414 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1415#ifdef HAS_64K_LIMIT
1416 if (newlen >= 0x10000)
1417 newlen = 0xFFFF;
1418#endif
1419 }
1420 else
1421 s = SvPVX_mutable(sv);
aeb18a1e 1422
bd81e77b
NC
1423 if (newlen > SvLEN(sv)) { /* need more room? */
1424 newlen = PERL_STRLEN_ROUNDUP(newlen);
1425 if (SvLEN(sv) && s) {
1426#ifdef MYMALLOC
1427 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1428 if (newlen <= l) {
1429 SvLEN_set(sv, l);
1430 return s;
1431 } else
1432#endif
10edeb5d 1433 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1434 }
1435 else {
10edeb5d 1436 s = (char*)safemalloc(newlen);
bd81e77b
NC
1437 if (SvPVX_const(sv) && SvCUR(sv)) {
1438 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1439 }
1440 }
1441 SvPV_set(sv, s);
1442 SvLEN_set(sv, newlen);
1443 }
1444 return s;
1445}
aeb18a1e 1446
bd81e77b
NC
1447/*
1448=for apidoc sv_setiv
932e9ff9 1449
bd81e77b
NC
1450Copies an integer into the given SV, upgrading first if necessary.
1451Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1452
bd81e77b
NC
1453=cut
1454*/
463ee0b2 1455
bd81e77b
NC
1456void
1457Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1458{
97aff369 1459 dVAR;
bd81e77b
NC
1460 SV_CHECK_THINKFIRST_COW_DROP(sv);
1461 switch (SvTYPE(sv)) {
1462 case SVt_NULL:
1463 sv_upgrade(sv, SVt_IV);
1464 break;
1465 case SVt_NV:
1466 sv_upgrade(sv, SVt_PVNV);
1467 break;
1468 case SVt_RV:
1469 case SVt_PV:
1470 sv_upgrade(sv, SVt_PVIV);
1471 break;
463ee0b2 1472
bd81e77b
NC
1473 case SVt_PVGV:
1474 case SVt_PVAV:
1475 case SVt_PVHV:
1476 case SVt_PVCV:
1477 case SVt_PVFM:
1478 case SVt_PVIO:
1479 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1480 OP_DESC(PL_op));
42d0e0b7 1481 default: NOOP;
bd81e77b
NC
1482 }
1483 (void)SvIOK_only(sv); /* validate number */
1484 SvIV_set(sv, i);
1485 SvTAINT(sv);
1486}
932e9ff9 1487
bd81e77b
NC
1488/*
1489=for apidoc sv_setiv_mg
d33b2eba 1490
bd81e77b 1491Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1492
bd81e77b
NC
1493=cut
1494*/
d33b2eba 1495
bd81e77b
NC
1496void
1497Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1498{
1499 sv_setiv(sv,i);
1500 SvSETMAGIC(sv);
1501}
727879eb 1502
bd81e77b
NC
1503/*
1504=for apidoc sv_setuv
d33b2eba 1505
bd81e77b
NC
1506Copies an unsigned integer into the given SV, upgrading first if necessary.
1507Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1508
bd81e77b
NC
1509=cut
1510*/
d33b2eba 1511
bd81e77b
NC
1512void
1513Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1514{
1515 /* With these two if statements:
1516 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1517
bd81e77b
NC
1518 without
1519 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1520
bd81e77b
NC
1521 If you wish to remove them, please benchmark to see what the effect is
1522 */
1523 if (u <= (UV)IV_MAX) {
1524 sv_setiv(sv, (IV)u);
1525 return;
1526 }
1527 sv_setiv(sv, 0);
1528 SvIsUV_on(sv);
1529 SvUV_set(sv, u);
1530}
d33b2eba 1531
bd81e77b
NC
1532/*
1533=for apidoc sv_setuv_mg
727879eb 1534
bd81e77b 1535Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1536
bd81e77b
NC
1537=cut
1538*/
5e2fc214 1539
bd81e77b
NC
1540void
1541Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1542{
bd81e77b
NC
1543 sv_setuv(sv,u);
1544 SvSETMAGIC(sv);
1545}
5e2fc214 1546
954c1994 1547/*
bd81e77b 1548=for apidoc sv_setnv
954c1994 1549
bd81e77b
NC
1550Copies a double into the given SV, upgrading first if necessary.
1551Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1552
1553=cut
1554*/
1555
63f97190 1556void
bd81e77b 1557Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1558{
97aff369 1559 dVAR;
bd81e77b
NC
1560 SV_CHECK_THINKFIRST_COW_DROP(sv);
1561 switch (SvTYPE(sv)) {
79072805 1562 case SVt_NULL:
79072805 1563 case SVt_IV:
bd81e77b 1564 sv_upgrade(sv, SVt_NV);
79072805 1565 break;
ed6116ce 1566 case SVt_RV:
79072805 1567 case SVt_PV:
79072805 1568 case SVt_PVIV:
bd81e77b 1569 sv_upgrade(sv, SVt_PVNV);
79072805 1570 break;
bd4b1eb5 1571
bd4b1eb5 1572 case SVt_PVGV:
bd81e77b
NC
1573 case SVt_PVAV:
1574 case SVt_PVHV:
79072805 1575 case SVt_PVCV:
bd81e77b
NC
1576 case SVt_PVFM:
1577 case SVt_PVIO:
1578 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1579 OP_NAME(PL_op));
42d0e0b7 1580 default: NOOP;
2068cd4d 1581 }
bd81e77b
NC
1582 SvNV_set(sv, num);
1583 (void)SvNOK_only(sv); /* validate number */
1584 SvTAINT(sv);
79072805
LW
1585}
1586
645c22ef 1587/*
bd81e77b 1588=for apidoc sv_setnv_mg
645c22ef 1589
bd81e77b 1590Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1591
1592=cut
1593*/
1594
bd81e77b
NC
1595void
1596Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1597{
bd81e77b
NC
1598 sv_setnv(sv,num);
1599 SvSETMAGIC(sv);
79072805
LW
1600}
1601
bd81e77b
NC
1602/* Print an "isn't numeric" warning, using a cleaned-up,
1603 * printable version of the offending string
1604 */
954c1994 1605
bd81e77b
NC
1606STATIC void
1607S_not_a_number(pTHX_ SV *sv)
79072805 1608{
97aff369 1609 dVAR;
bd81e77b
NC
1610 SV *dsv;
1611 char tmpbuf[64];
1612 const char *pv;
94463019
JH
1613
1614 if (DO_UTF8(sv)) {
396482e1 1615 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1616 pv = sv_uni_display(dsv, sv, 10, 0);
1617 } else {
1618 char *d = tmpbuf;
551405c4 1619 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1620 /* each *s can expand to 4 chars + "...\0",
1621 i.e. need room for 8 chars */
ecdeb87c 1622
00b6aa41
AL
1623 const char *s = SvPVX_const(sv);
1624 const char * const end = s + SvCUR(sv);
1625 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1626 int ch = *s & 0xFF;
1627 if (ch & 128 && !isPRINT_LC(ch)) {
1628 *d++ = 'M';
1629 *d++ = '-';
1630 ch &= 127;
1631 }
1632 if (ch == '\n') {
1633 *d++ = '\\';
1634 *d++ = 'n';
1635 }
1636 else if (ch == '\r') {
1637 *d++ = '\\';
1638 *d++ = 'r';
1639 }
1640 else if (ch == '\f') {
1641 *d++ = '\\';
1642 *d++ = 'f';
1643 }
1644 else if (ch == '\\') {
1645 *d++ = '\\';
1646 *d++ = '\\';
1647 }
1648 else if (ch == '\0') {
1649 *d++ = '\\';
1650 *d++ = '0';
1651 }
1652 else if (isPRINT_LC(ch))
1653 *d++ = ch;
1654 else {
1655 *d++ = '^';
1656 *d++ = toCTRL(ch);
1657 }
1658 }
1659 if (s < end) {
1660 *d++ = '.';
1661 *d++ = '.';
1662 *d++ = '.';
1663 }
1664 *d = '\0';
1665 pv = tmpbuf;
a0d0e21e 1666 }
a0d0e21e 1667
533c011a 1668 if (PL_op)
9014280d 1669 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1670 "Argument \"%s\" isn't numeric in %s", pv,
1671 OP_DESC(PL_op));
a0d0e21e 1672 else
9014280d 1673 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1674 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1675}
1676
c2988b20
NC
1677/*
1678=for apidoc looks_like_number
1679
645c22ef
DM
1680Test if the content of an SV looks like a number (or is a number).
1681C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1682non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1683
1684=cut
1685*/
1686
1687I32
1688Perl_looks_like_number(pTHX_ SV *sv)
1689{
a3b680e6 1690 register const char *sbegin;
c2988b20
NC
1691 STRLEN len;
1692
1693 if (SvPOK(sv)) {
3f7c398e 1694 sbegin = SvPVX_const(sv);
c2988b20
NC
1695 len = SvCUR(sv);
1696 }
1697 else if (SvPOKp(sv))
83003860 1698 sbegin = SvPV_const(sv, len);
c2988b20 1699 else
e0ab1c0e 1700 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1701 return grok_number(sbegin, len, NULL);
1702}
25da4f38 1703
19f6321d
NC
1704STATIC bool
1705S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1706{
1707 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1708 SV *const buffer = sv_newmortal();
1709
1710 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1711 is on. */
1712 SvFAKE_off(gv);
1713 gv_efullname3(buffer, gv, "*");
1714 SvFLAGS(gv) |= wasfake;
1715
675c862f
AL
1716 /* We know that all GVs stringify to something that is not-a-number,
1717 so no need to test that. */
1718 if (ckWARN(WARN_NUMERIC))
1719 not_a_number(buffer);
1720 /* We just want something true to return, so that S_sv_2iuv_common
1721 can tail call us and return true. */
19f6321d 1722 return TRUE;
675c862f
AL
1723}
1724
1725STATIC char *
19f6321d 1726S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1727{
1728 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1729 SV *const buffer = sv_newmortal();
1730
1731 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1732 is on. */
1733 SvFAKE_off(gv);
1734 gv_efullname3(buffer, gv, "*");
1735 SvFLAGS(gv) |= wasfake;
1736
1737 assert(SvPOK(buffer));
a6d61a6c
NC
1738 if (len) {
1739 *len = SvCUR(buffer);
1740 }
675c862f 1741 return SvPVX(buffer);
180488f8
NC
1742}
1743
25da4f38
IZ
1744/* Actually, ISO C leaves conversion of UV to IV undefined, but
1745 until proven guilty, assume that things are not that bad... */
1746
645c22ef
DM
1747/*
1748 NV_PRESERVES_UV:
1749
1750 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1751 an IV (an assumption perl has been based on to date) it becomes necessary
1752 to remove the assumption that the NV always carries enough precision to
1753 recreate the IV whenever needed, and that the NV is the canonical form.
1754 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1755 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1756 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1757 1) to distinguish between IV/UV/NV slots that have cached a valid
1758 conversion where precision was lost and IV/UV/NV slots that have a
1759 valid conversion which has lost no precision
645c22ef 1760 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1761 would lose precision, the precise conversion (or differently
1762 imprecise conversion) is also performed and cached, to prevent
1763 requests for different numeric formats on the same SV causing
1764 lossy conversion chains. (lossless conversion chains are perfectly
1765 acceptable (still))
1766
1767
1768 flags are used:
1769 SvIOKp is true if the IV slot contains a valid value
1770 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1771 SvNOKp is true if the NV slot contains a valid value
1772 SvNOK is true only if the NV value is accurate
1773
1774 so
645c22ef 1775 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1776 IV(or UV) would lose accuracy over a direct conversion from PV to
1777 IV(or UV). If it would, cache both conversions, return NV, but mark
1778 SV as IOK NOKp (ie not NOK).
1779
645c22ef 1780 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1781 NV would lose accuracy over a direct conversion from PV to NV. If it
1782 would, cache both conversions, flag similarly.
1783
1784 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1785 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1786 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1787 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1788 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1789
645c22ef
DM
1790 The benefit of this is that operations such as pp_add know that if
1791 SvIOK is true for both left and right operands, then integer addition
1792 can be used instead of floating point (for cases where the result won't
1793 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1794 loss of precision compared with integer addition.
1795
1796 * making IV and NV equal status should make maths accurate on 64 bit
1797 platforms
1798 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1799 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1800 looking for SvIOK and checking for overflow will not outweigh the
1801 fp to integer speedup)
1802 * will slow down integer operations (callers of SvIV) on "inaccurate"
1803 values, as the change from SvIOK to SvIOKp will cause a call into
1804 sv_2iv each time rather than a macro access direct to the IV slot
1805 * should speed up number->string conversion on integers as IV is
645c22ef 1806 favoured when IV and NV are equally accurate
28e5dec8
JH
1807
1808 ####################################################################
645c22ef
DM
1809 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1810 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1811 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1812 ####################################################################
1813
645c22ef 1814 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1815 performance ratio.
1816*/
1817
1818#ifndef NV_PRESERVES_UV
645c22ef
DM
1819# define IS_NUMBER_UNDERFLOW_IV 1
1820# define IS_NUMBER_UNDERFLOW_UV 2
1821# define IS_NUMBER_IV_AND_UV 2
1822# define IS_NUMBER_OVERFLOW_IV 4
1823# define IS_NUMBER_OVERFLOW_UV 5
1824
1825/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1826
1827/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1828STATIC int
645c22ef 1829S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1830{
97aff369 1831 dVAR;
b57a0404 1832 PERL_UNUSED_ARG(numtype); /* Used only under DEBUGGING? */
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 {
6f207bd3 2053 NOOP; /* 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 {
6f207bd3 2068 NOOP; /* 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 {
675c862f 2112 if (isGV_with_GP(sv))
a0933d07 2113 return glob_2number((GV *)sv);
180488f8 2114
af359546
NC
2115 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2116 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2117 report_uninit(sv);
2118 }
25da4f38
IZ
2119 if (SvTYPE(sv) < SVt_IV)
2120 /* Typically the caller expects that sv_any is not NULL now. */
2121 sv_upgrade(sv, SVt_IV);
af359546
NC
2122 /* Return 0 from the caller. */
2123 return TRUE;
2124 }
2125 return FALSE;
2126}
2127
2128/*
2129=for apidoc sv_2iv_flags
2130
2131Return the integer value of an SV, doing any necessary string
2132conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2133Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2134
2135=cut
2136*/
2137
2138IV
2139Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2140{
97aff369 2141 dVAR;
af359546 2142 if (!sv)
a0d0e21e 2143 return 0;
cecf5685
NC
2144 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2145 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e
NC
2146 cache IVs just in case. In practice it seems that they never
2147 actually anywhere accessible by user Perl code, let alone get used
2148 in anything other than a string context. */
af359546
NC
2149 if (flags & SV_GMAGIC)
2150 mg_get(sv);
2151 if (SvIOKp(sv))
2152 return SvIVX(sv);
2153 if (SvNOKp(sv)) {
2154 return I_V(SvNVX(sv));
2155 }
71c558c3
NC
2156 if (SvPOKp(sv) && SvLEN(sv)) {
2157 UV value;
2158 const int numtype
2159 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2160
2161 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2162 == IS_NUMBER_IN_UV) {
2163 /* It's definitely an integer */
2164 if (numtype & IS_NUMBER_NEG) {
2165 if (value < (UV)IV_MIN)
2166 return -(IV)value;
2167 } else {
2168 if (value < (UV)IV_MAX)
2169 return (IV)value;
2170 }
2171 }
2172 if (!numtype) {
2173 if (ckWARN(WARN_NUMERIC))
2174 not_a_number(sv);
2175 }
2176 return I_V(Atof(SvPVX_const(sv)));
2177 }
1c7ff15e
NC
2178 if (SvROK(sv)) {
2179 goto return_rok;
af359546 2180 }
1c7ff15e
NC
2181 assert(SvTYPE(sv) >= SVt_PVMG);
2182 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2183 } else if (SvTHINKFIRST(sv)) {
af359546 2184 if (SvROK(sv)) {
1c7ff15e 2185 return_rok:
af359546
NC
2186 if (SvAMAGIC(sv)) {
2187 SV * const tmpstr=AMG_CALLun(sv,numer);
2188 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2189 return SvIV(tmpstr);
2190 }
2191 }
2192 return PTR2IV(SvRV(sv));
2193 }
2194 if (SvIsCOW(sv)) {
2195 sv_force_normal_flags(sv, 0);
2196 }
2197 if (SvREADONLY(sv) && !SvOK(sv)) {
2198 if (ckWARN(WARN_UNINITIALIZED))
2199 report_uninit(sv);
2200 return 0;
2201 }
2202 }
2203 if (!SvIOKp(sv)) {
2204 if (S_sv_2iuv_common(aTHX_ sv))
2205 return 0;
79072805 2206 }
1d7c1841
GS
2207 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2208 PTR2UV(sv),SvIVX(sv)));
25da4f38 2209 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2210}
2211
645c22ef 2212/*
891f9566 2213=for apidoc sv_2uv_flags
645c22ef
DM
2214
2215Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2216conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2217Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2218
2219=cut
2220*/
2221
ff68c719 2222UV
891f9566 2223Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2224{
97aff369 2225 dVAR;
ff68c719 2226 if (!sv)
2227 return 0;
cecf5685
NC
2228 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2229 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2230 cache IVs just in case. */
891f9566
YST
2231 if (flags & SV_GMAGIC)
2232 mg_get(sv);
ff68c719 2233 if (SvIOKp(sv))
2234 return SvUVX(sv);
2235 if (SvNOKp(sv))
2236 return U_V(SvNVX(sv));
71c558c3
NC
2237 if (SvPOKp(sv) && SvLEN(sv)) {
2238 UV value;
2239 const int numtype
2240 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2241
2242 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2243 == IS_NUMBER_IN_UV) {
2244 /* It's definitely an integer */
2245 if (!(numtype & IS_NUMBER_NEG))
2246 return value;
2247 }
2248 if (!numtype) {
2249 if (ckWARN(WARN_NUMERIC))
2250 not_a_number(sv);
2251 }
2252 return U_V(Atof(SvPVX_const(sv)));
2253 }
1c7ff15e
NC
2254 if (SvROK(sv)) {
2255 goto return_rok;
3fe9a6f1 2256 }
1c7ff15e
NC
2257 assert(SvTYPE(sv) >= SVt_PVMG);
2258 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2259 } else if (SvTHINKFIRST(sv)) {
ff68c719 2260 if (SvROK(sv)) {
1c7ff15e 2261 return_rok:
deb46114
NC
2262 if (SvAMAGIC(sv)) {
2263 SV *const tmpstr = AMG_CALLun(sv,numer);
2264 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2265 return SvUV(tmpstr);
2266 }
2267 }
2268 return PTR2UV(SvRV(sv));
ff68c719 2269 }
765f542d
NC
2270 if (SvIsCOW(sv)) {
2271 sv_force_normal_flags(sv, 0);
8a818333 2272 }
0336b60e 2273 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2274 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2275 report_uninit(sv);
ff68c719 2276 return 0;
2277 }
2278 }
af359546
NC
2279 if (!SvIOKp(sv)) {
2280 if (S_sv_2iuv_common(aTHX_ sv))
2281 return 0;
ff68c719 2282 }
25da4f38 2283
1d7c1841
GS
2284 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2285 PTR2UV(sv),SvUVX(sv)));
25da4f38 2286 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2287}
2288
645c22ef
DM
2289/*
2290=for apidoc sv_2nv
2291
2292Return the num value of an SV, doing any necessary string or integer
2293conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2294macros.
2295
2296=cut
2297*/
2298
65202027 2299NV
864dbfa3 2300Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2301{
97aff369 2302 dVAR;
79072805
LW
2303 if (!sv)
2304 return 0.0;
cecf5685
NC
2305 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2306 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2307 cache IVs just in case. */
463ee0b2
LW
2308 mg_get(sv);
2309 if (SvNOKp(sv))
2310 return SvNVX(sv);
0aa395f8 2311 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2312 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2313 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2314 not_a_number(sv);
3f7c398e 2315 return Atof(SvPVX_const(sv));
a0d0e21e 2316 }
25da4f38 2317 if (SvIOKp(sv)) {
1c846c1f 2318 if (SvIsUV(sv))
65202027 2319 return (NV)SvUVX(sv);
25da4f38 2320 else
65202027 2321 return (NV)SvIVX(sv);
47a72cb8
NC
2322 }
2323 if (SvROK(sv)) {
2324 goto return_rok;
2325 }
2326 assert(SvTYPE(sv) >= SVt_PVMG);
2327 /* This falls through to the report_uninit near the end of the
2328 function. */
2329 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2330 if (SvROK(sv)) {
47a72cb8 2331 return_rok:
deb46114
NC
2332 if (SvAMAGIC(sv)) {
2333 SV *const tmpstr = AMG_CALLun(sv,numer);
2334 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2335 return SvNV(tmpstr);
2336 }
2337 }
2338 return PTR2NV(SvRV(sv));
a0d0e21e 2339 }
765f542d
NC
2340 if (SvIsCOW(sv)) {
2341 sv_force_normal_flags(sv, 0);
8a818333 2342 }
0336b60e 2343 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2344 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2345 report_uninit(sv);
ed6116ce
LW
2346 return 0.0;
2347 }
79072805
LW
2348 }
2349 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2350 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2351 sv_upgrade(sv, SVt_NV);
906f284f 2352#ifdef USE_LONG_DOUBLE
097ee67d 2353 DEBUG_c({
f93f4e46 2354 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2355 PerlIO_printf(Perl_debug_log,
2356 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2357 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2358 RESTORE_NUMERIC_LOCAL();
2359 });
65202027 2360#else
572bbb43 2361 DEBUG_c({
f93f4e46 2362 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2363 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2364 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2365 RESTORE_NUMERIC_LOCAL();
2366 });
572bbb43 2367#endif
79072805
LW
2368 }
2369 else if (SvTYPE(sv) < SVt_PVNV)
2370 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2371 if (SvNOKp(sv)) {
2372 return SvNVX(sv);
61604483 2373 }
59d8ce62 2374 if (SvIOKp(sv)) {
9d6ce603 2375 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2376#ifdef NV_PRESERVES_UV
2377 SvNOK_on(sv);
2378#else
2379 /* Only set the public NV OK flag if this NV preserves the IV */
2380 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2381 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2382 : (SvIVX(sv) == I_V(SvNVX(sv))))
2383 SvNOK_on(sv);
2384 else
2385 SvNOKp_on(sv);
2386#endif
93a17b20 2387 }
748a9306 2388 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2389 UV value;
3f7c398e 2390 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2391 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2392 not_a_number(sv);
28e5dec8 2393#ifdef NV_PRESERVES_UV
c2988b20
NC
2394 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2395 == IS_NUMBER_IN_UV) {
5e045b90 2396 /* It's definitely an integer */
9d6ce603 2397 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2398 } else
3f7c398e 2399 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2400 SvNOK_on(sv);
2401#else
3f7c398e 2402 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2403 /* Only set the public NV OK flag if this NV preserves the value in
2404 the PV at least as well as an IV/UV would.
2405 Not sure how to do this 100% reliably. */
2406 /* if that shift count is out of range then Configure's test is
2407 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2408 UV_BITS */
2409 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2410 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2411 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2412 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2413 /* Can't use strtol etc to convert this string, so don't try.
2414 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2415 SvNOK_on(sv);
2416 } else {
2417 /* value has been set. It may not be precise. */
2418 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2419 /* 2s complement assumption for (UV)IV_MIN */
2420 SvNOK_on(sv); /* Integer is too negative. */
2421 } else {
2422 SvNOKp_on(sv);
2423 SvIOKp_on(sv);
6fa402ec 2424
c2988b20 2425 if (numtype & IS_NUMBER_NEG) {
45977657 2426 SvIV_set(sv, -(IV)value);
c2988b20 2427 } else if (value <= (UV)IV_MAX) {
45977657 2428 SvIV_set(sv, (IV)value);
c2988b20 2429 } else {
607fa7f2 2430 SvUV_set(sv, value);
c2988b20
NC
2431 SvIsUV_on(sv);
2432 }
2433
2434 if (numtype & IS_NUMBER_NOT_INT) {
2435 /* I believe that even if the original PV had decimals,
2436 they are lost beyond the limit of the FP precision.
2437 However, neither is canonical, so both only get p
2438 flags. NWC, 2000/11/25 */
2439 /* Both already have p flags, so do nothing */
2440 } else {
66a1b24b 2441 const NV nv = SvNVX(sv);
c2988b20
NC
2442 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2443 if (SvIVX(sv) == I_V(nv)) {
2444 SvNOK_on(sv);
c2988b20 2445 } else {
c2988b20
NC
2446 /* It had no "." so it must be integer. */
2447 }
00b6aa41 2448 SvIOK_on(sv);
c2988b20
NC
2449 } else {
2450 /* between IV_MAX and NV(UV_MAX).
2451 Could be slightly > UV_MAX */
6fa402ec 2452
c2988b20
NC
2453 if (numtype & IS_NUMBER_NOT_INT) {
2454 /* UV and NV both imprecise. */
2455 } else {
66a1b24b 2456 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2457
2458 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2459 SvNOK_on(sv);
c2988b20 2460 }
00b6aa41 2461 SvIOK_on(sv);
c2988b20
NC
2462 }
2463 }
2464 }
2465 }
2466 }
28e5dec8 2467#endif /* NV_PRESERVES_UV */
93a17b20 2468 }
79072805 2469 else {
f7877b28 2470 if (isGV_with_GP(sv)) {
19f6321d 2471 glob_2number((GV *)sv);
180488f8
NC
2472 return 0.0;
2473 }
2474
041457d9 2475 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2476 report_uninit(sv);
7e25a7e9
NC
2477 assert (SvTYPE(sv) >= SVt_NV);
2478 /* Typically the caller expects that sv_any is not NULL now. */
2479 /* XXX Ilya implies that this is a bug in callers that assume this
2480 and ideally should be fixed. */
a0d0e21e 2481 return 0.0;
79072805 2482 }
572bbb43 2483#if defined(USE_LONG_DOUBLE)
097ee67d 2484 DEBUG_c({
f93f4e46 2485 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2486 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2487 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2488 RESTORE_NUMERIC_LOCAL();
2489 });
65202027 2490#else
572bbb43 2491 DEBUG_c({
f93f4e46 2492 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2493 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2494 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2495 RESTORE_NUMERIC_LOCAL();
2496 });
572bbb43 2497#endif
463ee0b2 2498 return SvNVX(sv);
79072805
LW
2499}
2500
645c22ef
DM
2501/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2502 * UV as a string towards the end of buf, and return pointers to start and
2503 * end of it.
2504 *
2505 * We assume that buf is at least TYPE_CHARS(UV) long.
2506 */
2507
864dbfa3 2508static char *
aec46f14 2509S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2510{
25da4f38 2511 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2512 char * const ebuf = ptr;
25da4f38 2513 int sign;
25da4f38
IZ
2514
2515 if (is_uv)
2516 sign = 0;
2517 else if (iv >= 0) {
2518 uv = iv;
2519 sign = 0;
2520 } else {
2521 uv = -iv;
2522 sign = 1;
2523 }
2524 do {
eb160463 2525 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2526 } while (uv /= 10);
2527 if (sign)
2528 *--ptr = '-';
2529 *peob = ebuf;
2530 return ptr;
2531}
2532
645c22ef
DM
2533/*
2534=for apidoc sv_2pv_flags
2535
ff276b08 2536Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2537If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2538if necessary.
2539Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2540usually end up here too.
2541
2542=cut
2543*/
2544
8d6d96c1
HS
2545char *
2546Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2547{
97aff369 2548 dVAR;
79072805 2549 register char *s;
79072805 2550
463ee0b2 2551 if (!sv) {
cdb061a3
NC
2552 if (lp)
2553 *lp = 0;
73d840c0 2554 return (char *)"";
463ee0b2 2555 }
8990e307 2556 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2557 if (flags & SV_GMAGIC)
2558 mg_get(sv);
463ee0b2 2559 if (SvPOKp(sv)) {
cdb061a3
NC
2560 if (lp)
2561 *lp = SvCUR(sv);
10516c54
NC
2562 if (flags & SV_MUTABLE_RETURN)
2563 return SvPVX_mutable(sv);
4d84ee25
NC
2564 if (flags & SV_CONST_RETURN)
2565 return (char *)SvPVX_const(sv);
463ee0b2
LW
2566 return SvPVX(sv);
2567 }
75dfc8ec
NC
2568 if (SvIOKp(sv) || SvNOKp(sv)) {
2569 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2570 STRLEN len;
2571
2572 if (SvIOKp(sv)) {
e80fed9d 2573 len = SvIsUV(sv)
d9fad198
JH
2574 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2575 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2576 } else {
e8ada2d0
NC
2577 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2578 len = strlen(tbuf);
75dfc8ec 2579 }
b5b886f0
NC
2580 assert(!SvROK(sv));
2581 {
75dfc8ec
NC
2582 dVAR;
2583
2584#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2585 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2586 tbuf[0] = '0';
2587 tbuf[1] = 0;
75dfc8ec
NC
2588 len = 1;
2589 }
2590#endif
2591 SvUPGRADE(sv, SVt_PV);
2592 if (lp)
2593 *lp = len;
2594 s = SvGROW_mutable(sv, len + 1);
2595 SvCUR_set(sv, len);
2596 SvPOKp_on(sv);
10edeb5d 2597 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2598 }
463ee0b2 2599 }
1c7ff15e
NC
2600 if (SvROK(sv)) {
2601 goto return_rok;
2602 }
2603 assert(SvTYPE(sv) >= SVt_PVMG);
2604 /* This falls through to the report_uninit near the end of the
2605 function. */
2606 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2607 if (SvROK(sv)) {
1c7ff15e 2608 return_rok:
deb46114
NC
2609 if (SvAMAGIC(sv)) {
2610 SV *const tmpstr = AMG_CALLun(sv,string);
2611 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2612 /* Unwrap this: */
2613 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2614 */
2615
2616 char *pv;
2617 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2618 if (flags & SV_CONST_RETURN) {
2619 pv = (char *) SvPVX_const(tmpstr);
2620 } else {
2621 pv = (flags & SV_MUTABLE_RETURN)
2622 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2623 }
2624 if (lp)
2625 *lp = SvCUR(tmpstr);
50adf7d2 2626 } else {
deb46114 2627 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2628 }
deb46114
NC
2629 if (SvUTF8(tmpstr))
2630 SvUTF8_on(sv);
2631 else
2632 SvUTF8_off(sv);
2633 return pv;
50adf7d2 2634 }
deb46114
NC
2635 }
2636 {
fafee734
NC
2637 STRLEN len;
2638 char *retval;
2639 char *buffer;
f9277f47 2640 MAGIC *mg;
d8eae41e
NC
2641 const SV *const referent = (SV*)SvRV(sv);
2642
2643 if (!referent) {
fafee734
NC
2644 len = 7;
2645 retval = buffer = savepvn("NULLREF", len);
042dae7a
NC
2646 } else if (SvTYPE(referent) == SVt_PVMG
2647 && ((SvFLAGS(referent) &
2648 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2649 == (SVs_OBJECT|SVs_SMG))
de8c5301
YO
2650 && (mg = mg_find(referent, PERL_MAGIC_qr)))
2651 {
2652 char *str = NULL;
2653 I32 haseval = 0;
60df1e07 2654 U32 flags = 0;
de8c5301
YO
2655 (str) = CALLREG_AS_STR(mg,lp,&flags,&haseval);
2656 if (flags & 1)
2657 SvUTF8_on(sv);
2658 else
2659 SvUTF8_off(sv);
2660 PL_reginterp_cnt += haseval;
2661 return str;
d8eae41e
NC
2662 } else {
2663 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2664 const STRLEN typelen = strlen(typestr);
2665 UV addr = PTR2UV(referent);
2666 const char *stashname = NULL;
2667 STRLEN stashnamelen = 0; /* hush, gcc */
2668 const char *buffer_end;
d8eae41e 2669
d8eae41e 2670 if (SvOBJECT(referent)) {
fafee734
NC
2671 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2672
2673 if (name) {
2674 stashname = HEK_KEY(name);
2675 stashnamelen = HEK_LEN(name);
2676
2677 if (HEK_UTF8(name)) {
2678 SvUTF8_on(sv);
2679 } else {
2680 SvUTF8_off(sv);
2681 }
2682 } else {
2683 stashname = "__ANON__";
2684 stashnamelen = 8;
2685 }
2686 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2687 + 2 * sizeof(UV) + 2 /* )\0 */;
2688 } else {
2689 len = typelen + 3 /* (0x */
2690 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2691 }
fafee734
NC
2692
2693 Newx(buffer, len, char);
2694 buffer_end = retval = buffer + len;
2695
2696 /* Working backwards */
2697 *--retval = '\0';
2698 *--retval = ')';
2699 do {
2700 *--retval = PL_hexdigit[addr & 15];
2701 } while (addr >>= 4);
2702 *--retval = 'x';
2703 *--retval = '0';
2704 *--retval = '(';
2705
2706 retval -= typelen;
2707 memcpy(retval, typestr, typelen);
2708
2709 if (stashname) {
2710 *--retval = '=';
2711 retval -= stashnamelen;
2712 memcpy(retval, stashname, stashnamelen);
2713 }
2714 /* retval may not neccesarily have reached the start of the
2715 buffer here. */
2716 assert (retval >= buffer);
2717
2718 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2719 }
042dae7a 2720 if (lp)
fafee734
NC
2721 *lp = len;
2722 SAVEFREEPV(buffer);
2723 return retval;
463ee0b2 2724 }
79072805 2725 }
0336b60e 2726 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2727 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2728 report_uninit(sv);
cdb061a3
NC
2729 if (lp)
2730 *lp = 0;
73d840c0 2731 return (char *)"";
79072805 2732 }
79072805 2733 }
28e5dec8
JH
2734 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2735 /* I'm assuming that if both IV and NV are equally valid then
2736 converting the IV is going to be more efficient */
e1ec3a88 2737 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2738 char buf[TYPE_CHARS(UV)];
2739 char *ebuf, *ptr;
2740
2741 if (SvTYPE(sv) < SVt_PVIV)
2742 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2743 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2744 /* inlined from sv_setpvn */
2745 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2746 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2747 SvCUR_set(sv, ebuf - ptr);
2748 s = SvEND(sv);
2749 *s = '\0';
28e5dec8
JH
2750 }
2751 else if (SvNOKp(sv)) {
c81271c3 2752 const int olderrno = errno;
79072805
LW
2753 if (SvTYPE(sv) < SVt_PVNV)
2754 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2755 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2756 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2757 /* some Xenix systems wipe out errno here */
79072805 2758#ifdef apollo
463ee0b2 2759 if (SvNVX(sv) == 0.0)
d1307786 2760 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2761 else
2762#endif /*apollo*/
bbce6d69 2763 {
2d4389e4 2764 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2765 }
79072805 2766 errno = olderrno;
a0d0e21e
LW
2767#ifdef FIXNEGATIVEZERO
2768 if (*s == '-' && s[1] == '0' && !s[2])
d1307786 2769 my_strlcpy(s, "0", SvLEN(s));
a0d0e21e 2770#endif
79072805
LW
2771 while (*s) s++;
2772#ifdef hcx
2773 if (s[-1] == '.')
46fc3d4c 2774 *--s = '\0';
79072805
LW
2775#endif
2776 }
79072805 2777 else {
675c862f 2778 if (isGV_with_GP(sv))
19f6321d 2779 return glob_2pv((GV *)sv, lp);
180488f8 2780
041457d9 2781 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2782 report_uninit(sv);
cdb061a3 2783 if (lp)
00b6aa41 2784 *lp = 0;
25da4f38
IZ
2785 if (SvTYPE(sv) < SVt_PV)
2786 /* Typically the caller expects that sv_any is not NULL now. */
2787 sv_upgrade(sv, SVt_PV);
73d840c0 2788 return (char *)"";
79072805 2789 }
cdb061a3 2790 {
823a54a3 2791 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2792 if (lp)
2793 *lp = len;
2794 SvCUR_set(sv, len);
2795 }
79072805 2796 SvPOK_on(sv);
1d7c1841 2797 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2798 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2799 if (flags & SV_CONST_RETURN)
2800 return (char *)SvPVX_const(sv);
10516c54
NC
2801 if (flags & SV_MUTABLE_RETURN)
2802 return SvPVX_mutable(sv);
463ee0b2
LW
2803 return SvPVX(sv);
2804}
2805
645c22ef 2806/*
6050d10e
JP
2807=for apidoc sv_copypv
2808
2809Copies a stringified representation of the source SV into the
2810destination SV. Automatically performs any necessary mg_get and
54f0641b 2811coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2812UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2813sv_2pv[_flags] but operates directly on an SV instead of just the
2814string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2815would lose the UTF-8'ness of the PV.
2816
2817=cut
2818*/
2819
2820void
2821Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2822{
446eaa42 2823 STRLEN len;
53c1dcc0 2824 const char * const s = SvPV_const(ssv,len);
cb50f42d 2825 sv_setpvn(dsv,s,len);
446eaa42 2826 if (SvUTF8(ssv))
cb50f42d 2827 SvUTF8_on(dsv);
446eaa42 2828 else
cb50f42d 2829 SvUTF8_off(dsv);
6050d10e
JP
2830}
2831
2832/*
645c22ef
DM
2833=for apidoc sv_2pvbyte
2834
2835Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2836to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2837side-effect.
2838
2839Usually accessed via the C<SvPVbyte> macro.
2840
2841=cut
2842*/
2843
7340a771
GS
2844char *
2845Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2846{
0875d2fe 2847 sv_utf8_downgrade(sv,0);
97972285 2848 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2849}
2850
645c22ef 2851/*
035cbb0e
RGS
2852=for apidoc sv_2pvutf8
2853
2854Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2855to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2856
2857Usually accessed via the C<SvPVutf8> macro.
2858
2859=cut
2860*/
645c22ef 2861
7340a771
GS
2862char *
2863Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2864{
035cbb0e
RGS
2865 sv_utf8_upgrade(sv);
2866 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2867}
1c846c1f 2868
7ee2227d 2869
645c22ef
DM
2870/*
2871=for apidoc sv_2bool
2872
2873This function is only called on magical items, and is only used by
8cf8f3d1 2874sv_true() or its macro equivalent.
645c22ef
DM
2875
2876=cut
2877*/
2878
463ee0b2 2879bool
864dbfa3 2880Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2881{
97aff369 2882 dVAR;
5b295bef 2883 SvGETMAGIC(sv);
463ee0b2 2884
a0d0e21e
LW
2885 if (!SvOK(sv))
2886 return 0;
2887 if (SvROK(sv)) {
fabdb6c0
AL
2888 if (SvAMAGIC(sv)) {
2889 SV * const tmpsv = AMG_CALLun(sv,bool_);
2890 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2891 return (bool)SvTRUE(tmpsv);
2892 }
2893 return SvRV(sv) != 0;
a0d0e21e 2894 }
463ee0b2 2895 if (SvPOKp(sv)) {
53c1dcc0
AL
2896 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2897 if (Xpvtmp &&
339049b0 2898 (*sv->sv_u.svu_pv > '0' ||
11343788 2899 Xpvtmp->xpv_cur > 1 ||
339049b0 2900 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2901 return 1;
2902 else
2903 return 0;
2904 }
2905 else {
2906 if (SvIOKp(sv))
2907 return SvIVX(sv) != 0;
2908 else {
2909 if (SvNOKp(sv))
2910 return SvNVX(sv) != 0.0;
180488f8 2911 else {
f7877b28 2912 if (isGV_with_GP(sv))
180488f8
NC
2913 return TRUE;
2914 else
2915 return FALSE;
2916 }
463ee0b2
LW
2917 }
2918 }
79072805
LW
2919}
2920
c461cf8f
JH
2921/*
2922=for apidoc sv_utf8_upgrade
2923
78ea37eb 2924Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2925Forces the SV to string form if it is not already.
4411f3b6
NIS
2926Always sets the SvUTF8 flag to avoid future validity checks even
2927if all the bytes have hibit clear.
c461cf8f 2928
13a6c0e0
JH
2929This is not as a general purpose byte encoding to Unicode interface:
2930use the Encode extension for that.
2931
8d6d96c1
HS
2932=for apidoc sv_utf8_upgrade_flags
2933
78ea37eb 2934Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2935Forces the SV to string form if it is not already.
8d6d96c1
HS
2936Always sets the SvUTF8 flag to avoid future validity checks even
2937if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2938will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2939C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2940
13a6c0e0
JH
2941This is not as a general purpose byte encoding to Unicode interface:
2942use the Encode extension for that.
2943
8d6d96c1
HS
2944=cut
2945*/
2946
2947STRLEN
2948Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2949{
97aff369 2950 dVAR;
808c356f
RGS
2951 if (sv == &PL_sv_undef)
2952 return 0;
e0e62c2a
NIS
2953 if (!SvPOK(sv)) {
2954 STRLEN len = 0;
d52b7888
NC
2955 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2956 (void) sv_2pv_flags(sv,&len, flags);
2957 if (SvUTF8(sv))
2958 return len;
2959 } else {
2960 (void) SvPV_force(sv,len);
2961 }
e0e62c2a 2962 }
4411f3b6 2963
f5cee72b 2964 if (SvUTF8(sv)) {
5fec3b1d 2965 return SvCUR(sv);
f5cee72b 2966 }
5fec3b1d 2967
765f542d
NC
2968 if (SvIsCOW(sv)) {
2969 sv_force_normal_flags(sv, 0);
db42d148
NIS
2970 }
2971
88632417 2972 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2973 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2974 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2975 /* This function could be much more efficient if we
2976 * had a FLAG in SVs to signal if there are any hibit
2977 * chars in the PV. Given that there isn't such a flag
2978 * make the loop as fast as possible. */
00b6aa41 2979 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 2980 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 2981 const U8 *t = s;
c4e7c712
NC
2982
2983 while (t < e) {
53c1dcc0 2984 const U8 ch = *t++;
00b6aa41
AL
2985 /* Check for hi bit */
2986 if (!NATIVE_IS_INVARIANT(ch)) {
2987 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
2988 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
2989
2990 SvPV_free(sv); /* No longer using what was there before. */
2991 SvPV_set(sv, (char*)recoded);
2992 SvCUR_set(sv, len - 1);
2993 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 2994 break;
00b6aa41 2995 }
c4e7c712
NC
2996 }
2997 /* Mark as UTF-8 even if no hibit - saves scanning loop */
2998 SvUTF8_on(sv);
560a288e 2999 }
4411f3b6 3000 return SvCUR(sv);
560a288e
GS
3001}
3002
c461cf8f
JH
3003/*
3004=for apidoc sv_utf8_downgrade
3005
78ea37eb
TS
3006Attempts to convert the PV of an SV from characters to bytes.
3007If the PV contains a character beyond byte, this conversion will fail;
3008in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3009true, croaks.
3010
13a6c0e0
JH
3011This is not as a general purpose Unicode to byte encoding interface:
3012use the Encode extension for that.
3013
c461cf8f
JH
3014=cut
3015*/
3016
560a288e
GS
3017bool
3018Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3019{
97aff369 3020 dVAR;
78ea37eb 3021 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3022 if (SvCUR(sv)) {
03cfe0ae 3023 U8 *s;
652088fc 3024 STRLEN len;
fa301091 3025
765f542d
NC
3026 if (SvIsCOW(sv)) {
3027 sv_force_normal_flags(sv, 0);
3028 }
03cfe0ae
NIS
3029 s = (U8 *) SvPV(sv, len);
3030 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3031 if (fail_ok)
3032 return FALSE;
3033 else {
3034 if (PL_op)
3035 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3036 OP_DESC(PL_op));
fa301091
JH
3037 else
3038 Perl_croak(aTHX_ "Wide character");
3039 }
4b3603a4 3040 }
b162af07 3041 SvCUR_set(sv, len);
67e989fb 3042 }
560a288e 3043 }
ffebcc3e 3044 SvUTF8_off(sv);
560a288e
GS
3045 return TRUE;
3046}
3047
c461cf8f
JH
3048/*
3049=for apidoc sv_utf8_encode
3050
78ea37eb
TS
3051Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3052flag off so that it looks like octets again.
c461cf8f
JH
3053
3054=cut
3055*/
3056
560a288e
GS
3057void
3058Perl_sv_utf8_encode(pTHX_ register SV *sv)
3059{
4c94c214
NC
3060 if (SvIsCOW(sv)) {
3061 sv_force_normal_flags(sv, 0);
3062 }
3063 if (SvREADONLY(sv)) {
3064 Perl_croak(aTHX_ PL_no_modify);
3065 }
a5f5288a 3066 (void) sv_utf8_upgrade(sv);
560a288e
GS
3067 SvUTF8_off(sv);
3068}
3069
4411f3b6
NIS
3070/*
3071=for apidoc sv_utf8_decode
3072
78ea37eb
TS
3073If the PV of the SV is an octet sequence in UTF-8
3074and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3075so that it looks like a character. If the PV contains only single-byte
3076characters, the C<SvUTF8> flag stays being off.
3077Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3078
3079=cut
3080*/
3081
560a288e
GS
3082bool
3083Perl_sv_utf8_decode(pTHX_ register SV *sv)
3084{
78ea37eb 3085 if (SvPOKp(sv)) {
93524f2b
NC
3086 const U8 *c;
3087 const U8 *e;
9cbac4c7 3088
645c22ef
DM
3089 /* The octets may have got themselves encoded - get them back as
3090 * bytes
3091 */
3092 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3093 return FALSE;
3094
3095 /* it is actually just a matter of turning the utf8 flag on, but
3096 * we want to make sure everything inside is valid utf8 first.
3097 */
93524f2b 3098 c = (const U8 *) SvPVX_const(sv);
63cd0674 3099 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3100 return FALSE;
93524f2b 3101 e = (const U8 *) SvEND(sv);
511c2ff0 3102 while (c < e) {
b64e5050 3103 const U8 ch = *c++;
c4d5f83a 3104 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3105 SvUTF8_on(sv);
3106 break;
3107 }
560a288e 3108 }
560a288e
GS
3109 }
3110 return TRUE;
3111}
3112
954c1994
GS
3113/*
3114=for apidoc sv_setsv
3115
645c22ef
DM
3116Copies the contents of the source SV C<ssv> into the destination SV
3117C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3118function if the source SV needs to be reused. Does not handle 'set' magic.
3119Loosely speaking, it performs a copy-by-value, obliterating any previous
3120content of the destination.
3121
3122You probably want to use one of the assortment of wrappers, such as
3123C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3124C<SvSetMagicSV_nosteal>.
3125
8d6d96c1
HS
3126=for apidoc sv_setsv_flags
3127
645c22ef
DM
3128Copies the contents of the source SV C<ssv> into the destination SV
3129C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3130function if the source SV needs to be reused. Does not handle 'set' magic.
3131Loosely speaking, it performs a copy-by-value, obliterating any previous
3132content of the destination.
3133If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3134C<ssv> if appropriate, else not. If the C<flags> parameter has the
3135C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3136and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3137
3138You probably want to use one of the assortment of wrappers, such as
3139C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3140C<SvSetMagicSV_nosteal>.
3141
3142This is the primary function for copying scalars, and most other
3143copy-ish functions and macros use this underneath.
8d6d96c1
HS
3144
3145=cut
3146*/
3147
5d0301b7 3148static void
2eb42952 3149S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3150{
3151 if (dtype != SVt_PVGV) {
3152 const char * const name = GvNAME(sstr);
3153 const STRLEN len = GvNAMELEN(sstr);
0d092c36 3154 {
f7877b28
NC
3155 if (dtype >= SVt_PV) {
3156 SvPV_free(dstr);
3157 SvPV_set(dstr, 0);
3158 SvLEN_set(dstr, 0);
3159 SvCUR_set(dstr, 0);
3160 }
0d092c36 3161 SvUPGRADE(dstr, SVt_PVGV);
dedf8e73 3162 (void)SvOK_off(dstr);
2e5b91de
NC
3163 /* FIXME - why are we doing this, then turning it off and on again
3164 below? */
3165 isGV_with_GP_on(dstr);
f7877b28 3166 }
5d0301b7
NC
3167 GvSTASH(dstr) = GvSTASH(sstr);
3168 if (GvSTASH(dstr))
3169 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3170 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3171 SvFAKE_on(dstr); /* can coerce to non-glob */
3172 }
3173
3174#ifdef GV_UNIQUE_CHECK
3175 if (GvUNIQUE((GV*)dstr)) {
3176 Perl_croak(aTHX_ PL_no_modify);
3177 }
3178#endif
3179
f7877b28 3180 gp_free((GV*)dstr);
2e5b91de 3181 isGV_with_GP_off(dstr);
5d0301b7 3182 (void)SvOK_off(dstr);
2e5b91de 3183 isGV_with_GP_on(dstr);
dedf8e73 3184 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3185 GvGP(dstr) = gp_ref(GvGP(sstr));
3186 if (SvTAINTED(sstr))
3187 SvTAINT(dstr);
3188 if (GvIMPORTED(dstr) != GVf_IMPORTED
3189 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3190 {
3191 GvIMPORTED_on(dstr);
3192 }
3193 GvMULTI_on(dstr);
3194 return;
3195}
3196
b8473700 3197static void
2eb42952 3198S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3199 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3200 SV *dref = NULL;
3201 const int intro = GvINTRO(dstr);
2440974c 3202 SV **location;
3386d083 3203 U8 import_flag = 0;
27242d61
NC
3204 const U32 stype = SvTYPE(sref);
3205
b8473700
NC
3206
3207#ifdef GV_UNIQUE_CHECK
3208 if (GvUNIQUE((GV*)dstr)) {
3209 Perl_croak(aTHX_ PL_no_modify);
3210 }
3211#endif
3212
3213 if (intro) {
3214 GvINTRO_off(dstr); /* one-shot flag */
3215 GvLINE(dstr) = CopLINE(PL_curcop);
3216 GvEGV(dstr) = (GV*)dstr;
3217 }
3218 GvMULTI_on(dstr);
27242d61 3219 switch (stype) {
b8473700 3220 case SVt_PVCV:
27242d61
NC
3221 location = (SV **) &GvCV(dstr);
3222 import_flag = GVf_IMPORTED_CV;
3223 goto common;
3224 case SVt_PVHV:
3225 location = (SV **) &GvHV(dstr);
3226 import_flag = GVf_IMPORTED_HV;
3227 goto common;
3228 case SVt_PVAV:
3229 location = (SV **) &GvAV(dstr);
3230 import_flag = GVf_IMPORTED_AV;
3231 goto common;
3232 case SVt_PVIO:
3233 location = (SV **) &GvIOp(dstr);
3234 goto common;
3235 case SVt_PVFM:
3236 location = (SV **) &GvFORM(dstr);
3237 default:
3238 location = &GvSV(dstr);
3239 import_flag = GVf_IMPORTED_SV;
3240 common:
b8473700 3241 if (intro) {
27242d61
NC
3242 if (stype == SVt_PVCV) {
3243 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3244 SvREFCNT_dec(GvCV(dstr));
3245 GvCV(dstr) = NULL;
3246 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3247 PL_sub_generation++;
3248 }
b8473700 3249 }
27242d61 3250 SAVEGENERICSV(*location);
b8473700
NC
3251 }
3252 else
27242d61
NC
3253 dref = *location;
3254 if (stype == SVt_PVCV && *location != sref) {
3255 CV* const cv = (CV*)*location;
b8473700
NC
3256 if (cv) {
3257 if (!GvCVGEN((GV*)dstr) &&
3258 (CvROOT(cv) || CvXSUB(cv)))
3259 {
3260 /* Redefining a sub - warning is mandatory if
3261 it was a const and its value changed. */
3262 if (CvCONST(cv) && CvCONST((CV*)sref)
3263 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3264 NOOP;
b8473700
NC
3265 /* They are 2 constant subroutines generated from
3266 the same constant. This probably means that
3267 they are really the "same" proxy subroutine
3268 instantiated in 2 places. Most likely this is
3269 when a constant is exported twice. Don't warn.
3270 */
3271 }
3272 else if (ckWARN(WARN_REDEFINE)
3273 || (CvCONST(cv)
3274 && (!CvCONST((CV*)sref)
3275 || sv_cmp(cv_const_sv(cv),
3276 cv_const_sv((CV*)sref))))) {
3277 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3278 (const char *)
3279 (CvCONST(cv)
3280 ? "Constant subroutine %s::%s redefined"
3281 : "Subroutine %s::%s redefined"),
b8473700
NC
3282 HvNAME_get(GvSTASH((GV*)dstr)),
3283 GvENAME((GV*)dstr));
3284 }
3285 }
3286 if (!intro)
cbf82dd0
NC
3287 cv_ckproto_len(cv, (GV*)dstr,
3288 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3289 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3290 }
b8473700
NC
3291 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3292 GvASSUMECV_on(dstr);
3293 PL_sub_generation++;
3294 }
2440974c 3295 *location = sref;
3386d083
NC
3296 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3297 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3298 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3299 }
3300 break;
3301 }
b37c2d43 3302 SvREFCNT_dec(dref);
b8473700
NC
3303 if (SvTAINTED(sstr))
3304 SvTAINT(dstr);
3305 return;
3306}
3307
8d6d96c1
HS
3308void
3309Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3310{
97aff369 3311 dVAR;
8990e307
LW
3312 register U32 sflags;
3313 register int dtype;
42d0e0b7 3314 register svtype stype;
463ee0b2 3315
79072805
LW
3316 if (sstr == dstr)
3317 return;
29f4f0ab
NC
3318
3319 if (SvIS_FREED(dstr)) {
3320 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
be2597df 3321 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
29f4f0ab 3322 }
765f542d 3323 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3324 if (!sstr)
3280af22 3325 sstr = &PL_sv_undef;
29f4f0ab 3326 if (SvIS_FREED(sstr)) {
6c9570dc
MHM
3327 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3328 (void*)sstr, (void*)dstr);
29f4f0ab 3329 }
8990e307
LW
3330 stype = SvTYPE(sstr);
3331 dtype = SvTYPE(dstr);
79072805 3332
52944de8 3333 (void)SvAMAGIC_off(dstr);
7a5fa8a2 3334 if ( SvVOK(dstr) )
ece467f9
JP
3335 {
3336 /* need to nuke the magic */
3337 mg_free(dstr);
3338 SvRMAGICAL_off(dstr);
3339 }
9e7bc3e8 3340
463ee0b2 3341 /* There's a lot of redundancy below but we're going for speed here */
79072805 3342
8990e307 3343 switch (stype) {
79072805 3344 case SVt_NULL:
aece5585 3345 undef_sstr:
20408e3c
GS
3346 if (dtype != SVt_PVGV) {
3347 (void)SvOK_off(dstr);
3348 return;
3349 }
3350 break;
463ee0b2 3351 case SVt_IV:
aece5585
GA
3352 if (SvIOK(sstr)) {
3353 switch (dtype) {
3354 case SVt_NULL:
8990e307 3355 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3356 break;
3357 case SVt_NV:
aece5585
GA
3358 case SVt_RV:
3359 case SVt_PV:
a0d0e21e 3360 sv_upgrade(dstr, SVt_PVIV);
aece5585 3361 break;
010be86b
NC
3362 case SVt_PVGV:
3363 goto end_of_first_switch;
aece5585
GA
3364 }
3365 (void)SvIOK_only(dstr);
45977657 3366 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3367 if (SvIsUV(sstr))
3368 SvIsUV_on(dstr);
37c25af0
NC
3369 /* SvTAINTED can only be true if the SV has taint magic, which in
3370 turn means that the SV type is PVMG (or greater). This is the
3371 case statement for SVt_IV, so this cannot be true (whatever gcov
3372 may say). */
3373 assert(!SvTAINTED(sstr));
aece5585 3374 return;
8990e307 3375 }
aece5585
GA
3376 goto undef_sstr;
3377
463ee0b2 3378 case SVt_NV:
aece5585
GA
3379 if (SvNOK(sstr)) {
3380 switch (dtype) {
3381 case SVt_NULL:
3382 case SVt_IV:
8990e307 3383 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3384 break;
3385 case SVt_RV:
3386 case SVt_PV:
3387 case SVt_PVIV:
a0d0e21e 3388 sv_upgrade(dstr, SVt_PVNV);
aece5585 3389 break;
010be86b
NC
3390 case SVt_PVGV:
3391 goto end_of_first_switch;
aece5585 3392 }
9d6ce603 3393 SvNV_set(dstr, SvNVX(sstr));
aece5585 3394 (void)SvNOK_only(dstr);
37c25af0
NC
3395 /* SvTAINTED can only be true if the SV has taint magic, which in
3396 turn means that the SV type is PVMG (or greater). This is the
3397 case statement for SVt_NV, so this cannot be true (whatever gcov
3398 may say). */
3399 assert(!SvTAINTED(sstr));
aece5585 3400 return;
8990e307 3401 }
aece5585
GA
3402 goto undef_sstr;
3403
ed6116ce 3404 case SVt_RV:
8990e307 3405 if (dtype < SVt_RV)
ed6116ce 3406 sv_upgrade(dstr, SVt_RV);
ed6116ce 3407 break;
fc36a67e 3408 case SVt_PVFM:
f8c7b90f 3409#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3410 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3411 if (dtype < SVt_PVIV)
3412 sv_upgrade(dstr, SVt_PVIV);
3413 break;
3414 }
3415 /* Fall through */
3416#endif
3417 case SVt_PV:
8990e307 3418 if (dtype < SVt_PV)
463ee0b2 3419 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3420 break;
3421 case SVt_PVIV:
8990e307 3422 if (dtype < SVt_PVIV)
463ee0b2 3423 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3424 break;
3425 case SVt_PVNV:
8990e307 3426 if (dtype < SVt_PVNV)
463ee0b2 3427 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3428 break;
489f7bfe 3429 default:
a3b680e6
AL
3430 {
3431 const char * const type = sv_reftype(sstr,0);
533c011a 3432 if (PL_op)
a3b680e6 3433 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3434 else
a3b680e6
AL
3435 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3436 }
4633a7c4
LW
3437 break;
3438
cecf5685 3439 /* case SVt_BIND: */
39cb70dc 3440 case SVt_PVLV:
79072805 3441 case SVt_PVGV:
cecf5685 3442 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
d4c19fe8 3443 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3444 return;
79072805 3445 }
cecf5685 3446 /* SvVALID means that this PVGV is playing at being an FBM. */
5f66b61c 3447 /*FALLTHROUGH*/
79072805 3448
489f7bfe 3449 case SVt_PVMG:
8d6d96c1 3450 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3451 mg_get(sstr);
1d9c78c6 3452 if (SvTYPE(sstr) != stype) {
973f89ab 3453 stype = SvTYPE(sstr);
cecf5685 3454 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3455 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3456 return;
3457 }
973f89ab
CS
3458 }
3459 }
ded42b9f 3460 if (stype == SVt_PVLV)
862a34c6 3461 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3462 else
42d0e0b7 3463 SvUPGRADE(dstr, (svtype)stype);
79072805 3464 }
010be86b 3465 end_of_first_switch:
79072805 3466
ff920335
NC
3467 /* dstr may have been upgraded. */
3468 dtype = SvTYPE(dstr);
8990e307
LW
3469 sflags = SvFLAGS(sstr);
3470
ba2fdce6 3471 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
85324b4d
NC
3472 /* Assigning to a subroutine sets the prototype. */
3473 if (SvOK(sstr)) {
3474 STRLEN len;
3475 const char *const ptr = SvPV_const(sstr, len);
3476
3477 SvGROW(dstr, len + 1);
3478 Copy(ptr, SvPVX(dstr), len + 1, char);
3479 SvCUR_set(dstr, len);
fcddd32e 3480 SvPOK_only(dstr);
ba2fdce6 3481 SvFLAGS(dstr) |= sflags & SVf_UTF8;
85324b4d
NC
3482 } else {
3483 SvOK_off(dstr);
3484 }
ba2fdce6
NC
3485 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3486 const char * const type = sv_reftype(dstr,0);
3487 if (PL_op)
3488 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3489 else
3490 Perl_croak(aTHX_ "Cannot copy to %s", type);
85324b4d 3491 } else if (sflags & SVf_ROK) {
cecf5685
NC
3492 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3493 && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
acaa9288
NC
3494 sstr = SvRV(sstr);
3495 if (sstr == dstr) {
3496 if (GvIMPORTED(dstr) != GVf_IMPORTED
3497 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3498 {
3499 GvIMPORTED_on(dstr);
3500 }
3501 GvMULTI_on(dstr);
3502 return;
3503 }
d4c19fe8 3504 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3505 return;
3506 }
3507
8990e307 3508 if (dtype >= SVt_PV) {
b8c701c1 3509 if (dtype == SVt_PVGV) {
d4c19fe8 3510 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3511 return;
3512 }
3f7c398e 3513 if (SvPVX_const(dstr)) {
8bd4d4c5 3514 SvPV_free(dstr);
b162af07
SP
3515 SvLEN_set(dstr, 0);
3516 SvCUR_set(dstr, 0);
a0d0e21e 3517 }
8990e307 3518 }
a0d0e21e 3519 (void)SvOK_off(dstr);
b162af07 3520 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
96d4b0ee 3521 SvFLAGS(dstr) |= sflags & SVf_ROK;
dfd48732
NC
3522 assert(!(sflags & SVp_NOK));
3523 assert(!(sflags & SVp_IOK));
3524 assert(!(sflags & SVf_NOK));
3525 assert(!(sflags & SVf_IOK));
ed6116ce 3526 }
cecf5685 3527 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
c0c44674
NC
3528 if (!(sflags & SVf_OK)) {
3529 if (ckWARN(WARN_MISC))
3530 Perl_warner(aTHX_ packWARN(WARN_MISC),
3531 "Undefined value assigned to typeglob");
3532 }
3533 else {
3534 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3535 if (dstr != (SV*)gv) {
3536 if (GvGP(dstr))
3537 gp_free((GV*)dstr);
3538 GvGP(dstr) = gp_ref(GvGP(gv));
3539 }
3540 }
3541 }
8990e307 3542 else if (sflags & SVp_POK) {
765f542d 3543 bool isSwipe = 0;
79072805
LW
3544
3545 /*
3546 * Check to see if we can just swipe the string. If so, it's a
3547 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3548 * It might even be a win on short strings if SvPVX_const(dstr)
3549 * has to be allocated and SvPVX_const(sstr) has to be freed.
34482cd6
NC
3550 * Likewise if we can set up COW rather than doing an actual copy, we
3551 * drop to the else clause, as the swipe code and the COW setup code
3552 * have much in common.
79072805
LW
3553 */
3554
120fac95
NC
3555 /* Whichever path we take through the next code, we want this true,
3556 and doing it now facilitates the COW check. */
3557 (void)SvPOK_only(dstr);
3558
765f542d 3559 if (
34482cd6
NC
3560 /* If we're already COW then this clause is not true, and if COW
3561 is allowed then we drop down to the else and make dest COW
3562 with us. If caller hasn't said that we're allowed to COW
3563 shared hash keys then we don't do the COW setup, even if the
3564 source scalar is a shared hash key scalar. */
3565 (((flags & SV_COW_SHARED_HASH_KEYS)
3566 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
3567 : 1 /* If making a COW copy is forbidden then the behaviour we
3568 desire is as if the source SV isn't actually already
3569 COW, even if it is. So we act as if the source flags
3570 are not COW, rather than actually testing them. */
3571 )
f8c7b90f 3572#ifndef PERL_OLD_COPY_ON_WRITE
34482cd6
NC
3573 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
3574 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
3575 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
3576 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
3577 but in turn, it's somewhat dead code, never expected to go
3578 live, but more kept as a placeholder on how to do it better
3579 in a newer implementation. */
3580 /* If we are COW and dstr is a suitable target then we drop down
3581 into the else and make dest a COW of us. */
b8f9541a
NC
3582 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3583#endif
3584 )
765f542d 3585 &&
765f542d
NC
3586 !(isSwipe =
3587 (sflags & SVs_TEMP) && /* slated for free anyway? */
3588 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3589 (!(flags & SV_NOSTEAL)) &&
3590 /* and we're allowed to steal temps */
765f542d
NC
3591 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3592 SvLEN(sstr) && /* and really is a string */
645c22ef 3593 /* and won't be needed again, potentially */
765f542d 3594 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3595#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3596 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3597 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3598 && SvTYPE(sstr) >= SVt_PVIV)
3599#endif
3600 ) {
3601 /* Failed the swipe test, and it's not a shared hash key either.
3602 Have to copy the string. */
3603 STRLEN len = SvCUR(sstr);
3604 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3605 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3606 SvCUR_set(dstr, len);
3607 *SvEND(dstr) = '\0';
765f542d 3608 } else {
f8c7b90f 3609 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3610 be true in here. */
765f542d
NC
3611 /* Either it's a shared hash key, or it's suitable for
3612 copy-on-write or we can swipe the string. */
46187eeb 3613 if (DEBUG_C_TEST) {
ed252734 3614 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3615 sv_dump(sstr);
3616 sv_dump(dstr);
46187eeb 3617 }
f8c7b90f 3618#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3619 if (!isSwipe) {
3620 /* I believe I should acquire a global SV mutex if
3621 it's a COW sv (not a shared hash key) to stop
3622 it going un copy-on-write.
3623 If the source SV has gone un copy on write between up there
3624 and down here, then (assert() that) it is of the correct
3625 form to make it copy on write again */
3626 if ((sflags & (SVf_FAKE | SVf_READONLY))
3627 != (SVf_FAKE | SVf_READONLY)) {
3628 SvREADONLY_on(sstr);
3629 SvFAKE_on(sstr);
3630 /* Make the source SV into a loop of 1.
3631 (about to become 2) */
a29f6d03 3632 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3633 }
3634 }
3635#endif
3636 /* Initial code is common. */
94010e71
NC
3637 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3638 SvPV_free(dstr);
79072805 3639 }
765f542d 3640
765f542d
NC
3641 if (!isSwipe) {
3642 /* making another shared SV. */
3643 STRLEN cur = SvCUR(sstr);
3644 STRLEN len = SvLEN(sstr);
f8c7b90f 3645#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3646 if (len) {
b8f9541a 3647 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3648 /* SvIsCOW_normal */
3649 /* splice us in between source and next-after-source. */
a29f6d03
NC
3650 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3651 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3652 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3653 } else
3654#endif
3655 {
765f542d 3656 /* SvIsCOW_shared_hash */
46187eeb
NC
3657 DEBUG_C(PerlIO_printf(Perl_debug_log,
3658 "Copy on write: Sharing hash\n"));
b8f9541a 3659
bdd68bc3 3660 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3661 SvPV_set(dstr,
d1db91c6 3662 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3663 }
87a1ef3d
SP
3664 SvLEN_set(dstr, len);
3665 SvCUR_set(dstr, cur);
765f542d
NC
3666 SvREADONLY_on(dstr);
3667 SvFAKE_on(dstr);
3668 /* Relesase a global SV mutex. */
3669 }
3670 else
765f542d 3671 { /* Passes the swipe test. */
78d1e721 3672 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3673 SvLEN_set(dstr, SvLEN(sstr));
3674 SvCUR_set(dstr, SvCUR(sstr));
3675
3676 SvTEMP_off(dstr);
3677 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3678 SvPV_set(sstr, NULL);
765f542d
NC
3679 SvLEN_set(sstr, 0);
3680 SvCUR_set(sstr, 0);
3681 SvTEMP_off(sstr);
3682 }
3683 }
8990e307 3684 if (sflags & SVp_NOK) {
9d6ce603 3685 SvNV_set(dstr, SvNVX(sstr));
79072805 3686 }
8990e307 3687 if (sflags & SVp_IOK) {
88555484 3688 SvOOK_off(dstr);
23525414
NC
3689 SvIV_set(dstr, SvIVX(sstr));
3690 /* Must do this otherwise some other overloaded use of 0x80000000
3691 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3692 if (sflags & SVf_IVisUV)
25da4f38 3693 SvIsUV_on(dstr);
79072805 3694 }
96d4b0ee 3695 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4f2da183 3696 {
b0a11fe1 3697 const MAGIC * const smg = SvVSTRING_mg(sstr);
4f2da183
NC
3698 if (smg) {
3699 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3700 smg->mg_ptr, smg->mg_len);
3701 SvRMAGICAL_on(dstr);
3702 }
7a5fa8a2 3703 }
79072805 3704 }
5d581361 3705 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3706 (void)SvOK_off(dstr);
96d4b0ee 3707 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
5d581361
NC
3708 if (sflags & SVp_IOK) {
3709 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3710 SvIV_set(dstr, SvIVX(sstr));
3711 }
3332b3c1 3712 if (sflags & SVp_NOK) {
9d6ce603 3713 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3714 }
3715 }
79072805 3716 else {
f7877b28 3717 if (isGV_with_GP(sstr)) {
180488f8
NC
3718 /* This stringification rule for globs is spread in 3 places.
3719 This feels bad. FIXME. */
3720 const U32 wasfake = sflags & SVf_FAKE;
3721
3722 /* FAKE globs can get coerced, so need to turn this off
3723 temporarily if it is on. */
3724 SvFAKE_off(sstr);
3725 gv_efullname3(dstr, (GV *)sstr, "*");
3726 SvFLAGS(sstr) |= wasfake;
3727 }
20408e3c
GS
3728 else
3729 (void)SvOK_off(dstr);
a0d0e21e 3730 }
27c9684d
AP
3731 if (SvTAINTED(sstr))
3732 SvTAINT(dstr);
79072805
LW
3733}
3734
954c1994
GS
3735/*
3736=for apidoc sv_setsv_mg
3737
3738Like C<sv_setsv>, but also handles 'set' magic.
3739
3740=cut
3741*/
3742
79072805 3743void
864dbfa3 3744Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3745{
3746 sv_setsv(dstr,sstr);
3747 SvSETMAGIC(dstr);
3748}
3749
f8c7b90f 3750#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3751SV *
3752Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3753{
3754 STRLEN cur = SvCUR(sstr);
3755 STRLEN len = SvLEN(sstr);
3756 register char *new_pv;
3757
3758 if (DEBUG_C_TEST) {
3759 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
6c9570dc 3760 (void*)sstr, (void*)dstr);
ed252734
NC
3761 sv_dump(sstr);
3762 if (dstr)
3763 sv_dump(dstr);
3764 }
3765
3766 if (dstr) {
3767 if (SvTHINKFIRST(dstr))
3768 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3769 else if (SvPVX_const(dstr))
3770 Safefree(SvPVX_const(dstr));
ed252734
NC
3771 }
3772 else
3773 new_SV(dstr);
862a34c6 3774 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3775
3776 assert (SvPOK(sstr));
3777 assert (SvPOKp(sstr));
3778 assert (!SvIOK(sstr));
3779 assert (!SvIOKp(sstr));
3780 assert (!SvNOK(sstr));
3781 assert (!SvNOKp(sstr));
3782
3783 if (SvIsCOW(sstr)) {
3784
3785 if (SvLEN(sstr) == 0) {
3786 /* source is a COW shared hash key. */
ed252734
NC
3787 DEBUG_C(PerlIO_printf(Perl_debug_log,
3788 "Fast copy on write: Sharing hash\n"));
d1db91c6 3789 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3790 goto common_exit;
3791 }
3792 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3793 } else {
3794 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3795 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3796 SvREADONLY_on(sstr);
3797 SvFAKE_on(sstr);
3798 DEBUG_C(PerlIO_printf(Perl_debug_log,
3799 "Fast copy on write: Converting sstr to COW\n"));
3800 SV_COW_NEXT_SV_SET(dstr, sstr);
3801 }
3802 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3803 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3804
3805 common_exit:
3806 SvPV_set(dstr, new_pv);
3807 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3808 if (SvUTF8(sstr))
3809 SvUTF8_on(dstr);
87a1ef3d
SP
3810 SvLEN_set(dstr, len);
3811 SvCUR_set(dstr, cur);
ed252734
NC
3812 if (DEBUG_C_TEST) {
3813 sv_dump(dstr);
3814 }
3815 return dstr;
3816}
3817#endif
3818
954c1994
GS
3819/*
3820=for apidoc sv_setpvn
3821
3822Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3823bytes to be copied. If the C<ptr> argument is NULL the SV will become
3824undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3825
3826=cut
3827*/
3828
ef50df4b 3829void
864dbfa3 3830Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3831{
97aff369 3832 dVAR;
c6f8c383 3833 register char *dptr;
22c522df 3834
765f542d 3835 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3836 if (!ptr) {
a0d0e21e 3837 (void)SvOK_off(sv);
463ee0b2
LW
3838 return;
3839 }
22c522df
JH
3840 else {
3841 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3842 const IV iv = len;
9c5ffd7c
JH
3843 if (iv < 0)
3844 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3845 }
862a34c6 3846 SvUPGRADE(sv, SVt_PV);
c6f8c383 3847
5902b6a9 3848 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3849 Move(ptr,dptr,len,char);
3850 dptr[len] = '\0';
79072805 3851 SvCUR_set(sv, len);
1aa99e6b 3852 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3853 SvTAINT(sv);
79072805
LW
3854}
3855
954c1994
GS
3856/*
3857=for apidoc sv_setpvn_mg
3858
3859Like C<sv_setpvn>, but also handles 'set' magic.
3860
3861=cut
3862*/
3863
79072805 3864void
864dbfa3 3865Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3866{
3867 sv_setpvn(sv,ptr,len);
3868 SvSETMAGIC(sv);
3869}
3870
954c1994
GS
3871/*
3872=for apidoc sv_setpv
3873
3874Copies a string into an SV. The string must be null-terminated. Does not
3875handle 'set' magic. See C<sv_setpv_mg>.
3876
3877=cut
3878*/
3879
ef50df4b 3880void
864dbfa3 3881Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3882{
97aff369 3883 dVAR;
79072805
LW
3884 register STRLEN len;
3885
765f542d 3886 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3887 if (!ptr) {
a0d0e21e 3888 (void)SvOK_off(sv);
463ee0b2
LW
3889 return;
3890 }
79072805 3891 len = strlen(ptr);
862a34c6 3892 SvUPGRADE(sv, SVt_PV);
c6f8c383 3893
79072805 3894 SvGROW(sv, len + 1);
463ee0b2 3895 Move(ptr,SvPVX(sv),len+1,char);
79072805 3896 SvCUR_set(sv, len);
1aa99e6b 3897 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3898 SvTAINT(sv);
3899}
3900
954c1994
GS
3901/*
3902=for apidoc sv_setpv_mg
3903
3904Like C<sv_setpv>, but also handles 'set' magic.
3905
3906=cut
3907*/
3908
463ee0b2 3909void
864dbfa3 3910Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3911{
3912 sv_setpv(sv,ptr);
3913 SvSETMAGIC(sv);
3914}
3915
954c1994 3916/*
47518d95 3917=for apidoc sv_usepvn_flags
954c1994 3918
794a0d33
JH
3919Tells an SV to use C<ptr> to find its string value. Normally the
3920string is stored inside the SV but sv_usepvn allows the SV to use an
3921outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
3922by C<malloc>. The string length, C<len>, must be supplied. By default
3923this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
3924so that pointer should not be freed or used by the programmer after
3925giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
3926that pointer (e.g. ptr + 1) be used.
3927
3928If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3929SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 3930will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 3931C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3932
3933=cut
3934*/
3935
ef50df4b 3936void
47518d95 3937Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3938{
97aff369 3939 dVAR;
1936d2a7 3940 STRLEN allocate;
765f542d 3941 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3942 SvUPGRADE(sv, SVt_PV);
463ee0b2 3943 if (!ptr) {
a0d0e21e 3944 (void)SvOK_off(sv);
47518d95
NC
3945 if (flags & SV_SMAGIC)
3946 SvSETMAGIC(sv);
463ee0b2
LW
3947 return;
3948 }
3f7c398e 3949 if (SvPVX_const(sv))
8bd4d4c5 3950 SvPV_free(sv);
1936d2a7 3951
0b7042f9 3952#ifdef DEBUGGING
2e90b4cd
NC
3953 if (flags & SV_HAS_TRAILING_NUL)
3954 assert(ptr[len] == '\0');
0b7042f9 3955#endif
2e90b4cd 3956
c1c21316 3957 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3958 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
cbf82dd0
NC
3959 if (flags & SV_HAS_TRAILING_NUL) {
3960 /* It's long enough - do nothing.
3961 Specfically Perl_newCONSTSUB is relying on this. */
3962 } else {
69d25b4f 3963#ifdef DEBUGGING
69d25b4f 3964 /* Force a move to shake out bugs in callers. */
10edeb5d 3965 char *new_ptr = (char*)safemalloc(allocate);
69d25b4f
NC
3966 Copy(ptr, new_ptr, len, char);
3967 PoisonFree(ptr,len,char);
3968 Safefree(ptr);
3969 ptr = new_ptr;
69d25b4f 3970#else
10edeb5d 3971 ptr = (char*) saferealloc (ptr, allocate);
69d25b4f 3972#endif
cbf82dd0 3973 }
f880fe2f 3974 SvPV_set(sv, ptr);
463ee0b2 3975 SvCUR_set(sv, len);
1936d2a7 3976 SvLEN_set(sv, allocate);
c1c21316
NC
3977 if (!(flags & SV_HAS_TRAILING_NUL)) {
3978 *SvEND(sv) = '\0';
3979 }
1aa99e6b 3980 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3981 SvTAINT(sv);
47518d95
NC
3982 if (flags & SV_SMAGIC)
3983 SvSETMAGIC(sv);
ef50df4b
GS
3984}
3985
f8c7b90f 3986#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3987/* Need to do this *after* making the SV normal, as we need the buffer
3988 pointer to remain valid until after we've copied it. If we let go too early,
3989 another thread could invalidate it by unsharing last of the same hash key
3990 (which it can do by means other than releasing copy-on-write Svs)
3991 or by changing the other copy-on-write SVs in the loop. */
3992STATIC void
5302ffd4 3993S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
765f542d 3994{
5302ffd4 3995 { /* this SV was SvIsCOW_normal(sv) */
765f542d 3996 /* we need to find the SV pointing to us. */
cf5629ad 3997 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3998
765f542d
NC
3999 if (current == sv) {
4000 /* The SV we point to points back to us (there were only two of us
4001 in the loop.)
4002 Hence other SV is no longer copy on write either. */
4003 SvFAKE_off(after);
4004 SvREADONLY_off(after);
4005 } else {
4006 /* We need to follow the pointers around the loop. */
4007 SV *next;
4008 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4009 assert (next);
4010 current = next;
4011 /* don't loop forever if the structure is bust, and we have
4012 a pointer into a closed loop. */
4013 assert (current != after);
3f7c398e 4014 assert (SvPVX_const(current) == pvx);
765f542d
NC
4015 }
4016 /* Make the SV before us point to the SV after us. */
a29f6d03 4017 SV_COW_NEXT_SV_SET(current, after);
765f542d 4018 }
765f542d
NC
4019 }
4020}
765f542d 4021#endif
645c22ef
DM
4022/*
4023=for apidoc sv_force_normal_flags
4024
4025Undo various types of fakery on an SV: if the PV is a shared string, make
4026a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4027an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4028we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4029then a copy-on-write scalar drops its PV buffer (if any) and becomes
4030SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4031set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4032C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4033with flags set to 0.
645c22ef
DM
4034
4035=cut
4036*/
4037
6fc92669 4038void
840a7b70 4039Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4040{
97aff369 4041 dVAR;
f8c7b90f 4042#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4043 if (SvREADONLY(sv)) {
4044 /* At this point I believe I should acquire a global SV mutex. */
4045 if (SvFAKE(sv)) {
b64e5050 4046 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4047 const STRLEN len = SvLEN(sv);
4048 const STRLEN cur = SvCUR(sv);
5302ffd4
NC
4049 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4050 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4051 we'll fail an assertion. */
4052 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4053
46187eeb
NC
4054 if (DEBUG_C_TEST) {
4055 PerlIO_printf(Perl_debug_log,
4056 "Copy on write: Force normal %ld\n",
4057 (long) flags);
e419cbc5 4058 sv_dump(sv);
46187eeb 4059 }
765f542d
NC
4060 SvFAKE_off(sv);
4061 SvREADONLY_off(sv);
9f653bb5 4062 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4063 SvPV_set(sv, NULL);
87a1ef3d 4064 SvLEN_set(sv, 0);
765f542d
NC
4065 if (flags & SV_COW_DROP_PV) {
4066 /* OK, so we don't need to copy our buffer. */
4067 SvPOK_off(sv);
4068 } else {
4069 SvGROW(sv, cur + 1);
4070 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4071 SvCUR_set(sv, cur);
765f542d
NC
4072 *SvEND(sv) = '\0';
4073 }
5302ffd4
NC
4074 if (len) {
4075 sv_release_COW(sv, pvx, next);
4076 } else {
4077 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4078 }
46187eeb 4079 if (DEBUG_C_TEST) {
e419cbc5 4080 sv_dump(sv);
46187eeb 4081 }
765f542d 4082 }
923e4eb5 4083 else if (IN_PERL_RUNTIME)
765f542d
NC
4084 Perl_croak(aTHX_ PL_no_modify);
4085 /* At this point I believe that I can drop the global SV mutex. */
4086 }
4087#else
2213622d 4088 if (SvREADONLY(sv)) {
1c846c1f 4089 if (SvFAKE(sv)) {
b64e5050 4090 const char * const pvx = SvPVX_const(sv);
66a1b24b 4091 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4092 SvFAKE_off(sv);
4093 SvREADONLY_off(sv);
bd61b366 4094 SvPV_set(sv, NULL);
66a1b24b 4095 SvLEN_set(sv, 0);
1c846c1f 4096 SvGROW(sv, len + 1);
706aa1c9 4097 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4098 *SvEND(sv) = '\0';
bdd68bc3 4099 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4100 }
923e4eb5 4101 else if (IN_PERL_RUNTIME)
cea2e8a9 4102 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4103 }
765f542d 4104#endif
2213622d 4105 if (SvROK(sv))
840a7b70 4106 sv_unref_flags(sv, flags);
6fc92669
GS
4107 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4108 sv_unglob(sv);
0f15f207 4109}
1c846c1f 4110
645c22ef 4111/*
954c1994
GS
4112=for apidoc sv_chop
4113
1c846c1f 4114Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4115SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4116the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4117string. Uses the "OOK hack".
3f7c398e 4118Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4119refer to the same chunk of data.
954c1994
GS
4120
4121=cut
4122*/
4123
79072805 4124void
f54cb97a 4125Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4126{
4127 register STRLEN delta;
a0d0e21e 4128 if (!ptr || !SvPOKp(sv))
79072805 4129 return;
3f7c398e 4130 delta = ptr - SvPVX_const(sv);
2213622d 4131 SV_CHECK_THINKFIRST(sv);
79072805
LW
4132 if (SvTYPE(sv) < SVt_PVIV)
4133 sv_upgrade(sv,SVt_PVIV);
4134
4135 if (!SvOOK(sv)) {
50483b2c 4136 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4137 const char *pvx = SvPVX_const(sv);
a28509cc 4138 const STRLEN len = SvCUR(sv);
50483b2c 4139 SvGROW(sv, len + 1);
706aa1c9 4140 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4141 *SvEND(sv) = '\0';
4142 }
45977657 4143 SvIV_set(sv, 0);
a4bfb290
AB
4144 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4145 and we do that anyway inside the SvNIOK_off
4146 */
7a5fa8a2 4147 SvFLAGS(sv) |= SVf_OOK;
79072805 4148 }
a4bfb290 4149 SvNIOK_off(sv);
b162af07
SP
4150 SvLEN_set(sv, SvLEN(sv) - delta);
4151 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4152 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4153 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4154}
4155
954c1994
GS
4156/*
4157=for apidoc sv_catpvn
4158
4159Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4160C<len> indicates number of bytes to copy. If the SV has the UTF-8
4161status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4162Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4163
8d6d96c1
HS
4164=for apidoc sv_catpvn_flags
4165
4166Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4167C<len> indicates number of bytes to copy. If the SV has the UTF-8
4168status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4169If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4170appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4171in terms of this function.
4172
4173=cut
4174*/
4175
4176void
4177Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4178{
97aff369 4179 dVAR;
8d6d96c1 4180 STRLEN dlen;
fabdb6c0 4181 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4182
8d6d96c1
HS
4183 SvGROW(dsv, dlen + slen + 1);
4184 if (sstr == dstr)
3f7c398e 4185 sstr = SvPVX_const(dsv);
8d6d96c1 4186 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4187 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4188 *SvEND(dsv) = '\0';
4189 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4190 SvTAINT(dsv);
bddd5118
NC
4191 if (flags & SV_SMAGIC)
4192 SvSETMAGIC(dsv);
79072805
LW
4193}
4194
954c1994 4195/*
954c1994
GS
4196=for apidoc sv_catsv
4197
13e8c8e3
JH
4198Concatenates the string from SV C<ssv> onto the end of the string in
4199SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4200not 'set' magic. See C<sv_catsv_mg>.
954c1994 4201
8d6d96c1
HS
4202=for apidoc sv_catsv_flags
4203
4204Concatenates the string from SV C<ssv> onto the end of the string in
4205SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4206bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4207and C<sv_catsv_nomg> are implemented in terms of this function.
4208
4209=cut */
4210
ef50df4b 4211void
8d6d96c1 4212Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4213{
97aff369 4214 dVAR;
bddd5118 4215 if (ssv) {
00b6aa41
AL
4216 STRLEN slen;
4217 const char *spv = SvPV_const(ssv, slen);
4218 if (spv) {
bddd5118
NC
4219 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4220 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4221 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4222 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4223 dsv->sv_flags doesn't have that bit set.
4fd84b44 4224 Andy Dougherty 12 Oct 2001
bddd5118
NC
4225 */
4226 const I32 sutf8 = DO_UTF8(ssv);
4227 I32 dutf8;
13e8c8e3 4228
bddd5118
NC
4229 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4230 mg_get(dsv);
4231 dutf8 = DO_UTF8(dsv);
8d6d96c1 4232
bddd5118
NC
4233 if (dutf8 != sutf8) {
4234 if (dutf8) {
4235 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4236 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4237
bddd5118
NC
4238 sv_utf8_upgrade(csv);
4239 spv = SvPV_const(csv, slen);
4240 }
4241 else
4242 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4243 }
bddd5118 4244 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4245 }
560a288e 4246 }
bddd5118
NC
4247 if (flags & SV_SMAGIC)
4248 SvSETMAGIC(dsv);
79072805
LW
4249}
4250
954c1994 4251/*
954c1994
GS
4252=for apidoc sv_catpv
4253
4254Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4255If the SV has the UTF-8 status set, then the bytes appended should be
4256valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4257
d5ce4a7c 4258=cut */
954c1994 4259
ef50df4b 4260void
0c981600 4261Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4262{
97aff369 4263 dVAR;
79072805 4264 register STRLEN len;
463ee0b2 4265 STRLEN tlen;
748a9306 4266 char *junk;
79072805 4267
0c981600 4268 if (!ptr)
79072805 4269 return;
748a9306 4270 junk = SvPV_force(sv, tlen);
0c981600 4271 len = strlen(ptr);
463ee0b2 4272 SvGROW(sv, tlen + len + 1);
0c981600 4273 if (ptr == junk)
3f7c398e 4274 ptr = SvPVX_const(sv);
0c981600 4275 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4276 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4277 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4278 SvTAINT(sv);
79072805
LW
4279}
4280
954c1994
GS
4281/*
4282=for apidoc sv_catpv_mg
4283
4284Like C<sv_catpv>, but also handles 'set' magic.
4285
4286=cut
4287*/
4288
ef50df4b 4289void
0c981600 4290Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4291{
0c981600 4292 sv_catpv(sv,ptr);
ef50df4b
GS
4293 SvSETMAGIC(sv);
4294}
4295
645c22ef
DM
4296/*
4297=for apidoc newSV
4298
561b68a9
SH
4299Creates a new SV. A non-zero C<len> parameter indicates the number of
4300bytes of preallocated string space the SV should have. An extra byte for a
4301trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4302space is allocated.) The reference count for the new SV is set to 1.
4303
4304In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4305parameter, I<x>, a debug aid which allowed callers to identify themselves.
4306This aid has been superseded by a new build option, PERL_MEM_LOG (see
4307L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4308modules supporting older perls.
645c22ef
DM
4309
4310=cut
4311*/
4312
79072805 4313SV *
864dbfa3 4314Perl_newSV(pTHX_ STRLEN len)
79072805 4315{
97aff369 4316 dVAR;
79072805 4317 register SV *sv;
1c846c1f 4318
4561caa4 4319 new_SV(sv);
79072805
LW
4320 if (len) {
4321 sv_upgrade(sv, SVt_PV);
4322 SvGROW(sv, len + 1);
4323 }
4324 return sv;
4325}
954c1994 4326/*
92110913 4327=for apidoc sv_magicext
954c1994 4328
68795e93 4329Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4330supplied vtable and returns a pointer to the magic added.
92110913 4331
2d8d5d5a
SH
4332Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4333In particular, you can add magic to SvREADONLY SVs, and add more than
4334one instance of the same 'how'.
645c22ef 4335
2d8d5d5a
SH
4336If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4337stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4338special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4339to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4340
2d8d5d5a 4341(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4342
4343=cut
4344*/
92110913 4345MAGIC *
53d44271 4346Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, const MGVTBL *vtable,
92110913 4347 const char* name, I32 namlen)
79072805 4348{
97aff369 4349 dVAR;
79072805 4350 MAGIC* mg;
68795e93 4351
92110913 4352 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4353 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4354 }
a02a5408 4355 Newxz(mg, 1, MAGIC);
79072805 4356 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4357 SvMAGIC_set(sv, mg);
75f9d97a 4358
05f95b08
SB
4359 /* Sometimes a magic contains a reference loop, where the sv and
4360 object refer to each other. To prevent a reference loop that
4361 would prevent such objects being freed, we look for such loops
4362 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4363
4364 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4365 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4366
4367 */
14befaf4
DM
4368 if (!obj || obj == sv ||
4369 how == PERL_MAGIC_arylen ||
4370 how == PERL_MAGIC_qr ||
8d2f4536 4371 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4372 (SvTYPE(obj) == SVt_PVGV &&
4373 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4374 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4375 GvFORM(obj) == (CV*)sv)))
75f9d97a 4376 {
8990e307 4377 mg->mg_obj = obj;
75f9d97a 4378 }
85e6fe83 4379 else {
b37c2d43 4380 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4381 mg->mg_flags |= MGf_REFCOUNTED;
4382 }
b5ccf5f2
YST
4383
4384 /* Normal self-ties simply pass a null object, and instead of
4385 using mg_obj directly, use the SvTIED_obj macro to produce a
4386 new RV as needed. For glob "self-ties", we are tieing the PVIO
4387 with an RV obj pointing to the glob containing the PVIO. In
4388 this case, to avoid a reference loop, we need to weaken the
4389 reference.
4390 */
4391
4392 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4393 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4394 {
4395 sv_rvweaken(obj);
4396 }
4397
79072805 4398 mg->mg_type = how;
565764a8 4399 mg->mg_len = namlen;
9cbac4c7 4400 if (name) {
92110913 4401 if (namlen > 0)
1edc1566 4402 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4403 else if (namlen == HEf_SVKEY)
b37c2d43 4404 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4405 else
92110913 4406 mg->mg_ptr = (char *) name;
9cbac4c7 4407 }
53d44271 4408 mg->mg_virtual = (MGVTBL *) vtable;
68795e93 4409
92110913
NIS
4410 mg_magical(sv);
4411 if (SvGMAGICAL(sv))
4412 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4413 return mg;
4414}
4415
4416/*
4417=for apidoc sv_magic
1c846c1f 4418
92110913
NIS
4419Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4420then adds a new magic item of type C<how> to the head of the magic list.
4421
2d8d5d5a
SH
4422See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4423handling of the C<name> and C<namlen> arguments.
4424
4509d3fb
SB
4425You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4426to add more than one instance of the same 'how'.
4427
92110913
NIS
4428=cut
4429*/
4430
4431void
4432Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4433{
97aff369 4434 dVAR;
53d44271 4435 const MGVTBL *vtable;
92110913 4436 MAGIC* mg;
92110913 4437
f8c7b90f 4438#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4439 if (SvIsCOW(sv))
4440 sv_force_normal_flags(sv, 0);
4441#endif
92110913 4442 if (SvREADONLY(sv)) {
d8084ca5
DM
4443 if (
4444 /* its okay to attach magic to shared strings; the subsequent
4445 * upgrade to PVMG will unshare the string */
4446 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4447
4448 && IN_PERL_RUNTIME
92110913
NIS
4449 && how != PERL_MAGIC_regex_global
4450 && how != PERL_MAGIC_bm
4451 && how != PERL_MAGIC_fm
4452 && how != PERL_MAGIC_sv
e6469971 4453 && how != PERL_MAGIC_backref
92110913
NIS
4454 )
4455 {
4456 Perl_croak(aTHX_ PL_no_modify);
4457 }
4458 }
4459 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4460 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4461 /* sv_magic() refuses to add a magic of the same 'how' as an
4462 existing one
92110913 4463 */
2a509ed3 4464 if (how == PERL_MAGIC_taint) {
92110913 4465 mg->mg_len |= 1;
2a509ed3
NC
4466 /* Any scalar which already had taint magic on which someone
4467 (erroneously?) did SvIOK_on() or similar will now be
4468 incorrectly sporting public "OK" flags. */
4469 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4470 }
92110913
NIS
4471 return;
4472 }
4473 }
68795e93 4474
79072805 4475 switch (how) {
14befaf4 4476 case PERL_MAGIC_sv:
92110913 4477 vtable = &PL_vtbl_sv;
79072805 4478 break;
14befaf4 4479 case PERL_MAGIC_overload:
92110913 4480 vtable = &PL_vtbl_amagic;
a0d0e21e 4481 break;
14befaf4 4482 case PERL_MAGIC_overload_elem:
92110913 4483 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4484 break;
14befaf4 4485 case PERL_MAGIC_overload_table:
92110913 4486 vtable = &PL_vtbl_ovrld;
a0d0e21e 4487 break;
14befaf4 4488 case PERL_MAGIC_bm:
92110913 4489 vtable = &PL_vtbl_bm;
79072805 4490 break;
14befaf4 4491 case PERL_MAGIC_regdata:
92110913 4492 vtable = &PL_vtbl_regdata;
6cef1e77 4493 break;
14befaf4 4494 case PERL_MAGIC_regdatum:
92110913 4495 vtable = &PL_vtbl_regdatum;
6cef1e77 4496 break;
14befaf4 4497 case PERL_MAGIC_env:
92110913 4498 vtable = &PL_vtbl_env;
79072805 4499 break;
14befaf4 4500 case PERL_MAGIC_fm:
92110913 4501 vtable = &PL_vtbl_fm;
55497cff 4502 break;
14befaf4 4503 case PERL_MAGIC_envelem:
92110913 4504 vtable = &PL_vtbl_envelem;
79072805 4505 break;
14befaf4 4506 case PERL_MAGIC_regex_global:
92110913 4507 vtable = &PL_vtbl_mglob;
93a17b20 4508 break;
14befaf4 4509 case PERL_MAGIC_isa:
92110913 4510 vtable = &PL_vtbl_isa;
463ee0b2 4511 break;
14befaf4 4512 case PERL_MAGIC_isaelem:
92110913 4513 vtable = &PL_vtbl_isaelem;
463ee0b2 4514 break;
14befaf4 4515 case PERL_MAGIC_nkeys:
92110913 4516 vtable = &PL_vtbl_nkeys;
16660edb 4517 break;
14befaf4 4518 case PERL_MAGIC_dbfile:
aec46f14 4519 vtable = NULL;
93a17b20 4520 break;
14befaf4 4521 case PERL_MAGIC_dbline:
92110913 4522 vtable = &PL_vtbl_dbline;
79072805 4523 break;
36477c24 4524#ifdef USE_LOCALE_COLLATE
14befaf4 4525 case PERL_MAGIC_collxfrm:
92110913 4526 vtable = &PL_vtbl_collxfrm;
bbce6d69 4527 break;
36477c24 4528#endif /* USE_LOCALE_COLLATE */
14befaf4 4529 case PERL_MAGIC_tied:
92110913 4530 vtable = &PL_vtbl_pack;
463ee0b2 4531 break;
14befaf4
DM
4532 case PERL_MAGIC_tiedelem:
4533 case PERL_MAGIC_tiedscalar:
92110913 4534 vtable = &PL_vtbl_packelem;
463ee0b2 4535 break;
14befaf4 4536 case PERL_MAGIC_qr:
92110913 4537 vtable = &PL_vtbl_regexp;
c277df42 4538 break;
b3ca2e83
NC
4539 case PERL_MAGIC_hints:
4540 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4541 case PERL_MAGIC_sig:
92110913 4542 vtable = &PL_vtbl_sig;
79072805 4543 break;
14befaf4 4544 case PERL_MAGIC_sigelem:
92110913 4545 vtable = &PL_vtbl_sigelem;
79072805 4546 break;
14befaf4 4547 case PERL_MAGIC_taint:
92110913 4548 vtable = &PL_vtbl_taint;
463ee0b2 4549 break;
14befaf4 4550 case PERL_MAGIC_uvar:
92110913 4551 vtable = &PL_vtbl_uvar;
79072805 4552 break;
14befaf4 4553 case PERL_MAGIC_vec:
92110913 4554 vtable = &PL_vtbl_vec;
79072805 4555 break;
a3874608 4556 case PERL_MAGIC_arylen_p:
bfcb3514 4557 case PERL_MAGIC_rhash:
8d2f4536 4558 case PERL_MAGIC_symtab:
ece467f9 4559 case PERL_MAGIC_vstring:
aec46f14 4560 vtable = NULL;
ece467f9 4561 break;
7e8c5dac
HS
4562 case PERL_MAGIC_utf8:
4563 vtable = &PL_vtbl_utf8;
4564 break;
14befaf4 4565 case PERL_MAGIC_substr:
92110913 4566 vtable = &PL_vtbl_substr;
79072805 4567 break;
14befaf4 4568 case PERL_MAGIC_defelem:
92110913 4569 vtable = &PL_vtbl_defelem;
5f05dabc 4570 break;
14befaf4 4571 case PERL_MAGIC_arylen:
92110913 4572 vtable = &PL_vtbl_arylen;
79072805 4573 break;
14befaf4 4574 case PERL_MAGIC_pos:
92110913 4575 vtable = &PL_vtbl_pos;
a0d0e21e 4576 break;
14befaf4 4577 case PERL_MAGIC_backref:
92110913 4578 vtable = &PL_vtbl_backref;
810b8aa5 4579 break;
b3ca2e83
NC
4580 case PERL_MAGIC_hintselem:
4581 vtable = &PL_vtbl_hintselem;
4582 break;
14befaf4
DM
4583 case PERL_MAGIC_ext:
4584 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4585 /* Useful for attaching extension internal data to perl vars. */
4586 /* Note that multiple extensions may clash if magical scalars */
4587 /* etc holding private data from one are passed to another. */
aec46f14 4588 vtable = NULL;
a0d0e21e 4589 break;
79072805 4590 default:
14befaf4 4591 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4592 }
68795e93 4593
92110913 4594 /* Rest of work is done else where */
aec46f14 4595 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4596
92110913
NIS
4597 switch (how) {
4598 case PERL_MAGIC_taint:
4599 mg->mg_len = 1;
4600 break;
4601 case PERL_MAGIC_ext:
4602 case PERL_MAGIC_dbfile:
4603 SvRMAGICAL_on(sv);
4604 break;
4605 }
463ee0b2
LW
4606}
4607
c461cf8f
JH
4608/*
4609=for apidoc sv_unmagic
4610
645c22ef 4611Removes all magic of type C<type> from an SV.
c461cf8f
JH
4612
4613=cut
4614*/
4615
463ee0b2 4616int
864dbfa3 4617Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4618{
4619 MAGIC* mg;
4620 MAGIC** mgp;
91bba347 4621 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4622 return 0;
064cf529 4623 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4624 for (mg = *mgp; mg; mg = *mgp) {
4625 if (mg->mg_type == type) {
e1ec3a88 4626 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4627 *mgp = mg->mg_moremagic;
1d7c1841 4628 if (vtbl && vtbl->svt_free)
fc0dc3b3 4629 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4630 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4631 if (mg->mg_len > 0)
1edc1566 4632 Safefree(mg->mg_ptr);
565764a8 4633 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4634 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4635 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4636 Safefree(mg->mg_ptr);
9cbac4c7 4637 }
a0d0e21e
LW
4638 if (mg->mg_flags & MGf_REFCOUNTED)
4639 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4640 Safefree(mg);
4641 }
4642 else
4643 mgp = &mg->mg_moremagic;
79072805 4644 }
91bba347 4645 if (!SvMAGIC(sv)) {
463ee0b2 4646 SvMAGICAL_off(sv);
c268c2a6 4647 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4648 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4649 }
4650
4651 return 0;
79072805
LW
4652}
4653
c461cf8f
JH
4654/*
4655=for apidoc sv_rvweaken
4656
645c22ef
DM
4657Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4658referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4659push a back-reference to this RV onto the array of backreferences
1e73acc8
AS
4660associated with that magic. If the RV is magical, set magic will be
4661called after the RV is cleared.
c461cf8f
JH
4662
4663=cut
4664*/
4665
810b8aa5 4666SV *
864dbfa3 4667Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4668{
4669 SV *tsv;
4670 if (!SvOK(sv)) /* let undefs pass */
4671 return sv;
4672 if (!SvROK(sv))
cea2e8a9 4673 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4674 else if (SvWEAKREF(sv)) {
810b8aa5 4675 if (ckWARN(WARN_MISC))
9014280d 4676 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4677 return sv;
4678 }
4679 tsv = SvRV(sv);
e15faf7d 4680 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4681 SvWEAKREF_on(sv);
1c846c1f 4682 SvREFCNT_dec(tsv);
810b8aa5
GS
4683 return sv;
4684}
4685
645c22ef
DM
4686/* Give tsv backref magic if it hasn't already got it, then push a
4687 * back-reference to sv onto the array associated with the backref magic.
4688 */
4689
e15faf7d
NC
4690void
4691Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4692{
97aff369 4693 dVAR;
810b8aa5 4694 AV *av;
86f55936
NC
4695
4696 if (SvTYPE(tsv) == SVt_PVHV) {
4697 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4698
4699 av = *avp;
4700 if (!av) {
4701 /* There is no AV in the offical place - try a fixup. */
4702 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4703
4704 if (mg) {
4705 /* Aha. They've got it stowed in magic. Bring it back. */
4706 av = (AV*)mg->mg_obj;
4707 /* Stop mg_free decreasing the refernce count. */
4708 mg->mg_obj = NULL;
4709 /* Stop mg_free even calling the destructor, given that
4710 there's no AV to free up. */
4711 mg->mg_virtual = 0;
4712 sv_unmagic(tsv, PERL_MAGIC_backref);
4713 } else {
4714 av = newAV();
4715 AvREAL_off(av);
b37c2d43 4716 SvREFCNT_inc_simple_void(av);
86f55936
NC
4717 }
4718 *avp = av;
4719 }
4720 } else {
4721 const MAGIC *const mg
4722 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4723 if (mg)
4724 av = (AV*)mg->mg_obj;
4725 else {
4726 av = newAV();
4727 AvREAL_off(av);
4728 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4729 /* av now has a refcnt of 2, which avoids it getting freed
4730 * before us during global cleanup. The extra ref is removed
4731 * by magic_killbackrefs() when tsv is being freed */
4732 }
810b8aa5 4733 }
d91d49e8 4734 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4735 av_extend(av, AvFILLp(av)+1);
4736 }
4737 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4738}
4739
645c22ef
DM
4740/* delete a back-reference to ourselves from the backref magic associated
4741 * with the SV we point to.
4742 */
4743
1c846c1f 4744STATIC void
e15faf7d 4745S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4746{
97aff369 4747 dVAR;
86f55936 4748 AV *av = NULL;
810b8aa5
GS
4749 SV **svp;
4750 I32 i;
86f55936
NC
4751
4752 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4753 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4754 /* We mustn't attempt to "fix up" the hash here by moving the
4755 backreference array back to the hv_aux structure, as that is stored
4756 in the main HvARRAY(), and hfreentries assumes that no-one
4757 reallocates HvARRAY() while it is running. */
86f55936
NC
4758 }
4759 if (!av) {
4760 const MAGIC *const mg
4761 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4762 if (mg)
4763 av = (AV *)mg->mg_obj;
4764 }
4765 if (!av) {
e15faf7d
NC
4766 if (PL_in_clean_all)
4767 return;
cea2e8a9 4768 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4769 }
4770
4771 if (SvIS_FREED(av))
4772 return;
4773
810b8aa5 4774 svp = AvARRAY(av);
6a76db8b
NC
4775 /* We shouldn't be in here more than once, but for paranoia reasons lets
4776 not assume this. */
4777 for (i = AvFILLp(av); i >= 0; i--) {
4778 if (svp[i] == sv) {
4779 const SSize_t fill = AvFILLp(av);
4780 if (i != fill) {
4781 /* We weren't the last entry.
4782 An unordered list has this property that you can take the
4783 last element off the end to fill the hole, and it's still
4784 an unordered list :-)
4785 */
4786 svp[i] = svp[fill];
4787 }
a0714e2c 4788 svp[fill] = NULL;
6a76db8b
NC
4789 AvFILLp(av) = fill - 1;
4790 }
4791 }
810b8aa5
GS
4792}
4793
86f55936
NC
4794int
4795Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4796{
4797 SV **svp = AvARRAY(av);
4798
4799 PERL_UNUSED_ARG(sv);
4800
4801 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4802 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4803 if (svp && !SvIS_FREED(av)) {
4804 SV *const *const last = svp + AvFILLp(av);
4805
4806 while (svp <= last) {
4807 if (*svp) {
4808 SV *const referrer = *svp;
4809 if (SvWEAKREF(referrer)) {
4810 /* XXX Should we check that it hasn't changed? */
4811 SvRV_set(referrer, 0);
4812 SvOK_off(referrer);
4813 SvWEAKREF_off(referrer);
1e73acc8 4814 SvSETMAGIC(referrer);
86f55936
NC
4815 } else if (SvTYPE(referrer) == SVt_PVGV ||
4816 SvTYPE(referrer) == SVt_PVLV) {
4817 /* You lookin' at me? */
4818 assert(GvSTASH(referrer));
4819 assert(GvSTASH(referrer) == (HV*)sv);
4820 GvSTASH(referrer) = 0;
4821 } else {
4822 Perl_croak(aTHX_
4823 "panic: magic_killbackrefs (flags=%"UVxf")",
4824 (UV)SvFLAGS(referrer));
4825 }
4826
a0714e2c 4827 *svp = NULL;
86f55936
NC
4828 }
4829 svp++;
4830 }
4831 }
4832 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4833 return 0;
4834}
4835
954c1994
GS
4836/*
4837=for apidoc sv_insert
4838
4839Inserts a string at the specified offset/length within the SV. Similar to
4840the Perl substr() function.
4841
4842=cut
4843*/
4844
79072805 4845void
e1ec3a88 4846Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4847{
97aff369 4848 dVAR;
79072805
LW
4849 register char *big;
4850 register char *mid;
4851 register char *midend;
4852 register char *bigend;
4853 register I32 i;
6ff81951 4854 STRLEN curlen;
1c846c1f 4855
79072805 4856
8990e307 4857 if (!bigstr)
cea2e8a9 4858 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4859 SvPV_force(bigstr, curlen);
60fa28ff 4860 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4861 if (offset + len > curlen) {
4862 SvGROW(bigstr, offset+len+1);
93524f2b 4863 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4864 SvCUR_set(bigstr, offset+len);
4865 }
79072805 4866
69b47968 4867 SvTAINT(bigstr);
79072805
LW
4868 i = littlelen - len;
4869 if (i > 0) { /* string might grow */
a0d0e21e 4870 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4871 mid = big + offset + len;
4872 midend = bigend = big + SvCUR(bigstr);
4873 bigend += i;
4874 *bigend = '\0';
4875 while (midend > mid) /* shove everything down */
4876 *--bigend = *--midend;
4877 Move(little,big+offset,littlelen,char);
b162af07 4878 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4879 SvSETMAGIC(bigstr);
4880 return;
4881 }
4882 else if (i == 0) {
463ee0b2 4883 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4884 SvSETMAGIC(bigstr);
4885 return;
4886 }
4887
463ee0b2 4888 big = SvPVX(bigstr);
79072805
LW
4889 mid = big + offset;
4890 midend = mid + len;
4891 bigend = big + SvCUR(bigstr);
4892
4893 if (midend > bigend)
cea2e8a9 4894 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4895
4896 if (mid - big > bigend - midend) { /* faster to shorten from end */
4897 if (littlelen) {
4898 Move(little, mid, littlelen,char);
4899 mid += littlelen;
4900 }
4901 i = bigend - midend;
4902 if (i > 0) {
4903 Move(midend, mid, i,char);
4904 mid += i;
4905 }
4906 *mid = '\0';
4907 SvCUR_set(bigstr, mid - big);
4908 }
155aba94 4909 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4910 midend -= littlelen;
4911 mid = midend;
4912 sv_chop(bigstr,midend-i);
4913 big += i;
4914 while (i--)
4915 *--midend = *--big;
4916 if (littlelen)
4917 Move(little, mid, littlelen,char);
4918 }
4919 else if (littlelen) {
4920 midend -= littlelen;
4921 sv_chop(bigstr,midend);
4922 Move(little,midend,littlelen,char);
4923 }
4924 else {
4925 sv_chop(bigstr,midend);
4926 }
4927 SvSETMAGIC(bigstr);
4928}
4929
c461cf8f
JH
4930/*
4931=for apidoc sv_replace
4932
4933Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4934The target SV physically takes over ownership of the body of the source SV
4935and inherits its flags; however, the target keeps any magic it owns,
4936and any magic in the source is discarded.
ff276b08 4937Note that this is a rather specialist SV copying operation; most of the
645c22ef 4938time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4939
4940=cut
4941*/
79072805
LW
4942
4943void
864dbfa3 4944Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4945{
97aff369 4946 dVAR;
a3b680e6 4947 const U32 refcnt = SvREFCNT(sv);
765f542d 4948 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4949 if (SvREFCNT(nsv) != 1) {
7437becc 4950 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4951 UVuf " != 1)", (UV) SvREFCNT(nsv));
4952 }
93a17b20 4953 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4954 if (SvMAGICAL(nsv))
4955 mg_free(nsv);
4956 else
4957 sv_upgrade(nsv, SVt_PVMG);
b162af07 4958 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4959 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4960 SvMAGICAL_off(sv);
b162af07 4961 SvMAGIC_set(sv, NULL);
93a17b20 4962 }
79072805
LW
4963 SvREFCNT(sv) = 0;
4964 sv_clear(sv);
477f5d66 4965 assert(!SvREFCNT(sv));
fd0854ff
DM
4966#ifdef DEBUG_LEAKING_SCALARS
4967 sv->sv_flags = nsv->sv_flags;
4968 sv->sv_any = nsv->sv_any;
4969 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4970 sv->sv_u = nsv->sv_u;
fd0854ff 4971#else
79072805 4972 StructCopy(nsv,sv,SV);
fd0854ff 4973#endif
7b2c381c
NC
4974 /* Currently could join these into one piece of pointer arithmetic, but
4975 it would be unclear. */
4976 if(SvTYPE(sv) == SVt_IV)
4977 SvANY(sv)
339049b0 4978 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4979 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4980 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4981 }
4982
fd0854ff 4983
f8c7b90f 4984#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4985 if (SvIsCOW_normal(nsv)) {
4986 /* We need to follow the pointers around the loop to make the
4987 previous SV point to sv, rather than nsv. */
4988 SV *next;
4989 SV *current = nsv;
4990 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4991 assert(next);
4992 current = next;
3f7c398e 4993 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4994 }
4995 /* Make the SV before us point to the SV after us. */
4996 if (DEBUG_C_TEST) {
4997 PerlIO_printf(Perl_debug_log, "previous is\n");
4998 sv_dump(current);
a29f6d03
NC
4999 PerlIO_printf(Perl_debug_log,
5000 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
5001 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5002 }
a29f6d03 5003 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
5004 }
5005#endif
79072805 5006 SvREFCNT(sv) = refcnt;
1edc1566 5007 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 5008 SvREFCNT(nsv) = 0;
463ee0b2 5009 del_SV(nsv);
79072805
LW
5010}
5011
c461cf8f
JH
5012/*
5013=for apidoc sv_clear
5014
645c22ef
DM
5015Clear an SV: call any destructors, free up any memory used by the body,
5016and free the body itself. The SV's head is I<not> freed, although
5017its type is set to all 1's so that it won't inadvertently be assumed
5018to be live during global destruction etc.
5019This function should only be called when REFCNT is zero. Most of the time
5020you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5021instead.
c461cf8f
JH
5022
5023=cut
5024*/
5025
79072805 5026void
864dbfa3 5027Perl_sv_clear(pTHX_ register SV *sv)
79072805 5028{
27da23d5 5029 dVAR;
82bb6deb 5030 const U32 type = SvTYPE(sv);
8edfc514
NC
5031 const struct body_details *const sv_type_details
5032 = bodies_by_type + type;
82bb6deb 5033
79072805
LW
5034 assert(sv);
5035 assert(SvREFCNT(sv) == 0);
5036
d2a0f284
JC
5037 if (type <= SVt_IV) {
5038 /* See the comment in sv.h about the collusion between this early
5039 return and the overloading of the NULL and IV slots in the size
5040 table. */
82bb6deb 5041 return;
d2a0f284 5042 }
82bb6deb 5043
ed6116ce 5044 if (SvOBJECT(sv)) {
3280af22 5045 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5046 dSP;
893645bd 5047 HV* stash;
d460ef45 5048 do {
b464bac0 5049 CV* destructor;
4e8e7886 5050 stash = SvSTASH(sv);
32251b26 5051 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5052 if (destructor) {
1b6737cc 5053 SV* const tmpref = newRV(sv);
5cc433a6 5054 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5055 ENTER;
e788e7d3 5056 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5057 EXTEND(SP, 2);
5058 PUSHMARK(SP);
5cc433a6 5059 PUSHs(tmpref);
4e8e7886 5060 PUTBACK;
44389ee9 5061 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5062
5063
d3acc0f7 5064 POPSTACK;
3095d977 5065 SPAGAIN;
4e8e7886 5066 LEAVE;
5cc433a6
AB
5067 if(SvREFCNT(tmpref) < 2) {
5068 /* tmpref is not kept alive! */
5069 SvREFCNT(sv)--;
b162af07 5070 SvRV_set(tmpref, NULL);
5cc433a6
AB
5071 SvROK_off(tmpref);
5072 }
5073 SvREFCNT_dec(tmpref);
4e8e7886
GS
5074 }
5075 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5076
6f44e0a4
JP
5077
5078 if (SvREFCNT(sv)) {
5079 if (PL_in_clean_objs)
cea2e8a9 5080 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5081 HvNAME_get(stash));
6f44e0a4
JP
5082 /* DESTROY gave object new lease on life */
5083 return;
5084 }
a0d0e21e 5085 }
4e8e7886 5086
a0d0e21e 5087 if (SvOBJECT(sv)) {
4e8e7886 5088 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5089 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5090 if (type != SVt_PVIO)
3280af22 5091 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5092 }
463ee0b2 5093 }
82bb6deb 5094 if (type >= SVt_PVMG) {
cecf5685 5095 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
73d95100 5096 SvREFCNT_dec(SvOURSTASH(sv));
e736a858 5097 } else if (SvMAGIC(sv))
524189f1 5098 mg_free(sv);
00b1698f 5099 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5100 SvREFCNT_dec(SvSTASH(sv));
5101 }
82bb6deb 5102 switch (type) {
cecf5685 5103 /* case SVt_BIND: */
8990e307 5104 case SVt_PVIO:
df0bd2f4
GS
5105 if (IoIFP(sv) &&
5106 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5107 IoIFP(sv) != PerlIO_stdout() &&
5108 IoIFP(sv) != PerlIO_stderr())
93578b34 5109 {
f2b5be74 5110 io_close((IO*)sv, FALSE);
93578b34 5111 }
1d7c1841 5112 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5113 PerlDir_close(IoDIRP(sv));
1d7c1841 5114 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5115 Safefree(IoTOP_NAME(sv));
5116 Safefree(IoFMT_NAME(sv));
5117 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5118 goto freescalar;
79072805 5119 case SVt_PVCV:
748a9306 5120 case SVt_PVFM:
85e6fe83 5121 cv_undef((CV*)sv);
a0d0e21e 5122 goto freescalar;
79072805 5123 case SVt_PVHV:
86f55936 5124 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5125 hv_undef((HV*)sv);
a0d0e21e 5126 break;
79072805 5127 case SVt_PVAV:
85e6fe83 5128 av_undef((AV*)sv);
a0d0e21e 5129 break;
02270b4e 5130 case SVt_PVLV:
dd28f7bb
DM
5131 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5132 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5133 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5134 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5135 }
5136 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5137 SvREFCNT_dec(LvTARG(sv));
a0d0e21e 5138 case SVt_PVGV:
cecf5685
NC
5139 if (isGV_with_GP(sv)) {
5140 gp_free((GV*)sv);
5141 if (GvNAME_HEK(sv))
5142 unshare_hek(GvNAME_HEK(sv));
893645bd
NC
5143 /* If we're in a stash, we don't own a reference to it. However it does
5144 have a back reference to us, which needs to be cleared. */
cecf5685
NC
5145 if (!SvVALID(sv) && GvSTASH(sv))
5146 sv_del_backref((SV*)GvSTASH(sv), sv);
5147 }
79072805 5148 case SVt_PVMG:
79072805
LW
5149 case SVt_PVNV:
5150 case SVt_PVIV:
a0d0e21e 5151 freescalar:
5228ca4e
NC
5152 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5153 if (SvOOK(sv)) {
93524f2b 5154 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5155 /* Don't even bother with turning off the OOK flag. */
5156 }
79072805 5157 case SVt_PV:
a0d0e21e 5158 case SVt_RV:
810b8aa5 5159 if (SvROK(sv)) {
b37c2d43 5160 SV * const target = SvRV(sv);
810b8aa5 5161 if (SvWEAKREF(sv))
e15faf7d 5162 sv_del_backref(target, sv);
810b8aa5 5163 else
e15faf7d 5164 SvREFCNT_dec(target);
810b8aa5 5165 }
f8c7b90f 5166#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5167 else if (SvPVX_const(sv)) {
765f542d
NC
5168 if (SvIsCOW(sv)) {
5169 /* I believe I need to grab the global SV mutex here and
5170 then recheck the COW status. */
46187eeb
NC
5171 if (DEBUG_C_TEST) {
5172 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5173 sv_dump(sv);
46187eeb 5174 }
5302ffd4
NC
5175 if (SvLEN(sv)) {
5176 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5177 } else {
5178 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5179 }
5180
765f542d
NC
5181 /* And drop it here. */
5182 SvFAKE_off(sv);
5183 } else if (SvLEN(sv)) {
3f7c398e 5184 Safefree(SvPVX_const(sv));
765f542d
NC
5185 }
5186 }
5187#else
3f7c398e 5188 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5189 Safefree(SvPVX_mutable(sv));
3f7c398e 5190 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5191 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5192 SvFAKE_off(sv);
5193 }
765f542d 5194#endif
79072805
LW
5195 break;
5196 case SVt_NV:
79072805
LW
5197 break;
5198 }
5199
893645bd
NC
5200 SvFLAGS(sv) &= SVf_BREAK;
5201 SvFLAGS(sv) |= SVTYPEMASK;
5202
8edfc514 5203 if (sv_type_details->arena) {
b9502f15 5204 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5205 &PL_body_roots[type]);
5206 }
d2a0f284 5207 else if (sv_type_details->body_size) {
8edfc514
NC
5208 my_safefree(SvANY(sv));
5209 }
79072805
LW
5210}
5211
645c22ef
DM
5212/*
5213=for apidoc sv_newref
5214
5215Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5216instead.
5217
5218=cut
5219*/
5220
79072805 5221SV *
864dbfa3 5222Perl_sv_newref(pTHX_ SV *sv)
79072805 5223{
96a5add6 5224 PERL_UNUSED_CONTEXT;
463ee0b2 5225 if (sv)
4db098f4 5226 (SvREFCNT(sv))++;
79072805
LW
5227 return sv;
5228}
5229
c461cf8f
JH
5230/*
5231=for apidoc sv_free
5232
645c22ef
DM
5233Decrement an SV's reference count, and if it drops to zero, call
5234C<sv_clear> to invoke destructors and free up any memory used by
5235the body; finally, deallocate the SV's head itself.
5236Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5237
5238=cut
5239*/
5240
79072805 5241void
864dbfa3 5242Perl_sv_free(pTHX_ SV *sv)
79072805 5243{
27da23d5 5244 dVAR;
79072805
LW
5245 if (!sv)
5246 return;
a0d0e21e
LW
5247 if (SvREFCNT(sv) == 0) {
5248 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5249 /* this SV's refcnt has been artificially decremented to
5250 * trigger cleanup */
a0d0e21e 5251 return;
3280af22 5252 if (PL_in_clean_all) /* All is fair */
1edc1566 5253 return;
d689ffdd
JP
5254 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5255 /* make sure SvREFCNT(sv)==0 happens very seldom */
5256 SvREFCNT(sv) = (~(U32)0)/2;
5257 return;
5258 }
41e4abd8 5259 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5260 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5261 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5262 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5263#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5264 Perl_dump_sv_child(aTHX_ sv);
5265#endif
5266 }
79072805
LW
5267 return;
5268 }
4db098f4 5269 if (--(SvREFCNT(sv)) > 0)
8990e307 5270 return;
8c4d3c90
NC
5271 Perl_sv_free2(aTHX_ sv);
5272}
5273
5274void
5275Perl_sv_free2(pTHX_ SV *sv)
5276{
27da23d5 5277 dVAR;
463ee0b2
LW
5278#ifdef DEBUGGING
5279 if (SvTEMP(sv)) {
0453d815 5280 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5281 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5282 "Attempt to free temp prematurely: SV 0x%"UVxf
5283 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5284 return;
79072805 5285 }
463ee0b2 5286#endif
d689ffdd
JP
5287 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5288 /* make sure SvREFCNT(sv)==0 happens very seldom */
5289 SvREFCNT(sv) = (~(U32)0)/2;
5290 return;
5291 }
79072805 5292 sv_clear(sv);
477f5d66
CS
5293 if (! SvREFCNT(sv))
5294 del_SV(sv);
79072805
LW
5295}
5296
954c1994
GS
5297/*
5298=for apidoc sv_len
5299
645c22ef
DM
5300Returns the length of the string in the SV. Handles magic and type
5301coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5302
5303=cut
5304*/
5305
79072805 5306STRLEN
864dbfa3 5307Perl_sv_len(pTHX_ register SV *sv)
79072805 5308{
463ee0b2 5309 STRLEN len;
79072805
LW
5310
5311 if (!sv)
5312 return 0;
5313
8990e307 5314 if (SvGMAGICAL(sv))
565764a8 5315 len = mg_length(sv);
8990e307 5316 else
4d84ee25 5317 (void)SvPV_const(sv, len);
463ee0b2 5318 return len;
79072805
LW
5319}
5320
c461cf8f
JH
5321/*
5322=for apidoc sv_len_utf8
5323
5324Returns the number of characters in the string in an SV, counting wide
1e54db1a 5325UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5326
5327=cut
5328*/
5329
7e8c5dac
HS
5330/*
5331 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5332 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5333 * (Note that the mg_len is not the length of the mg_ptr field.
5334 * This allows the cache to store the character length of the string without
5335 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5336 *
7e8c5dac
HS
5337 */
5338
a0ed51b3 5339STRLEN
864dbfa3 5340Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5341{
a0ed51b3
LW
5342 if (!sv)
5343 return 0;
5344
a0ed51b3 5345 if (SvGMAGICAL(sv))
b76347f2 5346 return mg_length(sv);
a0ed51b3 5347 else
b76347f2 5348 {
26346457 5349 STRLEN len;
e62f0680 5350 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5351
26346457
NC
5352 if (PL_utf8cache) {
5353 STRLEN ulen;
5354 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5355
5356 if (mg && mg->mg_len != -1) {
5357 ulen = mg->mg_len;
5358 if (PL_utf8cache < 0) {
5359 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5360 if (real != ulen) {
5361 /* Need to turn the assertions off otherwise we may
5362 recurse infinitely while printing error messages.
5363 */
5364 SAVEI8(PL_utf8cache);
5365 PL_utf8cache = 0;
f5992bc4
RB
5366 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
5367 " real %"UVuf" for %"SVf,
be2597df 5368 (UV) ulen, (UV) real, SVfARG(sv));
26346457
NC
5369 }
5370 }
5371 }
5372 else {
5373 ulen = Perl_utf8_length(aTHX_ s, s + len);
5374 if (!SvREADONLY(sv)) {
5375 if (!mg) {
5376 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5377 &PL_vtbl_utf8, 0, 0);
5378 }
cb9e20bb 5379 assert(mg);
26346457 5380 mg->mg_len = ulen;
cb9e20bb 5381 }
cb9e20bb 5382 }
26346457 5383 return ulen;
7e8c5dac 5384 }
26346457 5385 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5386 }
5387}
5388
9564a3bd
NC
5389/* Walk forwards to find the byte corresponding to the passed in UTF-8
5390 offset. */
bdf30dd6 5391static STRLEN
721e86b6 5392S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
bdf30dd6
NC
5393 STRLEN uoffset)
5394{
5395 const U8 *s = start;
5396
5397 while (s < send && uoffset--)
5398 s += UTF8SKIP(s);
5399 if (s > send) {
5400 /* This is the existing behaviour. Possibly it should be a croak, as
5401 it's actually a bounds error */
5402 s = send;
5403 }
5404 return s - start;
5405}
5406
9564a3bd
NC
5407/* Given the length of the string in both bytes and UTF-8 characters, decide
5408 whether to walk forwards or backwards to find the byte corresponding to
5409 the passed in UTF-8 offset. */
c336ad0b 5410static STRLEN
721e86b6 5411S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
c336ad0b
NC
5412 STRLEN uoffset, STRLEN uend)
5413{
5414 STRLEN backw = uend - uoffset;
5415 if (uoffset < 2 * backw) {
25a8a4ef 5416 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5417 forward (that's where the 2 * backw comes from).
5418 (The real figure of course depends on the UTF-8 data.) */
721e86b6 5419 return sv_pos_u2b_forwards(start, send, uoffset);
c336ad0b
NC
5420 }
5421
5422 while (backw--) {
5423 send--;
5424 while (UTF8_IS_CONTINUATION(*send))
5425 send--;
5426 }
5427 return send - start;
5428}
5429
9564a3bd
NC
5430/* For the string representation of the given scalar, find the byte
5431 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5432 give another position in the string, *before* the sought offset, which
5433 (which is always true, as 0, 0 is a valid pair of positions), which should
5434 help reduce the amount of linear searching.
5435 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5436 will be used to reduce the amount of linear searching. The cache will be
5437 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5438static STRLEN
5439S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5440 const U8 *const send, STRLEN uoffset,
5441 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5442 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5443 bool found = FALSE;
5444
75c33c12
NC
5445 assert (uoffset >= uoffset0);
5446
c336ad0b 5447 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5448 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5449 if ((*mgp)->mg_ptr) {
5450 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5451 if (cache[0] == uoffset) {
5452 /* An exact match. */
5453 return cache[1];
5454 }
ab455f60
NC
5455 if (cache[2] == uoffset) {
5456 /* An exact match. */
5457 return cache[3];
5458 }
668af93f
NC
5459
5460 if (cache[0] < uoffset) {
d8b2e1f9
NC
5461 /* The cache already knows part of the way. */
5462 if (cache[0] > uoffset0) {
5463 /* The cache knows more than the passed in pair */
5464 uoffset0 = cache[0];
5465 boffset0 = cache[1];
5466 }
5467 if ((*mgp)->mg_len != -1) {
5468 /* And we know the end too. */
5469 boffset = boffset0
721e86b6 5470 + sv_pos_u2b_midway(start + boffset0, send,
d8b2e1f9
NC
5471 uoffset - uoffset0,
5472 (*mgp)->mg_len - uoffset0);
5473 } else {
5474 boffset = boffset0
721e86b6 5475 + sv_pos_u2b_forwards(start + boffset0,
d8b2e1f9
NC
5476 send, uoffset - uoffset0);
5477 }
dd7c5fd3
NC
5478 }
5479 else if (cache[2] < uoffset) {
5480 /* We're between the two cache entries. */
5481 if (cache[2] > uoffset0) {
5482 /* and the cache knows more than the passed in pair */
5483 uoffset0 = cache[2];
5484 boffset0 = cache[3];
5485 }
5486
668af93f 5487 boffset = boffset0
721e86b6 5488 + sv_pos_u2b_midway(start + boffset0,
668af93f
NC
5489 start + cache[1],
5490 uoffset - uoffset0,
5491 cache[0] - uoffset0);
dd7c5fd3
NC
5492 } else {
5493 boffset = boffset0
721e86b6 5494 + sv_pos_u2b_midway(start + boffset0,
dd7c5fd3
NC
5495 start + cache[3],
5496 uoffset - uoffset0,
5497 cache[2] - uoffset0);
d8b2e1f9 5498 }
668af93f 5499 found = TRUE;
d8b2e1f9
NC
5500 }
5501 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5502 /* If we can take advantage of a passed in offset, do so. */
5503 /* In fact, offset0 is either 0, or less than offset, so don't
5504 need to worry about the other possibility. */
5505 boffset = boffset0
721e86b6 5506 + sv_pos_u2b_midway(start + boffset0, send,
75c33c12
NC
5507 uoffset - uoffset0,
5508 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5509 found = TRUE;
5510 }
28ccbf94 5511 }
c336ad0b
NC
5512
5513 if (!found || PL_utf8cache < 0) {
75c33c12 5514 const STRLEN real_boffset
721e86b6 5515 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
75c33c12
NC
5516 send, uoffset - uoffset0);
5517
c336ad0b
NC
5518 if (found && PL_utf8cache < 0) {
5519 if (real_boffset != boffset) {
5520 /* Need to turn the assertions off otherwise we may recurse
5521 infinitely while printing error messages. */
5522 SAVEI8(PL_utf8cache);
5523 PL_utf8cache = 0;
f5992bc4
RB
5524 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
5525 " real %"UVuf" for %"SVf,
be2597df 5526 (UV) boffset, (UV) real_boffset, SVfARG(sv));
c336ad0b
NC
5527 }
5528 }
5529 boffset = real_boffset;
28ccbf94 5530 }
0905937d 5531
ab455f60 5532 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5533 return boffset;
5534}
5535
9564a3bd
NC
5536
5537/*
5538=for apidoc sv_pos_u2b
5539
5540Converts the value pointed to by offsetp from a count of UTF-8 chars from
5541the start of the string, to a count of the equivalent number of bytes; if
5542lenp is non-zero, it does the same to lenp, but this time starting from
5543the offset, rather than from the start of the string. Handles magic and
5544type coercion.
5545
5546=cut
5547*/
5548
5549/*
5550 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5551 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5552 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5553 *
5554 */
5555
a0ed51b3 5556void
864dbfa3 5557Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5558{
245d4a47 5559 const U8 *start;
a0ed51b3
LW
5560 STRLEN len;
5561
5562 if (!sv)
5563 return;
5564
245d4a47 5565 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5566 if (len) {
bdf30dd6
NC
5567 STRLEN uoffset = (STRLEN) *offsetp;
5568 const U8 * const send = start + len;
0905937d 5569 MAGIC *mg = NULL;
721e86b6 5570 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
28ccbf94 5571 uoffset, 0, 0);
bdf30dd6
NC
5572
5573 *offsetp = (I32) boffset;
5574
5575 if (lenp) {
28ccbf94 5576 /* Convert the relative offset to absolute. */
721e86b6
AL
5577 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5578 const STRLEN boffset2
5579 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
28ccbf94 5580 uoffset, boffset) - boffset;
bdf30dd6 5581
28ccbf94 5582 *lenp = boffset2;
bdf30dd6 5583 }
7e8c5dac
HS
5584 }
5585 else {
5586 *offsetp = 0;
5587 if (lenp)
5588 *lenp = 0;
a0ed51b3 5589 }
e23c8137 5590
a0ed51b3
LW
5591 return;
5592}
5593
9564a3bd
NC
5594/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5595 byte length pairing. The (byte) length of the total SV is passed in too,
5596 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5597 may not have updated SvCUR, so we can't rely on reading it directly.
5598
5599 The proffered utf8/byte length pairing isn't used if the cache already has
5600 two pairs, and swapping either for the proffered pair would increase the
5601 RMS of the intervals between known byte offsets.
5602
5603 The cache itself consists of 4 STRLEN values
5604 0: larger UTF-8 offset
5605 1: corresponding byte offset
5606 2: smaller UTF-8 offset
5607 3: corresponding byte offset
5608
5609 Unused cache pairs have the value 0, 0.
5610 Keeping the cache "backwards" means that the invariant of
5611 cache[0] >= cache[2] is maintained even with empty slots, which means that
5612 the code that uses it doesn't need to worry if only 1 entry has actually
5613 been set to non-zero. It also makes the "position beyond the end of the
5614 cache" logic much simpler, as the first slot is always the one to start
5615 from.
645c22ef 5616*/
ec07b5e0 5617static void
ab455f60
NC
5618S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5619 STRLEN blen)
ec07b5e0
NC
5620{
5621 STRLEN *cache;
5622 if (SvREADONLY(sv))
5623 return;
5624
5625 if (!*mgp) {
5626 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5627 0);
5628 (*mgp)->mg_len = -1;
5629 }
5630 assert(*mgp);
5631
5632 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5633 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5634 (*mgp)->mg_ptr = (char *) cache;
5635 }
5636 assert(cache);
5637
5638 if (PL_utf8cache < 0) {
ef816a78 5639 const U8 *start = (const U8 *) SvPVX_const(sv);
6448472a 5640 const STRLEN realutf8 = utf8_length(start, start + byte);
ec07b5e0
NC
5641
5642 if (realutf8 != utf8) {
5643 /* Need to turn the assertions off otherwise we may recurse
5644 infinitely while printing error messages. */
5645 SAVEI8(PL_utf8cache);
5646 PL_utf8cache = 0;
f5992bc4 5647 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
be2597df 5648 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
ec07b5e0
NC
5649 }
5650 }
ab455f60
NC
5651
5652 /* Cache is held with the later position first, to simplify the code
5653 that deals with unbounded ends. */
5654
5655 ASSERT_UTF8_CACHE(cache);
5656 if (cache[1] == 0) {
5657 /* Cache is totally empty */
5658 cache[0] = utf8;
5659 cache[1] = byte;
5660 } else if (cache[3] == 0) {
5661 if (byte > cache[1]) {
5662 /* New one is larger, so goes first. */
5663 cache[2] = cache[0];
5664 cache[3] = cache[1];
5665 cache[0] = utf8;
5666 cache[1] = byte;
5667 } else {
5668 cache[2] = utf8;
5669 cache[3] = byte;
5670 }
5671 } else {
5672#define THREEWAY_SQUARE(a,b,c,d) \
5673 ((float)((d) - (c))) * ((float)((d) - (c))) \
5674 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5675 + ((float)((b) - (a))) * ((float)((b) - (a)))
5676
5677 /* Cache has 2 slots in use, and we know three potential pairs.
5678 Keep the two that give the lowest RMS distance. Do the
5679 calcualation in bytes simply because we always know the byte
5680 length. squareroot has the same ordering as the positive value,
5681 so don't bother with the actual square root. */
5682 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5683 if (byte > cache[1]) {
5684 /* New position is after the existing pair of pairs. */
5685 const float keep_earlier
5686 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5687 const float keep_later
5688 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5689
5690 if (keep_later < keep_earlier) {
5691 if (keep_later < existing) {
5692 cache[2] = cache[0];
5693 cache[3] = cache[1];
5694 cache[0] = utf8;
5695 cache[1] = byte;
5696 }
5697 }
5698 else {
5699 if (keep_earlier < existing) {
5700 cache[0] = utf8;
5701 cache[1] = byte;
5702 }
5703 }
5704 }
57d7fbf1
NC
5705 else if (byte > cache[3]) {
5706 /* New position is between the existing pair of pairs. */
5707 const float keep_earlier
5708 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5709 const float keep_later
5710 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5711
5712 if (keep_later < keep_earlier) {
5713 if (keep_later < existing) {
5714 cache[2] = utf8;
5715 cache[3] = byte;
5716 }
5717 }
5718 else {
5719 if (keep_earlier < existing) {
5720 cache[0] = utf8;
5721 cache[1] = byte;
5722 }
5723 }
5724 }
5725 else {
5726 /* New position is before the existing pair of pairs. */
5727 const float keep_earlier
5728 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5729 const float keep_later
5730 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5731
5732 if (keep_later < keep_earlier) {
5733 if (keep_later < existing) {
5734 cache[2] = utf8;
5735 cache[3] = byte;
5736 }
5737 }
5738 else {
5739 if (keep_earlier < existing) {
5740 cache[0] = cache[2];
5741 cache[1] = cache[3];
5742 cache[2] = utf8;
5743 cache[3] = byte;
5744 }
5745 }
5746 }
ab455f60 5747 }
0905937d 5748 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5749}
5750
ec07b5e0 5751/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5752 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5753 backward is half the speed of walking forward. */
ec07b5e0
NC
5754static STRLEN
5755S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5756 STRLEN endu)
5757{
5758 const STRLEN forw = target - s;
5759 STRLEN backw = end - target;
5760
5761 if (forw < 2 * backw) {
6448472a 5762 return utf8_length(s, target);
ec07b5e0
NC
5763 }
5764
5765 while (end > target) {
5766 end--;
5767 while (UTF8_IS_CONTINUATION(*end)) {
5768 end--;
5769 }
5770 endu--;
5771 }
5772 return endu;
5773}
5774
9564a3bd
NC
5775/*
5776=for apidoc sv_pos_b2u
5777
5778Converts the value pointed to by offsetp from a count of bytes from the
5779start of the string, to a count of the equivalent number of UTF-8 chars.
5780Handles magic and type coercion.
5781
5782=cut
5783*/
5784
5785/*
5786 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5787 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5788 * byte offsets.
5789 *
5790 */
a0ed51b3 5791void
7e8c5dac 5792Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5793{
83003860 5794 const U8* s;
ec07b5e0 5795 const STRLEN byte = *offsetp;
7087a21c 5796 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5797 STRLEN blen;
ec07b5e0
NC
5798 MAGIC* mg = NULL;
5799 const U8* send;
a922f900 5800 bool found = FALSE;
a0ed51b3
LW
5801
5802 if (!sv)
5803 return;
5804
ab455f60 5805 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5806
ab455f60 5807 if (blen < byte)
ec07b5e0 5808 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5809
ec07b5e0 5810 send = s + byte;
a67d7df9 5811
ffca234a
NC
5812 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5813 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5814 if (mg->mg_ptr) {
d4c19fe8 5815 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5816 if (cache[1] == byte) {
ec07b5e0
NC
5817 /* An exact match. */
5818 *offsetp = cache[0];
ec07b5e0 5819 return;
7e8c5dac 5820 }
ab455f60
NC
5821 if (cache[3] == byte) {
5822 /* An exact match. */
5823 *offsetp = cache[2];
5824 return;
5825 }
668af93f
NC
5826
5827 if (cache[1] < byte) {
ec07b5e0 5828 /* We already know part of the way. */
b9f984a5
NC
5829 if (mg->mg_len != -1) {
5830 /* Actually, we know the end too. */
5831 len = cache[0]
5832 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5833 s + blen, mg->mg_len - cache[0]);
b9f984a5 5834 } else {
6448472a 5835 len = cache[0] + utf8_length(s + cache[1], send);
b9f984a5 5836 }
7e8c5dac 5837 }
9f985e4c
NC
5838 else if (cache[3] < byte) {
5839 /* We're between the two cached pairs, so we do the calculation
5840 offset by the byte/utf-8 positions for the earlier pair,
5841 then add the utf-8 characters from the string start to
5842 there. */
5843 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5844 s + cache[1], cache[0] - cache[2])
5845 + cache[2];
5846
5847 }
5848 else { /* cache[3] > byte */
5849 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5850 cache[2]);
7e8c5dac 5851
7e8c5dac 5852 }
ec07b5e0 5853 ASSERT_UTF8_CACHE(cache);
a922f900 5854 found = TRUE;
ffca234a 5855 } else if (mg->mg_len != -1) {
ab455f60 5856 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5857 found = TRUE;
7e8c5dac 5858 }
a0ed51b3 5859 }
a922f900 5860 if (!found || PL_utf8cache < 0) {
6448472a 5861 const STRLEN real_len = utf8_length(s, send);
a922f900
NC
5862
5863 if (found && PL_utf8cache < 0) {
5864 if (len != real_len) {
5865 /* Need to turn the assertions off otherwise we may recurse
5866 infinitely while printing error messages. */
5867 SAVEI8(PL_utf8cache);
5868 PL_utf8cache = 0;
f5992bc4
RB
5869 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
5870 " real %"UVuf" for %"SVf,
be2597df 5871 (UV) len, (UV) real_len, SVfARG(sv));
a922f900
NC
5872 }
5873 }
5874 len = real_len;
ec07b5e0
NC
5875 }
5876 *offsetp = len;
5877
ab455f60 5878 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5879}
5880
954c1994
GS
5881/*
5882=for apidoc sv_eq
5883
5884Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5885identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5886coerce its args to strings if necessary.
954c1994
GS
5887
5888=cut
5889*/
5890
79072805 5891I32
e01b9e88 5892Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5893{
97aff369 5894 dVAR;
e1ec3a88 5895 const char *pv1;
463ee0b2 5896 STRLEN cur1;
e1ec3a88 5897 const char *pv2;
463ee0b2 5898 STRLEN cur2;
e01b9e88 5899 I32 eq = 0;
bd61b366 5900 char *tpv = NULL;
a0714e2c 5901 SV* svrecode = NULL;
79072805 5902
e01b9e88 5903 if (!sv1) {
79072805
LW
5904 pv1 = "";
5905 cur1 = 0;
5906 }
ced497e2
YST
5907 else {
5908 /* if pv1 and pv2 are the same, second SvPV_const call may
5909 * invalidate pv1, so we may need to make a copy */
5910 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
5911 pv1 = SvPV_const(sv1, cur1);
5912 sv1 = sv_2mortal(newSVpvn(pv1, cur1));
5913 if (SvUTF8(sv2)) SvUTF8_on(sv1);
5914 }
4d84ee25 5915 pv1 = SvPV_const(sv1, cur1);
ced497e2 5916 }
79072805 5917
e01b9e88
SC
5918 if (!sv2){
5919 pv2 = "";
5920 cur2 = 0;
92d29cee 5921 }
e01b9e88 5922 else
4d84ee25 5923 pv2 = SvPV_const(sv2, cur2);
79072805 5924
cf48d248 5925 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5926 /* Differing utf8ness.
5927 * Do not UTF8size the comparands as a side-effect. */
5928 if (PL_encoding) {
5929 if (SvUTF8(sv1)) {
553e1bcc
AT
5930 svrecode = newSVpvn(pv2, cur2);
5931 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5932 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5933 }
5934 else {
553e1bcc
AT
5935 svrecode = newSVpvn(pv1, cur1);
5936 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5937 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5938 }
5939 /* Now both are in UTF-8. */
0a1bd7ac
DM
5940 if (cur1 != cur2) {
5941 SvREFCNT_dec(svrecode);
799ef3cb 5942 return FALSE;
0a1bd7ac 5943 }
799ef3cb
JH
5944 }
5945 else {
5946 bool is_utf8 = TRUE;
5947
5948 if (SvUTF8(sv1)) {
5949 /* sv1 is the UTF-8 one,
5950 * if is equal it must be downgrade-able */
9d4ba2ae 5951 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5952 &cur1, &is_utf8);
5953 if (pv != pv1)
553e1bcc 5954 pv1 = tpv = pv;
799ef3cb
JH
5955 }
5956 else {
5957 /* sv2 is the UTF-8 one,
5958 * if is equal it must be downgrade-able */
9d4ba2ae 5959 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5960 &cur2, &is_utf8);
5961 if (pv != pv2)
553e1bcc 5962 pv2 = tpv = pv;
799ef3cb
JH
5963 }
5964 if (is_utf8) {
5965 /* Downgrade not possible - cannot be eq */
bf694877 5966 assert (tpv == 0);
799ef3cb
JH
5967 return FALSE;
5968 }
5969 }
cf48d248
JH
5970 }
5971
5972 if (cur1 == cur2)
765f542d 5973 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5974
b37c2d43 5975 SvREFCNT_dec(svrecode);
553e1bcc
AT
5976 if (tpv)
5977 Safefree(tpv);
cf48d248 5978
e01b9e88 5979 return eq;
79072805
LW
5980}
5981
954c1994
GS
5982/*
5983=for apidoc sv_cmp
5984
5985Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5986string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5987C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5988coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5989
5990=cut
5991*/
5992
79072805 5993I32
e01b9e88 5994Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5995{
97aff369 5996 dVAR;
560a288e 5997 STRLEN cur1, cur2;
e1ec3a88 5998 const char *pv1, *pv2;
bd61b366 5999 char *tpv = NULL;
cf48d248 6000 I32 cmp;
a0714e2c 6001 SV *svrecode = NULL;
560a288e 6002
e01b9e88
SC
6003 if (!sv1) {
6004 pv1 = "";
560a288e
GS
6005 cur1 = 0;
6006 }
e01b9e88 6007 else
4d84ee25 6008 pv1 = SvPV_const(sv1, cur1);
560a288e 6009
553e1bcc 6010 if (!sv2) {
e01b9e88 6011 pv2 = "";
560a288e
GS
6012 cur2 = 0;
6013 }
e01b9e88 6014 else
4d84ee25 6015 pv2 = SvPV_const(sv2, cur2);
79072805 6016
cf48d248 6017 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6018 /* Differing utf8ness.
6019 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6020 if (SvUTF8(sv1)) {
799ef3cb 6021 if (PL_encoding) {
553e1bcc
AT
6022 svrecode = newSVpvn(pv2, cur2);
6023 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6024 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6025 }
6026 else {
e1ec3a88 6027 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6028 }
cf48d248
JH
6029 }
6030 else {
799ef3cb 6031 if (PL_encoding) {
553e1bcc
AT
6032 svrecode = newSVpvn(pv1, cur1);
6033 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6034 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6035 }
6036 else {
e1ec3a88 6037 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6038 }
cf48d248
JH
6039 }
6040 }
6041
e01b9e88 6042 if (!cur1) {
cf48d248 6043 cmp = cur2 ? -1 : 0;
e01b9e88 6044 } else if (!cur2) {
cf48d248
JH
6045 cmp = 1;
6046 } else {
e1ec3a88 6047 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6048
6049 if (retval) {
cf48d248 6050 cmp = retval < 0 ? -1 : 1;
e01b9e88 6051 } else if (cur1 == cur2) {
cf48d248
JH
6052 cmp = 0;
6053 } else {
6054 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6055 }
cf48d248 6056 }
16660edb 6057
b37c2d43 6058 SvREFCNT_dec(svrecode);
553e1bcc
AT
6059 if (tpv)
6060 Safefree(tpv);
cf48d248
JH
6061
6062 return cmp;
bbce6d69 6063}
16660edb 6064
c461cf8f
JH
6065/*
6066=for apidoc sv_cmp_locale
6067
645c22ef
DM
6068Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6069'use bytes' aware, handles get magic, and will coerce its args to strings
6070if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6071
6072=cut
6073*/
6074
bbce6d69 6075I32
864dbfa3 6076Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6077{
97aff369 6078 dVAR;
36477c24 6079#ifdef USE_LOCALE_COLLATE
16660edb 6080
bbce6d69 6081 char *pv1, *pv2;
6082 STRLEN len1, len2;
6083 I32 retval;
16660edb 6084
3280af22 6085 if (PL_collation_standard)
bbce6d69 6086 goto raw_compare;
16660edb 6087
bbce6d69 6088 len1 = 0;
8ac85365 6089 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6090 len2 = 0;
8ac85365 6091 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6092
bbce6d69 6093 if (!pv1 || !len1) {
6094 if (pv2 && len2)
6095 return -1;
6096 else
6097 goto raw_compare;
6098 }
6099 else {
6100 if (!pv2 || !len2)
6101 return 1;
6102 }
16660edb 6103
bbce6d69 6104 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6105
bbce6d69 6106 if (retval)
16660edb 6107 return retval < 0 ? -1 : 1;
6108
bbce6d69 6109 /*
6110 * When the result of collation is equality, that doesn't mean
6111 * that there are no differences -- some locales exclude some
6112 * characters from consideration. So to avoid false equalities,
6113 * we use the raw string as a tiebreaker.
6114 */
16660edb 6115
bbce6d69 6116 raw_compare:
5f66b61c 6117 /*FALLTHROUGH*/
16660edb 6118
36477c24 6119#endif /* USE_LOCALE_COLLATE */
16660edb 6120
bbce6d69 6121 return sv_cmp(sv1, sv2);
6122}
79072805 6123
645c22ef 6124
36477c24 6125#ifdef USE_LOCALE_COLLATE
645c22ef 6126
7a4c00b4 6127/*
645c22ef
DM
6128=for apidoc sv_collxfrm
6129
6130Add Collate Transform magic to an SV if it doesn't already have it.
6131
6132Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6133scalar data of the variable, but transformed to such a format that a normal
6134memory comparison can be used to compare the data according to the locale
6135settings.
6136
6137=cut
6138*/
6139
bbce6d69 6140char *
864dbfa3 6141Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6142{
97aff369 6143 dVAR;
7a4c00b4 6144 MAGIC *mg;
16660edb 6145
14befaf4 6146 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6147 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6148 const char *s;
6149 char *xf;
bbce6d69 6150 STRLEN len, xlen;
6151
7a4c00b4 6152 if (mg)
6153 Safefree(mg->mg_ptr);
93524f2b 6154 s = SvPV_const(sv, len);
bbce6d69 6155 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6156 if (SvREADONLY(sv)) {
6157 SAVEFREEPV(xf);
6158 *nxp = xlen;
3280af22 6159 return xf + sizeof(PL_collation_ix);
ff0cee69 6160 }
7a4c00b4 6161 if (! mg) {
d83f0a82
NC
6162#ifdef PERL_OLD_COPY_ON_WRITE
6163 if (SvIsCOW(sv))
6164 sv_force_normal_flags(sv, 0);
6165#endif
6166 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6167 0, 0);
7a4c00b4 6168 assert(mg);
bbce6d69 6169 }
7a4c00b4 6170 mg->mg_ptr = xf;
565764a8 6171 mg->mg_len = xlen;
7a4c00b4 6172 }
6173 else {
ff0cee69 6174 if (mg) {
6175 mg->mg_ptr = NULL;
565764a8 6176 mg->mg_len = -1;
ff0cee69 6177 }
bbce6d69 6178 }
6179 }
7a4c00b4 6180 if (mg && mg->mg_ptr) {
565764a8 6181 *nxp = mg->mg_len;
3280af22 6182 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6183 }
6184 else {
6185 *nxp = 0;
6186 return NULL;
16660edb 6187 }
79072805
LW
6188}
6189
36477c24 6190#endif /* USE_LOCALE_COLLATE */
bbce6d69 6191
c461cf8f
JH
6192/*
6193=for apidoc sv_gets
6194
6195Get a line from the filehandle and store it into the SV, optionally
6196appending to the currently-stored string.
6197
6198=cut
6199*/
6200
79072805 6201char *
864dbfa3 6202Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6203{
97aff369 6204 dVAR;
e1ec3a88 6205 const char *rsptr;
c07a80fd 6206 STRLEN rslen;
6207 register STDCHAR rslast;
6208 register STDCHAR *bp;
6209 register I32 cnt;
9c5ffd7c 6210 I32 i = 0;
8bfdd7d9 6211 I32 rspara = 0;
c07a80fd 6212
bc44a8a2
NC
6213 if (SvTHINKFIRST(sv))
6214 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6215 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6216 from <>.
6217 However, perlbench says it's slower, because the existing swipe code
6218 is faster than copy on write.
6219 Swings and roundabouts. */
862a34c6 6220 SvUPGRADE(sv, SVt_PV);
99491443 6221
ff68c719 6222 SvSCREAM_off(sv);
efd8b2ba
AE
6223
6224 if (append) {
6225 if (PerlIO_isutf8(fp)) {
6226 if (!SvUTF8(sv)) {
6227 sv_utf8_upgrade_nomg(sv);
6228 sv_pos_u2b(sv,&append,0);
6229 }
6230 } else if (SvUTF8(sv)) {
561b68a9 6231 SV * const tsv = newSV(0);
efd8b2ba
AE
6232 sv_gets(tsv, fp, 0);
6233 sv_utf8_upgrade_nomg(tsv);
6234 SvCUR_set(sv,append);
6235 sv_catsv(sv,tsv);
6236 sv_free(tsv);
6237 goto return_string_or_null;
6238 }
6239 }
6240
6241 SvPOK_only(sv);
6242 if (PerlIO_isutf8(fp))
6243 SvUTF8_on(sv);
c07a80fd 6244
923e4eb5 6245 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6246 /* we always read code in line mode */
6247 rsptr = "\n";
6248 rslen = 1;
6249 }
6250 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6251 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6252 of amount we are going to read -- may result in mallocing
6253 more memory than we really need if the layers below reduce
6254 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6255 */
e311fd51 6256 Stat_t st;
e468d35b 6257 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6258 const Off_t offset = PerlIO_tell(fp);
58f1856e 6259 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6260 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6261 }
6262 }
c07a80fd 6263 rsptr = NULL;
6264 rslen = 0;
6265 }
3280af22 6266 else if (RsRECORD(PL_rs)) {
e311fd51 6267 I32 bytesread;
5b2b9c68 6268 char *buffer;
acbd132f 6269 U32 recsize;
5b2b9c68
HM
6270
6271 /* Grab the size of the record we're getting */
acbd132f 6272 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6273 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6274 /* Go yank in */
6275#ifdef VMS
6276 /* VMS wants read instead of fread, because fread doesn't respect */
6277 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6278 /* doing, but we've got no other real choice - except avoid stdio
6279 as implementation - perhaps write a :vms layer ?
6280 */
5b2b9c68
HM
6281 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6282#else
6283 bytesread = PerlIO_read(fp, buffer, recsize);
6284#endif
27e6ca2d
AE
6285 if (bytesread < 0)
6286 bytesread = 0;
e311fd51 6287 SvCUR_set(sv, bytesread += append);
e670df4e 6288 buffer[bytesread] = '\0';
efd8b2ba 6289 goto return_string_or_null;
5b2b9c68 6290 }
3280af22 6291 else if (RsPARA(PL_rs)) {
c07a80fd 6292 rsptr = "\n\n";
6293 rslen = 2;
8bfdd7d9 6294 rspara = 1;
c07a80fd 6295 }
7d59b7e4
NIS
6296 else {
6297 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6298 if (PerlIO_isutf8(fp)) {
6299 rsptr = SvPVutf8(PL_rs, rslen);
6300 }
6301 else {
6302 if (SvUTF8(PL_rs)) {
6303 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6304 Perl_croak(aTHX_ "Wide character in $/");
6305 }
6306 }
93524f2b 6307 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6308 }
6309 }
6310
c07a80fd 6311 rslast = rslen ? rsptr[rslen - 1] : '\0';
6312
8bfdd7d9 6313 if (rspara) { /* have to do this both before and after */
79072805 6314 do { /* to make sure file boundaries work right */
760ac839 6315 if (PerlIO_eof(fp))
a0d0e21e 6316 return 0;
760ac839 6317 i = PerlIO_getc(fp);
79072805 6318 if (i != '\n') {
a0d0e21e
LW
6319 if (i == -1)
6320 return 0;
760ac839 6321 PerlIO_ungetc(fp,i);
79072805
LW
6322 break;
6323 }
6324 } while (i != EOF);
6325 }
c07a80fd 6326
760ac839
LW
6327 /* See if we know enough about I/O mechanism to cheat it ! */
6328
6329 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6330 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6331 enough here - and may even be a macro allowing compile
6332 time optimization.
6333 */
6334
6335 if (PerlIO_fast_gets(fp)) {
6336
6337 /*
6338 * We're going to steal some values from the stdio struct
6339 * and put EVERYTHING in the innermost loop into registers.
6340 */
6341 register STDCHAR *ptr;
6342 STRLEN bpx;
6343 I32 shortbuffered;
6344
16660edb 6345#if defined(VMS) && defined(PERLIO_IS_STDIO)
6346 /* An ungetc()d char is handled separately from the regular
6347 * buffer, so we getc() it back out and stuff it in the buffer.
6348 */
6349 i = PerlIO_getc(fp);
6350 if (i == EOF) return 0;
6351 *(--((*fp)->_ptr)) = (unsigned char) i;
6352 (*fp)->_cnt++;
6353#endif
c07a80fd 6354
c2960299 6355 /* Here is some breathtakingly efficient cheating */
c07a80fd 6356
a20bf0c3 6357 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6358 /* make sure we have the room */
7a5fa8a2 6359 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6360 /* Not room for all of it
7a5fa8a2 6361 if we are looking for a separator and room for some
e468d35b
NIS
6362 */
6363 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6364 /* just process what we have room for */
79072805
LW
6365 shortbuffered = cnt - SvLEN(sv) + append + 1;
6366 cnt -= shortbuffered;
6367 }
6368 else {
6369 shortbuffered = 0;
bbce6d69 6370 /* remember that cnt can be negative */
eb160463 6371 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6372 }
6373 }
7a5fa8a2 6374 else
79072805 6375 shortbuffered = 0;
3f7c398e 6376 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6377 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6378 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6379 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6380 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6381 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6382 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6383 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6384 for (;;) {
6385 screamer:
93a17b20 6386 if (cnt > 0) {
c07a80fd 6387 if (rslen) {
760ac839
LW
6388 while (cnt > 0) { /* this | eat */
6389 cnt--;
c07a80fd 6390 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6391 goto thats_all_folks; /* screams | sed :-) */
6392 }
6393 }
6394 else {
1c846c1f
NIS
6395 Copy(ptr, bp, cnt, char); /* this | eat */
6396 bp += cnt; /* screams | dust */
c07a80fd 6397 ptr += cnt; /* louder | sed :-) */
a5f75d66 6398 cnt = 0;
93a17b20 6399 }
79072805
LW
6400 }
6401
748a9306 6402 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6403 cnt = shortbuffered;
6404 shortbuffered = 0;
3f7c398e 6405 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6406 SvCUR_set(sv, bpx);
6407 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6408 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6409 continue;
6410 }
6411
16660edb 6412 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6413 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6414 PTR2UV(ptr),(long)cnt));
cc00df79 6415 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6416#if 0
16660edb 6417 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6418 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6419 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6420 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6421#endif
1c846c1f 6422 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6423 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6424 another abstraction. */
760ac839 6425 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6426#if 0
16660edb 6427 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6428 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6429 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6430 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6431#endif
a20bf0c3
JH
6432 cnt = PerlIO_get_cnt(fp);
6433 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6434 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6435 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6436
748a9306
LW
6437 if (i == EOF) /* all done for ever? */
6438 goto thats_really_all_folks;
6439
3f7c398e 6440 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6441 SvCUR_set(sv, bpx);
6442 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6443 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6444
eb160463 6445 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6446
c07a80fd 6447 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6448 goto thats_all_folks;
79072805
LW
6449 }
6450
6451thats_all_folks:
3f7c398e 6452 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6453 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6454 goto screamer; /* go back to the fray */
79072805
LW
6455thats_really_all_folks:
6456 if (shortbuffered)
6457 cnt += shortbuffered;
16660edb 6458 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6459 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6460 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6461 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6462 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6463 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6464 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6465 *bp = '\0';
3f7c398e 6466 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6467 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6468 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6469 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6470 }
6471 else
79072805 6472 {
6edd2cd5 6473 /*The big, slow, and stupid way. */
27da23d5 6474#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6475 STDCHAR *buf = NULL;
a02a5408 6476 Newx(buf, 8192, STDCHAR);
6edd2cd5 6477 assert(buf);
4d2c4e07 6478#else
6edd2cd5 6479 STDCHAR buf[8192];
4d2c4e07 6480#endif
79072805 6481
760ac839 6482screamer2:
c07a80fd 6483 if (rslen) {
00b6aa41 6484 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6485 bp = buf;
eb160463 6486 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6487 ; /* keep reading */
6488 cnt = bp - buf;
c07a80fd 6489 }
6490 else {
760ac839 6491 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6492 /* Accomodate broken VAXC compiler, which applies U8 cast to
6493 * both args of ?: operator, causing EOF to change into 255
6494 */
37be0adf 6495 if (cnt > 0)
cbe9e203
JH
6496 i = (U8)buf[cnt - 1];
6497 else
37be0adf 6498 i = EOF;
c07a80fd 6499 }
79072805 6500
cbe9e203
JH
6501 if (cnt < 0)
6502 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6503 if (append)
6504 sv_catpvn(sv, (char *) buf, cnt);
6505 else
6506 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6507
6508 if (i != EOF && /* joy */
6509 (!rslen ||
6510 SvCUR(sv) < rslen ||
3f7c398e 6511 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6512 {
6513 append = -1;
63e4d877
CS
6514 /*
6515 * If we're reading from a TTY and we get a short read,
6516 * indicating that the user hit his EOF character, we need
6517 * to notice it now, because if we try to read from the TTY
6518 * again, the EOF condition will disappear.
6519 *
6520 * The comparison of cnt to sizeof(buf) is an optimization
6521 * that prevents unnecessary calls to feof().
6522 *
6523 * - jik 9/25/96
6524 */
bb7a0f54 6525 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 6526 goto screamer2;
79072805 6527 }
6edd2cd5 6528
27da23d5 6529#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6530 Safefree(buf);
6531#endif
79072805
LW
6532 }
6533
8bfdd7d9 6534 if (rspara) { /* have to do this both before and after */
c07a80fd 6535 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6536 i = PerlIO_getc(fp);
79072805 6537 if (i != '\n') {
760ac839 6538 PerlIO_ungetc(fp,i);
79072805
LW
6539 break;
6540 }
6541 }
6542 }
c07a80fd 6543
efd8b2ba 6544return_string_or_null:
bd61b366 6545 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6546}
6547
954c1994
GS
6548/*
6549=for apidoc sv_inc
6550
645c22ef
DM
6551Auto-increment of the value in the SV, doing string to numeric conversion
6552if necessary. Handles 'get' magic.
954c1994
GS
6553
6554=cut
6555*/
6556
79072805 6557void
864dbfa3 6558Perl_sv_inc(pTHX_ register SV *sv)
79072805 6559{
97aff369 6560 dVAR;
79072805 6561 register char *d;
463ee0b2 6562 int flags;
79072805
LW
6563
6564 if (!sv)
6565 return;
5b295bef 6566 SvGETMAGIC(sv);
ed6116ce 6567 if (SvTHINKFIRST(sv)) {
765f542d
NC
6568 if (SvIsCOW(sv))
6569 sv_force_normal_flags(sv, 0);
0f15f207 6570 if (SvREADONLY(sv)) {
923e4eb5 6571 if (IN_PERL_RUNTIME)
cea2e8a9 6572 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6573 }
a0d0e21e 6574 if (SvROK(sv)) {
b5be31e9 6575 IV i;
9e7bc3e8
JD
6576 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6577 return;
56431972 6578 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6579 sv_unref(sv);
6580 sv_setiv(sv, i);
a0d0e21e 6581 }
ed6116ce 6582 }
8990e307 6583 flags = SvFLAGS(sv);
28e5dec8
JH
6584 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6585 /* It's (privately or publicly) a float, but not tested as an
6586 integer, so test it to see. */
d460ef45 6587 (void) SvIV(sv);
28e5dec8
JH
6588 flags = SvFLAGS(sv);
6589 }
6590 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6591 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6592#ifdef PERL_PRESERVE_IVUV
28e5dec8 6593 oops_its_int:
59d8ce62 6594#endif
25da4f38
IZ
6595 if (SvIsUV(sv)) {
6596 if (SvUVX(sv) == UV_MAX)
a1e868e7 6597 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6598 else
6599 (void)SvIOK_only_UV(sv);
607fa7f2 6600 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6601 } else {
6602 if (SvIVX(sv) == IV_MAX)
28e5dec8 6603 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6604 else {
6605 (void)SvIOK_only(sv);
45977657 6606 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6607 }
55497cff 6608 }
79072805
LW
6609 return;
6610 }
28e5dec8
JH
6611 if (flags & SVp_NOK) {
6612 (void)SvNOK_only(sv);
9d6ce603 6613 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6614 return;
6615 }
6616
3f7c398e 6617 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6618 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6619 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6620 (void)SvIOK_only(sv);
45977657 6621 SvIV_set(sv, 1);
79072805
LW
6622 return;
6623 }
463ee0b2 6624 d = SvPVX(sv);
79072805
LW
6625 while (isALPHA(*d)) d++;
6626 while (isDIGIT(*d)) d++;
6627 if (*d) {
28e5dec8 6628#ifdef PERL_PRESERVE_IVUV
d1be9408 6629 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6630 warnings. Probably ought to make the sv_iv_please() that does
6631 the conversion if possible, and silently. */
504618e9 6632 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6633 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6634 /* Need to try really hard to see if it's an integer.
6635 9.22337203685478e+18 is an integer.
6636 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6637 so $a="9.22337203685478e+18"; $a+0; $a++
6638 needs to be the same as $a="9.22337203685478e+18"; $a++
6639 or we go insane. */
d460ef45 6640
28e5dec8
JH
6641 (void) sv_2iv(sv);
6642 if (SvIOK(sv))
6643 goto oops_its_int;
6644
6645 /* sv_2iv *should* have made this an NV */
6646 if (flags & SVp_NOK) {
6647 (void)SvNOK_only(sv);
9d6ce603 6648 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6649 return;
6650 }
6651 /* I don't think we can get here. Maybe I should assert this
6652 And if we do get here I suspect that sv_setnv will croak. NWC
6653 Fall through. */
6654#if defined(USE_LONG_DOUBLE)
6655 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 6656 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6657#else
1779d84d 6658 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 6659 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6660#endif
6661 }
6662#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6663 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6664 return;
6665 }
6666 d--;
3f7c398e 6667 while (d >= SvPVX_const(sv)) {
79072805
LW
6668 if (isDIGIT(*d)) {
6669 if (++*d <= '9')
6670 return;
6671 *(d--) = '0';
6672 }
6673 else {
9d116dd7
JH
6674#ifdef EBCDIC
6675 /* MKS: The original code here died if letters weren't consecutive.
6676 * at least it didn't have to worry about non-C locales. The
6677 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6678 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6679 * [A-Za-z] are accepted by isALPHA in the C locale.
6680 */
6681 if (*d != 'z' && *d != 'Z') {
6682 do { ++*d; } while (!isALPHA(*d));
6683 return;
6684 }
6685 *(d--) -= 'z' - 'a';
6686#else
79072805
LW
6687 ++*d;
6688 if (isALPHA(*d))
6689 return;
6690 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6691#endif
79072805
LW
6692 }
6693 }
6694 /* oh,oh, the number grew */
6695 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6696 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6697 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6698 *d = d[-1];
6699 if (isDIGIT(d[1]))
6700 *d = '1';
6701 else
6702 *d = d[1];
6703}
6704
954c1994
GS
6705/*
6706=for apidoc sv_dec
6707
645c22ef
DM
6708Auto-decrement of the value in the SV, doing string to numeric conversion
6709if necessary. Handles 'get' magic.
954c1994
GS
6710
6711=cut
6712*/
6713
79072805 6714void
864dbfa3 6715Perl_sv_dec(pTHX_ register SV *sv)
79072805 6716{
97aff369 6717 dVAR;
463ee0b2
LW
6718 int flags;
6719
79072805
LW
6720 if (!sv)
6721 return;
5b295bef 6722 SvGETMAGIC(sv);
ed6116ce 6723 if (SvTHINKFIRST(sv)) {
765f542d
NC
6724 if (SvIsCOW(sv))
6725 sv_force_normal_flags(sv, 0);
0f15f207 6726 if (SvREADONLY(sv)) {
923e4eb5 6727 if (IN_PERL_RUNTIME)
cea2e8a9 6728 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6729 }
a0d0e21e 6730 if (SvROK(sv)) {
b5be31e9 6731 IV i;
9e7bc3e8
JD
6732 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6733 return;
56431972 6734 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6735 sv_unref(sv);
6736 sv_setiv(sv, i);
a0d0e21e 6737 }
ed6116ce 6738 }
28e5dec8
JH
6739 /* Unlike sv_inc we don't have to worry about string-never-numbers
6740 and keeping them magic. But we mustn't warn on punting */
8990e307 6741 flags = SvFLAGS(sv);
28e5dec8
JH
6742 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6743 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6744#ifdef PERL_PRESERVE_IVUV
28e5dec8 6745 oops_its_int:
59d8ce62 6746#endif
25da4f38
IZ
6747 if (SvIsUV(sv)) {
6748 if (SvUVX(sv) == 0) {
6749 (void)SvIOK_only(sv);
45977657 6750 SvIV_set(sv, -1);
25da4f38
IZ
6751 }
6752 else {
6753 (void)SvIOK_only_UV(sv);
f4eee32f 6754 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6755 }
25da4f38
IZ
6756 } else {
6757 if (SvIVX(sv) == IV_MIN)
65202027 6758 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6759 else {
6760 (void)SvIOK_only(sv);
45977657 6761 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6762 }
55497cff 6763 }
6764 return;
6765 }
28e5dec8 6766 if (flags & SVp_NOK) {
9d6ce603 6767 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6768 (void)SvNOK_only(sv);
6769 return;
6770 }
8990e307 6771 if (!(flags & SVp_POK)) {
ef088171
NC
6772 if ((flags & SVTYPEMASK) < SVt_PVIV)
6773 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6774 SvIV_set(sv, -1);
6775 (void)SvIOK_only(sv);
79072805
LW
6776 return;
6777 }
28e5dec8
JH
6778#ifdef PERL_PRESERVE_IVUV
6779 {
504618e9 6780 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6781 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6782 /* Need to try really hard to see if it's an integer.
6783 9.22337203685478e+18 is an integer.
6784 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6785 so $a="9.22337203685478e+18"; $a+0; $a--
6786 needs to be the same as $a="9.22337203685478e+18"; $a--
6787 or we go insane. */
d460ef45 6788
28e5dec8
JH
6789 (void) sv_2iv(sv);
6790 if (SvIOK(sv))
6791 goto oops_its_int;
6792
6793 /* sv_2iv *should* have made this an NV */
6794 if (flags & SVp_NOK) {
6795 (void)SvNOK_only(sv);
9d6ce603 6796 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6797 return;
6798 }
6799 /* I don't think we can get here. Maybe I should assert this
6800 And if we do get here I suspect that sv_setnv will croak. NWC
6801 Fall through. */
6802#if defined(USE_LONG_DOUBLE)
6803 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 6804 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6805#else
1779d84d 6806 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 6807 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6808#endif
6809 }
6810 }
6811#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6812 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6813}
6814
954c1994
GS
6815/*
6816=for apidoc sv_mortalcopy
6817
645c22ef 6818Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6819The new SV is marked as mortal. It will be destroyed "soon", either by an
6820explicit call to FREETMPS, or by an implicit call at places such as
6821statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6822
6823=cut
6824*/
6825
79072805
LW
6826/* Make a string that will exist for the duration of the expression
6827 * evaluation. Actually, it may have to last longer than that, but
6828 * hopefully we won't free it until it has been assigned to a
6829 * permanent location. */
6830
6831SV *
864dbfa3 6832Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6833{
97aff369 6834 dVAR;
463ee0b2 6835 register SV *sv;
b881518d 6836
4561caa4 6837 new_SV(sv);
79072805 6838 sv_setsv(sv,oldstr);
677b06e3
GS
6839 EXTEND_MORTAL(1);
6840 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6841 SvTEMP_on(sv);
6842 return sv;
6843}
6844
954c1994
GS
6845/*
6846=for apidoc sv_newmortal
6847
645c22ef 6848Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6849set to 1. It will be destroyed "soon", either by an explicit call to
6850FREETMPS, or by an implicit call at places such as statement boundaries.
6851See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6852
6853=cut
6854*/
6855
8990e307 6856SV *
864dbfa3 6857Perl_sv_newmortal(pTHX)
8990e307 6858{
97aff369 6859 dVAR;
8990e307
LW
6860 register SV *sv;
6861
4561caa4 6862 new_SV(sv);
8990e307 6863 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6864 EXTEND_MORTAL(1);
6865 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6866 return sv;
6867}
6868
954c1994
GS
6869/*
6870=for apidoc sv_2mortal
6871
d4236ebc
DM
6872Marks an existing SV as mortal. The SV will be destroyed "soon", either
6873by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6874statement boundaries. SvTEMP() is turned on which means that the SV's
6875string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6876and C<sv_mortalcopy>.
954c1994
GS
6877
6878=cut
6879*/
6880
79072805 6881SV *
864dbfa3 6882Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6883{
27da23d5 6884 dVAR;
79072805 6885 if (!sv)
7a5b473e 6886 return NULL;
d689ffdd 6887 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6888 return sv;
677b06e3
GS
6889 EXTEND_MORTAL(1);
6890 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6891 SvTEMP_on(sv);
79072805
LW
6892 return sv;
6893}
6894
954c1994
GS
6895/*
6896=for apidoc newSVpv
6897
6898Creates a new SV and copies a string into it. The reference count for the
6899SV is set to 1. If C<len> is zero, Perl will compute the length using
6900strlen(). For efficiency, consider using C<newSVpvn> instead.
6901
6902=cut
6903*/
6904
79072805 6905SV *
864dbfa3 6906Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6907{
97aff369 6908 dVAR;
463ee0b2 6909 register SV *sv;
79072805 6910
4561caa4 6911 new_SV(sv);
ddfa59c7 6912 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
79072805
LW
6913 return sv;
6914}
6915
954c1994
GS
6916/*
6917=for apidoc newSVpvn
6918
6919Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6920SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6921string. You are responsible for ensuring that the source string is at least
9e09f5f2 6922C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6923
6924=cut
6925*/
6926
9da1e3b5 6927SV *
864dbfa3 6928Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6929{
97aff369 6930 dVAR;
9da1e3b5
MUN
6931 register SV *sv;
6932
6933 new_SV(sv);
9da1e3b5
MUN
6934 sv_setpvn(sv,s,len);
6935 return sv;
6936}
6937
bd08039b
NC
6938
6939/*
926f8064 6940=for apidoc newSVhek
bd08039b
NC
6941
6942Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6943point to the shared string table where possible. Returns a new (undefined)
6944SV if the hek is NULL.
bd08039b
NC
6945
6946=cut
6947*/
6948
6949SV *
c1b02ed8 6950Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6951{
97aff369 6952 dVAR;
5aaec2b4
NC
6953 if (!hek) {
6954 SV *sv;
6955
6956 new_SV(sv);
6957 return sv;
6958 }
6959
bd08039b
NC
6960 if (HEK_LEN(hek) == HEf_SVKEY) {
6961 return newSVsv(*(SV**)HEK_KEY(hek));
6962 } else {
6963 const int flags = HEK_FLAGS(hek);
6964 if (flags & HVhek_WASUTF8) {
6965 /* Trouble :-)
6966 Andreas would like keys he put in as utf8 to come back as utf8
6967 */
6968 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6969 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6970 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6971
6972 SvUTF8_on (sv);
6973 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6974 return sv;
45e34800 6975 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
bd08039b
NC
6976 /* We don't have a pointer to the hv, so we have to replicate the
6977 flag into every HEK. This hv is using custom a hasing
6978 algorithm. Hence we can't return a shared string scalar, as
6979 that would contain the (wrong) hash value, and might get passed
45e34800
NC
6980 into an hv routine with a regular hash.
6981 Similarly, a hash that isn't using shared hash keys has to have
6982 the flag in every key so that we know not to try to call
6983 share_hek_kek on it. */
bd08039b 6984
b64e5050 6985 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6986 if (HEK_UTF8(hek))
6987 SvUTF8_on (sv);
6988 return sv;
6989 }
6990 /* This will be overwhelminly the most common case. */
409dfe77
NC
6991 {
6992 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6993 more efficient than sharepvn(). */
6994 SV *sv;
6995
6996 new_SV(sv);
6997 sv_upgrade(sv, SVt_PV);
6998 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
6999 SvCUR_set(sv, HEK_LEN(hek));
7000 SvLEN_set(sv, 0);
7001 SvREADONLY_on(sv);
7002 SvFAKE_on(sv);
7003 SvPOK_on(sv);
7004 if (HEK_UTF8(hek))
7005 SvUTF8_on(sv);
7006 return sv;
7007 }
bd08039b
NC
7008 }
7009}
7010
1c846c1f
NIS
7011/*
7012=for apidoc newSVpvn_share
7013
3f7c398e 7014Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7015table. If the string does not already exist in the table, it is created
7016first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7017slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7018otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7019is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7020hash lookup will avoid string compare.
1c846c1f
NIS
7021
7022=cut
7023*/
7024
7025SV *
c3654f1a 7026Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7027{
97aff369 7028 dVAR;
1c846c1f 7029 register SV *sv;
c3654f1a 7030 bool is_utf8 = FALSE;
a51caccf
NC
7031 const char *const orig_src = src;
7032
c3654f1a 7033 if (len < 0) {
77caf834 7034 STRLEN tmplen = -len;
c3654f1a 7035 is_utf8 = TRUE;
75a54232 7036 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7037 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7038 len = tmplen;
7039 }
1c846c1f 7040 if (!hash)
5afd6d42 7041 PERL_HASH(hash, src, len);
1c846c1f 7042 new_SV(sv);
bdd68bc3 7043 sv_upgrade(sv, SVt_PV);
f880fe2f 7044 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7045 SvCUR_set(sv, len);
b162af07 7046 SvLEN_set(sv, 0);
1c846c1f
NIS
7047 SvREADONLY_on(sv);
7048 SvFAKE_on(sv);
7049 SvPOK_on(sv);
c3654f1a
IH
7050 if (is_utf8)
7051 SvUTF8_on(sv);
a51caccf
NC
7052 if (src != orig_src)
7053 Safefree(src);
1c846c1f
NIS
7054 return sv;
7055}
7056
645c22ef 7057
cea2e8a9 7058#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7059
7060/* pTHX_ magic can't cope with varargs, so this is a no-context
7061 * version of the main function, (which may itself be aliased to us).
7062 * Don't access this version directly.
7063 */
7064
46fc3d4c 7065SV *
cea2e8a9 7066Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7067{
cea2e8a9 7068 dTHX;
46fc3d4c 7069 register SV *sv;
7070 va_list args;
46fc3d4c 7071 va_start(args, pat);
c5be433b 7072 sv = vnewSVpvf(pat, &args);
46fc3d4c 7073 va_end(args);
7074 return sv;
7075}
cea2e8a9 7076#endif
46fc3d4c 7077
954c1994
GS
7078/*
7079=for apidoc newSVpvf
7080
645c22ef 7081Creates a new SV and initializes it with the string formatted like
954c1994
GS
7082C<sprintf>.
7083
7084=cut
7085*/
7086
cea2e8a9
GS
7087SV *
7088Perl_newSVpvf(pTHX_ const char* pat, ...)
7089{
7090 register SV *sv;
7091 va_list args;
cea2e8a9 7092 va_start(args, pat);
c5be433b 7093 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7094 va_end(args);
7095 return sv;
7096}
46fc3d4c 7097
645c22ef
DM
7098/* backend for newSVpvf() and newSVpvf_nocontext() */
7099
79072805 7100SV *
c5be433b
GS
7101Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7102{
97aff369 7103 dVAR;
c5be433b
GS
7104 register SV *sv;
7105 new_SV(sv);
4608196e 7106 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7107 return sv;
7108}
7109
954c1994
GS
7110/*
7111=for apidoc newSVnv
7112
7113Creates a new SV and copies a floating point value into it.
7114The reference count for the SV is set to 1.
7115
7116=cut
7117*/
7118
c5be433b 7119SV *
65202027 7120Perl_newSVnv(pTHX_ NV n)
79072805 7121{
97aff369 7122 dVAR;
463ee0b2 7123 register SV *sv;
79072805 7124
4561caa4 7125 new_SV(sv);
79072805
LW
7126 sv_setnv(sv,n);
7127 return sv;
7128}
7129
954c1994
GS
7130/*
7131=for apidoc newSViv
7132
7133Creates a new SV and copies an integer into it. The reference count for the
7134SV is set to 1.
7135
7136=cut
7137*/
7138
79072805 7139SV *
864dbfa3 7140Perl_newSViv(pTHX_ IV i)
79072805 7141{
97aff369 7142 dVAR;
463ee0b2 7143 register SV *sv;
79072805 7144
4561caa4 7145 new_SV(sv);
79072805
LW
7146 sv_setiv(sv,i);
7147 return sv;
7148}
7149
954c1994 7150/*
1a3327fb
JH
7151=for apidoc newSVuv
7152
7153Creates a new SV and copies an unsigned integer into it.
7154The reference count for the SV is set to 1.
7155
7156=cut
7157*/
7158
7159SV *
7160Perl_newSVuv(pTHX_ UV u)
7161{
97aff369 7162 dVAR;
1a3327fb
JH
7163 register SV *sv;
7164
7165 new_SV(sv);
7166 sv_setuv(sv,u);
7167 return sv;
7168}
7169
7170/*
954c1994
GS
7171=for apidoc newRV_noinc
7172
7173Creates an RV wrapper for an SV. The reference count for the original
7174SV is B<not> incremented.
7175
7176=cut
7177*/
7178
2304df62 7179SV *
864dbfa3 7180Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7181{
97aff369 7182 dVAR;
2304df62
AD
7183 register SV *sv;
7184
4561caa4 7185 new_SV(sv);
2304df62 7186 sv_upgrade(sv, SVt_RV);
76e3520e 7187 SvTEMP_off(tmpRef);
b162af07 7188 SvRV_set(sv, tmpRef);
2304df62 7189 SvROK_on(sv);
2304df62
AD
7190 return sv;
7191}
7192
ff276b08 7193/* newRV_inc is the official function name to use now.
645c22ef
DM
7194 * newRV_inc is in fact #defined to newRV in sv.h
7195 */
7196
5f05dabc 7197SV *
7f466ec7 7198Perl_newRV(pTHX_ SV *sv)
5f05dabc 7199{
97aff369 7200 dVAR;
7f466ec7 7201 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7202}
5f05dabc 7203
954c1994
GS
7204/*
7205=for apidoc newSVsv
7206
7207Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7208(Uses C<sv_setsv>).
954c1994
GS
7209
7210=cut
7211*/
7212
79072805 7213SV *
864dbfa3 7214Perl_newSVsv(pTHX_ register SV *old)
79072805 7215{
97aff369 7216 dVAR;
463ee0b2 7217 register SV *sv;
79072805
LW
7218
7219 if (!old)
7a5b473e 7220 return NULL;
8990e307 7221 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7222 if (ckWARN_d(WARN_INTERNAL))
9014280d 7223 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7224 return NULL;
79072805 7225 }
4561caa4 7226 new_SV(sv);
e90aabeb
NC
7227 /* SV_GMAGIC is the default for sv_setv()
7228 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7229 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7230 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7231 return sv;
79072805
LW
7232}
7233
645c22ef
DM
7234/*
7235=for apidoc sv_reset
7236
7237Underlying implementation for the C<reset> Perl function.
7238Note that the perl-level function is vaguely deprecated.
7239
7240=cut
7241*/
7242
79072805 7243void
e1ec3a88 7244Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7245{
27da23d5 7246 dVAR;
4802d5d7 7247 char todo[PERL_UCHAR_MAX+1];
79072805 7248
49d8d3a1
MB
7249 if (!stash)
7250 return;
7251
79072805 7252 if (!*s) { /* reset ?? searches */
aec46f14 7253 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7254 if (mg) {
7255 PMOP *pm = (PMOP *) mg->mg_obj;
7256 while (pm) {
7257 pm->op_pmdynflags &= ~PMdf_USED;
7258 pm = pm->op_pmnext;
7259 }
79072805
LW
7260 }
7261 return;
7262 }
7263
7264 /* reset variables */
7265
7266 if (!HvARRAY(stash))
7267 return;
463ee0b2
LW
7268
7269 Zero(todo, 256, char);
79072805 7270 while (*s) {
b464bac0
AL
7271 I32 max;
7272 I32 i = (unsigned char)*s;
79072805
LW
7273 if (s[1] == '-') {
7274 s += 2;
7275 }
4802d5d7 7276 max = (unsigned char)*s++;
79072805 7277 for ( ; i <= max; i++) {
463ee0b2
LW
7278 todo[i] = 1;
7279 }
a0d0e21e 7280 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7281 HE *entry;
79072805 7282 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7283 entry;
7284 entry = HeNEXT(entry))
7285 {
b464bac0
AL
7286 register GV *gv;
7287 register SV *sv;
7288
1edc1566 7289 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7290 continue;
1edc1566 7291 gv = (GV*)HeVAL(entry);
79072805 7292 sv = GvSV(gv);
e203899d
NC
7293 if (sv) {
7294 if (SvTHINKFIRST(sv)) {
7295 if (!SvREADONLY(sv) && SvROK(sv))
7296 sv_unref(sv);
7297 /* XXX Is this continue a bug? Why should THINKFIRST
7298 exempt us from resetting arrays and hashes? */
7299 continue;
7300 }
7301 SvOK_off(sv);
7302 if (SvTYPE(sv) >= SVt_PV) {
7303 SvCUR_set(sv, 0);
bd61b366 7304 if (SvPVX_const(sv) != NULL)
e203899d
NC
7305 *SvPVX(sv) = '\0';
7306 SvTAINT(sv);
7307 }
79072805
LW
7308 }
7309 if (GvAV(gv)) {
7310 av_clear(GvAV(gv));
7311 }
bfcb3514 7312 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7313#if defined(VMS)
7314 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7315#else /* ! VMS */
463ee0b2 7316 hv_clear(GvHV(gv));
b0269e46
AB
7317# if defined(USE_ENVIRON_ARRAY)
7318 if (gv == PL_envgv)
7319 my_clearenv();
7320# endif /* USE_ENVIRON_ARRAY */
7321#endif /* VMS */
79072805
LW
7322 }
7323 }
7324 }
7325 }
7326}
7327
645c22ef
DM
7328/*
7329=for apidoc sv_2io
7330
7331Using various gambits, try to get an IO from an SV: the IO slot if its a
7332GV; or the recursive result if we're an RV; or the IO slot of the symbol
7333named after the PV if we're a string.
7334
7335=cut
7336*/
7337
46fc3d4c 7338IO*
864dbfa3 7339Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7340{
7341 IO* io;
7342 GV* gv;
7343
7344 switch (SvTYPE(sv)) {
7345 case SVt_PVIO:
7346 io = (IO*)sv;
7347 break;
7348 case SVt_PVGV:
7349 gv = (GV*)sv;
7350 io = GvIO(gv);
7351 if (!io)
cea2e8a9 7352 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7353 break;
7354 default:
7355 if (!SvOK(sv))
cea2e8a9 7356 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7357 if (SvROK(sv))
7358 return sv_2io(SvRV(sv));
f776e3cd 7359 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7360 if (gv)
7361 io = GvIO(gv);
7362 else
7363 io = 0;
7364 if (!io)
be2597df 7365 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
46fc3d4c 7366 break;
7367 }
7368 return io;
7369}
7370
645c22ef
DM
7371/*
7372=for apidoc sv_2cv
7373
7374Using various gambits, try to get a CV from an SV; in addition, try if
7375possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7376The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7377
7378=cut
7379*/
7380
79072805 7381CV *
864dbfa3 7382Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7383{
27da23d5 7384 dVAR;
a0714e2c 7385 GV *gv = NULL;
601f1833 7386 CV *cv = NULL;
79072805 7387
85dec29a
NC
7388 if (!sv) {
7389 *st = NULL;
7390 *gvp = NULL;
7391 return NULL;
7392 }
79072805 7393 switch (SvTYPE(sv)) {
79072805
LW
7394 case SVt_PVCV:
7395 *st = CvSTASH(sv);
a0714e2c 7396 *gvp = NULL;
79072805
LW
7397 return (CV*)sv;
7398 case SVt_PVHV:
7399 case SVt_PVAV:
ef58ba18 7400 *st = NULL;
a0714e2c 7401 *gvp = NULL;
601f1833 7402 return NULL;
8990e307
LW
7403 case SVt_PVGV:
7404 gv = (GV*)sv;
a0d0e21e 7405 *gvp = gv;
8990e307
LW
7406 *st = GvESTASH(gv);
7407 goto fix_gv;
7408
79072805 7409 default:
5b295bef 7410 SvGETMAGIC(sv);
a0d0e21e 7411 if (SvROK(sv)) {
823a54a3 7412 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7413 tryAMAGICunDEREF(to_cv);
7414
62f274bf
GS
7415 sv = SvRV(sv);
7416 if (SvTYPE(sv) == SVt_PVCV) {
7417 cv = (CV*)sv;
a0714e2c 7418 *gvp = NULL;
62f274bf
GS
7419 *st = CvSTASH(cv);
7420 return cv;
7421 }
7422 else if(isGV(sv))
7423 gv = (GV*)sv;
7424 else
cea2e8a9 7425 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7426 }
62f274bf 7427 else if (isGV(sv))
79072805
LW
7428 gv = (GV*)sv;
7429 else
7a5fd60d 7430 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7431 *gvp = gv;
ef58ba18
NC
7432 if (!gv) {
7433 *st = NULL;
601f1833 7434 return NULL;
ef58ba18 7435 }
e26df76a
NC
7436 /* Some flags to gv_fetchsv mean don't really create the GV */
7437 if (SvTYPE(gv) != SVt_PVGV) {
7438 *st = NULL;
7439 return NULL;
7440 }
79072805 7441 *st = GvESTASH(gv);
8990e307 7442 fix_gv:
8ebc5c01 7443 if (lref && !GvCVu(gv)) {
4633a7c4 7444 SV *tmpsv;
748a9306 7445 ENTER;
561b68a9 7446 tmpsv = newSV(0);
bd61b366 7447 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7448 /* XXX this is probably not what they think they're getting.
7449 * It has the same effect as "sub name;", i.e. just a forward
7450 * declaration! */
774d564b 7451 newSUB(start_subparse(FALSE, 0),
4633a7c4 7452 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7453 NULL, NULL);
748a9306 7454 LEAVE;
8ebc5c01 7455 if (!GvCVu(gv))
35c1215d 7456 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
be2597df 7457 SVfARG(sv));
8990e307 7458 }
8ebc5c01 7459 return GvCVu(gv);
79072805
LW
7460 }
7461}
7462
c461cf8f
JH
7463/*
7464=for apidoc sv_true
7465
7466Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7467Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7468instead use an in-line version.
c461cf8f
JH
7469
7470=cut
7471*/
7472
79072805 7473I32
864dbfa3 7474Perl_sv_true(pTHX_ register SV *sv)
79072805 7475{
8990e307
LW
7476 if (!sv)
7477 return 0;
79072805 7478 if (SvPOK(sv)) {
823a54a3
AL
7479 register const XPV* const tXpv = (XPV*)SvANY(sv);
7480 if (tXpv &&
c2f1de04 7481 (tXpv->xpv_cur > 1 ||
339049b0 7482 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7483 return 1;
7484 else
7485 return 0;
7486 }
7487 else {
7488 if (SvIOK(sv))
463ee0b2 7489 return SvIVX(sv) != 0;
79072805
LW
7490 else {
7491 if (SvNOK(sv))
463ee0b2 7492 return SvNVX(sv) != 0.0;
79072805 7493 else
463ee0b2 7494 return sv_2bool(sv);
79072805
LW
7495 }
7496 }
7497}
79072805 7498
645c22ef 7499/*
c461cf8f
JH
7500=for apidoc sv_pvn_force
7501
7502Get a sensible string out of the SV somehow.
645c22ef
DM
7503A private implementation of the C<SvPV_force> macro for compilers which
7504can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7505
8d6d96c1
HS
7506=for apidoc sv_pvn_force_flags
7507
7508Get a sensible string out of the SV somehow.
7509If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7510appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7511implemented in terms of this function.
645c22ef
DM
7512You normally want to use the various wrapper macros instead: see
7513C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7514
7515=cut
7516*/
7517
7518char *
7519Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7520{
97aff369 7521 dVAR;
6fc92669 7522 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7523 sv_force_normal_flags(sv, 0);
1c846c1f 7524
a0d0e21e 7525 if (SvPOK(sv)) {
13c5b33c
NC
7526 if (lp)
7527 *lp = SvCUR(sv);
a0d0e21e
LW
7528 }
7529 else {
a3b680e6 7530 char *s;
13c5b33c
NC
7531 STRLEN len;
7532
4d84ee25 7533 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7534 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7535 if (PL_op)
7536 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7537 ref, OP_NAME(PL_op));
4d84ee25 7538 else
b64e5050 7539 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7540 }
b64e5050 7541 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7542 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7543 OP_NAME(PL_op));
b64e5050 7544 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7545 if (lp)
7546 *lp = len;
7547
3f7c398e 7548 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7549 if (SvROK(sv))
7550 sv_unref(sv);
862a34c6 7551 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7552 SvGROW(sv, len + 1);
706aa1c9 7553 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7554 SvCUR_set(sv, len);
7555 *SvEND(sv) = '\0';
7556 }
7557 if (!SvPOK(sv)) {
7558 SvPOK_on(sv); /* validate pointer */
7559 SvTAINT(sv);
1d7c1841 7560 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7561 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7562 }
7563 }
4d84ee25 7564 return SvPVX_mutable(sv);
a0d0e21e
LW
7565}
7566
645c22ef 7567/*
645c22ef
DM
7568=for apidoc sv_pvbyten_force
7569
0feed65a 7570The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7571
7572=cut
7573*/
7574
7340a771
GS
7575char *
7576Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7577{
46ec2f14 7578 sv_pvn_force(sv,lp);
ffebcc3e 7579 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7580 *lp = SvCUR(sv);
7581 return SvPVX(sv);
7340a771
GS
7582}
7583
645c22ef 7584/*
c461cf8f
JH
7585=for apidoc sv_pvutf8n_force
7586
0feed65a 7587The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7588
7589=cut
7590*/
7591
7340a771
GS
7592char *
7593Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7594{
46ec2f14 7595 sv_pvn_force(sv,lp);
560a288e 7596 sv_utf8_upgrade(sv);
46ec2f14
TS
7597 *lp = SvCUR(sv);
7598 return SvPVX(sv);
7340a771
GS
7599}
7600
c461cf8f
JH
7601/*
7602=for apidoc sv_reftype
7603
7604Returns a string describing what the SV is a reference to.
7605
7606=cut
7607*/
7608
2b388283 7609const char *
bfed75c6 7610Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7611{
07409e01
NC
7612 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7613 inside return suggests a const propagation bug in g++. */
c86bf373 7614 if (ob && SvOBJECT(sv)) {
1b6737cc 7615 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7616 return name ? name : (char *) "__ANON__";
c86bf373 7617 }
a0d0e21e
LW
7618 else {
7619 switch (SvTYPE(sv)) {
7620 case SVt_NULL:
7621 case SVt_IV:
7622 case SVt_NV:
7623 case SVt_RV:
7624 case SVt_PV:
7625 case SVt_PVIV:
7626 case SVt_PVNV:
7627 case SVt_PVMG:
1cb0ed9b 7628 if (SvVOK(sv))
439cb1c4 7629 return "VSTRING";
a0d0e21e
LW
7630 if (SvROK(sv))
7631 return "REF";
7632 else
7633 return "SCALAR";
1cb0ed9b 7634
07409e01 7635 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7636 /* tied lvalues should appear to be
7637 * scalars for backwards compatitbility */
7638 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7639 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7640 case SVt_PVAV: return "ARRAY";
7641 case SVt_PVHV: return "HASH";
7642 case SVt_PVCV: return "CODE";
7643 case SVt_PVGV: return "GLOB";
1d2dff63 7644 case SVt_PVFM: return "FORMAT";
27f9d8f3 7645 case SVt_PVIO: return "IO";
cecf5685 7646 case SVt_BIND: return "BIND";
a0d0e21e
LW
7647 default: return "UNKNOWN";
7648 }
7649 }
7650}
7651
954c1994
GS
7652/*
7653=for apidoc sv_isobject
7654
7655Returns a boolean indicating whether the SV is an RV pointing to a blessed
7656object. If the SV is not an RV, or if the object is not blessed, then this
7657will return false.
7658
7659=cut
7660*/
7661
463ee0b2 7662int
864dbfa3 7663Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7664{
68dc0745 7665 if (!sv)
7666 return 0;
5b295bef 7667 SvGETMAGIC(sv);
85e6fe83
LW
7668 if (!SvROK(sv))
7669 return 0;
7670 sv = (SV*)SvRV(sv);
7671 if (!SvOBJECT(sv))
7672 return 0;
7673 return 1;
7674}
7675
954c1994
GS
7676/*
7677=for apidoc sv_isa
7678
7679Returns a boolean indicating whether the SV is blessed into the specified
7680class. This does not check for subtypes; use C<sv_derived_from> to verify
7681an inheritance relationship.
7682
7683=cut
7684*/
7685
85e6fe83 7686int
864dbfa3 7687Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7688{
bfcb3514 7689 const char *hvname;
68dc0745 7690 if (!sv)
7691 return 0;
5b295bef 7692 SvGETMAGIC(sv);
ed6116ce 7693 if (!SvROK(sv))
463ee0b2 7694 return 0;
ed6116ce
LW
7695 sv = (SV*)SvRV(sv);
7696 if (!SvOBJECT(sv))
463ee0b2 7697 return 0;
bfcb3514
NC
7698 hvname = HvNAME_get(SvSTASH(sv));
7699 if (!hvname)
e27ad1f2 7700 return 0;
463ee0b2 7701
bfcb3514 7702 return strEQ(hvname, name);
463ee0b2
LW
7703}
7704
954c1994
GS
7705/*
7706=for apidoc newSVrv
7707
7708Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7709it will be upgraded to one. If C<classname> is non-null then the new SV will
7710be blessed in the specified package. The new SV is returned and its
7711reference count is 1.
7712
7713=cut
7714*/
7715
463ee0b2 7716SV*
864dbfa3 7717Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7718{
97aff369 7719 dVAR;
463ee0b2
LW
7720 SV *sv;
7721
4561caa4 7722 new_SV(sv);
51cf62d8 7723
765f542d 7724 SV_CHECK_THINKFIRST_COW_DROP(rv);
52944de8 7725 (void)SvAMAGIC_off(rv);
51cf62d8 7726
0199fce9 7727 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7728 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7729 SvREFCNT(rv) = 0;
7730 sv_clear(rv);
7731 SvFLAGS(rv) = 0;
7732 SvREFCNT(rv) = refcnt;
0199fce9 7733
dc5494d2
NC
7734 sv_upgrade(rv, SVt_RV);
7735 } else if (SvROK(rv)) {
7736 SvREFCNT_dec(SvRV(rv));
7737 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7738 sv_upgrade(rv, SVt_RV);
7739 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7740 SvPV_free(rv);
0199fce9
JD
7741 SvCUR_set(rv, 0);
7742 SvLEN_set(rv, 0);
7743 }
51cf62d8 7744
0c34ef67 7745 SvOK_off(rv);
b162af07 7746 SvRV_set(rv, sv);
ed6116ce 7747 SvROK_on(rv);
463ee0b2 7748
a0d0e21e 7749 if (classname) {
da51bb9b 7750 HV* const stash = gv_stashpv(classname, GV_ADD);
a0d0e21e
LW
7751 (void)sv_bless(rv, stash);
7752 }
7753 return sv;
7754}
7755
954c1994
GS
7756/*
7757=for apidoc sv_setref_pv
7758
7759Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7760argument will be upgraded to an RV. That RV will be modified to point to
7761the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7762into the SV. The C<classname> argument indicates the package for the
bd61b366 7763blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7764will have a reference count of 1, and the RV will be returned.
954c1994
GS
7765
7766Do not use with other Perl types such as HV, AV, SV, CV, because those
7767objects will become corrupted by the pointer copy process.
7768
7769Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7770
7771=cut
7772*/
7773
a0d0e21e 7774SV*
864dbfa3 7775Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7776{
97aff369 7777 dVAR;
189b2af5 7778 if (!pv) {
3280af22 7779 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7780 SvSETMAGIC(rv);
7781 }
a0d0e21e 7782 else
56431972 7783 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7784 return rv;
7785}
7786
954c1994
GS
7787/*
7788=for apidoc sv_setref_iv
7789
7790Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7791argument will be upgraded to an RV. That RV will be modified to point to
7792the new SV. The C<classname> argument indicates the package for the
bd61b366 7793blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7794will have a reference count of 1, and the RV will be returned.
954c1994
GS
7795
7796=cut
7797*/
7798
a0d0e21e 7799SV*
864dbfa3 7800Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7801{
7802 sv_setiv(newSVrv(rv,classname), iv);
7803 return rv;
7804}
7805
954c1994 7806/*
e1c57cef
JH
7807=for apidoc sv_setref_uv
7808
7809Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7810argument will be upgraded to an RV. That RV will be modified to point to
7811the new SV. The C<classname> argument indicates the package for the
bd61b366 7812blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7813will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7814
7815=cut
7816*/
7817
7818SV*
7819Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7820{
7821 sv_setuv(newSVrv(rv,classname), uv);
7822 return rv;
7823}
7824
7825/*
954c1994
GS
7826=for apidoc sv_setref_nv
7827
7828Copies a double into a new SV, optionally blessing the SV. The C<rv>
7829argument will be upgraded to an RV. That RV will be modified to point to
7830the new SV. The C<classname> argument indicates the package for the
bd61b366 7831blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7832will have a reference count of 1, and the RV will be returned.
954c1994
GS
7833
7834=cut
7835*/
7836
a0d0e21e 7837SV*
65202027 7838Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7839{
7840 sv_setnv(newSVrv(rv,classname), nv);
7841 return rv;
7842}
463ee0b2 7843
954c1994
GS
7844/*
7845=for apidoc sv_setref_pvn
7846
7847Copies a string into a new SV, optionally blessing the SV. The length of the
7848string must be specified with C<n>. The C<rv> argument will be upgraded to
7849an RV. That RV will be modified to point to the new SV. The C<classname>
7850argument indicates the package for the blessing. Set C<classname> to
bd61b366 7851C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7852of 1, and the RV will be returned.
954c1994
GS
7853
7854Note that C<sv_setref_pv> copies the pointer while this copies the string.
7855
7856=cut
7857*/
7858
a0d0e21e 7859SV*
1b6737cc 7860Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7861{
7862 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7863 return rv;
7864}
7865
954c1994
GS
7866/*
7867=for apidoc sv_bless
7868
7869Blesses an SV into a specified package. The SV must be an RV. The package
7870must be designated by its stash (see C<gv_stashpv()>). The reference count
7871of the SV is unaffected.
7872
7873=cut
7874*/
7875
a0d0e21e 7876SV*
864dbfa3 7877Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7878{
97aff369 7879 dVAR;
76e3520e 7880 SV *tmpRef;
a0d0e21e 7881 if (!SvROK(sv))
cea2e8a9 7882 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7883 tmpRef = SvRV(sv);
7884 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7885 if (SvREADONLY(tmpRef))
cea2e8a9 7886 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7887 if (SvOBJECT(tmpRef)) {
7888 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7889 --PL_sv_objcount;
76e3520e 7890 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7891 }
a0d0e21e 7892 }
76e3520e
GS
7893 SvOBJECT_on(tmpRef);
7894 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7895 ++PL_sv_objcount;
862a34c6 7896 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7897 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7898
2e3febc6
CS
7899 if (Gv_AMG(stash))
7900 SvAMAGIC_on(sv);
7901 else
52944de8 7902 (void)SvAMAGIC_off(sv);
a0d0e21e 7903
1edbfb88
AB
7904 if(SvSMAGICAL(tmpRef))
7905 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7906 mg_set(tmpRef);
7907
7908
ecdeb87c 7909
a0d0e21e
LW
7910 return sv;
7911}
7912
645c22ef 7913/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7914 */
7915
76e3520e 7916STATIC void
cea2e8a9 7917S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7918{
97aff369 7919 dVAR;
850fabdf 7920 void *xpvmg;
b37c2d43 7921 SV * const temp = sv_newmortal();
850fabdf 7922
a0d0e21e
LW
7923 assert(SvTYPE(sv) == SVt_PVGV);
7924 SvFAKE_off(sv);
180488f8
NC
7925 gv_efullname3(temp, (GV *) sv, "*");
7926
f7877b28 7927 if (GvGP(sv)) {
1edc1566 7928 gp_free((GV*)sv);
f7877b28 7929 }
e826b3c7 7930 if (GvSTASH(sv)) {
e15faf7d 7931 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7932 GvSTASH(sv) = NULL;
e826b3c7 7933 }
a5f75d66 7934 GvMULTI_off(sv);
acda4c6a
NC
7935 if (GvNAME_HEK(sv)) {
7936 unshare_hek(GvNAME_HEK(sv));
7937 }
2e5b91de 7938 isGV_with_GP_off(sv);
850fabdf
GS
7939
7940 /* need to keep SvANY(sv) in the right arena */
7941 xpvmg = new_XPVMG();
7942 StructCopy(SvANY(sv), xpvmg, XPVMG);
7943 del_XPVGV(SvANY(sv));
7944 SvANY(sv) = xpvmg;
7945
a0d0e21e
LW
7946 SvFLAGS(sv) &= ~SVTYPEMASK;
7947 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7948
7949 /* Intentionally not calling any local SET magic, as this isn't so much a
7950 set operation as merely an internal storage change. */
7951 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7952}
7953
954c1994 7954/*
840a7b70 7955=for apidoc sv_unref_flags
954c1994
GS
7956
7957Unsets the RV status of the SV, and decrements the reference count of
7958whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7959as a reversal of C<newSVrv>. The C<cflags> argument can contain
7960C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7961(otherwise the decrementing is conditional on the reference count being
7962different from one or the reference being a readonly SV).
7889fe52 7963See C<SvROK_off>.
954c1994
GS
7964
7965=cut
7966*/
7967
ed6116ce 7968void
e15faf7d 7969Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7970{
b64e5050 7971 SV* const target = SvRV(ref);
810b8aa5 7972
e15faf7d
NC
7973 if (SvWEAKREF(ref)) {
7974 sv_del_backref(target, ref);
7975 SvWEAKREF_off(ref);
7976 SvRV_set(ref, NULL);
810b8aa5
GS
7977 return;
7978 }
e15faf7d
NC
7979 SvRV_set(ref, NULL);
7980 SvROK_off(ref);
7981 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7982 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7983 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7984 SvREFCNT_dec(target);
840a7b70 7985 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7986 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7987}
8990e307 7988
840a7b70 7989/*
645c22ef
DM
7990=for apidoc sv_untaint
7991
7992Untaint an SV. Use C<SvTAINTED_off> instead.
7993=cut
7994*/
7995
bbce6d69 7996void
864dbfa3 7997Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7998{
13f57bf8 7999 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 8000 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 8001 if (mg)
565764a8 8002 mg->mg_len &= ~1;
36477c24 8003 }
bbce6d69 8004}
8005
645c22ef
DM
8006/*
8007=for apidoc sv_tainted
8008
8009Test an SV for taintedness. Use C<SvTAINTED> instead.
8010=cut
8011*/
8012
bbce6d69 8013bool
864dbfa3 8014Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8015{
13f57bf8 8016 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8017 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8018 if (mg && (mg->mg_len & 1) )
36477c24 8019 return TRUE;
8020 }
8021 return FALSE;
bbce6d69 8022}
8023
09540bc3
JH
8024/*
8025=for apidoc sv_setpviv
8026
8027Copies an integer into the given SV, also updating its string value.
8028Does not handle 'set' magic. See C<sv_setpviv_mg>.
8029
8030=cut
8031*/
8032
8033void
8034Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8035{
8036 char buf[TYPE_CHARS(UV)];
8037 char *ebuf;
b64e5050 8038 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8039
8040 sv_setpvn(sv, ptr, ebuf - ptr);
8041}
8042
8043/*
8044=for apidoc sv_setpviv_mg
8045
8046Like C<sv_setpviv>, but also handles 'set' magic.
8047
8048=cut
8049*/
8050
8051void
8052Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8053{
df7eb254 8054 sv_setpviv(sv, iv);
09540bc3
JH
8055 SvSETMAGIC(sv);
8056}
8057
cea2e8a9 8058#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8059
8060/* pTHX_ magic can't cope with varargs, so this is a no-context
8061 * version of the main function, (which may itself be aliased to us).
8062 * Don't access this version directly.
8063 */
8064
cea2e8a9
GS
8065void
8066Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8067{
8068 dTHX;
8069 va_list args;
8070 va_start(args, pat);
c5be433b 8071 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8072 va_end(args);
8073}
8074
645c22ef
DM
8075/* pTHX_ magic can't cope with varargs, so this is a no-context
8076 * version of the main function, (which may itself be aliased to us).
8077 * Don't access this version directly.
8078 */
cea2e8a9
GS
8079
8080void
8081Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8082{
8083 dTHX;
8084 va_list args;
8085 va_start(args, pat);
c5be433b 8086 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8087 va_end(args);
cea2e8a9
GS
8088}
8089#endif
8090
954c1994
GS
8091/*
8092=for apidoc sv_setpvf
8093
bffc3d17
SH
8094Works like C<sv_catpvf> but copies the text into the SV instead of
8095appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8096
8097=cut
8098*/
8099
46fc3d4c 8100void
864dbfa3 8101Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8102{
8103 va_list args;
46fc3d4c 8104 va_start(args, pat);
c5be433b 8105 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8106 va_end(args);
8107}
8108
bffc3d17
SH
8109/*
8110=for apidoc sv_vsetpvf
8111
8112Works like C<sv_vcatpvf> but copies the text into the SV instead of
8113appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8114
8115Usually used via its frontend C<sv_setpvf>.
8116
8117=cut
8118*/
645c22ef 8119
c5be433b
GS
8120void
8121Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8122{
4608196e 8123 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8124}
ef50df4b 8125
954c1994
GS
8126/*
8127=for apidoc sv_setpvf_mg
8128
8129Like C<sv_setpvf>, but also handles 'set' magic.
8130
8131=cut
8132*/
8133
ef50df4b 8134void
864dbfa3 8135Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8136{
8137 va_list args;
ef50df4b 8138 va_start(args, pat);
c5be433b 8139 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8140 va_end(args);
c5be433b
GS
8141}
8142
bffc3d17
SH
8143/*
8144=for apidoc sv_vsetpvf_mg
8145
8146Like C<sv_vsetpvf>, but also handles 'set' magic.
8147
8148Usually used via its frontend C<sv_setpvf_mg>.
8149
8150=cut
8151*/
645c22ef 8152
c5be433b
GS
8153void
8154Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8155{
4608196e 8156 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8157 SvSETMAGIC(sv);
8158}
8159
cea2e8a9 8160#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8161
8162/* pTHX_ magic can't cope with varargs, so this is a no-context
8163 * version of the main function, (which may itself be aliased to us).
8164 * Don't access this version directly.
8165 */
8166
cea2e8a9
GS
8167void
8168Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8169{
8170 dTHX;
8171 va_list args;
8172 va_start(args, pat);
c5be433b 8173 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8174 va_end(args);
8175}
8176
645c22ef
DM
8177/* pTHX_ magic can't cope with varargs, so this is a no-context
8178 * version of the main function, (which may itself be aliased to us).
8179 * Don't access this version directly.
8180 */
8181
cea2e8a9
GS
8182void
8183Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8184{
8185 dTHX;
8186 va_list args;
8187 va_start(args, pat);
c5be433b 8188 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8189 va_end(args);
cea2e8a9
GS
8190}
8191#endif
8192
954c1994
GS
8193/*
8194=for apidoc sv_catpvf
8195
d5ce4a7c
GA
8196Processes its arguments like C<sprintf> and appends the formatted
8197output to an SV. If the appended data contains "wide" characters
8198(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8199and characters >255 formatted with %c), the original SV might get
bffc3d17 8200upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8201C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8202valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8203
d5ce4a7c 8204=cut */
954c1994 8205
46fc3d4c 8206void
864dbfa3 8207Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8208{
8209 va_list args;
46fc3d4c 8210 va_start(args, pat);
c5be433b 8211 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8212 va_end(args);
8213}
8214
bffc3d17
SH
8215/*
8216=for apidoc sv_vcatpvf
8217
8218Processes its arguments like C<vsprintf> and appends the formatted output
8219to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8220
8221Usually used via its frontend C<sv_catpvf>.
8222
8223=cut
8224*/
645c22ef 8225
ef50df4b 8226void
c5be433b
GS
8227Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8228{
4608196e 8229 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8230}
8231
954c1994
GS
8232/*
8233=for apidoc sv_catpvf_mg
8234
8235Like C<sv_catpvf>, but also handles 'set' magic.
8236
8237=cut
8238*/
8239
c5be433b 8240void
864dbfa3 8241Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8242{
8243 va_list args;
ef50df4b 8244 va_start(args, pat);
c5be433b 8245 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8246 va_end(args);
c5be433b
GS
8247}
8248
bffc3d17
SH
8249/*
8250=for apidoc sv_vcatpvf_mg
8251
8252Like C<sv_vcatpvf>, but also handles 'set' magic.
8253
8254Usually used via its frontend C<sv_catpvf_mg>.
8255
8256=cut
8257*/
645c22ef 8258
c5be433b
GS
8259void
8260Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8261{
4608196e 8262 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8263 SvSETMAGIC(sv);
8264}
8265
954c1994
GS
8266/*
8267=for apidoc sv_vsetpvfn
8268
bffc3d17 8269Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8270appending it.
8271
bffc3d17 8272Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8273
954c1994
GS
8274=cut
8275*/
8276
46fc3d4c 8277void
7d5ea4e7 8278Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8279{
8280 sv_setpvn(sv, "", 0);
7d5ea4e7 8281 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8282}
8283
2d00ba3b 8284STATIC I32
9dd79c3f 8285S_expect_number(pTHX_ char** pattern)
211dfcf1 8286{
97aff369 8287 dVAR;
211dfcf1
HS
8288 I32 var = 0;
8289 switch (**pattern) {
8290 case '1': case '2': case '3':
8291 case '4': case '5': case '6':
8292 case '7': case '8': case '9':
2fba7546
GA
8293 var = *(*pattern)++ - '0';
8294 while (isDIGIT(**pattern)) {
5f66b61c 8295 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8296 if (tmp < var)
8297 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8298 var = tmp;
8299 }
211dfcf1
HS
8300 }
8301 return var;
8302}
211dfcf1 8303
c445ea15
AL
8304STATIC char *
8305S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8306{
a3b680e6 8307 const int neg = nv < 0;
4151a5fe 8308 UV uv;
4151a5fe
IZ
8309
8310 if (neg)
8311 nv = -nv;
8312 if (nv < UV_MAX) {
b464bac0 8313 char *p = endbuf;
4151a5fe 8314 nv += 0.5;
028f8eaa 8315 uv = (UV)nv;
4151a5fe
IZ
8316 if (uv & 1 && uv == nv)
8317 uv--; /* Round to even */
8318 do {
a3b680e6 8319 const unsigned dig = uv % 10;
4151a5fe
IZ
8320 *--p = '0' + dig;
8321 } while (uv /= 10);
8322 if (neg)
8323 *--p = '-';
8324 *len = endbuf - p;
8325 return p;
8326 }
bd61b366 8327 return NULL;
4151a5fe
IZ
8328}
8329
8330
954c1994
GS
8331/*
8332=for apidoc sv_vcatpvfn
8333
8334Processes its arguments like C<vsprintf> and appends the formatted output
8335to an SV. Uses an array of SVs if the C style variable argument list is
8336missing (NULL). When running with taint checks enabled, indicates via
8337C<maybe_tainted> if results are untrustworthy (often due to the use of
8338locales).
8339
bffc3d17 8340Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8341
954c1994
GS
8342=cut
8343*/
8344
8896765a
RB
8345
8346#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8347 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8348 vec_utf8 = DO_UTF8(vecsv);
8349
1ef29b0e
RGS
8350/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8351
46fc3d4c 8352void
7d5ea4e7 8353Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8354{
97aff369 8355 dVAR;
46fc3d4c 8356 char *p;
8357 char *q;
a3b680e6 8358 const char *patend;
fc36a67e 8359 STRLEN origlen;
46fc3d4c 8360 I32 svix = 0;
27da23d5 8361 static const char nullstr[] = "(null)";
a0714e2c 8362 SV *argsv = NULL;
b464bac0
AL
8363 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8364 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8365 SV *nsv = NULL;
4151a5fe
IZ
8366 /* Times 4: a decimal digit takes more than 3 binary digits.
8367 * NV_DIG: mantissa takes than many decimal digits.
8368 * Plus 32: Playing safe. */
8369 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8370 /* large enough for "%#.#f" --chip */
8371 /* what about long double NVs? --jhi */
db79b45b 8372
53c1dcc0
AL
8373 PERL_UNUSED_ARG(maybe_tainted);
8374
46fc3d4c 8375 /* no matter what, this is a string now */
fc36a67e 8376 (void)SvPV_force(sv, origlen);
46fc3d4c 8377
8896765a 8378 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8379 if (patlen == 0)
8380 return;
0dbb1585 8381 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8382 if (args) {
8383 const char * const s = va_arg(*args, char*);
8384 sv_catpv(sv, s ? s : nullstr);
8385 }
8386 else if (svix < svmax) {
8387 sv_catsv(sv, *svargs);
2d03de9c
AL
8388 }
8389 return;
0dbb1585 8390 }
8896765a
RB
8391 if (args && patlen == 3 && pat[0] == '%' &&
8392 pat[1] == '-' && pat[2] == 'p') {
6c9570dc 8393 argsv = (SV*)va_arg(*args, void*);
8896765a 8394 sv_catsv(sv, argsv);
8896765a 8395 return;
46fc3d4c 8396 }
8397
1d917b39 8398#ifndef USE_LONG_DOUBLE
4151a5fe 8399 /* special-case "%.<number>[gf]" */
7af36d83 8400 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8401 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8402 unsigned digits = 0;
8403 const char *pp;
8404
8405 pp = pat + 2;
8406 while (*pp >= '0' && *pp <= '9')
8407 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8408 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8409 NV nv;
8410
7af36d83 8411 if (svix < svmax)
4151a5fe
IZ
8412 nv = SvNV(*svargs);
8413 else
8414 return;
8415 if (*pp == 'g') {
2873255c
NC
8416 /* Add check for digits != 0 because it seems that some
8417 gconverts are buggy in this case, and we don't yet have
8418 a Configure test for this. */
8419 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8420 /* 0, point, slack */
2e59c212 8421 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8422 sv_catpv(sv, ebuf);
8423 if (*ebuf) /* May return an empty string for digits==0 */
8424 return;
8425 }
8426 } else if (!digits) {
8427 STRLEN l;
8428
8429 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8430 sv_catpvn(sv, p, l);
8431 return;
8432 }
8433 }
8434 }
8435 }
1d917b39 8436#endif /* !USE_LONG_DOUBLE */
4151a5fe 8437
2cf2cfc6 8438 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8439 has_utf8 = TRUE;
2cf2cfc6 8440
46fc3d4c 8441 patend = (char*)pat + patlen;
8442 for (p = (char*)pat; p < patend; p = q) {
8443 bool alt = FALSE;
8444 bool left = FALSE;
b22c7a20 8445 bool vectorize = FALSE;
211dfcf1 8446 bool vectorarg = FALSE;
2cf2cfc6 8447 bool vec_utf8 = FALSE;
46fc3d4c 8448 char fill = ' ';
8449 char plus = 0;
8450 char intsize = 0;
8451 STRLEN width = 0;
fc36a67e 8452 STRLEN zeros = 0;
46fc3d4c 8453 bool has_precis = FALSE;
8454 STRLEN precis = 0;
c445ea15 8455 const I32 osvix = svix;
2cf2cfc6 8456 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8457#ifdef HAS_LDBL_SPRINTF_BUG
8458 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8459 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8460 bool fix_ldbl_sprintf_bug = FALSE;
8461#endif
205f51d8 8462
46fc3d4c 8463 char esignbuf[4];
89ebb4a3 8464 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8465 STRLEN esignlen = 0;
8466
bd61b366 8467 const char *eptr = NULL;
fc36a67e 8468 STRLEN elen = 0;
a0714e2c 8469 SV *vecsv = NULL;
4608196e 8470 const U8 *vecstr = NULL;
b22c7a20 8471 STRLEN veclen = 0;
934abaf1 8472 char c = 0;
46fc3d4c 8473 int i;
9c5ffd7c 8474 unsigned base = 0;
8c8eb53c
RB
8475 IV iv = 0;
8476 UV uv = 0;
9e5b023a
JH
8477 /* we need a long double target in case HAS_LONG_DOUBLE but
8478 not USE_LONG_DOUBLE
8479 */
35fff930 8480#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8481 long double nv;
8482#else
65202027 8483 NV nv;
9e5b023a 8484#endif
46fc3d4c 8485 STRLEN have;
8486 STRLEN need;
8487 STRLEN gap;
7af36d83 8488 const char *dotstr = ".";
b22c7a20 8489 STRLEN dotstrlen = 1;
211dfcf1 8490 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8491 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8492 I32 epix = 0; /* explicit precision index */
8493 I32 evix = 0; /* explicit vector index */
eb3fce90 8494 bool asterisk = FALSE;
46fc3d4c 8495
211dfcf1 8496 /* echo everything up to the next format specification */
46fc3d4c 8497 for (q = p; q < patend && *q != '%'; ++q) ;
8498 if (q > p) {
db79b45b
JH
8499 if (has_utf8 && !pat_utf8)
8500 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8501 else
8502 sv_catpvn(sv, p, q - p);
46fc3d4c 8503 p = q;
8504 }
8505 if (q++ >= patend)
8506 break;
8507
211dfcf1
HS
8508/*
8509 We allow format specification elements in this order:
8510 \d+\$ explicit format parameter index
8511 [-+ 0#]+ flags
a472f209 8512 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8513 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8514 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8515 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8516 [hlqLV] size
8896765a
RB
8517 [%bcdefginopsuxDFOUX] format (mandatory)
8518*/
8519
8520 if (args) {
8521/*
8522 As of perl5.9.3, printf format checking is on by default.
8523 Internally, perl uses %p formats to provide an escape to
8524 some extended formatting. This block deals with those
8525 extensions: if it does not match, (char*)q is reset and
8526 the normal format processing code is used.
8527
8528 Currently defined extensions are:
8529 %p include pointer address (standard)
8530 %-p (SVf) include an SV (previously %_)
8531 %-<num>p include an SV with precision <num>
8532 %1p (VDf) include a v-string (as %vd)
8533 %<num>p reserved for future extensions
8534
8535 Robin Barker 2005-07-14
211dfcf1 8536*/
8896765a
RB
8537 char* r = q;
8538 bool sv = FALSE;
8539 STRLEN n = 0;
8540 if (*q == '-')
8541 sv = *q++;
c445ea15 8542 n = expect_number(&q);
8896765a
RB
8543 if (*q++ == 'p') {
8544 if (sv) { /* SVf */
8545 if (n) {
8546 precis = n;
8547 has_precis = TRUE;
8548 }
6c9570dc 8549 argsv = (SV*)va_arg(*args, void*);
8896765a
RB
8550 eptr = SvPVx_const(argsv, elen);
8551 if (DO_UTF8(argsv))
8552 is_utf8 = TRUE;
8553 goto string;
8554 }
8555#if vdNUMBER
8556 else if (n == vdNUMBER) { /* VDf */
8557 vectorize = TRUE;
8558 VECTORIZE_ARGS
8559 goto format_vd;
8560 }
8561#endif
8562 else if (n) {
8563 if (ckWARN_d(WARN_INTERNAL))
8564 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8565 "internal %%<num>p might conflict with future printf extensions");
8566 }
8567 }
8568 q = r;
8569 }
8570
c445ea15 8571 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8572 if (*q == '$') {
8573 ++q;
8574 efix = width;
8575 } else {
8576 goto gotwidth;
8577 }
8578 }
8579
fc36a67e 8580 /* FLAGS */
8581
46fc3d4c 8582 while (*q) {
8583 switch (*q) {
8584 case ' ':
8585 case '+':
9911cee9
TS
8586 if (plus == '+' && *q == ' ') /* '+' over ' ' */
8587 q++;
8588 else
8589 plus = *q++;
46fc3d4c 8590 continue;
8591
8592 case '-':
8593 left = TRUE;
8594 q++;
8595 continue;
8596
8597 case '0':
8598 fill = *q++;
8599 continue;
8600
8601 case '#':
8602 alt = TRUE;
8603 q++;
8604 continue;
8605
fc36a67e 8606 default:
8607 break;
8608 }
8609 break;
8610 }
46fc3d4c 8611
211dfcf1 8612 tryasterisk:
eb3fce90 8613 if (*q == '*') {
211dfcf1 8614 q++;
c445ea15 8615 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8616 if (*q++ != '$')
8617 goto unknown;
eb3fce90 8618 asterisk = TRUE;
211dfcf1
HS
8619 }
8620 if (*q == 'v') {
eb3fce90 8621 q++;
211dfcf1
HS
8622 if (vectorize)
8623 goto unknown;
9cbac4c7 8624 if ((vectorarg = asterisk)) {
211dfcf1
HS
8625 evix = ewix;
8626 ewix = 0;
8627 asterisk = FALSE;
8628 }
8629 vectorize = TRUE;
8630 goto tryasterisk;
eb3fce90
JH
8631 }
8632
211dfcf1 8633 if (!asterisk)
858a90f9 8634 {
7a5fa8a2 8635 if( *q == '0' )
f3583277 8636 fill = *q++;
c445ea15 8637 width = expect_number(&q);
858a90f9 8638 }
211dfcf1
HS
8639
8640 if (vectorize) {
8641 if (vectorarg) {
8642 if (args)
8643 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8644 else if (evix) {
8645 vecsv = (evix > 0 && evix <= svmax)
8646 ? svargs[evix-1] : &PL_sv_undef;
8647 } else {
8648 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8649 }
245d4a47 8650 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8651 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8652 bad with tied or overloaded values that return UTF8. */
211dfcf1 8653 if (DO_UTF8(vecsv))
2cf2cfc6 8654 is_utf8 = TRUE;
640283f5
NC
8655 else if (has_utf8) {
8656 vecsv = sv_mortalcopy(vecsv);
8657 sv_utf8_upgrade(vecsv);
8658 dotstr = SvPV_const(vecsv, dotstrlen);
8659 is_utf8 = TRUE;
8660 }
211dfcf1
HS
8661 }
8662 if (args) {
8896765a 8663 VECTORIZE_ARGS
eb3fce90 8664 }
7ad96abb 8665 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8666 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8667 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8668 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8669
8670 /* if this is a version object, we need to convert
8671 * back into v-string notation and then let the
8672 * vectorize happen normally
d7aa5382 8673 */
96b8f7ce
JP
8674 if (sv_derived_from(vecsv, "version")) {
8675 char *version = savesvpv(vecsv);
34ba6322
SP
8676 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8677 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8678 "vector argument not supported with alpha versions");
8679 goto unknown;
8680 }
96b8f7ce
JP
8681 vecsv = sv_newmortal();
8682 /* scan_vstring is expected to be called during
8683 * tokenization, so we need to fake up the end
8684 * of the buffer for it
8685 */
8686 PL_bufend = version + veclen;
8687 scan_vstring(version, vecsv);
8688 vecstr = (U8*)SvPV_const(vecsv, veclen);
8689 vec_utf8 = DO_UTF8(vecsv);
8690 Safefree(version);
d7aa5382 8691 }
211dfcf1
HS
8692 }
8693 else {
8694 vecstr = (U8*)"";
8695 veclen = 0;
8696 }
eb3fce90 8697 }
fc36a67e 8698
eb3fce90 8699 if (asterisk) {
fc36a67e 8700 if (args)
8701 i = va_arg(*args, int);
8702 else
eb3fce90
JH
8703 i = (ewix ? ewix <= svmax : svix < svmax) ?
8704 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8705 left |= (i < 0);
8706 width = (i < 0) ? -i : i;
fc36a67e 8707 }
211dfcf1 8708 gotwidth:
fc36a67e 8709
8710 /* PRECISION */
46fc3d4c 8711
fc36a67e 8712 if (*q == '.') {
8713 q++;
8714 if (*q == '*') {
211dfcf1 8715 q++;
c445ea15 8716 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8717 goto unknown;
8718 /* XXX: todo, support specified precision parameter */
8719 if (epix)
211dfcf1 8720 goto unknown;
46fc3d4c 8721 if (args)
8722 i = va_arg(*args, int);
8723 else
eb3fce90
JH
8724 i = (ewix ? ewix <= svmax : svix < svmax)
8725 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9911cee9
TS
8726 precis = i;
8727 has_precis = !(i < 0);
fc36a67e 8728 }
8729 else {
8730 precis = 0;
8731 while (isDIGIT(*q))
8732 precis = precis * 10 + (*q++ - '0');
9911cee9 8733 has_precis = TRUE;
fc36a67e 8734 }
fc36a67e 8735 }
46fc3d4c 8736
fc36a67e 8737 /* SIZE */
46fc3d4c 8738
fc36a67e 8739 switch (*q) {
c623ac67
GS
8740#ifdef WIN32
8741 case 'I': /* Ix, I32x, and I64x */
8742# ifdef WIN64
8743 if (q[1] == '6' && q[2] == '4') {
8744 q += 3;
8745 intsize = 'q';
8746 break;
8747 }
8748# endif
8749 if (q[1] == '3' && q[2] == '2') {
8750 q += 3;
8751 break;
8752 }
8753# ifdef WIN64
8754 intsize = 'q';
8755# endif
8756 q++;
8757 break;
8758#endif
9e5b023a 8759#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8760 case 'L': /* Ld */
5f66b61c 8761 /*FALLTHROUGH*/
e5c81feb 8762#ifdef HAS_QUAD
6f9bb7fd 8763 case 'q': /* qd */
9e5b023a 8764#endif
6f9bb7fd
GS
8765 intsize = 'q';
8766 q++;
8767 break;
8768#endif
fc36a67e 8769 case 'l':
9e5b023a 8770#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8771 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8772 intsize = 'q';
8773 q += 2;
46fc3d4c 8774 break;
cf2093f6 8775 }
fc36a67e 8776#endif
5f66b61c 8777 /*FALLTHROUGH*/
fc36a67e 8778 case 'h':
5f66b61c 8779 /*FALLTHROUGH*/
fc36a67e 8780 case 'V':
8781 intsize = *q++;
46fc3d4c 8782 break;
8783 }
8784
fc36a67e 8785 /* CONVERSION */
8786
211dfcf1
HS
8787 if (*q == '%') {
8788 eptr = q++;
8789 elen = 1;
26372e71
GA
8790 if (vectorize) {
8791 c = '%';
8792 goto unknown;
8793 }
211dfcf1
HS
8794 goto string;
8795 }
8796
26372e71 8797 if (!vectorize && !args) {
86c51f8b
NC
8798 if (efix) {
8799 const I32 i = efix-1;
8800 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8801 } else {
8802 argsv = (svix >= 0 && svix < svmax)
8803 ? svargs[svix++] : &PL_sv_undef;
8804 }
863811b2 8805 }
211dfcf1 8806
46fc3d4c 8807 switch (c = *q++) {
8808
8809 /* STRINGS */
8810
46fc3d4c 8811 case 'c':
26372e71
GA
8812 if (vectorize)
8813 goto unknown;
8814 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8815 if ((uv > 255 ||
8816 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8817 && !IN_BYTES) {
dfe13c55 8818 eptr = (char*)utf8buf;
9041c2e3 8819 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8820 is_utf8 = TRUE;
7e2040f0
GS
8821 }
8822 else {
8823 c = (char)uv;
8824 eptr = &c;
8825 elen = 1;
a0ed51b3 8826 }
46fc3d4c 8827 goto string;
8828
46fc3d4c 8829 case 's':
26372e71
GA
8830 if (vectorize)
8831 goto unknown;
8832 if (args) {
fc36a67e 8833 eptr = va_arg(*args, char*);
c635e13b 8834 if (eptr)
1d7c1841
GS
8835#ifdef MACOS_TRADITIONAL
8836 /* On MacOS, %#s format is used for Pascal strings */
8837 if (alt)
8838 elen = *eptr++;
8839 else
8840#endif
c635e13b 8841 elen = strlen(eptr);
8842 else {
27da23d5 8843 eptr = (char *)nullstr;
c635e13b 8844 elen = sizeof nullstr - 1;
8845 }
46fc3d4c 8846 }
211dfcf1 8847 else {
4d84ee25 8848 eptr = SvPVx_const(argsv, elen);
7e2040f0 8849 if (DO_UTF8(argsv)) {
59b61096 8850 I32 old_precis = precis;
a0ed51b3
LW
8851 if (has_precis && precis < elen) {
8852 I32 p = precis;
7e2040f0 8853 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8854 precis = p;
8855 }
8856 if (width) { /* fudge width (can't fudge elen) */
59b61096
AV
8857 if (has_precis && precis < elen)
8858 width += precis - old_precis;
8859 else
8860 width += elen - sv_len_utf8(argsv);
a0ed51b3 8861 }
2cf2cfc6 8862 is_utf8 = TRUE;
a0ed51b3
LW
8863 }
8864 }
fc36a67e 8865
46fc3d4c 8866 string:
8867 if (has_precis && elen > precis)
8868 elen = precis;
8869 break;
8870
8871 /* INTEGERS */
8872
fc36a67e 8873 case 'p':
be75b157 8874 if (alt || vectorize)
c2e66d9e 8875 goto unknown;
211dfcf1 8876 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8877 base = 16;
8878 goto integer;
8879
46fc3d4c 8880 case 'D':
29fe7a80 8881#ifdef IV_IS_QUAD
22f3ae8c 8882 intsize = 'q';
29fe7a80 8883#else
46fc3d4c 8884 intsize = 'l';
29fe7a80 8885#endif
5f66b61c 8886 /*FALLTHROUGH*/
46fc3d4c 8887 case 'd':
8888 case 'i':
8896765a
RB
8889#if vdNUMBER
8890 format_vd:
8891#endif
b22c7a20 8892 if (vectorize) {
ba210ebe 8893 STRLEN ulen;
211dfcf1
HS
8894 if (!veclen)
8895 continue;
2cf2cfc6
A
8896 if (vec_utf8)
8897 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8898 UTF8_ALLOW_ANYUV);
b22c7a20 8899 else {
e83d50c9 8900 uv = *vecstr;
b22c7a20
GS
8901 ulen = 1;
8902 }
8903 vecstr += ulen;
8904 veclen -= ulen;
e83d50c9
JP
8905 if (plus)
8906 esignbuf[esignlen++] = plus;
b22c7a20
GS
8907 }
8908 else if (args) {
46fc3d4c 8909 switch (intsize) {
8910 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8911 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8912 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8913 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8914#ifdef HAS_QUAD
8915 case 'q': iv = va_arg(*args, Quad_t); break;
8916#endif
46fc3d4c 8917 }
8918 }
8919 else {
b10c0dba 8920 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8921 switch (intsize) {
b10c0dba
MHM
8922 case 'h': iv = (short)tiv; break;
8923 case 'l': iv = (long)tiv; break;
8924 case 'V':
8925 default: iv = tiv; break;
cf2093f6 8926#ifdef HAS_QUAD
b10c0dba 8927 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8928#endif
46fc3d4c 8929 }
8930 }
e83d50c9
JP
8931 if ( !vectorize ) /* we already set uv above */
8932 {
8933 if (iv >= 0) {
8934 uv = iv;
8935 if (plus)
8936 esignbuf[esignlen++] = plus;
8937 }
8938 else {
8939 uv = -iv;
8940 esignbuf[esignlen++] = '-';
8941 }
46fc3d4c 8942 }
8943 base = 10;
8944 goto integer;
8945
fc36a67e 8946 case 'U':
29fe7a80 8947#ifdef IV_IS_QUAD
22f3ae8c 8948 intsize = 'q';
29fe7a80 8949#else
fc36a67e 8950 intsize = 'l';
29fe7a80 8951#endif
5f66b61c 8952 /*FALLTHROUGH*/
fc36a67e 8953 case 'u':
8954 base = 10;
8955 goto uns_integer;
8956
7ff06cc7 8957 case 'B':
4f19785b
WSI
8958 case 'b':
8959 base = 2;
8960 goto uns_integer;
8961
46fc3d4c 8962 case 'O':
29fe7a80 8963#ifdef IV_IS_QUAD
22f3ae8c 8964 intsize = 'q';
29fe7a80 8965#else
46fc3d4c 8966 intsize = 'l';
29fe7a80 8967#endif
5f66b61c 8968 /*FALLTHROUGH*/
46fc3d4c 8969 case 'o':
8970 base = 8;
8971 goto uns_integer;
8972
8973 case 'X':
46fc3d4c 8974 case 'x':
8975 base = 16;
46fc3d4c 8976
8977 uns_integer:
b22c7a20 8978 if (vectorize) {
ba210ebe 8979 STRLEN ulen;
b22c7a20 8980 vector:
211dfcf1
HS
8981 if (!veclen)
8982 continue;
2cf2cfc6
A
8983 if (vec_utf8)
8984 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8985 UTF8_ALLOW_ANYUV);
b22c7a20 8986 else {
a05b299f 8987 uv = *vecstr;
b22c7a20
GS
8988 ulen = 1;
8989 }
8990 vecstr += ulen;
8991 veclen -= ulen;
8992 }
8993 else if (args) {
46fc3d4c 8994 switch (intsize) {
8995 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8996 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8997 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8998 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8999#ifdef HAS_QUAD
9e3321a5 9000 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 9001#endif
46fc3d4c 9002 }
9003 }
9004 else {
b10c0dba 9005 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 9006 switch (intsize) {
b10c0dba
MHM
9007 case 'h': uv = (unsigned short)tuv; break;
9008 case 'l': uv = (unsigned long)tuv; break;
9009 case 'V':
9010 default: uv = tuv; break;
cf2093f6 9011#ifdef HAS_QUAD
b10c0dba 9012 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 9013#endif
46fc3d4c 9014 }
9015 }
9016
9017 integer:
4d84ee25
NC
9018 {
9019 char *ptr = ebuf + sizeof ebuf;
1387f30c
DD
9020 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9021 zeros = 0;
9022
4d84ee25
NC
9023 switch (base) {
9024 unsigned dig;
9025 case 16:
14eb61ab 9026 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
4d84ee25
NC
9027 do {
9028 dig = uv & 15;
9029 *--ptr = p[dig];
9030 } while (uv >>= 4);
1387f30c 9031 if (tempalt) {
4d84ee25
NC
9032 esignbuf[esignlen++] = '0';
9033 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9034 }
9035 break;
9036 case 8:
9037 do {
9038 dig = uv & 7;
9039 *--ptr = '0' + dig;
9040 } while (uv >>= 3);
9041 if (alt && *ptr != '0')
9042 *--ptr = '0';
9043 break;
9044 case 2:
9045 do {
9046 dig = uv & 1;
9047 *--ptr = '0' + dig;
9048 } while (uv >>= 1);
1387f30c 9049 if (tempalt) {
4d84ee25 9050 esignbuf[esignlen++] = '0';
7ff06cc7 9051 esignbuf[esignlen++] = c;
4d84ee25
NC
9052 }
9053 break;
9054 default: /* it had better be ten or less */
9055 do {
9056 dig = uv % base;
9057 *--ptr = '0' + dig;
9058 } while (uv /= base);
9059 break;
46fc3d4c 9060 }
4d84ee25
NC
9061 elen = (ebuf + sizeof ebuf) - ptr;
9062 eptr = ptr;
9063 if (has_precis) {
9064 if (precis > elen)
9065 zeros = precis - elen;
e6bb52fd
TS
9066 else if (precis == 0 && elen == 1 && *eptr == '0'
9067 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
4d84ee25 9068 elen = 0;
9911cee9
TS
9069
9070 /* a precision nullifies the 0 flag. */
9071 if (fill == '0')
9072 fill = ' ';
eda88b6d 9073 }
c10ed8b9 9074 }
46fc3d4c 9075 break;
9076
9077 /* FLOATING POINT */
9078
fc36a67e 9079 case 'F':
9080 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9081 /*FALLTHROUGH*/
46fc3d4c 9082 case 'e': case 'E':
fc36a67e 9083 case 'f':
46fc3d4c 9084 case 'g': case 'G':
26372e71
GA
9085 if (vectorize)
9086 goto unknown;
46fc3d4c 9087
9088 /* This is evil, but floating point is even more evil */
9089
9e5b023a
JH
9090 /* for SV-style calling, we can only get NV
9091 for C-style calling, we assume %f is double;
9092 for simplicity we allow any of %Lf, %llf, %qf for long double
9093 */
9094 switch (intsize) {
9095 case 'V':
9096#if defined(USE_LONG_DOUBLE)
9097 intsize = 'q';
9098#endif
9099 break;
8a2e3f14 9100/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9101 case 'l':
5f66b61c 9102 /*FALLTHROUGH*/
9e5b023a
JH
9103 default:
9104#if defined(USE_LONG_DOUBLE)
9105 intsize = args ? 0 : 'q';
9106#endif
9107 break;
9108 case 'q':
9109#if defined(HAS_LONG_DOUBLE)
9110 break;
9111#else
5f66b61c 9112 /*FALLTHROUGH*/
9e5b023a
JH
9113#endif
9114 case 'h':
9e5b023a
JH
9115 goto unknown;
9116 }
9117
9118 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9119 nv = (args) ?
35fff930
JH
9120#if LONG_DOUBLESIZE > DOUBLESIZE
9121 intsize == 'q' ?
205f51d8
AS
9122 va_arg(*args, long double) :
9123 va_arg(*args, double)
35fff930 9124#else
205f51d8 9125 va_arg(*args, double)
35fff930 9126#endif
9e5b023a 9127 : SvNVx(argsv);
fc36a67e 9128
9129 need = 0;
9130 if (c != 'e' && c != 'E') {
9131 i = PERL_INT_MIN;
9e5b023a
JH
9132 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9133 will cast our (long double) to (double) */
73b309ea 9134 (void)Perl_frexp(nv, &i);
fc36a67e 9135 if (i == PERL_INT_MIN)
cea2e8a9 9136 Perl_die(aTHX_ "panic: frexp");
c635e13b 9137 if (i > 0)
fc36a67e 9138 need = BIT_DIGITS(i);
9139 }
9140 need += has_precis ? precis : 6; /* known default */
20f6aaab 9141
fc36a67e 9142 if (need < width)
9143 need = width;
9144
20f6aaab
AS
9145#ifdef HAS_LDBL_SPRINTF_BUG
9146 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9147 with sfio - Allen <allens@cpan.org> */
9148
9149# ifdef DBL_MAX
9150# define MY_DBL_MAX DBL_MAX
9151# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9152# if DOUBLESIZE >= 8
9153# define MY_DBL_MAX 1.7976931348623157E+308L
9154# else
9155# define MY_DBL_MAX 3.40282347E+38L
9156# endif
9157# endif
9158
9159# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9160# define MY_DBL_MAX_BUG 1L
20f6aaab 9161# else
205f51d8 9162# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9163# endif
20f6aaab 9164
205f51d8
AS
9165# ifdef DBL_MIN
9166# define MY_DBL_MIN DBL_MIN
9167# else /* XXX guessing! -Allen */
9168# if DOUBLESIZE >= 8
9169# define MY_DBL_MIN 2.2250738585072014E-308L
9170# else
9171# define MY_DBL_MIN 1.17549435E-38L
9172# endif
9173# endif
20f6aaab 9174
205f51d8
AS
9175 if ((intsize == 'q') && (c == 'f') &&
9176 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9177 (need < DBL_DIG)) {
9178 /* it's going to be short enough that
9179 * long double precision is not needed */
9180
9181 if ((nv <= 0L) && (nv >= -0L))
9182 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9183 else {
9184 /* would use Perl_fp_class as a double-check but not
9185 * functional on IRIX - see perl.h comments */
9186
9187 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9188 /* It's within the range that a double can represent */
9189#if defined(DBL_MAX) && !defined(DBL_MIN)
9190 if ((nv >= ((long double)1/DBL_MAX)) ||
9191 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9192#endif
205f51d8 9193 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9194 }
205f51d8
AS
9195 }
9196 if (fix_ldbl_sprintf_bug == TRUE) {
9197 double temp;
9198
9199 intsize = 0;
9200 temp = (double)nv;
9201 nv = (NV)temp;
9202 }
20f6aaab 9203 }
205f51d8
AS
9204
9205# undef MY_DBL_MAX
9206# undef MY_DBL_MAX_BUG
9207# undef MY_DBL_MIN
9208
20f6aaab
AS
9209#endif /* HAS_LDBL_SPRINTF_BUG */
9210
46fc3d4c 9211 need += 20; /* fudge factor */
80252599
GS
9212 if (PL_efloatsize < need) {
9213 Safefree(PL_efloatbuf);
9214 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9215 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9216 PL_efloatbuf[0] = '\0';
46fc3d4c 9217 }
9218
4151a5fe
IZ
9219 if ( !(width || left || plus || alt) && fill != '0'
9220 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9221 /* See earlier comment about buggy Gconvert when digits,
9222 aka precis is 0 */
9223 if ( c == 'g' && precis) {
2e59c212 9224 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9225 /* May return an empty string for digits==0 */
9226 if (*PL_efloatbuf) {
9227 elen = strlen(PL_efloatbuf);
4151a5fe 9228 goto float_converted;
4150c189 9229 }
4151a5fe
IZ
9230 } else if ( c == 'f' && !precis) {
9231 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9232 break;
9233 }
9234 }
4d84ee25
NC
9235 {
9236 char *ptr = ebuf + sizeof ebuf;
9237 *--ptr = '\0';
9238 *--ptr = c;
9239 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9240#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9241 if (intsize == 'q') {
9242 /* Copy the one or more characters in a long double
9243 * format before the 'base' ([efgEFG]) character to
9244 * the format string. */
9245 static char const prifldbl[] = PERL_PRIfldbl;
9246 char const *p = prifldbl + sizeof(prifldbl) - 3;
9247 while (p >= prifldbl) { *--ptr = *p--; }
9248 }
65202027 9249#endif
4d84ee25
NC
9250 if (has_precis) {
9251 base = precis;
9252 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9253 *--ptr = '.';
9254 }
9255 if (width) {
9256 base = width;
9257 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9258 }
9259 if (fill == '0')
9260 *--ptr = fill;
9261 if (left)
9262 *--ptr = '-';
9263 if (plus)
9264 *--ptr = plus;
9265 if (alt)
9266 *--ptr = '#';
9267 *--ptr = '%';
9268
9269 /* No taint. Otherwise we are in the strange situation
9270 * where printf() taints but print($float) doesn't.
9271 * --jhi */
9e5b023a 9272#if defined(HAS_LONG_DOUBLE)
4150c189 9273 elen = ((intsize == 'q')
d9fad198
JH
9274 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
9275 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
9e5b023a 9276#else
4150c189 9277 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9278#endif
4d84ee25 9279 }
4151a5fe 9280 float_converted:
80252599 9281 eptr = PL_efloatbuf;
46fc3d4c 9282 break;
9283
fc36a67e 9284 /* SPECIAL */
9285
9286 case 'n':
26372e71
GA
9287 if (vectorize)
9288 goto unknown;
fc36a67e 9289 i = SvCUR(sv) - origlen;
26372e71 9290 if (args) {
c635e13b 9291 switch (intsize) {
9292 case 'h': *(va_arg(*args, short*)) = i; break;
9293 default: *(va_arg(*args, int*)) = i; break;
9294 case 'l': *(va_arg(*args, long*)) = i; break;
9295 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9296#ifdef HAS_QUAD
9297 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9298#endif
c635e13b 9299 }
fc36a67e 9300 }
9dd79c3f 9301 else
211dfcf1 9302 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9303 continue; /* not "break" */
9304
9305 /* UNKNOWN */
9306
46fc3d4c 9307 default:
fc36a67e 9308 unknown:
041457d9
DM
9309 if (!args
9310 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9311 && ckWARN(WARN_PRINTF))
9312 {
c4420975 9313 SV * const msg = sv_newmortal();
35c1215d
NC
9314 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9315 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9316 if (c) {
0f4b6630 9317 if (isPRINT(c))
1c846c1f 9318 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9319 "\"%%%c\"", c & 0xFF);
9320 else
9321 Perl_sv_catpvf(aTHX_ msg,
57def98f 9322 "\"%%\\%03"UVof"\"",
0f4b6630 9323 (UV)c & 0xFF);
0f4b6630 9324 } else
396482e1 9325 sv_catpvs(msg, "end of string");
be2597df 9326 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
c635e13b 9327 }
fb73857a 9328
9329 /* output mangled stuff ... */
9330 if (c == '\0')
9331 --q;
46fc3d4c 9332 eptr = p;
9333 elen = q - p;
fb73857a 9334
9335 /* ... right here, because formatting flags should not apply */
9336 SvGROW(sv, SvCUR(sv) + elen + 1);
9337 p = SvEND(sv);
4459522c 9338 Copy(eptr, p, elen, char);
fb73857a 9339 p += elen;
9340 *p = '\0';
3f7c398e 9341 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9342 svix = osvix;
fb73857a 9343 continue; /* not "break" */
46fc3d4c 9344 }
9345
cc61b222
TS
9346 if (is_utf8 != has_utf8) {
9347 if (is_utf8) {
9348 if (SvCUR(sv))
9349 sv_utf8_upgrade(sv);
9350 }
9351 else {
9352 const STRLEN old_elen = elen;
9353 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
9354 sv_utf8_upgrade(nsv);
9355 eptr = SvPVX_const(nsv);
9356 elen = SvCUR(nsv);
9357
9358 if (width) { /* fudge width (can't fudge elen) */
9359 width += elen - old_elen;
9360 }
9361 is_utf8 = TRUE;
9362 }
9363 }
9364
6c94ec8b 9365 have = esignlen + zeros + elen;
ed2b91d2
GA
9366 if (have < zeros)
9367 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9368
46fc3d4c 9369 need = (have > width ? have : width);
9370 gap = need - have;
9371
d2641cbd
PC
9372 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9373 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9374 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9375 p = SvEND(sv);
9376 if (esignlen && fill == '0') {
53c1dcc0 9377 int i;
eb160463 9378 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9379 *p++ = esignbuf[i];
9380 }
9381 if (gap && !left) {
9382 memset(p, fill, gap);
9383 p += gap;
9384 }
9385 if (esignlen && fill != '0') {
53c1dcc0 9386 int i;
eb160463 9387 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9388 *p++ = esignbuf[i];
9389 }
fc36a67e 9390 if (zeros) {
53c1dcc0 9391 int i;
fc36a67e 9392 for (i = zeros; i; i--)
9393 *p++ = '0';
9394 }
46fc3d4c 9395 if (elen) {
4459522c 9396 Copy(eptr, p, elen, char);
46fc3d4c 9397 p += elen;
9398 }
9399 if (gap && left) {
9400 memset(p, ' ', gap);
9401 p += gap;
9402 }
b22c7a20
GS
9403 if (vectorize) {
9404 if (veclen) {
4459522c 9405 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9406 p += dotstrlen;
9407 }
9408 else
9409 vectorize = FALSE; /* done iterating over vecstr */
9410 }
2cf2cfc6
A
9411 if (is_utf8)
9412 has_utf8 = TRUE;
9413 if (has_utf8)
7e2040f0 9414 SvUTF8_on(sv);
46fc3d4c 9415 *p = '\0';
3f7c398e 9416 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9417 if (vectorize) {
9418 esignlen = 0;
9419 goto vector;
9420 }
46fc3d4c 9421 }
9422}
51371543 9423
645c22ef
DM
9424/* =========================================================================
9425
9426=head1 Cloning an interpreter
9427
9428All the macros and functions in this section are for the private use of
9429the main function, perl_clone().
9430
9431The foo_dup() functions make an exact copy of an existing foo thinngy.
9432During the course of a cloning, a hash table is used to map old addresses
9433to new addresses. The table is created and manipulated with the
9434ptr_table_* functions.
9435
9436=cut
9437
9438============================================================================*/
9439
9440
1d7c1841
GS
9441#if defined(USE_ITHREADS)
9442
d4c19fe8 9443/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9444#ifndef GpREFCNT_inc
9445# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9446#endif
9447
9448
a41cc44e 9449/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
3e07292d
NC
9450 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
9451 If this changes, please unmerge ss_dup. */
d2d73c3e 9452#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9453#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9454#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9455#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9456#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9457#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9458#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9459#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9460#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9461#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9462#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9463#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9464#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9465#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9466
199e78b7
DM
9467/* clone a parser */
9468
9469yy_parser *
9470Perl_parser_dup(pTHX_ const yy_parser *proto, CLONE_PARAMS* param)
9471{
9472 yy_parser *parser;
9473
9474 if (!proto)
9475 return NULL;
9476
7c197c94
DM
9477 /* look for it in the table first */
9478 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
9479 if (parser)
9480 return parser;
9481
9482 /* create anew and remember what it is */
199e78b7 9483 Newxz(parser, 1, yy_parser);
7c197c94 9484 ptr_table_store(PL_ptr_table, proto, parser);
199e78b7
DM
9485
9486 parser->yyerrstatus = 0;
9487 parser->yychar = YYEMPTY; /* Cause a token to be read. */
9488
9489 /* XXX these not yet duped */
9490 parser->old_parser = NULL;
9491 parser->stack = NULL;
9492 parser->ps = NULL;
9493 parser->stack_size = 0;
9494 /* XXX parser->stack->state = 0; */
9495
9496 /* XXX eventually, just Copy() most of the parser struct ? */
9497
9498 parser->lex_brackets = proto->lex_brackets;
9499 parser->lex_casemods = proto->lex_casemods;
9500 parser->lex_brackstack = savepvn(proto->lex_brackstack,
9501 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
9502 parser->lex_casestack = savepvn(proto->lex_casestack,
9503 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
9504 parser->lex_defer = proto->lex_defer;
9505 parser->lex_dojoin = proto->lex_dojoin;
9506 parser->lex_expect = proto->lex_expect;
9507 parser->lex_formbrack = proto->lex_formbrack;
9508 parser->lex_inpat = proto->lex_inpat;
9509 parser->lex_inwhat = proto->lex_inwhat;
9510 parser->lex_op = proto->lex_op;
9511 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
9512 parser->lex_starts = proto->lex_starts;
9513 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
9514 parser->multi_close = proto->multi_close;
9515 parser->multi_open = proto->multi_open;
9516 parser->multi_start = proto->multi_start;
9517 parser->pending_ident = proto->pending_ident;
9518 parser->preambled = proto->preambled;
9519 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
9520
9521#ifdef PERL_MAD
9522 parser->endwhite = proto->endwhite;
9523 parser->faketokens = proto->faketokens;
9524 parser->lasttoke = proto->lasttoke;
9525 parser->nextwhite = proto->nextwhite;
9526 parser->realtokenstart = proto->realtokenstart;
9527 parser->skipwhite = proto->skipwhite;
9528 parser->thisclose = proto->thisclose;
9529 parser->thismad = proto->thismad;
9530 parser->thisopen = proto->thisopen;
9531 parser->thisstuff = proto->thisstuff;
9532 parser->thistoken = proto->thistoken;
9533 parser->thiswhite = proto->thiswhite;
9534#endif
9535 return parser;
9536}
9537
d2d73c3e 9538
d2d73c3e 9539/* duplicate a file handle */
645c22ef 9540
1d7c1841 9541PerlIO *
a8fc9800 9542Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9543{
9544 PerlIO *ret;
53c1dcc0
AL
9545
9546 PERL_UNUSED_ARG(type);
73d840c0 9547
1d7c1841
GS
9548 if (!fp)
9549 return (PerlIO*)NULL;
9550
9551 /* look for it in the table first */
9552 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9553 if (ret)
9554 return ret;
9555
9556 /* create anew and remember what it is */
ecdeb87c 9557 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9558 ptr_table_store(PL_ptr_table, fp, ret);
9559 return ret;
9560}
9561
645c22ef
DM
9562/* duplicate a directory handle */
9563
1d7c1841
GS
9564DIR *
9565Perl_dirp_dup(pTHX_ DIR *dp)
9566{
96a5add6 9567 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9568 if (!dp)
9569 return (DIR*)NULL;
9570 /* XXX TODO */
9571 return dp;
9572}
9573
ff276b08 9574/* duplicate a typeglob */
645c22ef 9575
1d7c1841 9576GP *
a8fc9800 9577Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9578{
9579 GP *ret;
b37c2d43 9580
1d7c1841
GS
9581 if (!gp)
9582 return (GP*)NULL;
9583 /* look for it in the table first */
9584 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9585 if (ret)
9586 return ret;
9587
9588 /* create anew and remember what it is */
a02a5408 9589 Newxz(ret, 1, GP);
1d7c1841
GS
9590 ptr_table_store(PL_ptr_table, gp, ret);
9591
9592 /* clone */
9593 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9594 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9595 ret->gp_io = io_dup_inc(gp->gp_io, param);
9596 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9597 ret->gp_av = av_dup_inc(gp->gp_av, param);
9598 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9599 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9600 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9601 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841 9602 ret->gp_line = gp->gp_line;
f4890806 9603 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
1d7c1841
GS
9604 return ret;
9605}
9606
645c22ef
DM
9607/* duplicate a chain of magic */
9608
1d7c1841 9609MAGIC *
a8fc9800 9610Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9611{
cb359b41
JH
9612 MAGIC *mgprev = (MAGIC*)NULL;
9613 MAGIC *mgret;
1d7c1841
GS
9614 if (!mg)
9615 return (MAGIC*)NULL;
9616 /* look for it in the table first */
9617 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9618 if (mgret)
9619 return mgret;
9620
9621 for (; mg; mg = mg->mg_moremagic) {
9622 MAGIC *nmg;
a02a5408 9623 Newxz(nmg, 1, MAGIC);
cb359b41 9624 if (mgprev)
1d7c1841 9625 mgprev->mg_moremagic = nmg;
cb359b41
JH
9626 else
9627 mgret = nmg;
1d7c1841
GS
9628 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9629 nmg->mg_private = mg->mg_private;
9630 nmg->mg_type = mg->mg_type;
9631 nmg->mg_flags = mg->mg_flags;
14befaf4 9632 if (mg->mg_type == PERL_MAGIC_qr) {
f8149455 9633 nmg->mg_obj = (SV*)CALLREGDUPE((REGEXP*)mg->mg_obj, param);
1d7c1841 9634 }
05bd4103 9635 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9636 /* The backref AV has its reference count deliberately bumped by
9637 1. */
9638 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9639 }
8d2f4536
NC
9640 else if (mg->mg_type == PERL_MAGIC_symtab) {
9641 nmg->mg_obj = mg->mg_obj;
9642 }
1d7c1841
GS
9643 else {
9644 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9645 ? sv_dup_inc(mg->mg_obj, param)
9646 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9647 }
9648 nmg->mg_len = mg->mg_len;
9649 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9650 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9651 if (mg->mg_len > 0) {
1d7c1841 9652 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9653 if (mg->mg_type == PERL_MAGIC_overload_table &&
9654 AMT_AMAGIC((AMT*)mg->mg_ptr))
9655 {
c445ea15 9656 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9657 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9658 I32 i;
9659 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9660 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9661 }
9662 }
9663 }
9664 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9665 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9666 }
68795e93
NIS
9667 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9668 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9669 }
1d7c1841
GS
9670 mgprev = nmg;
9671 }
9672 return mgret;
9673}
9674
4674ade5
NC
9675#endif /* USE_ITHREADS */
9676
645c22ef
DM
9677/* create a new pointer-mapping table */
9678
1d7c1841
GS
9679PTR_TBL_t *
9680Perl_ptr_table_new(pTHX)
9681{
9682 PTR_TBL_t *tbl;
96a5add6
AL
9683 PERL_UNUSED_CONTEXT;
9684
a02a5408 9685 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9686 tbl->tbl_max = 511;
9687 tbl->tbl_items = 0;
a02a5408 9688 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9689 return tbl;
9690}
9691
7119fd33
NC
9692#define PTR_TABLE_HASH(ptr) \
9693 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9694
93e68bfb
JC
9695/*
9696 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9697 following define) and at call to new_body_inline made below in
9698 Perl_ptr_table_store()
9699 */
9700
9701#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9702
645c22ef
DM
9703/* map an existing pointer using a table */
9704
7bf61b54 9705STATIC PTR_TBL_ENT_t *
b0e6ae5b 9706S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9707 PTR_TBL_ENT_t *tblent;
4373e329 9708 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9709 assert(tbl);
9710 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9711 for (; tblent; tblent = tblent->next) {
9712 if (tblent->oldval == sv)
7bf61b54 9713 return tblent;
1d7c1841 9714 }
d4c19fe8 9715 return NULL;
7bf61b54
NC
9716}
9717
9718void *
9719Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9720{
b0e6ae5b 9721 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9722 PERL_UNUSED_CONTEXT;
d4c19fe8 9723 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9724}
9725
645c22ef
DM
9726/* add a new entry to a pointer-mapping table */
9727
1d7c1841 9728void
44f8325f 9729Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9730{
0c9fdfe0 9731 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9732 PERL_UNUSED_CONTEXT;
1d7c1841 9733
7bf61b54
NC
9734 if (tblent) {
9735 tblent->newval = newsv;
9736 } else {
9737 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9738
d2a0f284
JC
9739 new_body_inline(tblent, PTE_SVSLOT);
9740
7bf61b54
NC
9741 tblent->oldval = oldsv;
9742 tblent->newval = newsv;
9743 tblent->next = tbl->tbl_ary[entry];
9744 tbl->tbl_ary[entry] = tblent;
9745 tbl->tbl_items++;
9746 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9747 ptr_table_split(tbl);
1d7c1841 9748 }
1d7c1841
GS
9749}
9750
645c22ef
DM
9751/* double the hash bucket size of an existing ptr table */
9752
1d7c1841
GS
9753void
9754Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9755{
9756 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9757 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9758 UV newsize = oldsize * 2;
9759 UV i;
96a5add6 9760 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9761
9762 Renew(ary, newsize, PTR_TBL_ENT_t*);
9763 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9764 tbl->tbl_max = --newsize;
9765 tbl->tbl_ary = ary;
9766 for (i=0; i < oldsize; i++, ary++) {
9767 PTR_TBL_ENT_t **curentp, **entp, *ent;
9768 if (!*ary)
9769 continue;
9770 curentp = ary + oldsize;
9771 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9772 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9773 *entp = ent->next;
9774 ent->next = *curentp;
9775 *curentp = ent;
9776 continue;
9777 }
9778 else
9779 entp = &ent->next;
9780 }
9781 }
9782}
9783
645c22ef
DM
9784/* remove all the entries from a ptr table */
9785
a0739874
DM
9786void
9787Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9788{
d5cefff9 9789 if (tbl && tbl->tbl_items) {
c445ea15 9790 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9791 UV riter = tbl->tbl_max;
a0739874 9792
d5cefff9
NC
9793 do {
9794 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9795
d5cefff9 9796 while (entry) {
00b6aa41 9797 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9798 entry = entry->next;
9799 del_pte(oentry);
9800 }
9801 } while (riter--);
a0739874 9802
d5cefff9
NC
9803 tbl->tbl_items = 0;
9804 }
a0739874
DM
9805}
9806
645c22ef
DM
9807/* clear and free a ptr table */
9808
a0739874
DM
9809void
9810Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9811{
9812 if (!tbl) {
9813 return;
9814 }
9815 ptr_table_clear(tbl);
9816 Safefree(tbl->tbl_ary);
9817 Safefree(tbl);
9818}
9819
4674ade5 9820#if defined(USE_ITHREADS)
5bd07a3d 9821
83841fad 9822void
eb86f8b3 9823Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9824{
9825 if (SvROK(sstr)) {
b162af07
SP
9826 SvRV_set(dstr, SvWEAKREF(sstr)
9827 ? sv_dup(SvRV(sstr), param)
9828 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9829
83841fad 9830 }
3f7c398e 9831 else if (SvPVX_const(sstr)) {
83841fad
NIS
9832 /* Has something there */
9833 if (SvLEN(sstr)) {
68795e93 9834 /* Normal PV - clone whole allocated space */
3f7c398e 9835 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9836 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9837 /* Not that normal - actually sstr is copy on write.
9838 But we are a true, independant SV, so: */
9839 SvREADONLY_off(dstr);
9840 SvFAKE_off(dstr);
9841 }
68795e93 9842 }
83841fad
NIS
9843 else {
9844 /* Special case - not normally malloced for some reason */
f7877b28
NC
9845 if (isGV_with_GP(sstr)) {
9846 /* Don't need to do anything here. */
9847 }
9848 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9849 /* A "shared" PV - clone it as "shared" PV */
9850 SvPV_set(dstr,
9851 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9852 param)));
83841fad
NIS
9853 }
9854 else {
9855 /* Some other special case - random pointer */
f880fe2f 9856 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9857 }
83841fad
NIS
9858 }
9859 }
9860 else {
4608196e 9861 /* Copy the NULL */
f880fe2f 9862 if (SvTYPE(dstr) == SVt_RV)
b162af07 9863 SvRV_set(dstr, NULL);
f880fe2f 9864 else
6136c704 9865 SvPV_set(dstr, NULL);
83841fad
NIS
9866 }
9867}
9868
662fb8b2
NC
9869/* duplicate an SV of any type (including AV, HV etc) */
9870
1d7c1841 9871SV *
eb86f8b3 9872Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9873{
27da23d5 9874 dVAR;
1d7c1841
GS
9875 SV *dstr;
9876
9877 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9878 return NULL;
1d7c1841
GS
9879 /* look for it in the table first */
9880 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9881 if (dstr)
9882 return dstr;
9883
0405e91e
AB
9884 if(param->flags & CLONEf_JOIN_IN) {
9885 /** We are joining here so we don't want do clone
9886 something that is bad **/
eb86f8b3
AL
9887 if (SvTYPE(sstr) == SVt_PVHV) {
9888 const char * const hvname = HvNAME_get(sstr);
9889 if (hvname)
9890 /** don't clone stashes if they already exist **/
9891 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9892 }
9893 }
9894
1d7c1841
GS
9895 /* create anew and remember what it is */
9896 new_SV(dstr);
fd0854ff
DM
9897
9898#ifdef DEBUG_LEAKING_SCALARS
9899 dstr->sv_debug_optype = sstr->sv_debug_optype;
9900 dstr->sv_debug_line = sstr->sv_debug_line;
9901 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9902 dstr->sv_debug_cloned = 1;
fd0854ff 9903 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9904#endif
9905
1d7c1841
GS
9906 ptr_table_store(PL_ptr_table, sstr, dstr);
9907
9908 /* clone */
9909 SvFLAGS(dstr) = SvFLAGS(sstr);
9910 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9911 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9912
9913#ifdef DEBUGGING
3f7c398e 9914 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9915 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
6c9570dc 9916 (void*)PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9917#endif
9918
9660f481
DM
9919 /* don't clone objects whose class has asked us not to */
9920 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9921 SvFLAGS(dstr) &= ~SVTYPEMASK;
9922 SvOBJECT_off(dstr);
9923 return dstr;
9924 }
9925
1d7c1841
GS
9926 switch (SvTYPE(sstr)) {
9927 case SVt_NULL:
9928 SvANY(dstr) = NULL;
9929 break;
9930 case SVt_IV:
339049b0 9931 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9932 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9933 break;
9934 case SVt_NV:
9935 SvANY(dstr) = new_XNV();
9d6ce603 9936 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9937 break;
9938 case SVt_RV:
339049b0 9939 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9940 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9941 break;
cecf5685 9942 /* case SVt_BIND: */
662fb8b2
NC
9943 default:
9944 {
9945 /* These are all the types that need complex bodies allocating. */
662fb8b2 9946 void *new_body;
2bcc16b3
NC
9947 const svtype sv_type = SvTYPE(sstr);
9948 const struct body_details *const sv_type_details
9949 = bodies_by_type + sv_type;
662fb8b2 9950
93e68bfb 9951 switch (sv_type) {
662fb8b2 9952 default:
bb263b4e 9953 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9954 break;
9955
662fb8b2
NC
9956 case SVt_PVGV:
9957 if (GvUNIQUE((GV*)sstr)) {
6f207bd3 9958 NOOP; /* Do sharing here, and fall through */
662fb8b2 9959 }
c22188b4
NC
9960 case SVt_PVIO:
9961 case SVt_PVFM:
9962 case SVt_PVHV:
9963 case SVt_PVAV:
662fb8b2 9964 case SVt_PVCV:
662fb8b2 9965 case SVt_PVLV:
662fb8b2 9966 case SVt_PVMG:
662fb8b2 9967 case SVt_PVNV:
662fb8b2 9968 case SVt_PVIV:
662fb8b2 9969 case SVt_PV:
d2a0f284 9970 assert(sv_type_details->body_size);
c22188b4 9971 if (sv_type_details->arena) {
d2a0f284 9972 new_body_inline(new_body, sv_type);
c22188b4 9973 new_body
b9502f15 9974 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9975 } else {
9976 new_body = new_NOARENA(sv_type_details);
9977 }
1d7c1841 9978 }
662fb8b2
NC
9979 assert(new_body);
9980 SvANY(dstr) = new_body;
9981
2bcc16b3 9982#ifndef PURIFY
b9502f15
NC
9983 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
9984 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 9985 sv_type_details->copy, char);
2bcc16b3
NC
9986#else
9987 Copy(((char*)SvANY(sstr)),
9988 ((char*)SvANY(dstr)),
d2a0f284 9989 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 9990#endif
662fb8b2 9991
f7877b28
NC
9992 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
9993 && !isGV_with_GP(dstr))
662fb8b2
NC
9994 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
9995
9996 /* The Copy above means that all the source (unduplicated) pointers
9997 are now in the destination. We can check the flags and the
9998 pointers in either, but it's possible that there's less cache
9999 missing by always going for the destination.
10000 FIXME - instrument and check that assumption */
f32993d6 10001 if (sv_type >= SVt_PVMG) {
885ffcb3 10002 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
73d95100 10003 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
e736a858 10004 } else if (SvMAGIC(dstr))
662fb8b2
NC
10005 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10006 if (SvSTASH(dstr))
10007 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10008 }
662fb8b2 10009
f32993d6
NC
10010 /* The cast silences a GCC warning about unhandled types. */
10011 switch ((int)sv_type) {
662fb8b2
NC
10012 case SVt_PV:
10013 break;
10014 case SVt_PVIV:
10015 break;
10016 case SVt_PVNV:
10017 break;
10018 case SVt_PVMG:
10019 break;
662fb8b2
NC
10020 case SVt_PVLV:
10021 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10022 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10023 LvTARG(dstr) = dstr;
10024 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10025 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10026 else
10027 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
662fb8b2 10028 case SVt_PVGV:
cecf5685
NC
10029 if(isGV_with_GP(sstr)) {
10030 if (GvNAME_HEK(dstr))
10031 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
39cb70dc
NC
10032 /* Don't call sv_add_backref here as it's going to be
10033 created as part of the magic cloning of the symbol
10034 table. */
f7877b28
NC
10035 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10036 at the point of this comment. */
39cb70dc 10037 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
f7877b28
NC
10038 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10039 (void)GpREFCNT_inc(GvGP(dstr));
10040 } else
10041 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10042 break;
10043 case SVt_PVIO:
10044 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10045 if (IoOFP(dstr) == IoIFP(sstr))
10046 IoOFP(dstr) = IoIFP(dstr);
10047 else
10048 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10049 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10050 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10051 /* I have no idea why fake dirp (rsfps)
10052 should be treated differently but otherwise
10053 we end up with leaks -- sky*/
10054 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10055 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10056 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10057 } else {
10058 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10059 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10060 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10061 if (IoDIRP(dstr)) {
10062 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10063 } else {
6f207bd3 10064 NOOP;
100ce7e1
NC
10065 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10066 }
662fb8b2
NC
10067 }
10068 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10069 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10070 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10071 break;
10072 case SVt_PVAV:
10073 if (AvARRAY((AV*)sstr)) {
10074 SV **dst_ary, **src_ary;
10075 SSize_t items = AvFILLp((AV*)sstr) + 1;
10076
10077 src_ary = AvARRAY((AV*)sstr);
a02a5408 10078 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2 10079 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
9c6bc640 10080 AvARRAY((AV*)dstr) = dst_ary;
662fb8b2
NC
10081 AvALLOC((AV*)dstr) = dst_ary;
10082 if (AvREAL((AV*)sstr)) {
10083 while (items-- > 0)
10084 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10085 }
10086 else {
10087 while (items-- > 0)
10088 *dst_ary++ = sv_dup(*src_ary++, param);
10089 }
10090 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10091 while (items-- > 0) {
10092 *dst_ary++ = &PL_sv_undef;
10093 }
bfcb3514 10094 }
662fb8b2 10095 else {
9c6bc640 10096 AvARRAY((AV*)dstr) = NULL;
662fb8b2 10097 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10098 }
662fb8b2
NC
10099 break;
10100 case SVt_PVHV:
7e265ef3
AL
10101 if (HvARRAY((HV*)sstr)) {
10102 STRLEN i = 0;
10103 const bool sharekeys = !!HvSHAREKEYS(sstr);
10104 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10105 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10106 char *darray;
10107 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
10108 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10109 char);
10110 HvARRAY(dstr) = (HE**)darray;
10111 while (i <= sxhv->xhv_max) {
10112 const HE * const source = HvARRAY(sstr)[i];
10113 HvARRAY(dstr)[i] = source
10114 ? he_dup(source, sharekeys, param) : 0;
10115 ++i;
10116 }
10117 if (SvOOK(sstr)) {
10118 HEK *hvname;
10119 const struct xpvhv_aux * const saux = HvAUX(sstr);
10120 struct xpvhv_aux * const daux = HvAUX(dstr);
10121 /* This flag isn't copied. */
10122 /* SvOOK_on(hv) attacks the IV flags. */
10123 SvFLAGS(dstr) |= SVf_OOK;
10124
10125 hvname = saux->xhv_name;
10126 daux->xhv_name = hvname ? hek_dup(hvname, param) : hvname;
10127
10128 daux->xhv_riter = saux->xhv_riter;
10129 daux->xhv_eiter = saux->xhv_eiter
10130 ? he_dup(saux->xhv_eiter,
10131 (bool)!!HvSHAREKEYS(sstr), param) : 0;
10132 daux->xhv_backreferences =
10133 saux->xhv_backreferences
86f55936 10134 ? (AV*) SvREFCNT_inc(
7e265ef3 10135 sv_dup((SV*)saux->xhv_backreferences, param))
86f55936 10136 : 0;
7e265ef3
AL
10137 /* Record stashes for possible cloning in Perl_clone(). */
10138 if (hvname)
10139 av_push(param->stashes, dstr);
662fb8b2 10140 }
662fb8b2 10141 }
7e265ef3 10142 else
797c7171 10143 HvARRAY((HV*)dstr) = NULL;
662fb8b2 10144 break;
662fb8b2 10145 case SVt_PVCV:
bb172083
NC
10146 if (!(param->flags & CLONEf_COPY_STACKS)) {
10147 CvDEPTH(dstr) = 0;
10148 }
10149 case SVt_PVFM:
662fb8b2
NC
10150 /* NOTE: not refcounted */
10151 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10152 OP_REFCNT_LOCK;
d04ba589
NC
10153 if (!CvISXSUB(dstr))
10154 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10155 OP_REFCNT_UNLOCK;
cfae286e 10156 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10157 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10158 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10159 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10160 }
10161 /* don't dup if copying back - CvGV isn't refcounted, so the
10162 * duped GV may never be freed. A bit of a hack! DAPM */
10163 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10164 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10165 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10166 CvOUTSIDE(dstr) =
10167 CvWEAKOUTSIDE(sstr)
10168 ? cv_dup( CvOUTSIDE(dstr), param)
10169 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10170 if (!CvISXSUB(dstr))
662fb8b2
NC
10171 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10172 break;
bfcb3514 10173 }
1d7c1841 10174 }
1d7c1841
GS
10175 }
10176
10177 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10178 ++PL_sv_objcount;
10179
10180 return dstr;
d2d73c3e 10181 }
1d7c1841 10182
645c22ef
DM
10183/* duplicate a context */
10184
1d7c1841 10185PERL_CONTEXT *
a8fc9800 10186Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10187{
10188 PERL_CONTEXT *ncxs;
10189
10190 if (!cxs)
10191 return (PERL_CONTEXT*)NULL;
10192
10193 /* look for it in the table first */
10194 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10195 if (ncxs)
10196 return ncxs;
10197
10198 /* create anew and remember what it is */
a02a5408 10199 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10200 ptr_table_store(PL_ptr_table, cxs, ncxs);
10201
10202 while (ix >= 0) {
c445ea15
AL
10203 PERL_CONTEXT * const cx = &cxs[ix];
10204 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10205 ncx->cx_type = cx->cx_type;
10206 if (CxTYPE(cx) == CXt_SUBST) {
10207 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10208 }
10209 else {
10210 ncx->blk_oldsp = cx->blk_oldsp;
10211 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10212 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10213 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10214 ncx->blk_oldpm = cx->blk_oldpm;
10215 ncx->blk_gimme = cx->blk_gimme;
10216 switch (CxTYPE(cx)) {
10217 case CXt_SUB:
10218 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10219 ? cv_dup_inc(cx->blk_sub.cv, param)
10220 : cv_dup(cx->blk_sub.cv,param));
cc8d50a7 10221 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10222 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10223 : NULL);
d2d73c3e 10224 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841 10225 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
cc8d50a7
NC
10226 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10227 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10228 ncx->blk_sub.retop = cx->blk_sub.retop;
d8d97e70
DM
10229 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
10230 cx->blk_sub.oldcomppad);
1d7c1841
GS
10231 break;
10232 case CXt_EVAL:
10233 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10234 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10235 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10236 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10237 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10238 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10239 break;
10240 case CXt_LOOP:
10241 ncx->blk_loop.label = cx->blk_loop.label;
10242 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
022eaa24 10243 ncx->blk_loop.my_op = cx->blk_loop.my_op;
1d7c1841
GS
10244 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10245 ? cx->blk_loop.iterdata
d2d73c3e 10246 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10247 ncx->blk_loop.oldcomppad
10248 = (PAD*)ptr_table_fetch(PL_ptr_table,
10249 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10250 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10251 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10252 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10253 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10254 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10255 break;
10256 case CXt_FORMAT:
d2d73c3e
AB
10257 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10258 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10259 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
cc8d50a7 10260 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10261 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10262 break;
10263 case CXt_BLOCK:
10264 case CXt_NULL:
10265 break;
10266 }
10267 }
10268 --ix;
10269 }
10270 return ncxs;
10271}
10272
645c22ef
DM
10273/* duplicate a stack info structure */
10274
1d7c1841 10275PERL_SI *
a8fc9800 10276Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10277{
10278 PERL_SI *nsi;
10279
10280 if (!si)
10281 return (PERL_SI*)NULL;
10282
10283 /* look for it in the table first */
10284 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10285 if (nsi)
10286 return nsi;
10287
10288 /* create anew and remember what it is */
a02a5408 10289 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10290 ptr_table_store(PL_ptr_table, si, nsi);
10291
d2d73c3e 10292 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10293 nsi->si_cxix = si->si_cxix;
10294 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10295 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10296 nsi->si_type = si->si_type;
d2d73c3e
AB
10297 nsi->si_prev = si_dup(si->si_prev, param);
10298 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10299 nsi->si_markoff = si->si_markoff;
10300
10301 return nsi;
10302}
10303
10304#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10305#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10306#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10307#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10308#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10309#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10310#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10311#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10312#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10313#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10314#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10315#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10316#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10317#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10318
10319/* XXXXX todo */
10320#define pv_dup_inc(p) SAVEPV(p)
10321#define pv_dup(p) SAVEPV(p)
10322#define svp_dup_inc(p,pp) any_dup(p,pp)
10323
645c22ef
DM
10324/* map any object to the new equivent - either something in the
10325 * ptr table, or something in the interpreter structure
10326 */
10327
1d7c1841 10328void *
53c1dcc0 10329Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10330{
10331 void *ret;
10332
10333 if (!v)
10334 return (void*)NULL;
10335
10336 /* look for it in the table first */
10337 ret = ptr_table_fetch(PL_ptr_table, v);
10338 if (ret)
10339 return ret;
10340
10341 /* see if it is part of the interpreter structure */
10342 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10343 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10344 else {
1d7c1841 10345 ret = v;
05ec9bb3 10346 }
1d7c1841
GS
10347
10348 return ret;
10349}
10350
645c22ef
DM
10351/* duplicate the save stack */
10352
1d7c1841 10353ANY *
a8fc9800 10354Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10355{
53d44271 10356 dVAR;
53c1dcc0
AL
10357 ANY * const ss = proto_perl->Tsavestack;
10358 const I32 max = proto_perl->Tsavestack_max;
10359 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10360 ANY *nss;
10361 SV *sv;
10362 GV *gv;
10363 AV *av;
10364 HV *hv;
10365 void* ptr;
10366 int intval;
10367 long longval;
10368 GP *gp;
10369 IV iv;
b24356f5 10370 I32 i;
c4e33207 10371 char *c = NULL;
1d7c1841 10372 void (*dptr) (void*);
acfe0abc 10373 void (*dxptr) (pTHX_ void*);
1d7c1841 10374
a02a5408 10375 Newxz(nss, max, ANY);
1d7c1841
GS
10376
10377 while (ix > 0) {
b24356f5
NC
10378 const I32 type = POPINT(ss,ix);
10379 TOPINT(nss,ix) = type;
10380 switch (type) {
3e07292d
NC
10381 case SAVEt_HELEM: /* hash element */
10382 sv = (SV*)POPPTR(ss,ix);
10383 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
10384 /* fall through */
1d7c1841 10385 case SAVEt_ITEM: /* normal string */
a41cc44e 10386 case SAVEt_SV: /* scalar reference */
1d7c1841 10387 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10388 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10389 /* fall through */
10390 case SAVEt_FREESV:
10391 case SAVEt_MORTALIZESV:
1d7c1841 10392 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10393 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10394 break;
05ec9bb3
NIS
10395 case SAVEt_SHARED_PVREF: /* char* in shared space */
10396 c = (char*)POPPTR(ss,ix);
10397 TOPPTR(nss,ix) = savesharedpv(c);
10398 ptr = POPPTR(ss,ix);
10399 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10400 break;
1d7c1841
GS
10401 case SAVEt_GENERIC_SVREF: /* generic sv */
10402 case SAVEt_SVREF: /* scalar reference */
10403 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10404 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10405 ptr = POPPTR(ss,ix);
10406 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10407 break;
a41cc44e 10408 case SAVEt_HV: /* hash reference */
1d7c1841 10409 case SAVEt_AV: /* array reference */
11b79775 10410 sv = (SV*) POPPTR(ss,ix);
337d28f5 10411 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
10412 /* fall through */
10413 case SAVEt_COMPPAD:
10414 case SAVEt_NSTAB:
667e2948 10415 sv = (SV*) POPPTR(ss,ix);
3e07292d 10416 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10417 break;
10418 case SAVEt_INT: /* int reference */
10419 ptr = POPPTR(ss,ix);
10420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10421 intval = (int)POPINT(ss,ix);
10422 TOPINT(nss,ix) = intval;
10423 break;
10424 case SAVEt_LONG: /* long reference */
10425 ptr = POPPTR(ss,ix);
10426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
3e07292d
NC
10427 /* fall through */
10428 case SAVEt_CLEARSV:
1d7c1841
GS
10429 longval = (long)POPLONG(ss,ix);
10430 TOPLONG(nss,ix) = longval;
10431 break;
10432 case SAVEt_I32: /* I32 reference */
10433 case SAVEt_I16: /* I16 reference */
10434 case SAVEt_I8: /* I8 reference */
88effcc9 10435 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10436 ptr = POPPTR(ss,ix);
10437 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
1ccabee8 10438 i = POPINT(ss,ix);
1d7c1841
GS
10439 TOPINT(nss,ix) = i;
10440 break;
10441 case SAVEt_IV: /* IV reference */
10442 ptr = POPPTR(ss,ix);
10443 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10444 iv = POPIV(ss,ix);
10445 TOPIV(nss,ix) = iv;
10446 break;
a41cc44e
NC
10447 case SAVEt_HPTR: /* HV* reference */
10448 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10449 case SAVEt_SPTR: /* SV* reference */
10450 ptr = POPPTR(ss,ix);
10451 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10452 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10453 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10454 break;
10455 case SAVEt_VPTR: /* random* reference */
10456 ptr = POPPTR(ss,ix);
10457 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10458 ptr = POPPTR(ss,ix);
10459 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10460 break;
b03d03b0 10461 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10462 case SAVEt_PPTR: /* char* reference */
10463 ptr = POPPTR(ss,ix);
10464 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10465 c = (char*)POPPTR(ss,ix);
10466 TOPPTR(nss,ix) = pv_dup(c);
10467 break;
1d7c1841
GS
10468 case SAVEt_GP: /* scalar reference */
10469 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10470 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10471 (void)GpREFCNT_inc(gp);
10472 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10473 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841 10474 break;
1d7c1841
GS
10475 case SAVEt_FREEOP:
10476 ptr = POPPTR(ss,ix);
10477 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10478 /* these are assumed to be refcounted properly */
53c1dcc0 10479 OP *o;
1d7c1841
GS
10480 switch (((OP*)ptr)->op_type) {
10481 case OP_LEAVESUB:
10482 case OP_LEAVESUBLV:
10483 case OP_LEAVEEVAL:
10484 case OP_LEAVE:
10485 case OP_SCOPE:
10486 case OP_LEAVEWRITE:
e977893f
GS
10487 TOPPTR(nss,ix) = ptr;
10488 o = (OP*)ptr;
d3c72c2a 10489 OP_REFCNT_LOCK;
e977893f 10490 OpREFCNT_inc(o);
d3c72c2a 10491 OP_REFCNT_UNLOCK;
1d7c1841
GS
10492 break;
10493 default:
5f66b61c 10494 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10495 break;
10496 }
10497 }
10498 else
5f66b61c 10499 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10500 break;
10501 case SAVEt_FREEPV:
10502 c = (char*)POPPTR(ss,ix);
10503 TOPPTR(nss,ix) = pv_dup_inc(c);
10504 break;
1d7c1841
GS
10505 case SAVEt_DELETE:
10506 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10507 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10508 c = (char*)POPPTR(ss,ix);
10509 TOPPTR(nss,ix) = pv_dup_inc(c);
3e07292d
NC
10510 /* fall through */
10511 case SAVEt_STACK_POS: /* Position on Perl stack */
1d7c1841
GS
10512 i = POPINT(ss,ix);
10513 TOPINT(nss,ix) = i;
10514 break;
10515 case SAVEt_DESTRUCTOR:
10516 ptr = POPPTR(ss,ix);
10517 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10518 dptr = POPDPTR(ss,ix);
8141890a
JH
10519 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10520 any_dup(FPTR2DPTR(void *, dptr),
10521 proto_perl));
1d7c1841
GS
10522 break;
10523 case SAVEt_DESTRUCTOR_X:
10524 ptr = POPPTR(ss,ix);
10525 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10526 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10527 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10528 any_dup(FPTR2DPTR(void *, dxptr),
10529 proto_perl));
1d7c1841
GS
10530 break;
10531 case SAVEt_REGCONTEXT:
10532 case SAVEt_ALLOC:
10533 i = POPINT(ss,ix);
10534 TOPINT(nss,ix) = i;
10535 ix -= i;
10536 break;
1d7c1841
GS
10537 case SAVEt_AELEM: /* array element */
10538 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10539 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10540 i = POPINT(ss,ix);
10541 TOPINT(nss,ix) = i;
10542 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10543 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841 10544 break;
1d7c1841
GS
10545 case SAVEt_OP:
10546 ptr = POPPTR(ss,ix);
10547 TOPPTR(nss,ix) = ptr;
10548 break;
10549 case SAVEt_HINTS:
10550 i = POPINT(ss,ix);
10551 TOPINT(nss,ix) = i;
b3ca2e83 10552 ptr = POPPTR(ss,ix);
080ac856 10553 if (ptr) {
7b6dd8c3 10554 HINTS_REFCNT_LOCK;
080ac856 10555 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10556 HINTS_REFCNT_UNLOCK;
10557 }
cbb1fbea 10558 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10559 if (i & HINT_LOCALIZE_HH) {
10560 hv = (HV*)POPPTR(ss,ix);
10561 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10562 }
1d7c1841 10563 break;
c3564e5c
GS
10564 case SAVEt_PADSV:
10565 longval = (long)POPLONG(ss,ix);
10566 TOPLONG(nss,ix) = longval;
10567 ptr = POPPTR(ss,ix);
10568 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10569 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10570 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10571 break;
a1bb4754 10572 case SAVEt_BOOL:
38d8b13e 10573 ptr = POPPTR(ss,ix);
b9609c01 10574 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10575 longval = (long)POPBOOL(ss,ix);
b9609c01 10576 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10577 break;
8bd2680e
MHM
10578 case SAVEt_SET_SVFLAGS:
10579 i = POPINT(ss,ix);
10580 TOPINT(nss,ix) = i;
10581 i = POPINT(ss,ix);
10582 TOPINT(nss,ix) = i;
10583 sv = (SV*)POPPTR(ss,ix);
10584 TOPPTR(nss,ix) = sv_dup(sv, param);
10585 break;
5bfb7d0e
NC
10586 case SAVEt_RE_STATE:
10587 {
10588 const struct re_save_state *const old_state
10589 = (struct re_save_state *)
10590 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10591 struct re_save_state *const new_state
10592 = (struct re_save_state *)
10593 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10594
10595 Copy(old_state, new_state, 1, struct re_save_state);
10596 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10597
10598 new_state->re_state_bostr
10599 = pv_dup(old_state->re_state_bostr);
10600 new_state->re_state_reginput
10601 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10602 new_state->re_state_regeol
10603 = pv_dup(old_state->re_state_regeol);
10604 new_state->re_state_regstartp
11b79775 10605 = (I32*) any_dup(old_state->re_state_regstartp, proto_perl);
5bfb7d0e 10606 new_state->re_state_regendp
11b79775 10607 = (I32*) any_dup(old_state->re_state_regendp, proto_perl);
5bfb7d0e 10608 new_state->re_state_reglastparen
11b79775
DD
10609 = (U32*) any_dup(old_state->re_state_reglastparen,
10610 proto_perl);
5bfb7d0e 10611 new_state->re_state_reglastcloseparen
11b79775 10612 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
5bfb7d0e 10613 proto_perl);
5bfb7d0e
NC
10614 /* XXX This just has to be broken. The old save_re_context
10615 code did SAVEGENERICPV(PL_reg_start_tmp);
10616 PL_reg_start_tmp is char **.
10617 Look above to what the dup code does for
10618 SAVEt_GENERIC_PVREF
10619 It can never have worked.
10620 So this is merely a faithful copy of the exiting bug: */
10621 new_state->re_state_reg_start_tmp
10622 = (char **) pv_dup((char *)
10623 old_state->re_state_reg_start_tmp);
10624 /* I assume that it only ever "worked" because no-one called
10625 (pseudo)fork while the regexp engine had re-entered itself.
10626 */
5bfb7d0e
NC
10627#ifdef PERL_OLD_COPY_ON_WRITE
10628 new_state->re_state_nrs
10629 = sv_dup(old_state->re_state_nrs, param);
10630#endif
10631 new_state->re_state_reg_magic
11b79775
DD
10632 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
10633 proto_perl);
5bfb7d0e 10634 new_state->re_state_reg_oldcurpm
11b79775
DD
10635 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
10636 proto_perl);
5bfb7d0e 10637 new_state->re_state_reg_curpm
11b79775
DD
10638 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
10639 proto_perl);
5bfb7d0e
NC
10640 new_state->re_state_reg_oldsaved
10641 = pv_dup(old_state->re_state_reg_oldsaved);
10642 new_state->re_state_reg_poscache
10643 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
10644 new_state->re_state_reg_starttry
10645 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
10646 break;
10647 }
68da3b2f
NC
10648 case SAVEt_COMPILE_WARNINGS:
10649 ptr = POPPTR(ss,ix);
10650 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10651 break;
7c197c94
DM
10652 case SAVEt_PARSER:
10653 ptr = POPPTR(ss,ix);
456084a8 10654 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
7c197c94 10655 break;
1d7c1841 10656 default:
147bc374
NC
10657 Perl_croak(aTHX_
10658 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
1d7c1841
GS
10659 }
10660 }
10661
bd81e77b
NC
10662 return nss;
10663}
10664
10665
10666/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10667 * flag to the result. This is done for each stash before cloning starts,
10668 * so we know which stashes want their objects cloned */
10669
10670static void
10671do_mark_cloneable_stash(pTHX_ SV *sv)
10672{
10673 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10674 if (hvname) {
10675 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10676 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10677 if (cloner && GvCV(cloner)) {
10678 dSP;
10679 UV status;
10680
10681 ENTER;
10682 SAVETMPS;
10683 PUSHMARK(SP);
10684 XPUSHs(sv_2mortal(newSVhek(hvname)));
10685 PUTBACK;
10686 call_sv((SV*)GvCV(cloner), G_SCALAR);
10687 SPAGAIN;
10688 status = POPu;
10689 PUTBACK;
10690 FREETMPS;
10691 LEAVE;
10692 if (status)
10693 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10694 }
10695 }
10696}
10697
10698
10699
10700/*
10701=for apidoc perl_clone
10702
10703Create and return a new interpreter by cloning the current one.
10704
10705perl_clone takes these flags as parameters:
10706
10707CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10708without it we only clone the data and zero the stacks,
10709with it we copy the stacks and the new perl interpreter is
10710ready to run at the exact same point as the previous one.
10711The pseudo-fork code uses COPY_STACKS while the
10712threads->new doesn't.
10713
10714CLONEf_KEEP_PTR_TABLE
10715perl_clone keeps a ptr_table with the pointer of the old
10716variable as a key and the new variable as a value,
10717this allows it to check if something has been cloned and not
10718clone it again but rather just use the value and increase the
10719refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10720the ptr_table using the function
10721C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10722reason to keep it around is if you want to dup some of your own
10723variable who are outside the graph perl scans, example of this
10724code is in threads.xs create
10725
10726CLONEf_CLONE_HOST
10727This is a win32 thing, it is ignored on unix, it tells perls
10728win32host code (which is c++) to clone itself, this is needed on
10729win32 if you want to run two threads at the same time,
10730if you just want to do some stuff in a separate perl interpreter
10731and then throw it away and return to the original one,
10732you don't need to do anything.
10733
10734=cut
10735*/
10736
10737/* XXX the above needs expanding by someone who actually understands it ! */
10738EXTERN_C PerlInterpreter *
10739perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10740
10741PerlInterpreter *
10742perl_clone(PerlInterpreter *proto_perl, UV flags)
10743{
10744 dVAR;
10745#ifdef PERL_IMPLICIT_SYS
10746
10747 /* perlhost.h so we need to call into it
10748 to clone the host, CPerlHost should have a c interface, sky */
10749
10750 if (flags & CLONEf_CLONE_HOST) {
10751 return perl_clone_host(proto_perl,flags);
10752 }
10753 return perl_clone_using(proto_perl, flags,
10754 proto_perl->IMem,
10755 proto_perl->IMemShared,
10756 proto_perl->IMemParse,
10757 proto_perl->IEnv,
10758 proto_perl->IStdIO,
10759 proto_perl->ILIO,
10760 proto_perl->IDir,
10761 proto_perl->ISock,
10762 proto_perl->IProc);
10763}
10764
10765PerlInterpreter *
10766perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10767 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10768 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10769 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10770 struct IPerlDir* ipD, struct IPerlSock* ipS,
10771 struct IPerlProc* ipP)
10772{
10773 /* XXX many of the string copies here can be optimized if they're
10774 * constants; they need to be allocated as common memory and just
10775 * their pointers copied. */
10776
10777 IV i;
10778 CLONE_PARAMS clone_params;
5f66b61c 10779 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10780
5f66b61c 10781 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10782 /* for each stash, determine whether its objects should be cloned */
10783 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10784 PERL_SET_THX(my_perl);
10785
10786# ifdef DEBUGGING
7e337ee0 10787 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10788 PL_op = NULL;
10789 PL_curcop = NULL;
bd81e77b
NC
10790 PL_markstack = 0;
10791 PL_scopestack = 0;
10792 PL_savestack = 0;
10793 PL_savestack_ix = 0;
10794 PL_savestack_max = -1;
10795 PL_sig_pending = 0;
10796 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10797# else /* !DEBUGGING */
10798 Zero(my_perl, 1, PerlInterpreter);
10799# endif /* DEBUGGING */
10800
10801 /* host pointers */
10802 PL_Mem = ipM;
10803 PL_MemShared = ipMS;
10804 PL_MemParse = ipMP;
10805 PL_Env = ipE;
10806 PL_StdIO = ipStd;
10807 PL_LIO = ipLIO;
10808 PL_Dir = ipD;
10809 PL_Sock = ipS;
10810 PL_Proc = ipP;
10811#else /* !PERL_IMPLICIT_SYS */
10812 IV i;
10813 CLONE_PARAMS clone_params;
10814 CLONE_PARAMS* param = &clone_params;
5f66b61c 10815 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10816 /* for each stash, determine whether its objects should be cloned */
10817 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10818 PERL_SET_THX(my_perl);
10819
10820# ifdef DEBUGGING
7e337ee0 10821 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10822 PL_op = NULL;
10823 PL_curcop = NULL;
bd81e77b
NC
10824 PL_markstack = 0;
10825 PL_scopestack = 0;
10826 PL_savestack = 0;
10827 PL_savestack_ix = 0;
10828 PL_savestack_max = -1;
10829 PL_sig_pending = 0;
10830 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10831# else /* !DEBUGGING */
10832 Zero(my_perl, 1, PerlInterpreter);
10833# endif /* DEBUGGING */
10834#endif /* PERL_IMPLICIT_SYS */
10835 param->flags = flags;
10836 param->proto_perl = proto_perl;
10837
7cb608b5
NC
10838 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10839
fdda85ca 10840 PL_body_arenas = NULL;
bd81e77b
NC
10841 Zero(&PL_body_roots, 1, PL_body_roots);
10842
10843 PL_nice_chunk = NULL;
10844 PL_nice_chunk_size = 0;
10845 PL_sv_count = 0;
10846 PL_sv_objcount = 0;
a0714e2c
SS
10847 PL_sv_root = NULL;
10848 PL_sv_arenaroot = NULL;
bd81e77b
NC
10849
10850 PL_debug = proto_perl->Idebug;
10851
10852 PL_hash_seed = proto_perl->Ihash_seed;
10853 PL_rehash_seed = proto_perl->Irehash_seed;
10854
10855#ifdef USE_REENTRANT_API
10856 /* XXX: things like -Dm will segfault here in perlio, but doing
10857 * PERL_SET_CONTEXT(proto_perl);
10858 * breaks too many other things
10859 */
10860 Perl_reentrant_init(aTHX);
10861#endif
10862
10863 /* create SV map for pointer relocation */
10864 PL_ptr_table = ptr_table_new();
10865
10866 /* initialize these special pointers as early as possible */
10867 SvANY(&PL_sv_undef) = NULL;
10868 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10869 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10870 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10871
10872 SvANY(&PL_sv_no) = new_XPVNV();
10873 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10874 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10875 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10876 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
10877 SvCUR_set(&PL_sv_no, 0);
10878 SvLEN_set(&PL_sv_no, 1);
10879 SvIV_set(&PL_sv_no, 0);
10880 SvNV_set(&PL_sv_no, 0);
10881 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10882
10883 SvANY(&PL_sv_yes) = new_XPVNV();
10884 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10885 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10886 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10887 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
10888 SvCUR_set(&PL_sv_yes, 1);
10889 SvLEN_set(&PL_sv_yes, 2);
10890 SvIV_set(&PL_sv_yes, 1);
10891 SvNV_set(&PL_sv_yes, 1);
10892 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10893
10894 /* create (a non-shared!) shared string table */
10895 PL_strtab = newHV();
10896 HvSHAREKEYS_off(PL_strtab);
10897 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10898 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10899
10900 PL_compiling = proto_perl->Icompiling;
10901
10902 /* These two PVs will be free'd special way so must set them same way op.c does */
10903 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10904 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10905
10906 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10907 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10908
10909 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10910 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
c28fe1ec 10911 if (PL_compiling.cop_hints_hash) {
cbb1fbea 10912 HINTS_REFCNT_LOCK;
c28fe1ec 10913 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
cbb1fbea
NC
10914 HINTS_REFCNT_UNLOCK;
10915 }
bd81e77b
NC
10916 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10917
10918 /* pseudo environmental stuff */
10919 PL_origargc = proto_perl->Iorigargc;
10920 PL_origargv = proto_perl->Iorigargv;
10921
10922 param->stashes = newAV(); /* Setup array of objects to call clone on */
10923
10924 /* Set tainting stuff before PerlIO_debug can possibly get called */
10925 PL_tainting = proto_perl->Itainting;
10926 PL_taint_warn = proto_perl->Itaint_warn;
10927
10928#ifdef PERLIO_LAYERS
10929 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10930 PerlIO_clone(aTHX_ proto_perl, param);
10931#endif
10932
10933 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10934 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10935 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10936 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10937 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10938 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10939
10940 /* switches */
10941 PL_minus_c = proto_perl->Iminus_c;
10942 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10943 PL_localpatches = proto_perl->Ilocalpatches;
10944 PL_splitstr = proto_perl->Isplitstr;
10945 PL_preprocess = proto_perl->Ipreprocess;
10946 PL_minus_n = proto_perl->Iminus_n;
10947 PL_minus_p = proto_perl->Iminus_p;
10948 PL_minus_l = proto_perl->Iminus_l;
10949 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10950 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10951 PL_minus_F = proto_perl->Iminus_F;
10952 PL_doswitches = proto_perl->Idoswitches;
10953 PL_dowarn = proto_perl->Idowarn;
10954 PL_doextract = proto_perl->Idoextract;
10955 PL_sawampersand = proto_perl->Isawampersand;
10956 PL_unsafe = proto_perl->Iunsafe;
10957 PL_inplace = SAVEPV(proto_perl->Iinplace);
10958 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10959 PL_perldb = proto_perl->Iperldb;
10960 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10961 PL_exit_flags = proto_perl->Iexit_flags;
10962
10963 /* magical thingies */
10964 /* XXX time(&PL_basetime) when asked for? */
10965 PL_basetime = proto_perl->Ibasetime;
10966 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
10967
10968 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
10969 PL_statusvalue = proto_perl->Istatusvalue;
10970#ifdef VMS
10971 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
10972#else
10973 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
10974#endif
10975 PL_encoding = sv_dup(proto_perl->Iencoding, param);
10976
10977 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
10978 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
10979 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
10980
84da74a7 10981
f9f4320a 10982 /* RE engine related */
84da74a7
YO
10983 Zero(&PL_reg_state, 1, struct re_save_state);
10984 PL_reginterp_cnt = 0;
10985 PL_regmatch_slab = NULL;
10986
bd81e77b
NC
10987 /* Clone the regex array */
10988 PL_regex_padav = newAV();
10989 {
10990 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 10991 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 10992 IV i;
7f466ec7 10993 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 10994 for(i = 1; i <= len; i++) {
7a5b473e
AL
10995 const SV * const regex = regexen[i];
10996 SV * const sv =
10997 SvREPADTMP(regex)
10998 ? sv_dup_inc(regex, param)
10999 : SvREFCNT_inc(
f8149455 11000 newSViv(PTR2IV(CALLREGDUPE(
7a5b473e
AL
11001 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11002 ;
60790534
DM
11003 if (SvFLAGS(regex) & SVf_BREAK)
11004 SvFLAGS(sv) |= SVf_BREAK; /* unrefcnted PL_curpm */
7a5b473e 11005 av_push(PL_regex_padav, sv);
bd81e77b
NC
11006 }
11007 }
11008 PL_regex_pad = AvARRAY(PL_regex_padav);
11009
11010 /* shortcuts to various I/O objects */
11011 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11012 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11013 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11014 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11015 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11016 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11017
bd81e77b
NC
11018 /* shortcuts to regexp stuff */
11019 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11020
bd81e77b
NC
11021 /* shortcuts to misc objects */
11022 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11023
bd81e77b
NC
11024 /* shortcuts to debugging objects */
11025 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11026 PL_DBline = gv_dup(proto_perl->IDBline, param);
11027 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11028 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11029 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11030 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11031 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11032 PL_lineary = av_dup(proto_perl->Ilineary, param);
11033 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11034
bd81e77b
NC
11035 /* symbol tables */
11036 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11037 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11038 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11039 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11040 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11041
11042 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11043 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11044 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
3c10abe3
AG
11045 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
11046 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
bd81e77b
NC
11047 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11048 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11049 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11050
11051 PL_sub_generation = proto_perl->Isub_generation;
11052
11053 /* funky return mechanisms */
11054 PL_forkprocess = proto_perl->Iforkprocess;
11055
11056 /* subprocess state */
11057 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11058
11059 /* internal state */
11060 PL_maxo = proto_perl->Imaxo;
11061 if (proto_perl->Iop_mask)
11062 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11063 else
bd61b366 11064 PL_op_mask = NULL;
bd81e77b
NC
11065 /* PL_asserting = proto_perl->Iasserting; */
11066
11067 /* current interpreter roots */
11068 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
d3c72c2a 11069 OP_REFCNT_LOCK;
bd81e77b 11070 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
d3c72c2a 11071 OP_REFCNT_UNLOCK;
bd81e77b
NC
11072 PL_main_start = proto_perl->Imain_start;
11073 PL_eval_root = proto_perl->Ieval_root;
11074 PL_eval_start = proto_perl->Ieval_start;
11075
11076 /* runtime control stuff */
11077 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11078 PL_copline = proto_perl->Icopline;
11079
11080 PL_filemode = proto_perl->Ifilemode;
11081 PL_lastfd = proto_perl->Ilastfd;
11082 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11083 PL_Argv = NULL;
bd61b366 11084 PL_Cmd = NULL;
bd81e77b 11085 PL_gensym = proto_perl->Igensym;
bd81e77b
NC
11086 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11087 PL_laststatval = proto_perl->Ilaststatval;
11088 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11089 PL_mess_sv = NULL;
bd81e77b
NC
11090
11091 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11092
11093 /* interpreter atexit processing */
11094 PL_exitlistlen = proto_perl->Iexitlistlen;
11095 if (PL_exitlistlen) {
11096 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11097 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11098 }
bd81e77b
NC
11099 else
11100 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11101
11102 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11103 if (PL_my_cxt_size) {
f16dd614
DM
11104 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11105 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
53d44271
JH
11106#ifdef PERL_GLOBAL_STRUCT_PRIVATE
11107 Newx(PL_my_cxt_keys, PL_my_cxt_size, char *);
11108 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
11109#endif
f16dd614 11110 }
53d44271 11111 else {
f16dd614 11112 PL_my_cxt_list = (void**)NULL;
53d44271
JH
11113#ifdef PERL_GLOBAL_STRUCT_PRIVATE
11114 PL_my_cxt_keys = (void**)NULL;
11115#endif
11116 }
bd81e77b
NC
11117 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11118 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11119 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11120
11121 PL_profiledata = NULL;
11122 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11123 /* PL_rsfp_filters entries have fake IoDIRP() */
11124 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11125
bd81e77b 11126 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11127
bd81e77b 11128 PAD_CLONE_VARS(proto_perl, param);
9660f481 11129
bd81e77b
NC
11130#ifdef HAVE_INTERP_INTERN
11131 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11132#endif
645c22ef 11133
bd81e77b
NC
11134 /* more statics moved here */
11135 PL_generation = proto_perl->Igeneration;
11136 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11137
bd81e77b
NC
11138 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11139 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11140
bd81e77b
NC
11141 PL_uid = proto_perl->Iuid;
11142 PL_euid = proto_perl->Ieuid;
11143 PL_gid = proto_perl->Igid;
11144 PL_egid = proto_perl->Iegid;
11145 PL_nomemok = proto_perl->Inomemok;
11146 PL_an = proto_perl->Ian;
11147 PL_evalseq = proto_perl->Ievalseq;
11148 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11149 PL_origalen = proto_perl->Iorigalen;
11150#ifdef PERL_USES_PL_PIDSTATUS
11151 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11152#endif
11153 PL_osname = SAVEPV(proto_perl->Iosname);
11154 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11155
bd81e77b 11156 PL_runops = proto_perl->Irunops;
6a78b4db 11157
bd81e77b 11158 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11159
bd81e77b
NC
11160#ifdef CSH
11161 PL_cshlen = proto_perl->Icshlen;
11162 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11163#endif
645c22ef 11164
199e78b7
DM
11165 PL_parser = parser_dup(proto_perl->Iparser, param);
11166
bd81e77b 11167 PL_lex_state = proto_perl->Ilex_state;
645c22ef 11168
5db06880
NC
11169#ifdef PERL_MAD
11170 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
5336380d 11171 PL_curforce = proto_perl->Icurforce;
5db06880 11172#else
bd81e77b
NC
11173 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11174 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11175 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11176#endif
c43294b8 11177
db4997f0
NC
11178 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11179 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11180 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11181 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11182 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11183 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11184 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11185 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11186 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
bd81e77b 11187 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
bd81e77b
NC
11188
11189 PL_expect = proto_perl->Iexpect;
11190
bd81e77b 11191 PL_multi_end = proto_perl->Imulti_end;
bd81e77b
NC
11192
11193 PL_error_count = proto_perl->Ierror_count;
11194 PL_subline = proto_perl->Isubline;
11195 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11196
db4997f0
NC
11197 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11198 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11199 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11200 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11201 PL_last_lop_op = proto_perl->Ilast_lop_op;
bd81e77b
NC
11202 PL_in_my = proto_perl->Iin_my;
11203 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11204#ifdef FCRYPT
11205 PL_cryptseen = proto_perl->Icryptseen;
11206#endif
1d7c1841 11207
bd81e77b 11208 PL_hints = proto_perl->Ihints;
1d7c1841 11209
bd81e77b 11210 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11211
bd81e77b
NC
11212#ifdef USE_LOCALE_COLLATE
11213 PL_collation_ix = proto_perl->Icollation_ix;
11214 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11215 PL_collation_standard = proto_perl->Icollation_standard;
11216 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11217 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11218#endif /* USE_LOCALE_COLLATE */
1d7c1841 11219
bd81e77b
NC
11220#ifdef USE_LOCALE_NUMERIC
11221 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11222 PL_numeric_standard = proto_perl->Inumeric_standard;
11223 PL_numeric_local = proto_perl->Inumeric_local;
11224 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11225#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11226
bd81e77b
NC
11227 /* utf8 character classes */
11228 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11229 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11230 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11231 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11232 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11233 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11234 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11235 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11236 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11237 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11238 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11239 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11240 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11241 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11242 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11243 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11244 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11245 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11246 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11247 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11248
bd81e77b
NC
11249 /* Did the locale setup indicate UTF-8? */
11250 PL_utf8locale = proto_perl->Iutf8locale;
11251 /* Unicode features (see perlrun/-C) */
11252 PL_unicode = proto_perl->Iunicode;
1d7c1841 11253
bd81e77b
NC
11254 /* Pre-5.8 signals control */
11255 PL_signals = proto_perl->Isignals;
1d7c1841 11256
bd81e77b
NC
11257 /* times() ticks per second */
11258 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11259
bd81e77b
NC
11260 /* Recursion stopper for PerlIO_find_layer */
11261 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11262
bd81e77b
NC
11263 /* sort() routine */
11264 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11265
bd81e77b
NC
11266 /* Not really needed/useful since the reenrant_retint is "volatile",
11267 * but do it for consistency's sake. */
11268 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11269
bd81e77b
NC
11270 /* Hooks to shared SVs and locks. */
11271 PL_sharehook = proto_perl->Isharehook;
11272 PL_lockhook = proto_perl->Ilockhook;
11273 PL_unlockhook = proto_perl->Iunlockhook;
11274 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11275
bd81e77b
NC
11276 PL_runops_std = proto_perl->Irunops_std;
11277 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11278
bd81e77b
NC
11279#ifdef THREADS_HAVE_PIDS
11280 PL_ppid = proto_perl->Ippid;
11281#endif
1d7c1841 11282
bd81e77b 11283 /* swatch cache */
5c284bb0 11284 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11285 PL_last_swash_klen = 0;
11286 PL_last_swash_key[0]= '\0';
11287 PL_last_swash_tmps = (U8*)NULL;
11288 PL_last_swash_slen = 0;
1d7c1841 11289
bd81e77b
NC
11290 PL_glob_index = proto_perl->Iglob_index;
11291 PL_srand_called = proto_perl->Isrand_called;
11b79775 11292 PL_uudmap[(U32) 'M'] = 0; /* reinits on demand */
bd61b366 11293 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11294
bd81e77b
NC
11295 if (proto_perl->Ipsig_pend) {
11296 Newxz(PL_psig_pend, SIG_SIZE, int);
11297 }
11298 else {
11299 PL_psig_pend = (int*)NULL;
11300 }
05ec9bb3 11301
bd81e77b
NC
11302 if (proto_perl->Ipsig_ptr) {
11303 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11304 Newxz(PL_psig_name, SIG_SIZE, SV*);
11305 for (i = 1; i < SIG_SIZE; i++) {
11306 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11307 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11308 }
11309 }
11310 else {
11311 PL_psig_ptr = (SV**)NULL;
11312 PL_psig_name = (SV**)NULL;
11313 }
05ec9bb3 11314
bd81e77b 11315 /* thrdvar.h stuff */
1d7c1841 11316
bd81e77b
NC
11317 if (flags & CLONEf_COPY_STACKS) {
11318 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11319 PL_tmps_ix = proto_perl->Ttmps_ix;
11320 PL_tmps_max = proto_perl->Ttmps_max;
11321 PL_tmps_floor = proto_perl->Ttmps_floor;
11322 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11323 i = 0;
11324 while (i <= PL_tmps_ix) {
11325 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11326 ++i;
11327 }
d2d73c3e 11328
bd81e77b
NC
11329 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11330 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11331 Newxz(PL_markstack, i, I32);
11332 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11333 - proto_perl->Tmarkstack);
11334 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11335 - proto_perl->Tmarkstack);
11336 Copy(proto_perl->Tmarkstack, PL_markstack,
11337 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11338
bd81e77b
NC
11339 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11340 * NOTE: unlike the others! */
11341 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11342 PL_scopestack_max = proto_perl->Tscopestack_max;
11343 Newxz(PL_scopestack, PL_scopestack_max, I32);
11344 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11345
bd81e77b
NC
11346 /* NOTE: si_dup() looks at PL_markstack */
11347 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11348
bd81e77b
NC
11349 /* PL_curstack = PL_curstackinfo->si_stack; */
11350 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11351 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11352
bd81e77b
NC
11353 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11354 PL_stack_base = AvARRAY(PL_curstack);
11355 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11356 - proto_perl->Tstack_base);
11357 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11358
bd81e77b
NC
11359 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11360 * NOTE: unlike the others! */
11361 PL_savestack_ix = proto_perl->Tsavestack_ix;
11362 PL_savestack_max = proto_perl->Tsavestack_max;
11363 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11364 PL_savestack = ss_dup(proto_perl, param);
11365 }
11366 else {
11367 init_stacks();
11368 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11369
11370 /* although we're not duplicating the tmps stack, we should still
11371 * add entries for any SVs on the tmps stack that got cloned by a
11372 * non-refcount means (eg a temp in @_); otherwise they will be
11373 * orphaned
11374 */
11375 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11376 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11377 proto_perl->Ttmps_stack[i]);
11378 if (nsv && !SvREFCNT(nsv)) {
11379 EXTEND_MORTAL(1);
b37c2d43 11380 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11381 }
11382 }
bd81e77b 11383 }
1d7c1841 11384
bd81e77b
NC
11385 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11386 PL_top_env = &PL_start_env;
1d7c1841 11387
bd81e77b 11388 PL_op = proto_perl->Top;
4a4c6fe3 11389
a0714e2c 11390 PL_Sv = NULL;
bd81e77b
NC
11391 PL_Xpv = (XPV*)NULL;
11392 PL_na = proto_perl->Tna;
1fcf4c12 11393
bd81e77b
NC
11394 PL_statbuf = proto_perl->Tstatbuf;
11395 PL_statcache = proto_perl->Tstatcache;
11396 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11397 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11398#ifdef HAS_TIMES
11399 PL_timesbuf = proto_perl->Ttimesbuf;
11400#endif
1d7c1841 11401
bd81e77b
NC
11402 PL_tainted = proto_perl->Ttainted;
11403 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11404 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11405 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11406 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11407 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11408 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11409 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11410 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11411 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11412
bd81e77b
NC
11413 PL_restartop = proto_perl->Trestartop;
11414 PL_in_eval = proto_perl->Tin_eval;
11415 PL_delaymagic = proto_perl->Tdelaymagic;
11416 PL_dirty = proto_perl->Tdirty;
11417 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11418
bd81e77b 11419 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11420 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11421 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11422 PL_lastgotoprobe = NULL;
bd81e77b 11423 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11424
bd81e77b
NC
11425 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11426 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11427 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11428 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11429 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11430 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11431
bd81e77b 11432 /* regex stuff */
1d7c1841 11433
bd81e77b
NC
11434 PL_screamfirst = NULL;
11435 PL_screamnext = NULL;
11436 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11437 PL_lastscream = NULL;
1d7c1841 11438
bd81e77b 11439 PL_watchaddr = NULL;
bd61b366 11440 PL_watchok = NULL;
1d7c1841 11441
bd81e77b 11442 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11443 PL_colorset = 0; /* reinits PL_colors[] */
11444 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11445
84da74a7 11446
1d7c1841 11447
bd81e77b
NC
11448 /* Pluggable optimizer */
11449 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11450
bd81e77b 11451 PL_stashcache = newHV();
1d7c1841 11452
bd81e77b
NC
11453 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11454 ptr_table_free(PL_ptr_table);
11455 PL_ptr_table = NULL;
11456 }
1d7c1841 11457
bd81e77b
NC
11458 /* Call the ->CLONE method, if it exists, for each of the stashes
11459 identified by sv_dup() above.
11460 */
11461 while(av_len(param->stashes) != -1) {
11462 HV* const stash = (HV*) av_shift(param->stashes);
11463 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11464 if (cloner && GvCV(cloner)) {
11465 dSP;
11466 ENTER;
11467 SAVETMPS;
11468 PUSHMARK(SP);
11469 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11470 PUTBACK;
11471 call_sv((SV*)GvCV(cloner), G_DISCARD);
11472 FREETMPS;
11473 LEAVE;
11474 }
1d7c1841 11475 }
1d7c1841 11476
bd81e77b 11477 SvREFCNT_dec(param->stashes);
1d7c1841 11478
bd81e77b
NC
11479 /* orphaned? eg threads->new inside BEGIN or use */
11480 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11481 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11482 SAVEFREESV(PL_compcv);
11483 }
dd2155a4 11484
bd81e77b
NC
11485 return my_perl;
11486}
1d7c1841 11487
bd81e77b 11488#endif /* USE_ITHREADS */
1d7c1841 11489
bd81e77b
NC
11490/*
11491=head1 Unicode Support
1d7c1841 11492
bd81e77b 11493=for apidoc sv_recode_to_utf8
1d7c1841 11494
bd81e77b
NC
11495The encoding is assumed to be an Encode object, on entry the PV
11496of the sv is assumed to be octets in that encoding, and the sv
11497will be converted into Unicode (and UTF-8).
1d7c1841 11498
bd81e77b
NC
11499If the sv already is UTF-8 (or if it is not POK), or if the encoding
11500is not a reference, nothing is done to the sv. If the encoding is not
11501an C<Encode::XS> Encoding object, bad things will happen.
11502(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11503
bd81e77b 11504The PV of the sv is returned.
1d7c1841 11505
bd81e77b 11506=cut */
1d7c1841 11507
bd81e77b
NC
11508char *
11509Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11510{
11511 dVAR;
11512 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11513 SV *uni;
11514 STRLEN len;
11515 const char *s;
11516 dSP;
11517 ENTER;
11518 SAVETMPS;
11519 save_re_context();
11520 PUSHMARK(sp);
11521 EXTEND(SP, 3);
11522 XPUSHs(encoding);
11523 XPUSHs(sv);
11524/*
11525 NI-S 2002/07/09
11526 Passing sv_yes is wrong - it needs to be or'ed set of constants
11527 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11528 remove converted chars from source.
1d7c1841 11529
bd81e77b 11530 Both will default the value - let them.
1d7c1841 11531
bd81e77b
NC
11532 XPUSHs(&PL_sv_yes);
11533*/
11534 PUTBACK;
11535 call_method("decode", G_SCALAR);
11536 SPAGAIN;
11537 uni = POPs;
11538 PUTBACK;
11539 s = SvPV_const(uni, len);
11540 if (s != SvPVX_const(sv)) {
11541 SvGROW(sv, len + 1);
11542 Move(s, SvPVX(sv), len + 1, char);
11543 SvCUR_set(sv, len);
11544 }
11545 FREETMPS;
11546 LEAVE;
11547 SvUTF8_on(sv);
11548 return SvPVX(sv);
389edf32 11549 }
bd81e77b
NC
11550 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11551}
1d7c1841 11552
bd81e77b
NC
11553/*
11554=for apidoc sv_cat_decode
1d7c1841 11555
bd81e77b
NC
11556The encoding is assumed to be an Encode object, the PV of the ssv is
11557assumed to be octets in that encoding and decoding the input starts
11558from the position which (PV + *offset) pointed to. The dsv will be
11559concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11560when the string tstr appears in decoding output or the input ends on
11561the PV of the ssv. The value which the offset points will be modified
11562to the last input position on the ssv.
1d7c1841 11563
bd81e77b 11564Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11565
bd81e77b
NC
11566=cut */
11567
11568bool
11569Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11570 SV *ssv, int *offset, char *tstr, int tlen)
11571{
11572 dVAR;
11573 bool ret = FALSE;
11574 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11575 SV *offsv;
11576 dSP;
11577 ENTER;
11578 SAVETMPS;
11579 save_re_context();
11580 PUSHMARK(sp);
11581 EXTEND(SP, 6);
11582 XPUSHs(encoding);
11583 XPUSHs(dsv);
11584 XPUSHs(ssv);
11585 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11586 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11587 PUTBACK;
11588 call_method("cat_decode", G_SCALAR);
11589 SPAGAIN;
11590 ret = SvTRUE(TOPs);
11591 *offset = SvIV(offsv);
11592 PUTBACK;
11593 FREETMPS;
11594 LEAVE;
389edf32 11595 }
bd81e77b
NC
11596 else
11597 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11598 return ret;
1d7c1841 11599
bd81e77b 11600}
1d7c1841 11601
bd81e77b
NC
11602/* ---------------------------------------------------------------------
11603 *
11604 * support functions for report_uninit()
11605 */
1d7c1841 11606
bd81e77b
NC
11607/* the maxiumum size of array or hash where we will scan looking
11608 * for the undefined element that triggered the warning */
1d7c1841 11609
bd81e77b 11610#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11611
bd81e77b
NC
11612/* Look for an entry in the hash whose value has the same SV as val;
11613 * If so, return a mortal copy of the key. */
1d7c1841 11614
bd81e77b
NC
11615STATIC SV*
11616S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11617{
11618 dVAR;
11619 register HE **array;
11620 I32 i;
6c3182a5 11621
bd81e77b
NC
11622 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11623 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11624 return NULL;
6c3182a5 11625
bd81e77b 11626 array = HvARRAY(hv);
6c3182a5 11627
bd81e77b
NC
11628 for (i=HvMAX(hv); i>0; i--) {
11629 register HE *entry;
11630 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11631 if (HeVAL(entry) != val)
11632 continue;
11633 if ( HeVAL(entry) == &PL_sv_undef ||
11634 HeVAL(entry) == &PL_sv_placeholder)
11635 continue;
11636 if (!HeKEY(entry))
a0714e2c 11637 return NULL;
bd81e77b
NC
11638 if (HeKLEN(entry) == HEf_SVKEY)
11639 return sv_mortalcopy(HeKEY_sv(entry));
11640 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11641 }
11642 }
a0714e2c 11643 return NULL;
bd81e77b 11644}
6c3182a5 11645
bd81e77b
NC
11646/* Look for an entry in the array whose value has the same SV as val;
11647 * If so, return the index, otherwise return -1. */
6c3182a5 11648
bd81e77b
NC
11649STATIC I32
11650S_find_array_subscript(pTHX_ AV *av, SV* val)
11651{
97aff369 11652 dVAR;
bd81e77b
NC
11653 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11654 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11655 return -1;
57c6e6d2 11656
4a021917
AL
11657 if (val != &PL_sv_undef) {
11658 SV ** const svp = AvARRAY(av);
11659 I32 i;
11660
11661 for (i=AvFILLp(av); i>=0; i--)
11662 if (svp[i] == val)
11663 return i;
bd81e77b
NC
11664 }
11665 return -1;
11666}
15a5279a 11667
bd81e77b
NC
11668/* S_varname(): return the name of a variable, optionally with a subscript.
11669 * If gv is non-zero, use the name of that global, along with gvtype (one
11670 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11671 * targ. Depending on the value of the subscript_type flag, return:
11672 */
bce260cd 11673
bd81e77b
NC
11674#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11675#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11676#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11677#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11678
bd81e77b
NC
11679STATIC SV*
11680S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11681 SV* keyname, I32 aindex, int subscript_type)
11682{
1d7c1841 11683
bd81e77b
NC
11684 SV * const name = sv_newmortal();
11685 if (gv) {
11686 char buffer[2];
11687 buffer[0] = gvtype;
11688 buffer[1] = 0;
1d7c1841 11689
bd81e77b 11690 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11691
bd81e77b 11692 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11693
bd81e77b
NC
11694 if ((unsigned int)SvPVX(name)[1] <= 26) {
11695 buffer[0] = '^';
11696 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11697
bd81e77b
NC
11698 /* Swap the 1 unprintable control character for the 2 byte pretty
11699 version - ie substr($name, 1, 1) = $buffer; */
11700 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11701 }
bd81e77b
NC
11702 }
11703 else {
11704 U32 unused;
11705 CV * const cv = find_runcv(&unused);
11706 SV *sv;
11707 AV *av;
1d7c1841 11708
bd81e77b 11709 if (!cv || !CvPADLIST(cv))
a0714e2c 11710 return NULL;
bd81e77b
NC
11711 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11712 sv = *av_fetch(av, targ, FALSE);
f8503592 11713 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
bd81e77b 11714 }
1d7c1841 11715
bd81e77b 11716 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11717 SV * const sv = newSV(0);
bd81e77b
NC
11718 *SvPVX(name) = '$';
11719 Perl_sv_catpvf(aTHX_ name, "{%s}",
11720 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11721 SvREFCNT_dec(sv);
11722 }
11723 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11724 *SvPVX(name) = '$';
11725 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11726 }
11727 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11728 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11729
bd81e77b
NC
11730 return name;
11731}
1d7c1841 11732
1d7c1841 11733
bd81e77b
NC
11734/*
11735=for apidoc find_uninit_var
1d7c1841 11736
bd81e77b
NC
11737Find the name of the undefined variable (if any) that caused the operator o
11738to issue a "Use of uninitialized value" warning.
11739If match is true, only return a name if it's value matches uninit_sv.
11740So roughly speaking, if a unary operator (such as OP_COS) generates a
11741warning, then following the direct child of the op may yield an
11742OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11743other hand, with OP_ADD there are two branches to follow, so we only print
11744the variable name if we get an exact match.
1d7c1841 11745
bd81e77b 11746The name is returned as a mortal SV.
1d7c1841 11747
bd81e77b
NC
11748Assumes that PL_op is the op that originally triggered the error, and that
11749PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11750
bd81e77b
NC
11751=cut
11752*/
1d7c1841 11753
bd81e77b
NC
11754STATIC SV *
11755S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11756{
11757 dVAR;
11758 SV *sv;
11759 AV *av;
11760 GV *gv;
11761 OP *o, *o2, *kid;
1d7c1841 11762
bd81e77b
NC
11763 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11764 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11765 return NULL;
1d7c1841 11766
bd81e77b 11767 switch (obase->op_type) {
1d7c1841 11768
bd81e77b
NC
11769 case OP_RV2AV:
11770 case OP_RV2HV:
11771 case OP_PADAV:
11772 case OP_PADHV:
11773 {
11774 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11775 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11776 I32 index = 0;
a0714e2c 11777 SV *keysv = NULL;
bd81e77b 11778 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11779
bd81e77b
NC
11780 if (pad) { /* @lex, %lex */
11781 sv = PAD_SVl(obase->op_targ);
a0714e2c 11782 gv = NULL;
bd81e77b
NC
11783 }
11784 else {
11785 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11786 /* @global, %global */
11787 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11788 if (!gv)
11789 break;
11790 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11791 }
11792 else /* @{expr}, %{expr} */
11793 return find_uninit_var(cUNOPx(obase)->op_first,
11794 uninit_sv, match);
11795 }
1d7c1841 11796
bd81e77b
NC
11797 /* attempt to find a match within the aggregate */
11798 if (hash) {
d4c19fe8 11799 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11800 if (keysv)
11801 subscript_type = FUV_SUBSCRIPT_HASH;
11802 }
11803 else {
e15d5972 11804 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11805 if (index >= 0)
11806 subscript_type = FUV_SUBSCRIPT_ARRAY;
11807 }
1d7c1841 11808
bd81e77b
NC
11809 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11810 break;
1d7c1841 11811
bd81e77b
NC
11812 return varname(gv, hash ? '%' : '@', obase->op_targ,
11813 keysv, index, subscript_type);
11814 }
1d7c1841 11815
bd81e77b
NC
11816 case OP_PADSV:
11817 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11818 break;
a0714e2c
SS
11819 return varname(NULL, '$', obase->op_targ,
11820 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11821
bd81e77b
NC
11822 case OP_GVSV:
11823 gv = cGVOPx_gv(obase);
11824 if (!gv || (match && GvSV(gv) != uninit_sv))
11825 break;
a0714e2c 11826 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11827
bd81e77b
NC
11828 case OP_AELEMFAST:
11829 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11830 if (match) {
11831 SV **svp;
11832 av = (AV*)PAD_SV(obase->op_targ);
11833 if (!av || SvRMAGICAL(av))
11834 break;
11835 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11836 if (!svp || *svp != uninit_sv)
11837 break;
11838 }
a0714e2c
SS
11839 return varname(NULL, '$', obase->op_targ,
11840 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11841 }
11842 else {
11843 gv = cGVOPx_gv(obase);
11844 if (!gv)
11845 break;
11846 if (match) {
11847 SV **svp;
11848 av = GvAV(gv);
11849 if (!av || SvRMAGICAL(av))
11850 break;
11851 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11852 if (!svp || *svp != uninit_sv)
11853 break;
11854 }
11855 return varname(gv, '$', 0,
a0714e2c 11856 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11857 }
11858 break;
1d7c1841 11859
bd81e77b
NC
11860 case OP_EXISTS:
11861 o = cUNOPx(obase)->op_first;
11862 if (!o || o->op_type != OP_NULL ||
11863 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11864 break;
11865 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11866
bd81e77b
NC
11867 case OP_AELEM:
11868 case OP_HELEM:
11869 if (PL_op == obase)
11870 /* $a[uninit_expr] or $h{uninit_expr} */
11871 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11872
a0714e2c 11873 gv = NULL;
bd81e77b
NC
11874 o = cBINOPx(obase)->op_first;
11875 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11876
bd81e77b 11877 /* get the av or hv, and optionally the gv */
a0714e2c 11878 sv = NULL;
bd81e77b
NC
11879 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11880 sv = PAD_SV(o->op_targ);
11881 }
11882 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11883 && cUNOPo->op_first->op_type == OP_GV)
11884 {
11885 gv = cGVOPx_gv(cUNOPo->op_first);
11886 if (!gv)
11887 break;
11888 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11889 }
11890 if (!sv)
11891 break;
11892
11893 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11894 /* index is constant */
11895 if (match) {
11896 if (SvMAGICAL(sv))
11897 break;
11898 if (obase->op_type == OP_HELEM) {
11899 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11900 if (!he || HeVAL(he) != uninit_sv)
11901 break;
11902 }
11903 else {
00b6aa41 11904 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11905 if (!svp || *svp != uninit_sv)
11906 break;
11907 }
11908 }
11909 if (obase->op_type == OP_HELEM)
11910 return varname(gv, '%', o->op_targ,
11911 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11912 else
a0714e2c 11913 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11914 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11915 }
11916 else {
11917 /* index is an expression;
11918 * attempt to find a match within the aggregate */
11919 if (obase->op_type == OP_HELEM) {
d4c19fe8 11920 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11921 if (keysv)
11922 return varname(gv, '%', o->op_targ,
11923 keysv, 0, FUV_SUBSCRIPT_HASH);
11924 }
11925 else {
d4c19fe8 11926 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11927 if (index >= 0)
11928 return varname(gv, '@', o->op_targ,
a0714e2c 11929 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11930 }
11931 if (match)
11932 break;
11933 return varname(gv,
11934 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
11935 ? '@' : '%',
a0714e2c 11936 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 11937 }
bd81e77b 11938 break;
dc507217 11939
bd81e77b
NC
11940 case OP_AASSIGN:
11941 /* only examine RHS */
11942 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 11943
bd81e77b
NC
11944 case OP_OPEN:
11945 o = cUNOPx(obase)->op_first;
11946 if (o->op_type == OP_PUSHMARK)
11947 o = o->op_sibling;
1d7c1841 11948
bd81e77b
NC
11949 if (!o->op_sibling) {
11950 /* one-arg version of open is highly magical */
a0ae6670 11951
bd81e77b
NC
11952 if (o->op_type == OP_GV) { /* open FOO; */
11953 gv = cGVOPx_gv(o);
11954 if (match && GvSV(gv) != uninit_sv)
11955 break;
11956 return varname(gv, '$', 0,
a0714e2c 11957 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
11958 }
11959 /* other possibilities not handled are:
11960 * open $x; or open my $x; should return '${*$x}'
11961 * open expr; should return '$'.expr ideally
11962 */
11963 break;
11964 }
11965 goto do_op;
ccfc67b7 11966
bd81e77b
NC
11967 /* ops where $_ may be an implicit arg */
11968 case OP_TRANS:
11969 case OP_SUBST:
11970 case OP_MATCH:
11971 if ( !(obase->op_flags & OPf_STACKED)) {
11972 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
11973 ? PAD_SVl(obase->op_targ)
11974 : DEFSV))
11975 {
11976 sv = sv_newmortal();
11977 sv_setpvn(sv, "$_", 2);
11978 return sv;
11979 }
11980 }
11981 goto do_op;
9f4817db 11982
bd81e77b
NC
11983 case OP_PRTF:
11984 case OP_PRINT:
11985 /* skip filehandle as it can't produce 'undef' warning */
11986 o = cUNOPx(obase)->op_first;
11987 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
11988 o = o->op_sibling->op_sibling;
11989 goto do_op2;
9f4817db 11990
9f4817db 11991
bd81e77b
NC
11992 case OP_RV2SV:
11993 case OP_CUSTOM:
11994 case OP_ENTERSUB:
11995 match = 1; /* XS or custom code could trigger random warnings */
11996 goto do_op;
9f4817db 11997
bd81e77b
NC
11998 case OP_SCHOMP:
11999 case OP_CHOMP:
12000 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12001 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12002 /*FALLTHROUGH*/
5d170f3a 12003
bd81e77b
NC
12004 default:
12005 do_op:
12006 if (!(obase->op_flags & OPf_KIDS))
12007 break;
12008 o = cUNOPx(obase)->op_first;
12009
12010 do_op2:
12011 if (!o)
12012 break;
f9893866 12013
bd81e77b
NC
12014 /* if all except one arg are constant, or have no side-effects,
12015 * or are optimized away, then it's unambiguous */
5f66b61c 12016 o2 = NULL;
bd81e77b 12017 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
12018 if (kid) {
12019 const OPCODE type = kid->op_type;
12020 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
12021 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12022 || (type == OP_PUSHMARK)
bd81e77b 12023 )
bd81e77b 12024 continue;
e15d5972 12025 }
bd81e77b 12026 if (o2) { /* more than one found */
5f66b61c 12027 o2 = NULL;
bd81e77b
NC
12028 break;
12029 }
12030 o2 = kid;
12031 }
12032 if (o2)
12033 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12034
bd81e77b
NC
12035 /* scan all args */
12036 while (o) {
12037 sv = find_uninit_var(o, uninit_sv, 1);
12038 if (sv)
12039 return sv;
12040 o = o->op_sibling;
d0063567 12041 }
bd81e77b 12042 break;
f9893866 12043 }
a0714e2c 12044 return NULL;
9f4817db
JH
12045}
12046
220e2d4e 12047
bd81e77b
NC
12048/*
12049=for apidoc report_uninit
68795e93 12050
bd81e77b 12051Print appropriate "Use of uninitialized variable" warning
220e2d4e 12052
bd81e77b
NC
12053=cut
12054*/
220e2d4e 12055
bd81e77b
NC
12056void
12057Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12058{
97aff369 12059 dVAR;
bd81e77b 12060 if (PL_op) {
a0714e2c 12061 SV* varname = NULL;
bd81e77b
NC
12062 if (uninit_sv) {
12063 varname = find_uninit_var(PL_op, uninit_sv,0);
12064 if (varname)
12065 sv_insert(varname, 0, 0, " ", 1);
12066 }
12067 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12068 varname ? SvPV_nolen_const(varname) : "",
12069 " in ", OP_DESC(PL_op));
220e2d4e 12070 }
a73e8557 12071 else
bd81e77b
NC
12072 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12073 "", "", "");
220e2d4e 12074}
f9893866 12075
241d1a3b
NC
12076/*
12077 * Local variables:
12078 * c-indentation-style: bsd
12079 * c-basic-offset: 4
12080 * indent-tabs-mode: t
12081 * End:
12082 *
37442d52
RGS
12083 * ex: set ts=8 sts=4 sw=4 noet:
12084 */