This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Change type of temporary padoff_du used in dUNDERBAR
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b94e2f88 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137 32#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 33/* if adding more checks watch out for the following tests:
e23c8137
JH
34 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
35 * lib/utf8.t lib/Unicode/Collate/t/index.t
36 * --jhi
37 */
6f207bd3 38# define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
39 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
40 assert((cache)[2] <= (cache)[3]); \
41 assert((cache)[3] <= (cache)[1]);} \
42 } STMT_END
e23c8137 43#else
6f207bd3 44# define ASSERT_UTF8_CACHE(cache) NOOP
e23c8137
JH
45#endif
46
f8c7b90f 47#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 48#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 49#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 50/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 51 on-write. */
765f542d 52#endif
645c22ef
DM
53
54/* ============================================================================
55
56=head1 Allocation and deallocation of SVs.
57
d2a0f284
JC
58An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
59sv, av, hv...) contains type and reference count information, and for
60many types, a pointer to the body (struct xrv, xpv, xpviv...), which
61contains fields specific to each type. Some types store all they need
62in the head, so don't have a body.
63
64In all but the most memory-paranoid configuations (ex: PURIFY), heads
65and bodies are allocated out of arenas, which by default are
66approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
67Sv-bodies are allocated by their sv-type, guaranteeing size
68consistency needed to allocate safely from arrays.
69
d2a0f284
JC
70For SV-heads, the first slot in each arena is reserved, and holds a
71link to the next arena, some flags, and a note of the number of slots.
72Snaked through each arena chain is a linked list of free items; when
73this becomes empty, an extra arena is allocated and divided up into N
74items which are threaded into the free list.
75
76SV-bodies are similar, but they use arena-sets by default, which
77separate the link and info from the arena itself, and reclaim the 1st
78slot in the arena. SV-bodies are further described later.
645c22ef
DM
79
80The following global variables are associated with arenas:
81
82 PL_sv_arenaroot pointer to list of SV arenas
83 PL_sv_root pointer to list of free SV structures
84
d2a0f284
JC
85 PL_body_arenas head of linked-list of body arenas
86 PL_body_roots[] array of pointers to list of free bodies of svtype
87 arrays are indexed by the svtype needed
93e68bfb 88
d2a0f284
JC
89A few special SV heads are not allocated from an arena, but are
90instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
91The size of arenas can be changed from the default by setting
92PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
93
94The SV arena serves the secondary purpose of allowing still-live SVs
95to be located and destroyed during final cleanup.
96
97At the lowest level, the macros new_SV() and del_SV() grab and free
98an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
99to return the SV to the free list with error checking.) new_SV() calls
100more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
101SVs in the free list have their SvTYPE field set to all ones.
102
ff276b08 103At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 104perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 105start of the interpreter.
645c22ef
DM
106
107Manipulation of any of the PL_*root pointers is protected by enclosing
108LOCK_SV_MUTEX; ... UNLOCK_SV_MUTEX calls which should Do the Right Thing
109if threads are enabled.
110
111The function visit() scans the SV arenas list, and calls a specified
112function for each SV it finds which is still live - ie which has an SvTYPE
113other than all 1's, and a non-zero SvREFCNT. visit() is used by the
114following functions (specified as [function that calls visit()] / [function
115called by visit() for each SV]):
116
117 sv_report_used() / do_report_used()
f2524eef 118 dump all remaining SVs (debugging aid)
645c22ef
DM
119
120 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
121 Attempt to free all objects pointed to by RVs,
122 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
123 try to do the same for all objects indirectly
124 referenced by typeglobs too. Called once from
125 perl_destruct(), prior to calling sv_clean_all()
126 below.
127
128 sv_clean_all() / do_clean_all()
129 SvREFCNT_dec(sv) each remaining SV, possibly
130 triggering an sv_free(). It also sets the
131 SVf_BREAK flag on the SV to indicate that the
132 refcnt has been artificially lowered, and thus
133 stopping sv_free() from giving spurious warnings
134 about SVs which unexpectedly have a refcnt
135 of zero. called repeatedly from perl_destruct()
136 until there are no SVs left.
137
93e68bfb 138=head2 Arena allocator API Summary
645c22ef
DM
139
140Private API to rest of sv.c
141
142 new_SV(), del_SV(),
143
144 new_XIV(), del_XIV(),
145 new_XNV(), del_XNV(),
146 etc
147
148Public API:
149
8cf8f3d1 150 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 151
645c22ef
DM
152=cut
153
154============================================================================ */
155
4561caa4
CS
156/*
157 * "A time to plant, and a time to uproot what was planted..."
158 */
159
77354fb4
NC
160/*
161 * nice_chunk and nice_chunk size need to be set
162 * and queried under the protection of sv_mutex
163 */
164void
165Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
166{
97aff369 167 dVAR;
77354fb4
NC
168 void *new_chunk;
169 U32 new_chunk_size;
170 LOCK_SV_MUTEX;
171 new_chunk = (void *)(chunk);
172 new_chunk_size = (chunk_size);
173 if (new_chunk_size > PL_nice_chunk_size) {
174 Safefree(PL_nice_chunk);
175 PL_nice_chunk = (char *) new_chunk;
176 PL_nice_chunk_size = new_chunk_size;
177 } else {
178 Safefree(chunk);
179 }
180 UNLOCK_SV_MUTEX;
181}
cac9b346 182
fd0854ff 183#ifdef DEBUG_LEAKING_SCALARS
22162ca8 184# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
185#else
186# define FREE_SV_DEBUG_FILE(sv)
187#endif
188
48614a46
NC
189#ifdef PERL_POISON
190# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
191/* Whilst I'd love to do this, it seems that things like to check on
192 unreferenced scalars
7e337ee0 193# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 194*/
7e337ee0
JH
195# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
196 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
197#else
198# define SvARENA_CHAIN(sv) SvANY(sv)
199# define POSION_SV_HEAD(sv)
200#endif
201
053fc874
GS
202#define plant_SV(p) \
203 STMT_START { \
fd0854ff 204 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
205 POSION_SV_HEAD(p); \
206 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
207 SvFLAGS(p) = SVTYPEMASK; \
208 PL_sv_root = (p); \
209 --PL_sv_count; \
210 } STMT_END
a0d0e21e 211
fba3b22e 212/* sv_mutex must be held while calling uproot_SV() */
053fc874
GS
213#define uproot_SV(p) \
214 STMT_START { \
215 (p) = PL_sv_root; \
bb7bbd9c 216 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
217 ++PL_sv_count; \
218 } STMT_END
219
645c22ef 220
cac9b346
NC
221/* make some more SVs by adding another arena */
222
223/* sv_mutex must be held while calling more_sv() */
224STATIC SV*
225S_more_sv(pTHX)
226{
97aff369 227 dVAR;
cac9b346
NC
228 SV* sv;
229
230 if (PL_nice_chunk) {
231 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 232 PL_nice_chunk = NULL;
cac9b346
NC
233 PL_nice_chunk_size = 0;
234 }
235 else {
236 char *chunk; /* must use New here to match call to */
d2a0f284 237 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 238 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
239 }
240 uproot_SV(sv);
241 return sv;
242}
243
645c22ef
DM
244/* new_SV(): return a new, empty SV head */
245
eba0f806
DM
246#ifdef DEBUG_LEAKING_SCALARS
247/* provide a real function for a debugger to play with */
248STATIC SV*
249S_new_SV(pTHX)
250{
251 SV* sv;
252
253 LOCK_SV_MUTEX;
254 if (PL_sv_root)
255 uproot_SV(sv);
256 else
cac9b346 257 sv = S_more_sv(aTHX);
eba0f806
DM
258 UNLOCK_SV_MUTEX;
259 SvANY(sv) = 0;
260 SvREFCNT(sv) = 1;
261 SvFLAGS(sv) = 0;
fd0854ff
DM
262 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
263 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
264 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
265 sv->sv_debug_inpad = 0;
266 sv->sv_debug_cloned = 0;
fd0854ff 267 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 268
eba0f806
DM
269 return sv;
270}
271# define new_SV(p) (p)=S_new_SV(aTHX)
272
273#else
274# define new_SV(p) \
053fc874
GS
275 STMT_START { \
276 LOCK_SV_MUTEX; \
277 if (PL_sv_root) \
278 uproot_SV(p); \
279 else \
cac9b346 280 (p) = S_more_sv(aTHX); \
053fc874
GS
281 UNLOCK_SV_MUTEX; \
282 SvANY(p) = 0; \
283 SvREFCNT(p) = 1; \
284 SvFLAGS(p) = 0; \
285 } STMT_END
eba0f806 286#endif
463ee0b2 287
645c22ef
DM
288
289/* del_SV(): return an empty SV head to the free list */
290
a0d0e21e 291#ifdef DEBUGGING
4561caa4 292
053fc874
GS
293#define del_SV(p) \
294 STMT_START { \
295 LOCK_SV_MUTEX; \
aea4f609 296 if (DEBUG_D_TEST) \
053fc874
GS
297 del_sv(p); \
298 else \
299 plant_SV(p); \
300 UNLOCK_SV_MUTEX; \
301 } STMT_END
a0d0e21e 302
76e3520e 303STATIC void
cea2e8a9 304S_del_sv(pTHX_ SV *p)
463ee0b2 305{
97aff369 306 dVAR;
aea4f609 307 if (DEBUG_D_TEST) {
4633a7c4 308 SV* sva;
a3b680e6 309 bool ok = 0;
3280af22 310 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
311 const SV * const sv = sva + 1;
312 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 313 if (p >= sv && p < svend) {
a0d0e21e 314 ok = 1;
c0ff570e
NC
315 break;
316 }
a0d0e21e
LW
317 }
318 if (!ok) {
0453d815 319 if (ckWARN_d(WARN_INTERNAL))
9014280d 320 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
321 "Attempt to free non-arena SV: 0x%"UVxf
322 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
323 return;
324 }
325 }
4561caa4 326 plant_SV(p);
463ee0b2 327}
a0d0e21e 328
4561caa4
CS
329#else /* ! DEBUGGING */
330
331#define del_SV(p) plant_SV(p)
332
333#endif /* DEBUGGING */
463ee0b2 334
645c22ef
DM
335
336/*
ccfc67b7
JH
337=head1 SV Manipulation Functions
338
645c22ef
DM
339=for apidoc sv_add_arena
340
341Given a chunk of memory, link it to the head of the list of arenas,
342and split it into a list of free SVs.
343
344=cut
345*/
346
4633a7c4 347void
864dbfa3 348Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 349{
97aff369 350 dVAR;
0bd48802 351 SV* const sva = (SV*)ptr;
463ee0b2
LW
352 register SV* sv;
353 register SV* svend;
4633a7c4
LW
354
355 /* The first SV in an arena isn't an SV. */
3280af22 356 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
357 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
358 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
359
3280af22
NIS
360 PL_sv_arenaroot = sva;
361 PL_sv_root = sva + 1;
4633a7c4
LW
362
363 svend = &sva[SvREFCNT(sva) - 1];
364 sv = sva + 1;
463ee0b2 365 while (sv < svend) {
48614a46 366 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 367#ifdef DEBUGGING
978b032e 368 SvREFCNT(sv) = 0;
03e36789
NC
369#endif
370 /* Must always set typemask because it's awlays checked in on cleanup
371 when the arenas are walked looking for objects. */
8990e307 372 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
373 sv++;
374 }
48614a46 375 SvARENA_CHAIN(sv) = 0;
03e36789
NC
376#ifdef DEBUGGING
377 SvREFCNT(sv) = 0;
378#endif
4633a7c4
LW
379 SvFLAGS(sv) = SVTYPEMASK;
380}
381
055972dc
DM
382/* visit(): call the named function for each non-free SV in the arenas
383 * whose flags field matches the flags/mask args. */
645c22ef 384
5226ed68 385STATIC I32
055972dc 386S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 387{
97aff369 388 dVAR;
4633a7c4 389 SV* sva;
5226ed68 390 I32 visited = 0;
8990e307 391
3280af22 392 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 393 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 394 register SV* sv;
4561caa4 395 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
396 if (SvTYPE(sv) != SVTYPEMASK
397 && (sv->sv_flags & mask) == flags
398 && SvREFCNT(sv))
399 {
acfe0abc 400 (FCALL)(aTHX_ sv);
5226ed68
JH
401 ++visited;
402 }
8990e307
LW
403 }
404 }
5226ed68 405 return visited;
8990e307
LW
406}
407
758a08c3
JH
408#ifdef DEBUGGING
409
645c22ef
DM
410/* called by sv_report_used() for each live SV */
411
412static void
acfe0abc 413do_report_used(pTHX_ SV *sv)
645c22ef
DM
414{
415 if (SvTYPE(sv) != SVTYPEMASK) {
416 PerlIO_printf(Perl_debug_log, "****\n");
417 sv_dump(sv);
418 }
419}
758a08c3 420#endif
645c22ef
DM
421
422/*
423=for apidoc sv_report_used
424
425Dump the contents of all SVs not yet freed. (Debugging aid).
426
427=cut
428*/
429
8990e307 430void
864dbfa3 431Perl_sv_report_used(pTHX)
4561caa4 432{
ff270d3a 433#ifdef DEBUGGING
055972dc 434 visit(do_report_used, 0, 0);
96a5add6
AL
435#else
436 PERL_UNUSED_CONTEXT;
ff270d3a 437#endif
4561caa4
CS
438}
439
645c22ef
DM
440/* called by sv_clean_objs() for each live SV */
441
442static void
e15faf7d 443do_clean_objs(pTHX_ SV *ref)
645c22ef 444{
97aff369 445 dVAR;
823a54a3
AL
446 if (SvROK(ref)) {
447 SV * const target = SvRV(ref);
448 if (SvOBJECT(target)) {
449 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
450 if (SvWEAKREF(ref)) {
451 sv_del_backref(target, ref);
452 SvWEAKREF_off(ref);
453 SvRV_set(ref, NULL);
454 } else {
455 SvROK_off(ref);
456 SvRV_set(ref, NULL);
457 SvREFCNT_dec(target);
458 }
645c22ef
DM
459 }
460 }
461
462 /* XXX Might want to check arrays, etc. */
463}
464
465/* called by sv_clean_objs() for each live SV */
466
467#ifndef DISABLE_DESTRUCTOR_KLUDGE
468static void
acfe0abc 469do_clean_named_objs(pTHX_ SV *sv)
645c22ef 470{
97aff369 471 dVAR;
f7877b28 472 if (SvTYPE(sv) == SVt_PVGV && isGV_with_GP(sv) && GvGP(sv)) {
c69033f2
NC
473 if ((
474#ifdef PERL_DONT_CREATE_GVSV
475 GvSV(sv) &&
476#endif
477 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
478 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
479 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
480 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
481 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
482 {
483 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 484 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
485 SvREFCNT_dec(sv);
486 }
487 }
488}
489#endif
490
491/*
492=for apidoc sv_clean_objs
493
494Attempt to destroy all objects not yet freed
495
496=cut
497*/
498
4561caa4 499void
864dbfa3 500Perl_sv_clean_objs(pTHX)
4561caa4 501{
97aff369 502 dVAR;
3280af22 503 PL_in_clean_objs = TRUE;
055972dc 504 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 505#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 506 /* some barnacles may yet remain, clinging to typeglobs */
055972dc 507 visit(do_clean_named_objs, SVt_PVGV, SVTYPEMASK);
4561caa4 508#endif
3280af22 509 PL_in_clean_objs = FALSE;
4561caa4
CS
510}
511
645c22ef
DM
512/* called by sv_clean_all() for each live SV */
513
514static void
acfe0abc 515do_clean_all(pTHX_ SV *sv)
645c22ef 516{
97aff369 517 dVAR;
645c22ef
DM
518 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
519 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b 520 if (PL_comppad == (AV*)sv) {
7d49f689 521 PL_comppad = NULL;
4608196e 522 PL_curpad = NULL;
0e705b3b 523 }
645c22ef
DM
524 SvREFCNT_dec(sv);
525}
526
527/*
528=for apidoc sv_clean_all
529
530Decrement the refcnt of each remaining SV, possibly triggering a
531cleanup. This function may have to be called multiple times to free
ff276b08 532SVs which are in complex self-referential hierarchies.
645c22ef
DM
533
534=cut
535*/
536
5226ed68 537I32
864dbfa3 538Perl_sv_clean_all(pTHX)
8990e307 539{
97aff369 540 dVAR;
5226ed68 541 I32 cleaned;
3280af22 542 PL_in_clean_all = TRUE;
055972dc 543 cleaned = visit(do_clean_all, 0,0);
3280af22 544 PL_in_clean_all = FALSE;
5226ed68 545 return cleaned;
8990e307 546}
463ee0b2 547
5e258f8c
JC
548/*
549 ARENASETS: a meta-arena implementation which separates arena-info
550 into struct arena_set, which contains an array of struct
551 arena_descs, each holding info for a single arena. By separating
552 the meta-info from the arena, we recover the 1st slot, formerly
553 borrowed for list management. The arena_set is about the size of an
554 arena, avoiding the needless malloc overhead of a naive linked-list
555
556 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
557 memory in the last arena-set (1/2 on average). In trade, we get
558 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284
JC
559 smaller types). The recovery of the wasted space allows use of
560 small arenas for large, rare body types,
5e258f8c 561*/
5e258f8c 562struct arena_desc {
398c677b
NC
563 char *arena; /* the raw storage, allocated aligned */
564 size_t size; /* its size ~4k typ */
565 int unit_type; /* useful for arena audits */
5e258f8c
JC
566 /* info for sv-heads (eventually)
567 int count, flags;
568 */
569};
570
e6148039
NC
571struct arena_set;
572
573/* Get the maximum number of elements in set[] such that struct arena_set
574 will fit within PERL_ARENA_SIZE, which is probabably just under 4K, and
575 therefore likely to be 1 aligned memory page. */
576
577#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
578 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
579
580struct arena_set {
581 struct arena_set* next;
582 int set_size; /* ie ARENAS_PER_SET */
583 int curr; /* index of next available arena-desc */
584 struct arena_desc set[ARENAS_PER_SET];
585};
586
645c22ef
DM
587/*
588=for apidoc sv_free_arenas
589
590Deallocate the memory used by all arenas. Note that all the individual SV
591heads and bodies within the arenas must already have been freed.
592
593=cut
594*/
4633a7c4 595void
864dbfa3 596Perl_sv_free_arenas(pTHX)
4633a7c4 597{
97aff369 598 dVAR;
4633a7c4
LW
599 SV* sva;
600 SV* svanext;
93e68bfb 601 int i;
4633a7c4
LW
602
603 /* Free arenas here, but be careful about fake ones. (We assume
604 contiguity of the fake ones with the corresponding real ones.) */
605
3280af22 606 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
607 svanext = (SV*) SvANY(sva);
608 while (svanext && SvFAKE(svanext))
609 svanext = (SV*) SvANY(svanext);
610
611 if (!SvFAKE(sva))
1df70142 612 Safefree(sva);
4633a7c4 613 }
93e68bfb 614
5e258f8c
JC
615 {
616 struct arena_set *next, *aroot = (struct arena_set*) PL_body_arenas;
617
618 for (; aroot; aroot = next) {
96a5add6 619 const int max = aroot->curr;
5e258f8c
JC
620 for (i=0; i<max; i++) {
621 assert(aroot->set[i].arena);
622 Safefree(aroot->set[i].arena);
623 }
624 next = aroot->next;
625 Safefree(aroot);
626 }
627 }
dc8220bf 628 PL_body_arenas = 0;
fdda85ca 629
232d1c15 630 for (i=0; i<PERL_ARENA_ROOTS_SIZE; i++)
93e68bfb 631 PL_body_roots[i] = 0;
93e68bfb 632
43c5f42d 633 Safefree(PL_nice_chunk);
bd61b366 634 PL_nice_chunk = NULL;
3280af22
NIS
635 PL_nice_chunk_size = 0;
636 PL_sv_arenaroot = 0;
637 PL_sv_root = 0;
4633a7c4
LW
638}
639
bd81e77b
NC
640/*
641 Here are mid-level routines that manage the allocation of bodies out
642 of the various arenas. There are 5 kinds of arenas:
29489e7c 643
bd81e77b
NC
644 1. SV-head arenas, which are discussed and handled above
645 2. regular body arenas
646 3. arenas for reduced-size bodies
647 4. Hash-Entry arenas
648 5. pte arenas (thread related)
29489e7c 649
bd81e77b
NC
650 Arena types 2 & 3 are chained by body-type off an array of
651 arena-root pointers, which is indexed by svtype. Some of the
652 larger/less used body types are malloced singly, since a large
653 unused block of them is wasteful. Also, several svtypes dont have
654 bodies; the data fits into the sv-head itself. The arena-root
655 pointer thus has a few unused root-pointers (which may be hijacked
656 later for arena types 4,5)
29489e7c 657
bd81e77b
NC
658 3 differs from 2 as an optimization; some body types have several
659 unused fields in the front of the structure (which are kept in-place
660 for consistency). These bodies can be allocated in smaller chunks,
661 because the leading fields arent accessed. Pointers to such bodies
662 are decremented to point at the unused 'ghost' memory, knowing that
663 the pointers are used with offsets to the real memory.
29489e7c 664
bd81e77b
NC
665 HE, HEK arenas are managed separately, with separate code, but may
666 be merge-able later..
667
668 PTE arenas are not sv-bodies, but they share these mid-level
669 mechanics, so are considered here. The new mid-level mechanics rely
670 on the sv_type of the body being allocated, so we just reserve one
671 of the unused body-slots for PTEs, then use it in those (2) PTE
672 contexts below (line ~10k)
673*/
674
bd26d9a3 675/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
676 TBD: export properly for hv.c: S_more_he().
677*/
678void*
679Perl_get_arena(pTHX_ int arena_size)
680{
5e258f8c 681 struct arena_desc* adesc;
476a1e16 682 struct arena_set *newroot, **aroot = (struct arena_set**) &PL_body_arenas;
5e258f8c
JC
683 int curr;
684
476a1e16
JC
685 /* shouldnt need this
686 if (!arena_size) arena_size = PERL_ARENA_SIZE;
687 */
5e258f8c
JC
688
689 /* may need new arena-set to hold new arena */
476a1e16 690 if (!*aroot || (*aroot)->curr >= (*aroot)->set_size) {
5e258f8c
JC
691 Newxz(newroot, 1, struct arena_set);
692 newroot->set_size = ARENAS_PER_SET;
476a1e16
JC
693 newroot->next = *aroot;
694 *aroot = newroot;
695 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", *aroot));
5e258f8c
JC
696 }
697
698 /* ok, now have arena-set with at least 1 empty/available arena-desc */
476a1e16
JC
699 curr = (*aroot)->curr++;
700 adesc = &((*aroot)->set[curr]);
5e258f8c
JC
701 assert(!adesc->arena);
702
5e258f8c
JC
703 Newxz(adesc->arena, arena_size, char);
704 adesc->size = arena_size;
d2a0f284
JC
705 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
706 curr, adesc->arena, arena_size));
5e258f8c
JC
707
708 return adesc->arena;
5e258f8c
JC
709}
710
53c1dcc0 711
bd81e77b 712/* return a thing to the free list */
29489e7c 713
bd81e77b
NC
714#define del_body(thing, root) \
715 STMT_START { \
00b6aa41 716 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
717 LOCK_SV_MUTEX; \
718 *thing_copy = *root; \
719 *root = (void*)thing_copy; \
720 UNLOCK_SV_MUTEX; \
721 } STMT_END
29489e7c 722
bd81e77b 723/*
d2a0f284
JC
724
725=head1 SV-Body Allocation
726
727Allocation of SV-bodies is similar to SV-heads, differing as follows;
728the allocation mechanism is used for many body types, so is somewhat
729more complicated, it uses arena-sets, and has no need for still-live
730SV detection.
731
732At the outermost level, (new|del)_X*V macros return bodies of the
733appropriate type. These macros call either (new|del)_body_type or
734(new|del)_body_allocated macro pairs, depending on specifics of the
735type. Most body types use the former pair, the latter pair is used to
736allocate body types with "ghost fields".
737
738"ghost fields" are fields that are unused in certain types, and
739consequently dont need to actually exist. They are declared because
740they're part of a "base type", which allows use of functions as
741methods. The simplest examples are AVs and HVs, 2 aggregate types
742which don't use the fields which support SCALAR semantics.
743
744For these types, the arenas are carved up into *_allocated size
745chunks, we thus avoid wasted memory for those unaccessed members.
746When bodies are allocated, we adjust the pointer back in memory by the
747size of the bit not allocated, so it's as if we allocated the full
748structure. (But things will all go boom if you write to the part that
749is "not there", because you'll be overwriting the last members of the
750preceding structure in memory.)
751
752We calculate the correction using the STRUCT_OFFSET macro. For
753example, if xpv_allocated is the same structure as XPV then the two
754OFFSETs sum to zero, and the pointer is unchanged. If the allocated
755structure is smaller (no initial NV actually allocated) then the net
756effect is to subtract the size of the NV from the pointer, to return a
757new pointer as if an initial NV were actually allocated.
758
759This is the same trick as was used for NV and IV bodies. Ironically it
760doesn't need to be used for NV bodies any more, because NV is now at
761the start of the structure. IV bodies don't need it either, because
762they are no longer allocated.
763
764In turn, the new_body_* allocators call S_new_body(), which invokes
765new_body_inline macro, which takes a lock, and takes a body off the
766linked list at PL_body_roots[sv_type], calling S_more_bodies() if
767necessary to refresh an empty list. Then the lock is released, and
768the body is returned.
769
770S_more_bodies calls get_arena(), and carves it up into an array of N
771bodies, which it strings into a linked list. It looks up arena-size
772and body-size from the body_details table described below, thus
773supporting the multiple body-types.
774
775If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
776the (new|del)_X*V macros are mapped directly to malloc/free.
777
778*/
779
780/*
781
782For each sv-type, struct body_details bodies_by_type[] carries
783parameters which control these aspects of SV handling:
784
785Arena_size determines whether arenas are used for this body type, and if
786so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
787zero, forcing individual mallocs and frees.
788
789Body_size determines how big a body is, and therefore how many fit into
790each arena. Offset carries the body-pointer adjustment needed for
791*_allocated body types, and is used in *_allocated macros.
792
793But its main purpose is to parameterize info needed in
794Perl_sv_upgrade(). The info here dramatically simplifies the function
795vs the implementation in 5.8.7, making it table-driven. All fields
796are used for this, except for arena_size.
797
798For the sv-types that have no bodies, arenas are not used, so those
799PL_body_roots[sv_type] are unused, and can be overloaded. In
800something of a special case, SVt_NULL is borrowed for HE arenas;
801PL_body_roots[SVt_NULL] is filled by S_more_he, but the
802bodies_by_type[SVt_NULL] slot is not used, as the table is not
803available in hv.c,
804
805PTEs also use arenas, but are never seen in Perl_sv_upgrade.
806Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
807they can just use the same allocation semantics. At first, PTEs were
808also overloaded to a non-body sv-type, but this yielded hard-to-find
809malloc bugs, so was simplified by claiming a new slot. This choice
810has no consequence at this time.
811
29489e7c
DM
812*/
813
bd81e77b 814struct body_details {
0fb58b32 815 U8 body_size; /* Size to allocate */
10666ae3 816 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 817 U8 offset;
10666ae3
NC
818 unsigned int type : 4; /* We have space for a sanity check. */
819 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
820 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
821 unsigned int arena : 1; /* Allocated from an arena */
822 size_t arena_size; /* Size of arena to allocate */
bd81e77b 823};
29489e7c 824
bd81e77b
NC
825#define HADNV FALSE
826#define NONV TRUE
29489e7c 827
d2a0f284 828
bd81e77b
NC
829#ifdef PURIFY
830/* With -DPURFIY we allocate everything directly, and don't use arenas.
831 This seems a rather elegant way to simplify some of the code below. */
832#define HASARENA FALSE
833#else
834#define HASARENA TRUE
835#endif
836#define NOARENA FALSE
29489e7c 837
d2a0f284
JC
838/* Size the arenas to exactly fit a given number of bodies. A count
839 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
840 simplifying the default. If count > 0, the arena is sized to fit
841 only that many bodies, allowing arenas to be used for large, rare
842 bodies (XPVFM, XPVIO) without undue waste. The arena size is
843 limited by PERL_ARENA_SIZE, so we can safely oversize the
844 declarations.
845 */
95db5f15
MB
846#define FIT_ARENA0(body_size) \
847 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
848#define FIT_ARENAn(count,body_size) \
849 ( count * body_size <= PERL_ARENA_SIZE) \
850 ? count * body_size \
851 : FIT_ARENA0 (body_size)
852#define FIT_ARENA(count,body_size) \
853 count \
854 ? FIT_ARENAn (count, body_size) \
855 : FIT_ARENA0 (body_size)
d2a0f284 856
bd81e77b 857/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 858
bd81e77b
NC
859typedef struct {
860 STRLEN xpv_cur;
861 STRLEN xpv_len;
862} xpv_allocated;
29489e7c 863
bd81e77b 864to make its members accessible via a pointer to (say)
29489e7c 865
bd81e77b
NC
866struct xpv {
867 NV xnv_nv;
868 STRLEN xpv_cur;
869 STRLEN xpv_len;
870};
29489e7c 871
bd81e77b 872*/
29489e7c 873
bd81e77b
NC
874#define relative_STRUCT_OFFSET(longer, shorter, member) \
875 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 876
bd81e77b
NC
877/* Calculate the length to copy. Specifically work out the length less any
878 final padding the compiler needed to add. See the comment in sv_upgrade
879 for why copying the padding proved to be a bug. */
29489e7c 880
bd81e77b
NC
881#define copy_length(type, last_member) \
882 STRUCT_OFFSET(type, last_member) \
883 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 884
bd81e77b 885static const struct body_details bodies_by_type[] = {
10666ae3
NC
886 { sizeof(HE), 0, 0, SVt_NULL,
887 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284
JC
888
889 /* IVs are in the head, so the allocation size is 0.
890 However, the slot is overloaded for PTEs. */
891 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
892 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 893 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
894 NOARENA /* IVS don't need an arena */,
895 /* But PTEs need to know the size of their arena */
896 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
897 },
898
bd81e77b 899 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 900 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
901 FIT_ARENA(0, sizeof(NV)) },
902
903 /* RVs are in the head now. */
10666ae3 904 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 905
bd81e77b 906 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
907 { sizeof(xpv_allocated),
908 copy_length(XPV, xpv_len)
909 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
910 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 911 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 912
bd81e77b 913 /* 12 */
d2a0f284
JC
914 { sizeof(xpviv_allocated),
915 copy_length(XPVIV, xiv_u)
916 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
917 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 918 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 919
bd81e77b 920 /* 20 */
10666ae3 921 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
922 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
923
bd81e77b 924 /* 28 */
10666ae3 925 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
926 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
927
bd81e77b 928 /* 36 */
10666ae3 929 { sizeof(XPVBM), sizeof(XPVBM), 0, SVt_PVBM, TRUE, HADNV,
d2a0f284
JC
930 HASARENA, FIT_ARENA(0, sizeof(XPVBM)) },
931
bd81e77b 932 /* 48 */
10666ae3 933 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
934 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
935
bd81e77b 936 /* 64 */
10666ae3 937 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
938 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
939
940 { sizeof(xpvav_allocated),
941 copy_length(XPVAV, xmg_stash)
942 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
943 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 944 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
945
946 { sizeof(xpvhv_allocated),
947 copy_length(XPVHV, xmg_stash)
948 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
949 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 950 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 951
c84c4652 952 /* 56 */
4115f141 953 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 954 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 955 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 956
4115f141 957 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 958 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 959 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
960
961 /* XPVIO is 84 bytes, fits 48x */
10666ae3 962 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 963 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 964};
29489e7c 965
d2a0f284
JC
966#define new_body_type(sv_type) \
967 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 968
bd81e77b
NC
969#define del_body_type(p, sv_type) \
970 del_body(p, &PL_body_roots[sv_type])
29489e7c 971
29489e7c 972
bd81e77b 973#define new_body_allocated(sv_type) \
d2a0f284 974 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 975 - bodies_by_type[sv_type].offset)
29489e7c 976
bd81e77b
NC
977#define del_body_allocated(p, sv_type) \
978 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 979
29489e7c 980
bd81e77b
NC
981#define my_safemalloc(s) (void*)safemalloc(s)
982#define my_safecalloc(s) (void*)safecalloc(s, 1)
983#define my_safefree(p) safefree((char*)p)
29489e7c 984
bd81e77b 985#ifdef PURIFY
29489e7c 986
bd81e77b
NC
987#define new_XNV() my_safemalloc(sizeof(XPVNV))
988#define del_XNV(p) my_safefree(p)
29489e7c 989
bd81e77b
NC
990#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
991#define del_XPVNV(p) my_safefree(p)
29489e7c 992
bd81e77b
NC
993#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
994#define del_XPVAV(p) my_safefree(p)
29489e7c 995
bd81e77b
NC
996#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
997#define del_XPVHV(p) my_safefree(p)
29489e7c 998
bd81e77b
NC
999#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1000#define del_XPVMG(p) my_safefree(p)
29489e7c 1001
bd81e77b
NC
1002#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1003#define del_XPVGV(p) my_safefree(p)
29489e7c 1004
bd81e77b 1005#else /* !PURIFY */
29489e7c 1006
bd81e77b
NC
1007#define new_XNV() new_body_type(SVt_NV)
1008#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1009
bd81e77b
NC
1010#define new_XPVNV() new_body_type(SVt_PVNV)
1011#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1012
bd81e77b
NC
1013#define new_XPVAV() new_body_allocated(SVt_PVAV)
1014#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1015
bd81e77b
NC
1016#define new_XPVHV() new_body_allocated(SVt_PVHV)
1017#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1018
bd81e77b
NC
1019#define new_XPVMG() new_body_type(SVt_PVMG)
1020#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1021
bd81e77b
NC
1022#define new_XPVGV() new_body_type(SVt_PVGV)
1023#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1024
bd81e77b 1025#endif /* PURIFY */
93e68bfb 1026
bd81e77b 1027/* no arena for you! */
93e68bfb 1028
bd81e77b 1029#define new_NOARENA(details) \
d2a0f284 1030 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1031#define new_NOARENAZ(details) \
d2a0f284
JC
1032 my_safecalloc((details)->body_size + (details)->offset)
1033
10666ae3
NC
1034#ifdef DEBUGGING
1035static bool done_sanity_check;
1036#endif
1037
d2a0f284
JC
1038STATIC void *
1039S_more_bodies (pTHX_ svtype sv_type)
1040{
1041 dVAR;
1042 void ** const root = &PL_body_roots[sv_type];
96a5add6 1043 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1044 const size_t body_size = bdp->body_size;
1045 char *start;
1046 const char *end;
1047
1048 assert(bdp->arena_size);
10666ae3
NC
1049
1050#ifdef DEBUGGING
1051 if (!done_sanity_check) {
ea471437 1052 unsigned int i = SVt_LAST;
10666ae3
NC
1053
1054 done_sanity_check = TRUE;
1055
1056 while (i--)
1057 assert (bodies_by_type[i].type == i);
1058 }
1059#endif
1060
d2a0f284
JC
1061 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size);
1062
1063 end = start + bdp->arena_size - body_size;
1064
d2a0f284
JC
1065 /* computed count doesnt reflect the 1st slot reservation */
1066 DEBUG_m(PerlIO_printf(Perl_debug_log,
1067 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1068 start, end, bdp->arena_size, sv_type, body_size,
1069 bdp->arena_size / body_size));
d2a0f284
JC
1070
1071 *root = (void *)start;
1072
1073 while (start < end) {
1074 char * const next = start + body_size;
1075 *(void**) start = (void *)next;
1076 start = next;
1077 }
1078 *(void **)start = 0;
1079
1080 return *root;
1081}
1082
1083/* grab a new thing from the free list, allocating more if necessary.
1084 The inline version is used for speed in hot routines, and the
1085 function using it serves the rest (unless PURIFY).
1086*/
1087#define new_body_inline(xpv, sv_type) \
1088 STMT_START { \
1089 void ** const r3wt = &PL_body_roots[sv_type]; \
1090 LOCK_SV_MUTEX; \
1091 xpv = *((void **)(r3wt)) \
d4c19fe8 1092 ? *((void **)(r3wt)) : more_bodies(sv_type); \
d2a0f284
JC
1093 *(r3wt) = *(void**)(xpv); \
1094 UNLOCK_SV_MUTEX; \
1095 } STMT_END
1096
1097#ifndef PURIFY
1098
1099STATIC void *
1100S_new_body(pTHX_ svtype sv_type)
1101{
1102 dVAR;
1103 void *xpv;
1104 new_body_inline(xpv, sv_type);
1105 return xpv;
1106}
1107
1108#endif
93e68bfb 1109
bd81e77b
NC
1110/*
1111=for apidoc sv_upgrade
93e68bfb 1112
bd81e77b
NC
1113Upgrade an SV to a more complex form. Generally adds a new body type to the
1114SV, then copies across as much information as possible from the old body.
1115You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1116
bd81e77b 1117=cut
93e68bfb 1118*/
93e68bfb 1119
bd81e77b
NC
1120void
1121Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type)
cac9b346 1122{
97aff369 1123 dVAR;
bd81e77b
NC
1124 void* old_body;
1125 void* new_body;
1126 const U32 old_type = SvTYPE(sv);
d2a0f284 1127 const struct body_details *new_type_details;
bd81e77b
NC
1128 const struct body_details *const old_type_details
1129 = bodies_by_type + old_type;
cac9b346 1130
bd81e77b
NC
1131 if (new_type != SVt_PV && SvIsCOW(sv)) {
1132 sv_force_normal_flags(sv, 0);
1133 }
cac9b346 1134
bd81e77b
NC
1135 if (old_type == new_type)
1136 return;
cac9b346 1137
bd81e77b
NC
1138 if (old_type > new_type)
1139 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1140 (int)old_type, (int)new_type);
cac9b346 1141
cac9b346 1142
bd81e77b 1143 old_body = SvANY(sv);
de042e1d 1144
bd81e77b
NC
1145 /* Copying structures onto other structures that have been neatly zeroed
1146 has a subtle gotcha. Consider XPVMG
cac9b346 1147
bd81e77b
NC
1148 +------+------+------+------+------+-------+-------+
1149 | NV | CUR | LEN | IV | MAGIC | STASH |
1150 +------+------+------+------+------+-------+-------+
1151 0 4 8 12 16 20 24 28
645c22ef 1152
bd81e77b
NC
1153 where NVs are aligned to 8 bytes, so that sizeof that structure is
1154 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1155
bd81e77b
NC
1156 +------+------+------+------+------+-------+-------+------+
1157 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1158 +------+------+------+------+------+-------+-------+------+
1159 0 4 8 12 16 20 24 28 32
08742458 1160
bd81e77b 1161 so what happens if you allocate memory for this structure:
30f9da9e 1162
bd81e77b
NC
1163 +------+------+------+------+------+-------+-------+------+------+...
1164 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1165 +------+------+------+------+------+-------+-------+------+------+...
1166 0 4 8 12 16 20 24 28 32 36
bfc44f79 1167
bd81e77b
NC
1168 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1169 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1170 started out as zero once, but it's quite possible that it isn't. So now,
1171 rather than a nicely zeroed GP, you have it pointing somewhere random.
1172 Bugs ensue.
bfc44f79 1173
bd81e77b
NC
1174 (In fact, GP ends up pointing at a previous GP structure, because the
1175 principle cause of the padding in XPVMG getting garbage is a copy of
1176 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 1177
bd81e77b
NC
1178 So we are careful and work out the size of used parts of all the
1179 structures. */
bfc44f79 1180
bd81e77b
NC
1181 switch (old_type) {
1182 case SVt_NULL:
1183 break;
1184 case SVt_IV:
1185 if (new_type < SVt_PVIV) {
1186 new_type = (new_type == SVt_NV)
1187 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1188 }
1189 break;
1190 case SVt_NV:
1191 if (new_type < SVt_PVNV) {
1192 new_type = SVt_PVNV;
bd81e77b
NC
1193 }
1194 break;
1195 case SVt_RV:
1196 break;
1197 case SVt_PV:
1198 assert(new_type > SVt_PV);
1199 assert(SVt_IV < SVt_PV);
1200 assert(SVt_NV < SVt_PV);
1201 break;
1202 case SVt_PVIV:
1203 break;
1204 case SVt_PVNV:
1205 break;
1206 case SVt_PVMG:
1207 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1208 there's no way that it can be safely upgraded, because perl.c
1209 expects to Safefree(SvANY(PL_mess_sv)) */
1210 assert(sv != PL_mess_sv);
1211 /* This flag bit is used to mean other things in other scalar types.
1212 Given that it only has meaning inside the pad, it shouldn't be set
1213 on anything that can get upgraded. */
00b1698f 1214 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1215 break;
1216 default:
1217 if (old_type_details->cant_upgrade)
c81225bc
NC
1218 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1219 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1220 }
2fa1109b 1221 new_type_details = bodies_by_type + new_type;
645c22ef 1222
bd81e77b
NC
1223 SvFLAGS(sv) &= ~SVTYPEMASK;
1224 SvFLAGS(sv) |= new_type;
932e9ff9 1225
ab4416c0
NC
1226 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1227 the return statements above will have triggered. */
1228 assert (new_type != SVt_NULL);
bd81e77b 1229 switch (new_type) {
bd81e77b
NC
1230 case SVt_IV:
1231 assert(old_type == SVt_NULL);
1232 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1233 SvIV_set(sv, 0);
1234 return;
1235 case SVt_NV:
1236 assert(old_type == SVt_NULL);
1237 SvANY(sv) = new_XNV();
1238 SvNV_set(sv, 0);
1239 return;
1240 case SVt_RV:
1241 assert(old_type == SVt_NULL);
1242 SvANY(sv) = &sv->sv_u.svu_rv;
1243 SvRV_set(sv, 0);
1244 return;
1245 case SVt_PVHV:
bd81e77b 1246 case SVt_PVAV:
d2a0f284 1247 assert(new_type_details->body_size);
c1ae03ae
NC
1248
1249#ifndef PURIFY
1250 assert(new_type_details->arena);
d2a0f284 1251 assert(new_type_details->arena_size);
c1ae03ae 1252 /* This points to the start of the allocated area. */
d2a0f284
JC
1253 new_body_inline(new_body, new_type);
1254 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1255 new_body = ((char *)new_body) - new_type_details->offset;
1256#else
1257 /* We always allocated the full length item with PURIFY. To do this
1258 we fake things so that arena is false for all 16 types.. */
1259 new_body = new_NOARENAZ(new_type_details);
1260#endif
1261 SvANY(sv) = new_body;
1262 if (new_type == SVt_PVAV) {
1263 AvMAX(sv) = -1;
1264 AvFILLp(sv) = -1;
1265 AvREAL_only(sv);
1266 }
aeb18a1e 1267
bd81e77b
NC
1268 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1269 The target created by newSVrv also is, and it can have magic.
1270 However, it never has SvPVX set.
1271 */
1272 if (old_type >= SVt_RV) {
1273 assert(SvPVX_const(sv) == 0);
1274 }
aeb18a1e 1275
bd81e77b
NC
1276 /* Could put this in the else clause below, as PVMG must have SvPVX
1277 0 already (the assertion above) */
6136c704 1278 SvPV_set(sv, NULL);
93e68bfb 1279
bd81e77b 1280 if (old_type >= SVt_PVMG) {
e736a858 1281 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1282 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
bd81e77b
NC
1283 }
1284 break;
93e68bfb 1285
93e68bfb 1286
bd81e77b
NC
1287 case SVt_PVIV:
1288 /* XXX Is this still needed? Was it ever needed? Surely as there is
1289 no route from NV to PVIV, NOK can never be true */
1290 assert(!SvNOKp(sv));
1291 assert(!SvNOK(sv));
1292 case SVt_PVIO:
1293 case SVt_PVFM:
1294 case SVt_PVBM:
1295 case SVt_PVGV:
1296 case SVt_PVCV:
1297 case SVt_PVLV:
1298 case SVt_PVMG:
1299 case SVt_PVNV:
1300 case SVt_PV:
93e68bfb 1301
d2a0f284 1302 assert(new_type_details->body_size);
bd81e77b
NC
1303 /* We always allocated the full length item with PURIFY. To do this
1304 we fake things so that arena is false for all 16 types.. */
1305 if(new_type_details->arena) {
1306 /* This points to the start of the allocated area. */
d2a0f284
JC
1307 new_body_inline(new_body, new_type);
1308 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1309 new_body = ((char *)new_body) - new_type_details->offset;
1310 } else {
1311 new_body = new_NOARENAZ(new_type_details);
1312 }
1313 SvANY(sv) = new_body;
5e2fc214 1314
bd81e77b 1315 if (old_type_details->copy) {
f9ba3d20
NC
1316 /* There is now the potential for an upgrade from something without
1317 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1318 int offset = old_type_details->offset;
1319 int length = old_type_details->copy;
1320
1321 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1322 const int difference
f9ba3d20
NC
1323 = new_type_details->offset - old_type_details->offset;
1324 offset += difference;
1325 length -= difference;
1326 }
1327 assert (length >= 0);
1328
1329 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1330 char);
bd81e77b
NC
1331 }
1332
1333#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1334 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1335 * correct 0.0 for us. Otherwise, if the old body didn't have an
1336 * NV slot, but the new one does, then we need to initialise the
1337 * freshly created NV slot with whatever the correct bit pattern is
1338 * for 0.0 */
1339 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1340 SvNV_set(sv, 0);
82048762 1341#endif
5e2fc214 1342
bd81e77b 1343 if (new_type == SVt_PVIO)
f2524eef 1344 IoPAGE_LEN(sv) = 60;
bd81e77b 1345 if (old_type < SVt_RV)
6136c704 1346 SvPV_set(sv, NULL);
bd81e77b
NC
1347 break;
1348 default:
afd78fd5
JH
1349 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1350 (unsigned long)new_type);
bd81e77b 1351 }
73171d91 1352
d2a0f284
JC
1353 if (old_type_details->arena) {
1354 /* If there was an old body, then we need to free it.
1355 Note that there is an assumption that all bodies of types that
1356 can be upgraded came from arenas. Only the more complex non-
1357 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1358#ifdef PURIFY
1359 my_safefree(old_body);
1360#else
1361 del_body((void*)((char*)old_body + old_type_details->offset),
1362 &PL_body_roots[old_type]);
1363#endif
1364 }
1365}
73171d91 1366
bd81e77b
NC
1367/*
1368=for apidoc sv_backoff
73171d91 1369
bd81e77b
NC
1370Remove any string offset. You should normally use the C<SvOOK_off> macro
1371wrapper instead.
73171d91 1372
bd81e77b 1373=cut
73171d91
NC
1374*/
1375
bd81e77b
NC
1376int
1377Perl_sv_backoff(pTHX_ register SV *sv)
1378{
96a5add6 1379 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1380 assert(SvOOK(sv));
1381 assert(SvTYPE(sv) != SVt_PVHV);
1382 assert(SvTYPE(sv) != SVt_PVAV);
1383 if (SvIVX(sv)) {
1384 const char * const s = SvPVX_const(sv);
1385 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1386 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1387 SvIV_set(sv, 0);
1388 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1389 }
1390 SvFLAGS(sv) &= ~SVf_OOK;
1391 return 0;
1392}
73171d91 1393
bd81e77b
NC
1394/*
1395=for apidoc sv_grow
73171d91 1396
bd81e77b
NC
1397Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1398upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1399Use the C<SvGROW> wrapper instead.
93e68bfb 1400
bd81e77b
NC
1401=cut
1402*/
93e68bfb 1403
bd81e77b
NC
1404char *
1405Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1406{
1407 register char *s;
93e68bfb 1408
5db06880
NC
1409 if (PL_madskills && newlen >= 0x100000) {
1410 PerlIO_printf(Perl_debug_log,
1411 "Allocation too large: %"UVxf"\n", (UV)newlen);
1412 }
bd81e77b
NC
1413#ifdef HAS_64K_LIMIT
1414 if (newlen >= 0x10000) {
1415 PerlIO_printf(Perl_debug_log,
1416 "Allocation too large: %"UVxf"\n", (UV)newlen);
1417 my_exit(1);
1418 }
1419#endif /* HAS_64K_LIMIT */
1420 if (SvROK(sv))
1421 sv_unref(sv);
1422 if (SvTYPE(sv) < SVt_PV) {
1423 sv_upgrade(sv, SVt_PV);
1424 s = SvPVX_mutable(sv);
1425 }
1426 else if (SvOOK(sv)) { /* pv is offset? */
1427 sv_backoff(sv);
1428 s = SvPVX_mutable(sv);
1429 if (newlen > SvLEN(sv))
1430 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1431#ifdef HAS_64K_LIMIT
1432 if (newlen >= 0x10000)
1433 newlen = 0xFFFF;
1434#endif
1435 }
1436 else
1437 s = SvPVX_mutable(sv);
aeb18a1e 1438
bd81e77b
NC
1439 if (newlen > SvLEN(sv)) { /* need more room? */
1440 newlen = PERL_STRLEN_ROUNDUP(newlen);
1441 if (SvLEN(sv) && s) {
1442#ifdef MYMALLOC
1443 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1444 if (newlen <= l) {
1445 SvLEN_set(sv, l);
1446 return s;
1447 } else
1448#endif
1449 s = saferealloc(s, newlen);
1450 }
1451 else {
1452 s = safemalloc(newlen);
1453 if (SvPVX_const(sv) && SvCUR(sv)) {
1454 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1455 }
1456 }
1457 SvPV_set(sv, s);
1458 SvLEN_set(sv, newlen);
1459 }
1460 return s;
1461}
aeb18a1e 1462
bd81e77b
NC
1463/*
1464=for apidoc sv_setiv
932e9ff9 1465
bd81e77b
NC
1466Copies an integer into the given SV, upgrading first if necessary.
1467Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1468
bd81e77b
NC
1469=cut
1470*/
463ee0b2 1471
bd81e77b
NC
1472void
1473Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1474{
97aff369 1475 dVAR;
bd81e77b
NC
1476 SV_CHECK_THINKFIRST_COW_DROP(sv);
1477 switch (SvTYPE(sv)) {
1478 case SVt_NULL:
1479 sv_upgrade(sv, SVt_IV);
1480 break;
1481 case SVt_NV:
1482 sv_upgrade(sv, SVt_PVNV);
1483 break;
1484 case SVt_RV:
1485 case SVt_PV:
1486 sv_upgrade(sv, SVt_PVIV);
1487 break;
463ee0b2 1488
bd81e77b
NC
1489 case SVt_PVGV:
1490 case SVt_PVAV:
1491 case SVt_PVHV:
1492 case SVt_PVCV:
1493 case SVt_PVFM:
1494 case SVt_PVIO:
1495 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1496 OP_DESC(PL_op));
1497 }
1498 (void)SvIOK_only(sv); /* validate number */
1499 SvIV_set(sv, i);
1500 SvTAINT(sv);
1501}
932e9ff9 1502
bd81e77b
NC
1503/*
1504=for apidoc sv_setiv_mg
d33b2eba 1505
bd81e77b 1506Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1507
bd81e77b
NC
1508=cut
1509*/
d33b2eba 1510
bd81e77b
NC
1511void
1512Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1513{
1514 sv_setiv(sv,i);
1515 SvSETMAGIC(sv);
1516}
727879eb 1517
bd81e77b
NC
1518/*
1519=for apidoc sv_setuv
d33b2eba 1520
bd81e77b
NC
1521Copies an unsigned integer into the given SV, upgrading first if necessary.
1522Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1523
bd81e77b
NC
1524=cut
1525*/
d33b2eba 1526
bd81e77b
NC
1527void
1528Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1529{
1530 /* With these two if statements:
1531 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1532
bd81e77b
NC
1533 without
1534 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1535
bd81e77b
NC
1536 If you wish to remove them, please benchmark to see what the effect is
1537 */
1538 if (u <= (UV)IV_MAX) {
1539 sv_setiv(sv, (IV)u);
1540 return;
1541 }
1542 sv_setiv(sv, 0);
1543 SvIsUV_on(sv);
1544 SvUV_set(sv, u);
1545}
d33b2eba 1546
bd81e77b
NC
1547/*
1548=for apidoc sv_setuv_mg
727879eb 1549
bd81e77b 1550Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1551
bd81e77b
NC
1552=cut
1553*/
5e2fc214 1554
bd81e77b
NC
1555void
1556Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1557{
1558 sv_setiv(sv, 0);
1559 SvIsUV_on(sv);
1560 sv_setuv(sv,u);
1561 SvSETMAGIC(sv);
1562}
5e2fc214 1563
954c1994 1564/*
bd81e77b 1565=for apidoc sv_setnv
954c1994 1566
bd81e77b
NC
1567Copies a double into the given SV, upgrading first if necessary.
1568Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1569
1570=cut
1571*/
1572
63f97190 1573void
bd81e77b 1574Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1575{
97aff369 1576 dVAR;
bd81e77b
NC
1577 SV_CHECK_THINKFIRST_COW_DROP(sv);
1578 switch (SvTYPE(sv)) {
79072805 1579 case SVt_NULL:
79072805 1580 case SVt_IV:
bd81e77b 1581 sv_upgrade(sv, SVt_NV);
79072805 1582 break;
ed6116ce 1583 case SVt_RV:
79072805 1584 case SVt_PV:
79072805 1585 case SVt_PVIV:
bd81e77b 1586 sv_upgrade(sv, SVt_PVNV);
79072805 1587 break;
bd4b1eb5 1588
bd4b1eb5 1589 case SVt_PVGV:
bd81e77b
NC
1590 case SVt_PVAV:
1591 case SVt_PVHV:
79072805 1592 case SVt_PVCV:
bd81e77b
NC
1593 case SVt_PVFM:
1594 case SVt_PVIO:
1595 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1596 OP_NAME(PL_op));
2068cd4d 1597 }
bd81e77b
NC
1598 SvNV_set(sv, num);
1599 (void)SvNOK_only(sv); /* validate number */
1600 SvTAINT(sv);
79072805
LW
1601}
1602
645c22ef 1603/*
bd81e77b 1604=for apidoc sv_setnv_mg
645c22ef 1605
bd81e77b 1606Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1607
1608=cut
1609*/
1610
bd81e77b
NC
1611void
1612Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1613{
bd81e77b
NC
1614 sv_setnv(sv,num);
1615 SvSETMAGIC(sv);
79072805
LW
1616}
1617
bd81e77b
NC
1618/* Print an "isn't numeric" warning, using a cleaned-up,
1619 * printable version of the offending string
1620 */
954c1994 1621
bd81e77b
NC
1622STATIC void
1623S_not_a_number(pTHX_ SV *sv)
79072805 1624{
97aff369 1625 dVAR;
bd81e77b
NC
1626 SV *dsv;
1627 char tmpbuf[64];
1628 const char *pv;
94463019
JH
1629
1630 if (DO_UTF8(sv)) {
396482e1 1631 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1632 pv = sv_uni_display(dsv, sv, 10, 0);
1633 } else {
1634 char *d = tmpbuf;
551405c4 1635 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1636 /* each *s can expand to 4 chars + "...\0",
1637 i.e. need room for 8 chars */
ecdeb87c 1638
00b6aa41
AL
1639 const char *s = SvPVX_const(sv);
1640 const char * const end = s + SvCUR(sv);
1641 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1642 int ch = *s & 0xFF;
1643 if (ch & 128 && !isPRINT_LC(ch)) {
1644 *d++ = 'M';
1645 *d++ = '-';
1646 ch &= 127;
1647 }
1648 if (ch == '\n') {
1649 *d++ = '\\';
1650 *d++ = 'n';
1651 }
1652 else if (ch == '\r') {
1653 *d++ = '\\';
1654 *d++ = 'r';
1655 }
1656 else if (ch == '\f') {
1657 *d++ = '\\';
1658 *d++ = 'f';
1659 }
1660 else if (ch == '\\') {
1661 *d++ = '\\';
1662 *d++ = '\\';
1663 }
1664 else if (ch == '\0') {
1665 *d++ = '\\';
1666 *d++ = '0';
1667 }
1668 else if (isPRINT_LC(ch))
1669 *d++ = ch;
1670 else {
1671 *d++ = '^';
1672 *d++ = toCTRL(ch);
1673 }
1674 }
1675 if (s < end) {
1676 *d++ = '.';
1677 *d++ = '.';
1678 *d++ = '.';
1679 }
1680 *d = '\0';
1681 pv = tmpbuf;
a0d0e21e 1682 }
a0d0e21e 1683
533c011a 1684 if (PL_op)
9014280d 1685 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1686 "Argument \"%s\" isn't numeric in %s", pv,
1687 OP_DESC(PL_op));
a0d0e21e 1688 else
9014280d 1689 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1690 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1691}
1692
c2988b20
NC
1693/*
1694=for apidoc looks_like_number
1695
645c22ef
DM
1696Test if the content of an SV looks like a number (or is a number).
1697C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1698non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1699
1700=cut
1701*/
1702
1703I32
1704Perl_looks_like_number(pTHX_ SV *sv)
1705{
a3b680e6 1706 register const char *sbegin;
c2988b20
NC
1707 STRLEN len;
1708
1709 if (SvPOK(sv)) {
3f7c398e 1710 sbegin = SvPVX_const(sv);
c2988b20
NC
1711 len = SvCUR(sv);
1712 }
1713 else if (SvPOKp(sv))
83003860 1714 sbegin = SvPV_const(sv, len);
c2988b20 1715 else
e0ab1c0e 1716 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1717 return grok_number(sbegin, len, NULL);
1718}
25da4f38 1719
19f6321d
NC
1720STATIC bool
1721S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1722{
1723 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1724 SV *const buffer = sv_newmortal();
1725
1726 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1727 is on. */
1728 SvFAKE_off(gv);
1729 gv_efullname3(buffer, gv, "*");
1730 SvFLAGS(gv) |= wasfake;
1731
675c862f
AL
1732 /* We know that all GVs stringify to something that is not-a-number,
1733 so no need to test that. */
1734 if (ckWARN(WARN_NUMERIC))
1735 not_a_number(buffer);
1736 /* We just want something true to return, so that S_sv_2iuv_common
1737 can tail call us and return true. */
19f6321d 1738 return TRUE;
675c862f
AL
1739}
1740
1741STATIC char *
19f6321d 1742S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1743{
1744 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1745 SV *const buffer = sv_newmortal();
1746
1747 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1748 is on. */
1749 SvFAKE_off(gv);
1750 gv_efullname3(buffer, gv, "*");
1751 SvFLAGS(gv) |= wasfake;
1752
1753 assert(SvPOK(buffer));
1754 *len = SvCUR(buffer);
1755 return SvPVX(buffer);
180488f8
NC
1756}
1757
25da4f38
IZ
1758/* Actually, ISO C leaves conversion of UV to IV undefined, but
1759 until proven guilty, assume that things are not that bad... */
1760
645c22ef
DM
1761/*
1762 NV_PRESERVES_UV:
1763
1764 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1765 an IV (an assumption perl has been based on to date) it becomes necessary
1766 to remove the assumption that the NV always carries enough precision to
1767 recreate the IV whenever needed, and that the NV is the canonical form.
1768 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1769 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1770 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1771 1) to distinguish between IV/UV/NV slots that have cached a valid
1772 conversion where precision was lost and IV/UV/NV slots that have a
1773 valid conversion which has lost no precision
645c22ef 1774 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1775 would lose precision, the precise conversion (or differently
1776 imprecise conversion) is also performed and cached, to prevent
1777 requests for different numeric formats on the same SV causing
1778 lossy conversion chains. (lossless conversion chains are perfectly
1779 acceptable (still))
1780
1781
1782 flags are used:
1783 SvIOKp is true if the IV slot contains a valid value
1784 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1785 SvNOKp is true if the NV slot contains a valid value
1786 SvNOK is true only if the NV value is accurate
1787
1788 so
645c22ef 1789 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1790 IV(or UV) would lose accuracy over a direct conversion from PV to
1791 IV(or UV). If it would, cache both conversions, return NV, but mark
1792 SV as IOK NOKp (ie not NOK).
1793
645c22ef 1794 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1795 NV would lose accuracy over a direct conversion from PV to NV. If it
1796 would, cache both conversions, flag similarly.
1797
1798 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1799 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1800 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1801 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1802 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1803
645c22ef
DM
1804 The benefit of this is that operations such as pp_add know that if
1805 SvIOK is true for both left and right operands, then integer addition
1806 can be used instead of floating point (for cases where the result won't
1807 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1808 loss of precision compared with integer addition.
1809
1810 * making IV and NV equal status should make maths accurate on 64 bit
1811 platforms
1812 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1813 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1814 looking for SvIOK and checking for overflow will not outweigh the
1815 fp to integer speedup)
1816 * will slow down integer operations (callers of SvIV) on "inaccurate"
1817 values, as the change from SvIOK to SvIOKp will cause a call into
1818 sv_2iv each time rather than a macro access direct to the IV slot
1819 * should speed up number->string conversion on integers as IV is
645c22ef 1820 favoured when IV and NV are equally accurate
28e5dec8
JH
1821
1822 ####################################################################
645c22ef
DM
1823 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1824 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1825 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1826 ####################################################################
1827
645c22ef 1828 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1829 performance ratio.
1830*/
1831
1832#ifndef NV_PRESERVES_UV
645c22ef
DM
1833# define IS_NUMBER_UNDERFLOW_IV 1
1834# define IS_NUMBER_UNDERFLOW_UV 2
1835# define IS_NUMBER_IV_AND_UV 2
1836# define IS_NUMBER_OVERFLOW_IV 4
1837# define IS_NUMBER_OVERFLOW_UV 5
1838
1839/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1840
1841/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1842STATIC int
645c22ef 1843S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1844{
97aff369 1845 dVAR;
3f7c398e 1846 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
28e5dec8
JH
1847 if (SvNVX(sv) < (NV)IV_MIN) {
1848 (void)SvIOKp_on(sv);
1849 (void)SvNOK_on(sv);
45977657 1850 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1851 return IS_NUMBER_UNDERFLOW_IV;
1852 }
1853 if (SvNVX(sv) > (NV)UV_MAX) {
1854 (void)SvIOKp_on(sv);
1855 (void)SvNOK_on(sv);
1856 SvIsUV_on(sv);
607fa7f2 1857 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1858 return IS_NUMBER_OVERFLOW_UV;
1859 }
c2988b20
NC
1860 (void)SvIOKp_on(sv);
1861 (void)SvNOK_on(sv);
1862 /* Can't use strtol etc to convert this string. (See truth table in
1863 sv_2iv */
1864 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1865 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1866 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1867 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1868 } else {
1869 /* Integer is imprecise. NOK, IOKp */
1870 }
1871 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1872 }
1873 SvIsUV_on(sv);
607fa7f2 1874 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1875 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1876 if (SvUVX(sv) == UV_MAX) {
1877 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1878 possibly be preserved by NV. Hence, it must be overflow.
1879 NOK, IOKp */
1880 return IS_NUMBER_OVERFLOW_UV;
1881 }
1882 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1883 } else {
1884 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1885 }
c2988b20 1886 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1887}
645c22ef
DM
1888#endif /* !NV_PRESERVES_UV*/
1889
af359546
NC
1890STATIC bool
1891S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1892 dVAR;
af359546 1893 if (SvNOKp(sv)) {
28e5dec8
JH
1894 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1895 * without also getting a cached IV/UV from it at the same time
1896 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1897 * IV or UV at same time to avoid this. */
1898 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1899
1900 if (SvTYPE(sv) == SVt_NV)
1901 sv_upgrade(sv, SVt_PVNV);
1902
28e5dec8
JH
1903 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1904 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1905 certainly cast into the IV range at IV_MAX, whereas the correct
1906 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1907 cases go to UV */
cab190d4
JD
1908#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1909 if (Perl_isnan(SvNVX(sv))) {
1910 SvUV_set(sv, 0);
1911 SvIsUV_on(sv);
fdbe6d7c 1912 return FALSE;
cab190d4 1913 }
cab190d4 1914#endif
28e5dec8 1915 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1916 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1917 if (SvNVX(sv) == (NV) SvIVX(sv)
1918#ifndef NV_PRESERVES_UV
1919 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1920 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1921 /* Don't flag it as "accurately an integer" if the number
1922 came from a (by definition imprecise) NV operation, and
1923 we're outside the range of NV integer precision */
1924#endif
1925 ) {
1926 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1927 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1928 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1929 PTR2UV(sv),
1930 SvNVX(sv),
1931 SvIVX(sv)));
1932
1933 } else {
1934 /* IV not precise. No need to convert from PV, as NV
1935 conversion would already have cached IV if it detected
1936 that PV->IV would be better than PV->NV->IV
1937 flags already correct - don't set public IOK. */
1938 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1939 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1940 PTR2UV(sv),
1941 SvNVX(sv),
1942 SvIVX(sv)));
1943 }
1944 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1945 but the cast (NV)IV_MIN rounds to a the value less (more
1946 negative) than IV_MIN which happens to be equal to SvNVX ??
1947 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1948 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1949 (NV)UVX == NVX are both true, but the values differ. :-(
1950 Hopefully for 2s complement IV_MIN is something like
1951 0x8000000000000000 which will be exact. NWC */
d460ef45 1952 }
25da4f38 1953 else {
607fa7f2 1954 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1955 if (
1956 (SvNVX(sv) == (NV) SvUVX(sv))
1957#ifndef NV_PRESERVES_UV
1958 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1959 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1960 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1961 /* Don't flag it as "accurately an integer" if the number
1962 came from a (by definition imprecise) NV operation, and
1963 we're outside the range of NV integer precision */
1964#endif
1965 )
1966 SvIOK_on(sv);
25da4f38 1967 SvIsUV_on(sv);
1c846c1f 1968 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1969 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1970 PTR2UV(sv),
57def98f
JH
1971 SvUVX(sv),
1972 SvUVX(sv)));
25da4f38 1973 }
748a9306
LW
1974 }
1975 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1976 UV value;
504618e9 1977 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1978 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1979 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1980 the same as the direct translation of the initial string
1981 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1982 be careful to ensure that the value with the .456 is around if the
1983 NV value is requested in the future).
1c846c1f 1984
af359546 1985 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1986 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1987 cache the NV if we are sure it's not needed.
25da4f38 1988 */
16b7a9a4 1989
c2988b20
NC
1990 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1991 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1992 == IS_NUMBER_IN_UV) {
5e045b90 1993 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1994 if (SvTYPE(sv) < SVt_PVIV)
1995 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1996 (void)SvIOK_on(sv);
c2988b20
NC
1997 } else if (SvTYPE(sv) < SVt_PVNV)
1998 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1999
f2524eef 2000 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2001 we aren't going to call atof() below. If NVs don't preserve UVs
2002 then the value returned may have more precision than atof() will
2003 return, even though value isn't perfectly accurate. */
2004 if ((numtype & (IS_NUMBER_IN_UV
2005#ifdef NV_PRESERVES_UV
2006 | IS_NUMBER_NOT_INT
2007#endif
2008 )) == IS_NUMBER_IN_UV) {
2009 /* This won't turn off the public IOK flag if it was set above */
2010 (void)SvIOKp_on(sv);
2011
2012 if (!(numtype & IS_NUMBER_NEG)) {
2013 /* positive */;
2014 if (value <= (UV)IV_MAX) {
45977657 2015 SvIV_set(sv, (IV)value);
c2988b20 2016 } else {
af359546 2017 /* it didn't overflow, and it was positive. */
607fa7f2 2018 SvUV_set(sv, value);
c2988b20
NC
2019 SvIsUV_on(sv);
2020 }
2021 } else {
2022 /* 2s complement assumption */
2023 if (value <= (UV)IV_MIN) {
45977657 2024 SvIV_set(sv, -(IV)value);
c2988b20
NC
2025 } else {
2026 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2027 I'm assuming it will be rare. */
c2988b20
NC
2028 if (SvTYPE(sv) < SVt_PVNV)
2029 sv_upgrade(sv, SVt_PVNV);
2030 SvNOK_on(sv);
2031 SvIOK_off(sv);
2032 SvIOKp_on(sv);
9d6ce603 2033 SvNV_set(sv, -(NV)value);
45977657 2034 SvIV_set(sv, IV_MIN);
c2988b20
NC
2035 }
2036 }
2037 }
2038 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2039 will be in the previous block to set the IV slot, and the next
2040 block to set the NV slot. So no else here. */
2041
2042 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2043 != IS_NUMBER_IN_UV) {
2044 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2045 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2046
c2988b20
NC
2047 if (! numtype && ckWARN(WARN_NUMERIC))
2048 not_a_number(sv);
28e5dec8 2049
65202027 2050#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2051 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2052 PTR2UV(sv), SvNVX(sv)));
65202027 2053#else
1779d84d 2054 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2055 PTR2UV(sv), SvNVX(sv)));
65202027 2056#endif
28e5dec8 2057
28e5dec8 2058#ifdef NV_PRESERVES_UV
af359546
NC
2059 (void)SvIOKp_on(sv);
2060 (void)SvNOK_on(sv);
2061 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2062 SvIV_set(sv, I_V(SvNVX(sv)));
2063 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2064 SvIOK_on(sv);
2065 } else {
6f207bd3 2066 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2067 }
2068 /* UV will not work better than IV */
2069 } else {
2070 if (SvNVX(sv) > (NV)UV_MAX) {
2071 SvIsUV_on(sv);
2072 /* Integer is inaccurate. NOK, IOKp, is UV */
2073 SvUV_set(sv, UV_MAX);
af359546
NC
2074 } else {
2075 SvUV_set(sv, U_V(SvNVX(sv)));
2076 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2077 NV preservse UV so can do correct comparison. */
2078 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2079 SvIOK_on(sv);
af359546 2080 } else {
6f207bd3 2081 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2082 }
2083 }
4b0c9573 2084 SvIsUV_on(sv);
af359546 2085 }
28e5dec8 2086#else /* NV_PRESERVES_UV */
c2988b20
NC
2087 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2088 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2089 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2090 grok_number above. The NV slot has just been set using
2091 Atof. */
560b0c46 2092 SvNOK_on(sv);
c2988b20
NC
2093 assert (SvIOKp(sv));
2094 } else {
2095 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2096 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2097 /* Small enough to preserve all bits. */
2098 (void)SvIOKp_on(sv);
2099 SvNOK_on(sv);
45977657 2100 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2101 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2102 SvIOK_on(sv);
2103 /* Assumption: first non-preserved integer is < IV_MAX,
2104 this NV is in the preserved range, therefore: */
2105 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2106 < (UV)IV_MAX)) {
32fdb065 2107 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
c2988b20
NC
2108 }
2109 } else {
2110 /* IN_UV NOT_INT
2111 0 0 already failed to read UV.
2112 0 1 already failed to read UV.
2113 1 0 you won't get here in this case. IV/UV
2114 slot set, public IOK, Atof() unneeded.
2115 1 1 already read UV.
2116 so there's no point in sv_2iuv_non_preserve() attempting
2117 to use atol, strtol, strtoul etc. */
40a17c4c 2118 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2119 }
2120 }
28e5dec8 2121#endif /* NV_PRESERVES_UV */
25da4f38 2122 }
af359546
NC
2123 }
2124 else {
675c862f 2125 if (isGV_with_GP(sv))
a0933d07 2126 return glob_2number((GV *)sv);
180488f8 2127
af359546
NC
2128 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2129 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2130 report_uninit(sv);
2131 }
25da4f38
IZ
2132 if (SvTYPE(sv) < SVt_IV)
2133 /* Typically the caller expects that sv_any is not NULL now. */
2134 sv_upgrade(sv, SVt_IV);
af359546
NC
2135 /* Return 0 from the caller. */
2136 return TRUE;
2137 }
2138 return FALSE;
2139}
2140
2141/*
2142=for apidoc sv_2iv_flags
2143
2144Return the integer value of an SV, doing any necessary string
2145conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2146Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2147
2148=cut
2149*/
2150
2151IV
2152Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2153{
97aff369 2154 dVAR;
af359546 2155 if (!sv)
a0d0e21e 2156 return 0;
af359546
NC
2157 if (SvGMAGICAL(sv)) {
2158 if (flags & SV_GMAGIC)
2159 mg_get(sv);
2160 if (SvIOKp(sv))
2161 return SvIVX(sv);
2162 if (SvNOKp(sv)) {
2163 return I_V(SvNVX(sv));
2164 }
71c558c3
NC
2165 if (SvPOKp(sv) && SvLEN(sv)) {
2166 UV value;
2167 const int numtype
2168 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2169
2170 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2171 == IS_NUMBER_IN_UV) {
2172 /* It's definitely an integer */
2173 if (numtype & IS_NUMBER_NEG) {
2174 if (value < (UV)IV_MIN)
2175 return -(IV)value;
2176 } else {
2177 if (value < (UV)IV_MAX)
2178 return (IV)value;
2179 }
2180 }
2181 if (!numtype) {
2182 if (ckWARN(WARN_NUMERIC))
2183 not_a_number(sv);
2184 }
2185 return I_V(Atof(SvPVX_const(sv)));
2186 }
1c7ff15e
NC
2187 if (SvROK(sv)) {
2188 goto return_rok;
af359546 2189 }
1c7ff15e
NC
2190 assert(SvTYPE(sv) >= SVt_PVMG);
2191 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2192 } else if (SvTHINKFIRST(sv)) {
af359546 2193 if (SvROK(sv)) {
1c7ff15e 2194 return_rok:
af359546
NC
2195 if (SvAMAGIC(sv)) {
2196 SV * const tmpstr=AMG_CALLun(sv,numer);
2197 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2198 return SvIV(tmpstr);
2199 }
2200 }
2201 return PTR2IV(SvRV(sv));
2202 }
2203 if (SvIsCOW(sv)) {
2204 sv_force_normal_flags(sv, 0);
2205 }
2206 if (SvREADONLY(sv) && !SvOK(sv)) {
2207 if (ckWARN(WARN_UNINITIALIZED))
2208 report_uninit(sv);
2209 return 0;
2210 }
2211 }
2212 if (!SvIOKp(sv)) {
2213 if (S_sv_2iuv_common(aTHX_ sv))
2214 return 0;
79072805 2215 }
1d7c1841
GS
2216 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2217 PTR2UV(sv),SvIVX(sv)));
25da4f38 2218 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2219}
2220
645c22ef 2221/*
891f9566 2222=for apidoc sv_2uv_flags
645c22ef
DM
2223
2224Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2225conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2226Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2227
2228=cut
2229*/
2230
ff68c719 2231UV
891f9566 2232Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2233{
97aff369 2234 dVAR;
ff68c719 2235 if (!sv)
2236 return 0;
2237 if (SvGMAGICAL(sv)) {
891f9566
YST
2238 if (flags & SV_GMAGIC)
2239 mg_get(sv);
ff68c719 2240 if (SvIOKp(sv))
2241 return SvUVX(sv);
2242 if (SvNOKp(sv))
2243 return U_V(SvNVX(sv));
71c558c3
NC
2244 if (SvPOKp(sv) && SvLEN(sv)) {
2245 UV value;
2246 const int numtype
2247 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2248
2249 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2250 == IS_NUMBER_IN_UV) {
2251 /* It's definitely an integer */
2252 if (!(numtype & IS_NUMBER_NEG))
2253 return value;
2254 }
2255 if (!numtype) {
2256 if (ckWARN(WARN_NUMERIC))
2257 not_a_number(sv);
2258 }
2259 return U_V(Atof(SvPVX_const(sv)));
2260 }
1c7ff15e
NC
2261 if (SvROK(sv)) {
2262 goto return_rok;
3fe9a6f1 2263 }
1c7ff15e
NC
2264 assert(SvTYPE(sv) >= SVt_PVMG);
2265 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2266 } else if (SvTHINKFIRST(sv)) {
ff68c719 2267 if (SvROK(sv)) {
1c7ff15e 2268 return_rok:
deb46114
NC
2269 if (SvAMAGIC(sv)) {
2270 SV *const tmpstr = AMG_CALLun(sv,numer);
2271 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2272 return SvUV(tmpstr);
2273 }
2274 }
2275 return PTR2UV(SvRV(sv));
ff68c719 2276 }
765f542d
NC
2277 if (SvIsCOW(sv)) {
2278 sv_force_normal_flags(sv, 0);
8a818333 2279 }
0336b60e 2280 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2281 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2282 report_uninit(sv);
ff68c719 2283 return 0;
2284 }
2285 }
af359546
NC
2286 if (!SvIOKp(sv)) {
2287 if (S_sv_2iuv_common(aTHX_ sv))
2288 return 0;
ff68c719 2289 }
25da4f38 2290
1d7c1841
GS
2291 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2292 PTR2UV(sv),SvUVX(sv)));
25da4f38 2293 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2294}
2295
645c22ef
DM
2296/*
2297=for apidoc sv_2nv
2298
2299Return the num value of an SV, doing any necessary string or integer
2300conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2301macros.
2302
2303=cut
2304*/
2305
65202027 2306NV
864dbfa3 2307Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2308{
97aff369 2309 dVAR;
79072805
LW
2310 if (!sv)
2311 return 0.0;
8990e307 2312 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2313 mg_get(sv);
2314 if (SvNOKp(sv))
2315 return SvNVX(sv);
0aa395f8 2316 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2317 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2318 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2319 not_a_number(sv);
3f7c398e 2320 return Atof(SvPVX_const(sv));
a0d0e21e 2321 }
25da4f38 2322 if (SvIOKp(sv)) {
1c846c1f 2323 if (SvIsUV(sv))
65202027 2324 return (NV)SvUVX(sv);
25da4f38 2325 else
65202027 2326 return (NV)SvIVX(sv);
47a72cb8
NC
2327 }
2328 if (SvROK(sv)) {
2329 goto return_rok;
2330 }
2331 assert(SvTYPE(sv) >= SVt_PVMG);
2332 /* This falls through to the report_uninit near the end of the
2333 function. */
2334 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2335 if (SvROK(sv)) {
47a72cb8 2336 return_rok:
deb46114
NC
2337 if (SvAMAGIC(sv)) {
2338 SV *const tmpstr = AMG_CALLun(sv,numer);
2339 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2340 return SvNV(tmpstr);
2341 }
2342 }
2343 return PTR2NV(SvRV(sv));
a0d0e21e 2344 }
765f542d
NC
2345 if (SvIsCOW(sv)) {
2346 sv_force_normal_flags(sv, 0);
8a818333 2347 }
0336b60e 2348 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2349 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2350 report_uninit(sv);
ed6116ce
LW
2351 return 0.0;
2352 }
79072805
LW
2353 }
2354 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2355 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2356 sv_upgrade(sv, SVt_NV);
906f284f 2357#ifdef USE_LONG_DOUBLE
097ee67d 2358 DEBUG_c({
f93f4e46 2359 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2360 PerlIO_printf(Perl_debug_log,
2361 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2362 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2363 RESTORE_NUMERIC_LOCAL();
2364 });
65202027 2365#else
572bbb43 2366 DEBUG_c({
f93f4e46 2367 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2368 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2369 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2370 RESTORE_NUMERIC_LOCAL();
2371 });
572bbb43 2372#endif
79072805
LW
2373 }
2374 else if (SvTYPE(sv) < SVt_PVNV)
2375 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2376 if (SvNOKp(sv)) {
2377 return SvNVX(sv);
61604483 2378 }
59d8ce62 2379 if (SvIOKp(sv)) {
9d6ce603 2380 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2381#ifdef NV_PRESERVES_UV
2382 SvNOK_on(sv);
2383#else
2384 /* Only set the public NV OK flag if this NV preserves the IV */
2385 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2386 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2387 : (SvIVX(sv) == I_V(SvNVX(sv))))
2388 SvNOK_on(sv);
2389 else
2390 SvNOKp_on(sv);
2391#endif
93a17b20 2392 }
748a9306 2393 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2394 UV value;
3f7c398e 2395 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2396 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2397 not_a_number(sv);
28e5dec8 2398#ifdef NV_PRESERVES_UV
c2988b20
NC
2399 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2400 == IS_NUMBER_IN_UV) {
5e045b90 2401 /* It's definitely an integer */
9d6ce603 2402 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2403 } else
3f7c398e 2404 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2405 SvNOK_on(sv);
2406#else
3f7c398e 2407 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2408 /* Only set the public NV OK flag if this NV preserves the value in
2409 the PV at least as well as an IV/UV would.
2410 Not sure how to do this 100% reliably. */
2411 /* if that shift count is out of range then Configure's test is
2412 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2413 UV_BITS */
2414 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2415 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2416 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2417 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2418 /* Can't use strtol etc to convert this string, so don't try.
2419 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2420 SvNOK_on(sv);
2421 } else {
2422 /* value has been set. It may not be precise. */
2423 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2424 /* 2s complement assumption for (UV)IV_MIN */
2425 SvNOK_on(sv); /* Integer is too negative. */
2426 } else {
2427 SvNOKp_on(sv);
2428 SvIOKp_on(sv);
6fa402ec 2429
c2988b20 2430 if (numtype & IS_NUMBER_NEG) {
45977657 2431 SvIV_set(sv, -(IV)value);
c2988b20 2432 } else if (value <= (UV)IV_MAX) {
45977657 2433 SvIV_set(sv, (IV)value);
c2988b20 2434 } else {
607fa7f2 2435 SvUV_set(sv, value);
c2988b20
NC
2436 SvIsUV_on(sv);
2437 }
2438
2439 if (numtype & IS_NUMBER_NOT_INT) {
2440 /* I believe that even if the original PV had decimals,
2441 they are lost beyond the limit of the FP precision.
2442 However, neither is canonical, so both only get p
2443 flags. NWC, 2000/11/25 */
2444 /* Both already have p flags, so do nothing */
2445 } else {
66a1b24b 2446 const NV nv = SvNVX(sv);
c2988b20
NC
2447 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2448 if (SvIVX(sv) == I_V(nv)) {
2449 SvNOK_on(sv);
c2988b20 2450 } else {
c2988b20
NC
2451 /* It had no "." so it must be integer. */
2452 }
00b6aa41 2453 SvIOK_on(sv);
c2988b20
NC
2454 } else {
2455 /* between IV_MAX and NV(UV_MAX).
2456 Could be slightly > UV_MAX */
6fa402ec 2457
c2988b20
NC
2458 if (numtype & IS_NUMBER_NOT_INT) {
2459 /* UV and NV both imprecise. */
2460 } else {
66a1b24b 2461 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2462
2463 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2464 SvNOK_on(sv);
c2988b20 2465 }
00b6aa41 2466 SvIOK_on(sv);
c2988b20
NC
2467 }
2468 }
2469 }
2470 }
2471 }
28e5dec8 2472#endif /* NV_PRESERVES_UV */
93a17b20 2473 }
79072805 2474 else {
f7877b28 2475 if (isGV_with_GP(sv)) {
19f6321d 2476 glob_2number((GV *)sv);
180488f8
NC
2477 return 0.0;
2478 }
2479
041457d9 2480 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2481 report_uninit(sv);
7e25a7e9
NC
2482 assert (SvTYPE(sv) >= SVt_NV);
2483 /* Typically the caller expects that sv_any is not NULL now. */
2484 /* XXX Ilya implies that this is a bug in callers that assume this
2485 and ideally should be fixed. */
a0d0e21e 2486 return 0.0;
79072805 2487 }
572bbb43 2488#if defined(USE_LONG_DOUBLE)
097ee67d 2489 DEBUG_c({
f93f4e46 2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2492 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2493 RESTORE_NUMERIC_LOCAL();
2494 });
65202027 2495#else
572bbb43 2496 DEBUG_c({
f93f4e46 2497 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2498 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2499 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2500 RESTORE_NUMERIC_LOCAL();
2501 });
572bbb43 2502#endif
463ee0b2 2503 return SvNVX(sv);
79072805
LW
2504}
2505
645c22ef
DM
2506/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2507 * UV as a string towards the end of buf, and return pointers to start and
2508 * end of it.
2509 *
2510 * We assume that buf is at least TYPE_CHARS(UV) long.
2511 */
2512
864dbfa3 2513static char *
aec46f14 2514S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2515{
25da4f38 2516 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2517 char * const ebuf = ptr;
25da4f38 2518 int sign;
25da4f38
IZ
2519
2520 if (is_uv)
2521 sign = 0;
2522 else if (iv >= 0) {
2523 uv = iv;
2524 sign = 0;
2525 } else {
2526 uv = -iv;
2527 sign = 1;
2528 }
2529 do {
eb160463 2530 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2531 } while (uv /= 10);
2532 if (sign)
2533 *--ptr = '-';
2534 *peob = ebuf;
2535 return ptr;
2536}
2537
9af30d34
NC
2538/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2539 * a regexp to its stringified form.
2540 */
2541
2542static char *
2543S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
97aff369 2544 dVAR;
00b6aa41 2545 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2546
2547 if (!mg->mg_ptr) {
2548 const char *fptr = "msix";
2549 char reflags[6];
2550 char ch;
2551 int left = 0;
2552 int right = 4;
00b6aa41 2553 bool need_newline = 0;
9af30d34
NC
2554 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2555
2556 while((ch = *fptr++)) {
2557 if(reganch & 1) {
2558 reflags[left++] = ch;
2559 }
2560 else {
2561 reflags[right--] = ch;
2562 }
2563 reganch >>= 1;
2564 }
2565 if(left != 4) {
2566 reflags[left] = '-';
2567 left = 5;
2568 }
2569
2570 mg->mg_len = re->prelen + 4 + left;
2571 /*
2572 * If /x was used, we have to worry about a regex ending with a
2573 * comment later being embedded within another regex. If so, we don't
2574 * want this regex's "commentization" to leak out to the right part of
2575 * the enclosing regex, we must cap it with a newline.
2576 *
2577 * So, if /x was used, we scan backwards from the end of the regex. If
2578 * we find a '#' before we find a newline, we need to add a newline
2579 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2580 * we don't need to add anything. -jfriedl
2581 */
2582 if (PMf_EXTENDED & re->reganch) {
2583 const char *endptr = re->precomp + re->prelen;
2584 while (endptr >= re->precomp) {
2585 const char c = *(endptr--);
2586 if (c == '\n')
2587 break; /* don't need another */
2588 if (c == '#') {
2589 /* we end while in a comment, so we need a newline */
2590 mg->mg_len++; /* save space for it */
2591 need_newline = 1; /* note to add it */
2592 break;
2593 }
2594 }
2595 }
2596
2597 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2598 mg->mg_ptr[0] = '(';
2599 mg->mg_ptr[1] = '?';
2600 Copy(reflags, mg->mg_ptr+2, left, char);
2601 *(mg->mg_ptr+left+2) = ':';
2602 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2603 if (need_newline)
2604 mg->mg_ptr[mg->mg_len - 2] = '\n';
2605 mg->mg_ptr[mg->mg_len - 1] = ')';
2606 mg->mg_ptr[mg->mg_len] = 0;
2607 }
2608 PL_reginterp_cnt += re->program[0].next_off;
2609
2610 if (re->reganch & ROPT_UTF8)
2611 SvUTF8_on(sv);
2612 else
2613 SvUTF8_off(sv);
2614 if (lp)
2615 *lp = mg->mg_len;
2616 return mg->mg_ptr;
2617}
2618
645c22ef
DM
2619/*
2620=for apidoc sv_2pv_flags
2621
ff276b08 2622Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2623If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2624if necessary.
2625Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2626usually end up here too.
2627
2628=cut
2629*/
2630
8d6d96c1
HS
2631char *
2632Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2633{
97aff369 2634 dVAR;
79072805 2635 register char *s;
79072805 2636
463ee0b2 2637 if (!sv) {
cdb061a3
NC
2638 if (lp)
2639 *lp = 0;
73d840c0 2640 return (char *)"";
463ee0b2 2641 }
8990e307 2642 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2643 if (flags & SV_GMAGIC)
2644 mg_get(sv);
463ee0b2 2645 if (SvPOKp(sv)) {
cdb061a3
NC
2646 if (lp)
2647 *lp = SvCUR(sv);
10516c54
NC
2648 if (flags & SV_MUTABLE_RETURN)
2649 return SvPVX_mutable(sv);
4d84ee25
NC
2650 if (flags & SV_CONST_RETURN)
2651 return (char *)SvPVX_const(sv);
463ee0b2
LW
2652 return SvPVX(sv);
2653 }
75dfc8ec
NC
2654 if (SvIOKp(sv) || SvNOKp(sv)) {
2655 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2656 STRLEN len;
2657
2658 if (SvIOKp(sv)) {
e8ada2d0
NC
2659 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2660 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2661 } else {
e8ada2d0
NC
2662 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2663 len = strlen(tbuf);
75dfc8ec 2664 }
b5b886f0
NC
2665 assert(!SvROK(sv));
2666 {
75dfc8ec
NC
2667 dVAR;
2668
2669#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2670 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2671 tbuf[0] = '0';
2672 tbuf[1] = 0;
75dfc8ec
NC
2673 len = 1;
2674 }
2675#endif
2676 SvUPGRADE(sv, SVt_PV);
2677 if (lp)
2678 *lp = len;
2679 s = SvGROW_mutable(sv, len + 1);
2680 SvCUR_set(sv, len);
2681 SvPOKp_on(sv);
e8ada2d0 2682 return memcpy(s, tbuf, len + 1);
75dfc8ec 2683 }
463ee0b2 2684 }
1c7ff15e
NC
2685 if (SvROK(sv)) {
2686 goto return_rok;
2687 }
2688 assert(SvTYPE(sv) >= SVt_PVMG);
2689 /* This falls through to the report_uninit near the end of the
2690 function. */
2691 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2692 if (SvROK(sv)) {
1c7ff15e 2693 return_rok:
deb46114
NC
2694 if (SvAMAGIC(sv)) {
2695 SV *const tmpstr = AMG_CALLun(sv,string);
2696 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2697 /* Unwrap this: */
2698 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2699 */
2700
2701 char *pv;
2702 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2703 if (flags & SV_CONST_RETURN) {
2704 pv = (char *) SvPVX_const(tmpstr);
2705 } else {
2706 pv = (flags & SV_MUTABLE_RETURN)
2707 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2708 }
2709 if (lp)
2710 *lp = SvCUR(tmpstr);
50adf7d2 2711 } else {
deb46114 2712 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2713 }
deb46114
NC
2714 if (SvUTF8(tmpstr))
2715 SvUTF8_on(sv);
2716 else
2717 SvUTF8_off(sv);
2718 return pv;
50adf7d2 2719 }
deb46114
NC
2720 }
2721 {
75dfc8ec 2722 SV *tsv;
f9277f47 2723 MAGIC *mg;
d8eae41e
NC
2724 const SV *const referent = (SV*)SvRV(sv);
2725
2726 if (!referent) {
396482e1 2727 tsv = sv_2mortal(newSVpvs("NULLREF"));
042dae7a
NC
2728 } else if (SvTYPE(referent) == SVt_PVMG
2729 && ((SvFLAGS(referent) &
2730 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2731 == (SVs_OBJECT|SVs_SMG))
2732 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
c445ea15 2733 return stringify_regexp(sv, mg, lp);
d8eae41e
NC
2734 } else {
2735 const char *const typestr = sv_reftype(referent, 0);
2736
2737 tsv = sv_newmortal();
2738 if (SvOBJECT(referent)) {
2739 const char *const name = HvNAME_get(SvSTASH(referent));
2740 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2741 name ? name : "__ANON__" , typestr,
2742 PTR2UV(referent));
2743 }
2744 else
2745 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2746 PTR2UV(referent));
c080367d 2747 }
042dae7a
NC
2748 if (lp)
2749 *lp = SvCUR(tsv);
2750 return SvPVX(tsv);
463ee0b2 2751 }
79072805 2752 }
0336b60e 2753 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2754 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2755 report_uninit(sv);
cdb061a3
NC
2756 if (lp)
2757 *lp = 0;
73d840c0 2758 return (char *)"";
79072805 2759 }
79072805 2760 }
28e5dec8
JH
2761 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2762 /* I'm assuming that if both IV and NV are equally valid then
2763 converting the IV is going to be more efficient */
e1ec3a88
AL
2764 const U32 isIOK = SvIOK(sv);
2765 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2766 char buf[TYPE_CHARS(UV)];
2767 char *ebuf, *ptr;
2768
2769 if (SvTYPE(sv) < SVt_PVIV)
2770 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2771 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2772 /* inlined from sv_setpvn */
2773 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2774 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2775 SvCUR_set(sv, ebuf - ptr);
2776 s = SvEND(sv);
2777 *s = '\0';
2778 if (isIOK)
2779 SvIOK_on(sv);
2780 else
2781 SvIOKp_on(sv);
2782 if (isUIOK)
2783 SvIsUV_on(sv);
2784 }
2785 else if (SvNOKp(sv)) {
c81271c3 2786 const int olderrno = errno;
79072805
LW
2787 if (SvTYPE(sv) < SVt_PVNV)
2788 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2789 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2790 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2791 /* some Xenix systems wipe out errno here */
79072805 2792#ifdef apollo
463ee0b2 2793 if (SvNVX(sv) == 0.0)
79072805
LW
2794 (void)strcpy(s,"0");
2795 else
2796#endif /*apollo*/
bbce6d69 2797 {
2d4389e4 2798 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2799 }
79072805 2800 errno = olderrno;
a0d0e21e
LW
2801#ifdef FIXNEGATIVEZERO
2802 if (*s == '-' && s[1] == '0' && !s[2])
2803 strcpy(s,"0");
2804#endif
79072805
LW
2805 while (*s) s++;
2806#ifdef hcx
2807 if (s[-1] == '.')
46fc3d4c 2808 *--s = '\0';
79072805
LW
2809#endif
2810 }
79072805 2811 else {
675c862f 2812 if (isGV_with_GP(sv))
19f6321d 2813 return glob_2pv((GV *)sv, lp);
180488f8 2814
041457d9 2815 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2816 report_uninit(sv);
cdb061a3 2817 if (lp)
00b6aa41 2818 *lp = 0;
25da4f38
IZ
2819 if (SvTYPE(sv) < SVt_PV)
2820 /* Typically the caller expects that sv_any is not NULL now. */
2821 sv_upgrade(sv, SVt_PV);
73d840c0 2822 return (char *)"";
79072805 2823 }
cdb061a3 2824 {
823a54a3 2825 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2826 if (lp)
2827 *lp = len;
2828 SvCUR_set(sv, len);
2829 }
79072805 2830 SvPOK_on(sv);
1d7c1841 2831 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2832 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2833 if (flags & SV_CONST_RETURN)
2834 return (char *)SvPVX_const(sv);
10516c54
NC
2835 if (flags & SV_MUTABLE_RETURN)
2836 return SvPVX_mutable(sv);
463ee0b2
LW
2837 return SvPVX(sv);
2838}
2839
645c22ef 2840/*
6050d10e
JP
2841=for apidoc sv_copypv
2842
2843Copies a stringified representation of the source SV into the
2844destination SV. Automatically performs any necessary mg_get and
54f0641b 2845coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2846UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2847sv_2pv[_flags] but operates directly on an SV instead of just the
2848string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2849would lose the UTF-8'ness of the PV.
2850
2851=cut
2852*/
2853
2854void
2855Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2856{
446eaa42 2857 STRLEN len;
53c1dcc0 2858 const char * const s = SvPV_const(ssv,len);
cb50f42d 2859 sv_setpvn(dsv,s,len);
446eaa42 2860 if (SvUTF8(ssv))
cb50f42d 2861 SvUTF8_on(dsv);
446eaa42 2862 else
cb50f42d 2863 SvUTF8_off(dsv);
6050d10e
JP
2864}
2865
2866/*
645c22ef
DM
2867=for apidoc sv_2pvbyte
2868
2869Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2870to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2871side-effect.
2872
2873Usually accessed via the C<SvPVbyte> macro.
2874
2875=cut
2876*/
2877
7340a771
GS
2878char *
2879Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2880{
0875d2fe 2881 sv_utf8_downgrade(sv,0);
97972285 2882 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2883}
2884
645c22ef 2885/*
035cbb0e
RGS
2886=for apidoc sv_2pvutf8
2887
2888Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2889to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2890
2891Usually accessed via the C<SvPVutf8> macro.
2892
2893=cut
2894*/
645c22ef 2895
7340a771
GS
2896char *
2897Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2898{
035cbb0e
RGS
2899 sv_utf8_upgrade(sv);
2900 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2901}
1c846c1f 2902
7ee2227d 2903
645c22ef
DM
2904/*
2905=for apidoc sv_2bool
2906
2907This function is only called on magical items, and is only used by
8cf8f3d1 2908sv_true() or its macro equivalent.
645c22ef
DM
2909
2910=cut
2911*/
2912
463ee0b2 2913bool
864dbfa3 2914Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2915{
97aff369 2916 dVAR;
5b295bef 2917 SvGETMAGIC(sv);
463ee0b2 2918
a0d0e21e
LW
2919 if (!SvOK(sv))
2920 return 0;
2921 if (SvROK(sv)) {
fabdb6c0
AL
2922 if (SvAMAGIC(sv)) {
2923 SV * const tmpsv = AMG_CALLun(sv,bool_);
2924 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2925 return (bool)SvTRUE(tmpsv);
2926 }
2927 return SvRV(sv) != 0;
a0d0e21e 2928 }
463ee0b2 2929 if (SvPOKp(sv)) {
53c1dcc0
AL
2930 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2931 if (Xpvtmp &&
339049b0 2932 (*sv->sv_u.svu_pv > '0' ||
11343788 2933 Xpvtmp->xpv_cur > 1 ||
339049b0 2934 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2935 return 1;
2936 else
2937 return 0;
2938 }
2939 else {
2940 if (SvIOKp(sv))
2941 return SvIVX(sv) != 0;
2942 else {
2943 if (SvNOKp(sv))
2944 return SvNVX(sv) != 0.0;
180488f8 2945 else {
f7877b28 2946 if (isGV_with_GP(sv))
180488f8
NC
2947 return TRUE;
2948 else
2949 return FALSE;
2950 }
463ee0b2
LW
2951 }
2952 }
79072805
LW
2953}
2954
c461cf8f
JH
2955/*
2956=for apidoc sv_utf8_upgrade
2957
78ea37eb 2958Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2959Forces the SV to string form if it is not already.
4411f3b6
NIS
2960Always sets the SvUTF8 flag to avoid future validity checks even
2961if all the bytes have hibit clear.
c461cf8f 2962
13a6c0e0
JH
2963This is not as a general purpose byte encoding to Unicode interface:
2964use the Encode extension for that.
2965
8d6d96c1
HS
2966=for apidoc sv_utf8_upgrade_flags
2967
78ea37eb 2968Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2969Forces the SV to string form if it is not already.
8d6d96c1
HS
2970Always sets the SvUTF8 flag to avoid future validity checks even
2971if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2972will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2973C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2974
13a6c0e0
JH
2975This is not as a general purpose byte encoding to Unicode interface:
2976use the Encode extension for that.
2977
8d6d96c1
HS
2978=cut
2979*/
2980
2981STRLEN
2982Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2983{
97aff369 2984 dVAR;
808c356f
RGS
2985 if (sv == &PL_sv_undef)
2986 return 0;
e0e62c2a
NIS
2987 if (!SvPOK(sv)) {
2988 STRLEN len = 0;
d52b7888
NC
2989 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2990 (void) sv_2pv_flags(sv,&len, flags);
2991 if (SvUTF8(sv))
2992 return len;
2993 } else {
2994 (void) SvPV_force(sv,len);
2995 }
e0e62c2a 2996 }
4411f3b6 2997
f5cee72b 2998 if (SvUTF8(sv)) {
5fec3b1d 2999 return SvCUR(sv);
f5cee72b 3000 }
5fec3b1d 3001
765f542d
NC
3002 if (SvIsCOW(sv)) {
3003 sv_force_normal_flags(sv, 0);
db42d148
NIS
3004 }
3005
88632417 3006 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 3007 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 3008 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
3009 /* This function could be much more efficient if we
3010 * had a FLAG in SVs to signal if there are any hibit
3011 * chars in the PV. Given that there isn't such a flag
3012 * make the loop as fast as possible. */
00b6aa41 3013 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3014 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3015 const U8 *t = s;
c4e7c712
NC
3016
3017 while (t < e) {
53c1dcc0 3018 const U8 ch = *t++;
00b6aa41
AL
3019 /* Check for hi bit */
3020 if (!NATIVE_IS_INVARIANT(ch)) {
3021 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3022 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3023
3024 SvPV_free(sv); /* No longer using what was there before. */
3025 SvPV_set(sv, (char*)recoded);
3026 SvCUR_set(sv, len - 1);
3027 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3028 break;
00b6aa41 3029 }
c4e7c712
NC
3030 }
3031 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3032 SvUTF8_on(sv);
560a288e 3033 }
4411f3b6 3034 return SvCUR(sv);
560a288e
GS
3035}
3036
c461cf8f
JH
3037/*
3038=for apidoc sv_utf8_downgrade
3039
78ea37eb
TS
3040Attempts to convert the PV of an SV from characters to bytes.
3041If the PV contains a character beyond byte, this conversion will fail;
3042in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3043true, croaks.
3044
13a6c0e0
JH
3045This is not as a general purpose Unicode to byte encoding interface:
3046use the Encode extension for that.
3047
c461cf8f
JH
3048=cut
3049*/
3050
560a288e
GS
3051bool
3052Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3053{
97aff369 3054 dVAR;
78ea37eb 3055 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3056 if (SvCUR(sv)) {
03cfe0ae 3057 U8 *s;
652088fc 3058 STRLEN len;
fa301091 3059
765f542d
NC
3060 if (SvIsCOW(sv)) {
3061 sv_force_normal_flags(sv, 0);
3062 }
03cfe0ae
NIS
3063 s = (U8 *) SvPV(sv, len);
3064 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3065 if (fail_ok)
3066 return FALSE;
3067 else {
3068 if (PL_op)
3069 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3070 OP_DESC(PL_op));
fa301091
JH
3071 else
3072 Perl_croak(aTHX_ "Wide character");
3073 }
4b3603a4 3074 }
b162af07 3075 SvCUR_set(sv, len);
67e989fb 3076 }
560a288e 3077 }
ffebcc3e 3078 SvUTF8_off(sv);
560a288e
GS
3079 return TRUE;
3080}
3081
c461cf8f
JH
3082/*
3083=for apidoc sv_utf8_encode
3084
78ea37eb
TS
3085Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3086flag off so that it looks like octets again.
c461cf8f
JH
3087
3088=cut
3089*/
3090
560a288e
GS
3091void
3092Perl_sv_utf8_encode(pTHX_ register SV *sv)
3093{
4411f3b6 3094 (void) sv_utf8_upgrade(sv);
4c94c214
NC
3095 if (SvIsCOW(sv)) {
3096 sv_force_normal_flags(sv, 0);
3097 }
3098 if (SvREADONLY(sv)) {
3099 Perl_croak(aTHX_ PL_no_modify);
3100 }
560a288e
GS
3101 SvUTF8_off(sv);
3102}
3103
4411f3b6
NIS
3104/*
3105=for apidoc sv_utf8_decode
3106
78ea37eb
TS
3107If the PV of the SV is an octet sequence in UTF-8
3108and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3109so that it looks like a character. If the PV contains only single-byte
3110characters, the C<SvUTF8> flag stays being off.
3111Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3112
3113=cut
3114*/
3115
560a288e
GS
3116bool
3117Perl_sv_utf8_decode(pTHX_ register SV *sv)
3118{
78ea37eb 3119 if (SvPOKp(sv)) {
93524f2b
NC
3120 const U8 *c;
3121 const U8 *e;
9cbac4c7 3122
645c22ef
DM
3123 /* The octets may have got themselves encoded - get them back as
3124 * bytes
3125 */
3126 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3127 return FALSE;
3128
3129 /* it is actually just a matter of turning the utf8 flag on, but
3130 * we want to make sure everything inside is valid utf8 first.
3131 */
93524f2b 3132 c = (const U8 *) SvPVX_const(sv);
63cd0674 3133 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3134 return FALSE;
93524f2b 3135 e = (const U8 *) SvEND(sv);
511c2ff0 3136 while (c < e) {
b64e5050 3137 const U8 ch = *c++;
c4d5f83a 3138 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3139 SvUTF8_on(sv);
3140 break;
3141 }
560a288e 3142 }
560a288e
GS
3143 }
3144 return TRUE;
3145}
3146
954c1994
GS
3147/*
3148=for apidoc sv_setsv
3149
645c22ef
DM
3150Copies the contents of the source SV C<ssv> into the destination SV
3151C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3152function if the source SV needs to be reused. Does not handle 'set' magic.
3153Loosely speaking, it performs a copy-by-value, obliterating any previous
3154content of the destination.
3155
3156You probably want to use one of the assortment of wrappers, such as
3157C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3158C<SvSetMagicSV_nosteal>.
3159
8d6d96c1
HS
3160=for apidoc sv_setsv_flags
3161
645c22ef
DM
3162Copies the contents of the source SV C<ssv> into the destination SV
3163C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3164function if the source SV needs to be reused. Does not handle 'set' magic.
3165Loosely speaking, it performs a copy-by-value, obliterating any previous
3166content of the destination.
3167If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3168C<ssv> if appropriate, else not. If the C<flags> parameter has the
3169C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3170and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3171
3172You probably want to use one of the assortment of wrappers, such as
3173C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3174C<SvSetMagicSV_nosteal>.
3175
3176This is the primary function for copying scalars, and most other
3177copy-ish functions and macros use this underneath.
8d6d96c1
HS
3178
3179=cut
3180*/
3181
5d0301b7 3182static void
2eb42952 3183S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3184{
3185 if (dtype != SVt_PVGV) {
3186 const char * const name = GvNAME(sstr);
3187 const STRLEN len = GvNAMELEN(sstr);
3188 /* don't upgrade SVt_PVLV: it can hold a glob */
f7877b28
NC
3189 if (dtype != SVt_PVLV) {
3190 if (dtype >= SVt_PV) {
3191 SvPV_free(dstr);
3192 SvPV_set(dstr, 0);
3193 SvLEN_set(dstr, 0);
3194 SvCUR_set(dstr, 0);
3195 }
5d0301b7 3196 sv_upgrade(dstr, SVt_PVGV);
dedf8e73
NC
3197 (void)SvOK_off(dstr);
3198 SvSCREAM_on(dstr);
f7877b28 3199 }
5d0301b7
NC
3200 GvSTASH(dstr) = GvSTASH(sstr);
3201 if (GvSTASH(dstr))
3202 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3203 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3204 SvFAKE_on(dstr); /* can coerce to non-glob */
3205 }
3206
3207#ifdef GV_UNIQUE_CHECK
3208 if (GvUNIQUE((GV*)dstr)) {
3209 Perl_croak(aTHX_ PL_no_modify);
3210 }
3211#endif
3212
f7877b28
NC
3213 gp_free((GV*)dstr);
3214 SvSCREAM_off(dstr);
5d0301b7 3215 (void)SvOK_off(dstr);
f7877b28 3216 SvSCREAM_on(dstr);
dedf8e73 3217 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3218 GvGP(dstr) = gp_ref(GvGP(sstr));
3219 if (SvTAINTED(sstr))
3220 SvTAINT(dstr);
3221 if (GvIMPORTED(dstr) != GVf_IMPORTED
3222 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3223 {
3224 GvIMPORTED_on(dstr);
3225 }
3226 GvMULTI_on(dstr);
3227 return;
3228}
3229
b8473700 3230static void
2eb42952 3231S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3232 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3233 SV *dref = NULL;
3234 const int intro = GvINTRO(dstr);
2440974c 3235 SV **location;
3386d083 3236 U8 import_flag = 0;
27242d61
NC
3237 const U32 stype = SvTYPE(sref);
3238
b8473700
NC
3239
3240#ifdef GV_UNIQUE_CHECK
3241 if (GvUNIQUE((GV*)dstr)) {
3242 Perl_croak(aTHX_ PL_no_modify);
3243 }
3244#endif
3245
3246 if (intro) {
3247 GvINTRO_off(dstr); /* one-shot flag */
3248 GvLINE(dstr) = CopLINE(PL_curcop);
3249 GvEGV(dstr) = (GV*)dstr;
3250 }
3251 GvMULTI_on(dstr);
27242d61 3252 switch (stype) {
b8473700 3253 case SVt_PVCV:
27242d61
NC
3254 location = (SV **) &GvCV(dstr);
3255 import_flag = GVf_IMPORTED_CV;
3256 goto common;
3257 case SVt_PVHV:
3258 location = (SV **) &GvHV(dstr);
3259 import_flag = GVf_IMPORTED_HV;
3260 goto common;
3261 case SVt_PVAV:
3262 location = (SV **) &GvAV(dstr);
3263 import_flag = GVf_IMPORTED_AV;
3264 goto common;
3265 case SVt_PVIO:
3266 location = (SV **) &GvIOp(dstr);
3267 goto common;
3268 case SVt_PVFM:
3269 location = (SV **) &GvFORM(dstr);
3270 default:
3271 location = &GvSV(dstr);
3272 import_flag = GVf_IMPORTED_SV;
3273 common:
b8473700 3274 if (intro) {
27242d61
NC
3275 if (stype == SVt_PVCV) {
3276 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3277 SvREFCNT_dec(GvCV(dstr));
3278 GvCV(dstr) = NULL;
3279 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3280 PL_sub_generation++;
3281 }
b8473700 3282 }
27242d61 3283 SAVEGENERICSV(*location);
b8473700
NC
3284 }
3285 else
27242d61
NC
3286 dref = *location;
3287 if (stype == SVt_PVCV && *location != sref) {
3288 CV* const cv = (CV*)*location;
b8473700
NC
3289 if (cv) {
3290 if (!GvCVGEN((GV*)dstr) &&
3291 (CvROOT(cv) || CvXSUB(cv)))
3292 {
3293 /* Redefining a sub - warning is mandatory if
3294 it was a const and its value changed. */
3295 if (CvCONST(cv) && CvCONST((CV*)sref)
3296 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3297 NOOP;
b8473700
NC
3298 /* They are 2 constant subroutines generated from
3299 the same constant. This probably means that
3300 they are really the "same" proxy subroutine
3301 instantiated in 2 places. Most likely this is
3302 when a constant is exported twice. Don't warn.
3303 */
3304 }
3305 else if (ckWARN(WARN_REDEFINE)
3306 || (CvCONST(cv)
3307 && (!CvCONST((CV*)sref)
3308 || sv_cmp(cv_const_sv(cv),
3309 cv_const_sv((CV*)sref))))) {
3310 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3311 CvCONST(cv)
3312 ? "Constant subroutine %s::%s redefined"
3313 : "Subroutine %s::%s redefined",
3314 HvNAME_get(GvSTASH((GV*)dstr)),
3315 GvENAME((GV*)dstr));
3316 }
3317 }
3318 if (!intro)
cbf82dd0
NC
3319 cv_ckproto_len(cv, (GV*)dstr,
3320 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3321 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3322 }
b8473700
NC
3323 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3324 GvASSUMECV_on(dstr);
3325 PL_sub_generation++;
3326 }
2440974c 3327 *location = sref;
3386d083
NC
3328 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3329 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3330 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3331 }
3332 break;
3333 }
b37c2d43 3334 SvREFCNT_dec(dref);
b8473700
NC
3335 if (SvTAINTED(sstr))
3336 SvTAINT(dstr);
3337 return;
3338}
3339
8d6d96c1
HS
3340void
3341Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3342{
97aff369 3343 dVAR;
8990e307
LW
3344 register U32 sflags;
3345 register int dtype;
3346 register int stype;
463ee0b2 3347
79072805
LW
3348 if (sstr == dstr)
3349 return;
765f542d 3350 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3351 if (!sstr)
3280af22 3352 sstr = &PL_sv_undef;
8990e307
LW
3353 stype = SvTYPE(sstr);
3354 dtype = SvTYPE(dstr);
79072805 3355
a0d0e21e 3356 SvAMAGIC_off(dstr);
7a5fa8a2 3357 if ( SvVOK(dstr) )
ece467f9
JP
3358 {
3359 /* need to nuke the magic */
3360 mg_free(dstr);
3361 SvRMAGICAL_off(dstr);
3362 }
9e7bc3e8 3363
463ee0b2 3364 /* There's a lot of redundancy below but we're going for speed here */
79072805 3365
8990e307 3366 switch (stype) {
79072805 3367 case SVt_NULL:
aece5585 3368 undef_sstr:
20408e3c
GS
3369 if (dtype != SVt_PVGV) {
3370 (void)SvOK_off(dstr);
3371 return;
3372 }
3373 break;
463ee0b2 3374 case SVt_IV:
aece5585
GA
3375 if (SvIOK(sstr)) {
3376 switch (dtype) {
3377 case SVt_NULL:
8990e307 3378 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3379 break;
3380 case SVt_NV:
aece5585
GA
3381 case SVt_RV:
3382 case SVt_PV:
a0d0e21e 3383 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
3384 break;
3385 }
3386 (void)SvIOK_only(dstr);
45977657 3387 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3388 if (SvIsUV(sstr))
3389 SvIsUV_on(dstr);
37c25af0
NC
3390 /* SvTAINTED can only be true if the SV has taint magic, which in
3391 turn means that the SV type is PVMG (or greater). This is the
3392 case statement for SVt_IV, so this cannot be true (whatever gcov
3393 may say). */
3394 assert(!SvTAINTED(sstr));
aece5585 3395 return;
8990e307 3396 }
aece5585
GA
3397 goto undef_sstr;
3398
463ee0b2 3399 case SVt_NV:
aece5585
GA
3400 if (SvNOK(sstr)) {
3401 switch (dtype) {
3402 case SVt_NULL:
3403 case SVt_IV:
8990e307 3404 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3405 break;
3406 case SVt_RV:
3407 case SVt_PV:
3408 case SVt_PVIV:
a0d0e21e 3409 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
3410 break;
3411 }
9d6ce603 3412 SvNV_set(dstr, SvNVX(sstr));
aece5585 3413 (void)SvNOK_only(dstr);
37c25af0
NC
3414 /* SvTAINTED can only be true if the SV has taint magic, which in
3415 turn means that the SV type is PVMG (or greater). This is the
3416 case statement for SVt_NV, so this cannot be true (whatever gcov
3417 may say). */
3418 assert(!SvTAINTED(sstr));
aece5585 3419 return;
8990e307 3420 }
aece5585
GA
3421 goto undef_sstr;
3422
ed6116ce 3423 case SVt_RV:
8990e307 3424 if (dtype < SVt_RV)
ed6116ce 3425 sv_upgrade(dstr, SVt_RV);
ed6116ce 3426 break;
fc36a67e 3427 case SVt_PVFM:
f8c7b90f 3428#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3429 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3430 if (dtype < SVt_PVIV)
3431 sv_upgrade(dstr, SVt_PVIV);
3432 break;
3433 }
3434 /* Fall through */
3435#endif
3436 case SVt_PV:
8990e307 3437 if (dtype < SVt_PV)
463ee0b2 3438 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3439 break;
3440 case SVt_PVIV:
8990e307 3441 if (dtype < SVt_PVIV)
463ee0b2 3442 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3443 break;
3444 case SVt_PVNV:
8990e307 3445 if (dtype < SVt_PVNV)
463ee0b2 3446 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3447 break;
489f7bfe 3448 default:
a3b680e6
AL
3449 {
3450 const char * const type = sv_reftype(sstr,0);
533c011a 3451 if (PL_op)
a3b680e6 3452 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3453 else
a3b680e6
AL
3454 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3455 }
4633a7c4
LW
3456 break;
3457
79072805 3458 case SVt_PVGV:
8990e307 3459 if (dtype <= SVt_PVGV) {
d4c19fe8 3460 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3461 return;
79072805 3462 }
5f66b61c 3463 /*FALLTHROUGH*/
79072805 3464
489f7bfe
NC
3465 case SVt_PVMG:
3466 case SVt_PVLV:
3467 case SVt_PVBM:
8d6d96c1 3468 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3469 mg_get(sstr);
eb160463 3470 if ((int)SvTYPE(sstr) != stype) {
973f89ab 3471 stype = SvTYPE(sstr);
b8c701c1 3472 if (stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3473 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3474 return;
3475 }
973f89ab
CS
3476 }
3477 }
ded42b9f 3478 if (stype == SVt_PVLV)
862a34c6 3479 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3480 else
862a34c6 3481 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3482 }
3483
ff920335
NC
3484 /* dstr may have been upgraded. */
3485 dtype = SvTYPE(dstr);
8990e307
LW
3486 sflags = SvFLAGS(sstr);
3487
3488 if (sflags & SVf_ROK) {
acaa9288
NC
3489 if (dtype == SVt_PVGV &&
3490 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
3491 sstr = SvRV(sstr);
3492 if (sstr == dstr) {
3493 if (GvIMPORTED(dstr) != GVf_IMPORTED
3494 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3495 {
3496 GvIMPORTED_on(dstr);
3497 }
3498 GvMULTI_on(dstr);
3499 return;
3500 }
d4c19fe8 3501 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3502 return;
3503 }
3504
8990e307 3505 if (dtype >= SVt_PV) {
b8c701c1 3506 if (dtype == SVt_PVGV) {
d4c19fe8 3507 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3508 return;
3509 }
3f7c398e 3510 if (SvPVX_const(dstr)) {
8bd4d4c5 3511 SvPV_free(dstr);
b162af07
SP
3512 SvLEN_set(dstr, 0);
3513 SvCUR_set(dstr, 0);
a0d0e21e 3514 }
8990e307 3515 }
a0d0e21e 3516 (void)SvOK_off(dstr);
b162af07 3517 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
dfd48732
NC
3518 SvFLAGS(dstr) |= sflags & (SVf_ROK|SVf_AMAGIC);
3519 assert(!(sflags & SVp_NOK));
3520 assert(!(sflags & SVp_IOK));
3521 assert(!(sflags & SVf_NOK));
3522 assert(!(sflags & SVf_IOK));
ed6116ce 3523 }
c0c44674
NC
3524 else if (dtype == SVt_PVGV) {
3525 if (!(sflags & SVf_OK)) {
3526 if (ckWARN(WARN_MISC))
3527 Perl_warner(aTHX_ packWARN(WARN_MISC),
3528 "Undefined value assigned to typeglob");
3529 }
3530 else {
3531 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3532 if (dstr != (SV*)gv) {
3533 if (GvGP(dstr))
3534 gp_free((GV*)dstr);
3535 GvGP(dstr) = gp_ref(GvGP(gv));
3536 }
3537 }
3538 }
8990e307 3539 else if (sflags & SVp_POK) {
765f542d 3540 bool isSwipe = 0;
79072805
LW
3541
3542 /*
3543 * Check to see if we can just swipe the string. If so, it's a
3544 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3545 * It might even be a win on short strings if SvPVX_const(dstr)
3546 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3547 */
3548
120fac95
NC
3549 /* Whichever path we take through the next code, we want this true,
3550 and doing it now facilitates the COW check. */
3551 (void)SvPOK_only(dstr);
3552
765f542d 3553 if (
b8f9541a
NC
3554 /* We're not already COW */
3555 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3556#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3557 /* or we are, but dstr isn't a suitable target. */
3558 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3559#endif
3560 )
765f542d 3561 &&
765f542d
NC
3562 !(isSwipe =
3563 (sflags & SVs_TEMP) && /* slated for free anyway? */
3564 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3565 (!(flags & SV_NOSTEAL)) &&
3566 /* and we're allowed to steal temps */
765f542d
NC
3567 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3568 SvLEN(sstr) && /* and really is a string */
645c22ef 3569 /* and won't be needed again, potentially */
765f542d 3570 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3571#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3572 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3573 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3574 && SvTYPE(sstr) >= SVt_PVIV)
3575#endif
3576 ) {
3577 /* Failed the swipe test, and it's not a shared hash key either.
3578 Have to copy the string. */
3579 STRLEN len = SvCUR(sstr);
3580 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3581 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3582 SvCUR_set(dstr, len);
3583 *SvEND(dstr) = '\0';
765f542d 3584 } else {
f8c7b90f 3585 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3586 be true in here. */
765f542d
NC
3587 /* Either it's a shared hash key, or it's suitable for
3588 copy-on-write or we can swipe the string. */
46187eeb 3589 if (DEBUG_C_TEST) {
ed252734 3590 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3591 sv_dump(sstr);
3592 sv_dump(dstr);
46187eeb 3593 }
f8c7b90f 3594#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3595 if (!isSwipe) {
3596 /* I believe I should acquire a global SV mutex if
3597 it's a COW sv (not a shared hash key) to stop
3598 it going un copy-on-write.
3599 If the source SV has gone un copy on write between up there
3600 and down here, then (assert() that) it is of the correct
3601 form to make it copy on write again */
3602 if ((sflags & (SVf_FAKE | SVf_READONLY))
3603 != (SVf_FAKE | SVf_READONLY)) {
3604 SvREADONLY_on(sstr);
3605 SvFAKE_on(sstr);
3606 /* Make the source SV into a loop of 1.
3607 (about to become 2) */
a29f6d03 3608 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3609 }
3610 }
3611#endif
3612 /* Initial code is common. */
94010e71
NC
3613 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3614 SvPV_free(dstr);
79072805 3615 }
765f542d 3616
765f542d
NC
3617 if (!isSwipe) {
3618 /* making another shared SV. */
3619 STRLEN cur = SvCUR(sstr);
3620 STRLEN len = SvLEN(sstr);
f8c7b90f 3621#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3622 if (len) {
b8f9541a 3623 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3624 /* SvIsCOW_normal */
3625 /* splice us in between source and next-after-source. */
a29f6d03
NC
3626 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3627 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3628 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3629 } else
3630#endif
3631 {
765f542d 3632 /* SvIsCOW_shared_hash */
46187eeb
NC
3633 DEBUG_C(PerlIO_printf(Perl_debug_log,
3634 "Copy on write: Sharing hash\n"));
b8f9541a 3635
bdd68bc3 3636 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3637 SvPV_set(dstr,
d1db91c6 3638 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3639 }
87a1ef3d
SP
3640 SvLEN_set(dstr, len);
3641 SvCUR_set(dstr, cur);
765f542d
NC
3642 SvREADONLY_on(dstr);
3643 SvFAKE_on(dstr);
3644 /* Relesase a global SV mutex. */
3645 }
3646 else
765f542d 3647 { /* Passes the swipe test. */
78d1e721 3648 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3649 SvLEN_set(dstr, SvLEN(sstr));
3650 SvCUR_set(dstr, SvCUR(sstr));
3651
3652 SvTEMP_off(dstr);
3653 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3654 SvPV_set(sstr, NULL);
765f542d
NC
3655 SvLEN_set(sstr, 0);
3656 SvCUR_set(sstr, 0);
3657 SvTEMP_off(sstr);
3658 }
3659 }
8990e307 3660 if (sflags & SVp_NOK) {
9d6ce603 3661 SvNV_set(dstr, SvNVX(sstr));
79072805 3662 }
8990e307 3663 if (sflags & SVp_IOK) {
23525414
NC
3664 SvRELEASE_IVX(dstr);
3665 SvIV_set(dstr, SvIVX(sstr));
3666 /* Must do this otherwise some other overloaded use of 0x80000000
3667 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3668 if (sflags & SVf_IVisUV)
25da4f38 3669 SvIsUV_on(dstr);
79072805 3670 }
dd2eae66
NC
3671 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8
3672 |SVf_AMAGIC);
4f2da183
NC
3673 {
3674 const MAGIC * const smg = SvVOK(sstr);
3675 if (smg) {
3676 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3677 smg->mg_ptr, smg->mg_len);
3678 SvRMAGICAL_on(dstr);
3679 }
7a5fa8a2 3680 }
79072805 3681 }
5d581361 3682 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3683 (void)SvOK_off(dstr);
dd2eae66
NC
3684 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK
3685 |SVf_AMAGIC);
5d581361
NC
3686 if (sflags & SVp_IOK) {
3687 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3688 SvIV_set(dstr, SvIVX(sstr));
3689 }
3332b3c1 3690 if (sflags & SVp_NOK) {
9d6ce603 3691 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3692 }
3693 }
79072805 3694 else {
f7877b28 3695 if (isGV_with_GP(sstr)) {
180488f8
NC
3696 /* This stringification rule for globs is spread in 3 places.
3697 This feels bad. FIXME. */
3698 const U32 wasfake = sflags & SVf_FAKE;
3699
3700 /* FAKE globs can get coerced, so need to turn this off
3701 temporarily if it is on. */
3702 SvFAKE_off(sstr);
3703 gv_efullname3(dstr, (GV *)sstr, "*");
3704 SvFLAGS(sstr) |= wasfake;
dd2eae66 3705 SvFLAGS(dstr) |= sflags & SVf_AMAGIC;
180488f8 3706 }
20408e3c
GS
3707 else
3708 (void)SvOK_off(dstr);
a0d0e21e 3709 }
27c9684d
AP
3710 if (SvTAINTED(sstr))
3711 SvTAINT(dstr);
79072805
LW
3712}
3713
954c1994
GS
3714/*
3715=for apidoc sv_setsv_mg
3716
3717Like C<sv_setsv>, but also handles 'set' magic.
3718
3719=cut
3720*/
3721
79072805 3722void
864dbfa3 3723Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3724{
3725 sv_setsv(dstr,sstr);
3726 SvSETMAGIC(dstr);
3727}
3728
f8c7b90f 3729#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3730SV *
3731Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3732{
3733 STRLEN cur = SvCUR(sstr);
3734 STRLEN len = SvLEN(sstr);
3735 register char *new_pv;
3736
3737 if (DEBUG_C_TEST) {
3738 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3739 sstr, dstr);
3740 sv_dump(sstr);
3741 if (dstr)
3742 sv_dump(dstr);
3743 }
3744
3745 if (dstr) {
3746 if (SvTHINKFIRST(dstr))
3747 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3748 else if (SvPVX_const(dstr))
3749 Safefree(SvPVX_const(dstr));
ed252734
NC
3750 }
3751 else
3752 new_SV(dstr);
862a34c6 3753 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3754
3755 assert (SvPOK(sstr));
3756 assert (SvPOKp(sstr));
3757 assert (!SvIOK(sstr));
3758 assert (!SvIOKp(sstr));
3759 assert (!SvNOK(sstr));
3760 assert (!SvNOKp(sstr));
3761
3762 if (SvIsCOW(sstr)) {
3763
3764 if (SvLEN(sstr) == 0) {
3765 /* source is a COW shared hash key. */
ed252734
NC
3766 DEBUG_C(PerlIO_printf(Perl_debug_log,
3767 "Fast copy on write: Sharing hash\n"));
d1db91c6 3768 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3769 goto common_exit;
3770 }
3771 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3772 } else {
3773 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3774 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3775 SvREADONLY_on(sstr);
3776 SvFAKE_on(sstr);
3777 DEBUG_C(PerlIO_printf(Perl_debug_log,
3778 "Fast copy on write: Converting sstr to COW\n"));
3779 SV_COW_NEXT_SV_SET(dstr, sstr);
3780 }
3781 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3782 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3783
3784 common_exit:
3785 SvPV_set(dstr, new_pv);
3786 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3787 if (SvUTF8(sstr))
3788 SvUTF8_on(dstr);
87a1ef3d
SP
3789 SvLEN_set(dstr, len);
3790 SvCUR_set(dstr, cur);
ed252734
NC
3791 if (DEBUG_C_TEST) {
3792 sv_dump(dstr);
3793 }
3794 return dstr;
3795}
3796#endif
3797
954c1994
GS
3798/*
3799=for apidoc sv_setpvn
3800
3801Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3802bytes to be copied. If the C<ptr> argument is NULL the SV will become
3803undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3804
3805=cut
3806*/
3807
ef50df4b 3808void
864dbfa3 3809Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3810{
97aff369 3811 dVAR;
c6f8c383 3812 register char *dptr;
22c522df 3813
765f542d 3814 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3815 if (!ptr) {
a0d0e21e 3816 (void)SvOK_off(sv);
463ee0b2
LW
3817 return;
3818 }
22c522df
JH
3819 else {
3820 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3821 const IV iv = len;
9c5ffd7c
JH
3822 if (iv < 0)
3823 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3824 }
862a34c6 3825 SvUPGRADE(sv, SVt_PV);
c6f8c383 3826
5902b6a9 3827 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3828 Move(ptr,dptr,len,char);
3829 dptr[len] = '\0';
79072805 3830 SvCUR_set(sv, len);
1aa99e6b 3831 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3832 SvTAINT(sv);
79072805
LW
3833}
3834
954c1994
GS
3835/*
3836=for apidoc sv_setpvn_mg
3837
3838Like C<sv_setpvn>, but also handles 'set' magic.
3839
3840=cut
3841*/
3842
79072805 3843void
864dbfa3 3844Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3845{
3846 sv_setpvn(sv,ptr,len);
3847 SvSETMAGIC(sv);
3848}
3849
954c1994
GS
3850/*
3851=for apidoc sv_setpv
3852
3853Copies a string into an SV. The string must be null-terminated. Does not
3854handle 'set' magic. See C<sv_setpv_mg>.
3855
3856=cut
3857*/
3858
ef50df4b 3859void
864dbfa3 3860Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3861{
97aff369 3862 dVAR;
79072805
LW
3863 register STRLEN len;
3864
765f542d 3865 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3866 if (!ptr) {
a0d0e21e 3867 (void)SvOK_off(sv);
463ee0b2
LW
3868 return;
3869 }
79072805 3870 len = strlen(ptr);
862a34c6 3871 SvUPGRADE(sv, SVt_PV);
c6f8c383 3872
79072805 3873 SvGROW(sv, len + 1);
463ee0b2 3874 Move(ptr,SvPVX(sv),len+1,char);
79072805 3875 SvCUR_set(sv, len);
1aa99e6b 3876 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3877 SvTAINT(sv);
3878}
3879
954c1994
GS
3880/*
3881=for apidoc sv_setpv_mg
3882
3883Like C<sv_setpv>, but also handles 'set' magic.
3884
3885=cut
3886*/
3887
463ee0b2 3888void
864dbfa3 3889Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3890{
3891 sv_setpv(sv,ptr);
3892 SvSETMAGIC(sv);
3893}
3894
954c1994 3895/*
47518d95 3896=for apidoc sv_usepvn_flags
954c1994 3897
794a0d33
JH
3898Tells an SV to use C<ptr> to find its string value. Normally the
3899string is stored inside the SV but sv_usepvn allows the SV to use an
3900outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
3901by C<malloc>. The string length, C<len>, must be supplied. By default
3902this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
3903so that pointer should not be freed or used by the programmer after
3904giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
3905that pointer (e.g. ptr + 1) be used.
3906
3907If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3908SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 3909will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 3910C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3911
3912=cut
3913*/
3914
ef50df4b 3915void
47518d95 3916Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3917{
97aff369 3918 dVAR;
1936d2a7 3919 STRLEN allocate;
765f542d 3920 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3921 SvUPGRADE(sv, SVt_PV);
463ee0b2 3922 if (!ptr) {
a0d0e21e 3923 (void)SvOK_off(sv);
47518d95
NC
3924 if (flags & SV_SMAGIC)
3925 SvSETMAGIC(sv);
463ee0b2
LW
3926 return;
3927 }
3f7c398e 3928 if (SvPVX_const(sv))
8bd4d4c5 3929 SvPV_free(sv);
1936d2a7 3930
2e90b4cd
NC
3931 if (flags & SV_HAS_TRAILING_NUL)
3932 assert(ptr[len] == '\0');
3933
c1c21316 3934 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3935 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
cbf82dd0
NC
3936 if (flags & SV_HAS_TRAILING_NUL) {
3937 /* It's long enough - do nothing.
3938 Specfically Perl_newCONSTSUB is relying on this. */
3939 } else {
69d25b4f 3940#ifdef DEBUGGING
69d25b4f
NC
3941 /* Force a move to shake out bugs in callers. */
3942 char *new_ptr = safemalloc(allocate);
3943 Copy(ptr, new_ptr, len, char);
3944 PoisonFree(ptr,len,char);
3945 Safefree(ptr);
3946 ptr = new_ptr;
69d25b4f 3947#else
c1c21316 3948 ptr = saferealloc (ptr, allocate);
69d25b4f 3949#endif
cbf82dd0 3950 }
f880fe2f 3951 SvPV_set(sv, ptr);
463ee0b2 3952 SvCUR_set(sv, len);
1936d2a7 3953 SvLEN_set(sv, allocate);
c1c21316
NC
3954 if (!(flags & SV_HAS_TRAILING_NUL)) {
3955 *SvEND(sv) = '\0';
3956 }
1aa99e6b 3957 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3958 SvTAINT(sv);
47518d95
NC
3959 if (flags & SV_SMAGIC)
3960 SvSETMAGIC(sv);
ef50df4b
GS
3961}
3962
f8c7b90f 3963#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3964/* Need to do this *after* making the SV normal, as we need the buffer
3965 pointer to remain valid until after we've copied it. If we let go too early,
3966 another thread could invalidate it by unsharing last of the same hash key
3967 (which it can do by means other than releasing copy-on-write Svs)
3968 or by changing the other copy-on-write SVs in the loop. */
3969STATIC void
bdd68bc3 3970S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3971{
3972 if (len) { /* this SV was SvIsCOW_normal(sv) */
3973 /* we need to find the SV pointing to us. */
cf5629ad 3974 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3975
765f542d
NC
3976 if (current == sv) {
3977 /* The SV we point to points back to us (there were only two of us
3978 in the loop.)
3979 Hence other SV is no longer copy on write either. */
3980 SvFAKE_off(after);
3981 SvREADONLY_off(after);
3982 } else {
3983 /* We need to follow the pointers around the loop. */
3984 SV *next;
3985 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3986 assert (next);
3987 current = next;
3988 /* don't loop forever if the structure is bust, and we have
3989 a pointer into a closed loop. */
3990 assert (current != after);
3f7c398e 3991 assert (SvPVX_const(current) == pvx);
765f542d
NC
3992 }
3993 /* Make the SV before us point to the SV after us. */
a29f6d03 3994 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3995 }
3996 } else {
bdd68bc3 3997 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3998 }
3999}
4000
4001int
4002Perl_sv_release_IVX(pTHX_ register SV *sv)
4003{
4004 if (SvIsCOW(sv))
4005 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
4006 SvOOK_off(sv);
4007 return 0;
765f542d
NC
4008}
4009#endif
645c22ef
DM
4010/*
4011=for apidoc sv_force_normal_flags
4012
4013Undo various types of fakery on an SV: if the PV is a shared string, make
4014a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4015an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4016we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4017then a copy-on-write scalar drops its PV buffer (if any) and becomes
4018SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4019set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4020C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4021with flags set to 0.
645c22ef
DM
4022
4023=cut
4024*/
4025
6fc92669 4026void
840a7b70 4027Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4028{
97aff369 4029 dVAR;
f8c7b90f 4030#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4031 if (SvREADONLY(sv)) {
4032 /* At this point I believe I should acquire a global SV mutex. */
4033 if (SvFAKE(sv)) {
b64e5050 4034 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4035 const STRLEN len = SvLEN(sv);
4036 const STRLEN cur = SvCUR(sv);
a28509cc 4037 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4038 if (DEBUG_C_TEST) {
4039 PerlIO_printf(Perl_debug_log,
4040 "Copy on write: Force normal %ld\n",
4041 (long) flags);
e419cbc5 4042 sv_dump(sv);
46187eeb 4043 }
765f542d
NC
4044 SvFAKE_off(sv);
4045 SvREADONLY_off(sv);
9f653bb5 4046 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4047 SvPV_set(sv, NULL);
87a1ef3d 4048 SvLEN_set(sv, 0);
765f542d
NC
4049 if (flags & SV_COW_DROP_PV) {
4050 /* OK, so we don't need to copy our buffer. */
4051 SvPOK_off(sv);
4052 } else {
4053 SvGROW(sv, cur + 1);
4054 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4055 SvCUR_set(sv, cur);
765f542d
NC
4056 *SvEND(sv) = '\0';
4057 }
bdd68bc3 4058 sv_release_COW(sv, pvx, len, next);
46187eeb 4059 if (DEBUG_C_TEST) {
e419cbc5 4060 sv_dump(sv);
46187eeb 4061 }
765f542d 4062 }
923e4eb5 4063 else if (IN_PERL_RUNTIME)
765f542d
NC
4064 Perl_croak(aTHX_ PL_no_modify);
4065 /* At this point I believe that I can drop the global SV mutex. */
4066 }
4067#else
2213622d 4068 if (SvREADONLY(sv)) {
1c846c1f 4069 if (SvFAKE(sv)) {
b64e5050 4070 const char * const pvx = SvPVX_const(sv);
66a1b24b 4071 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4072 SvFAKE_off(sv);
4073 SvREADONLY_off(sv);
bd61b366 4074 SvPV_set(sv, NULL);
66a1b24b 4075 SvLEN_set(sv, 0);
1c846c1f 4076 SvGROW(sv, len + 1);
706aa1c9 4077 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4078 *SvEND(sv) = '\0';
bdd68bc3 4079 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4080 }
923e4eb5 4081 else if (IN_PERL_RUNTIME)
cea2e8a9 4082 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4083 }
765f542d 4084#endif
2213622d 4085 if (SvROK(sv))
840a7b70 4086 sv_unref_flags(sv, flags);
6fc92669
GS
4087 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4088 sv_unglob(sv);
0f15f207 4089}
1c846c1f 4090
645c22ef 4091/*
954c1994
GS
4092=for apidoc sv_chop
4093
1c846c1f 4094Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4095SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4096the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4097string. Uses the "OOK hack".
3f7c398e 4098Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4099refer to the same chunk of data.
954c1994
GS
4100
4101=cut
4102*/
4103
79072805 4104void
f54cb97a 4105Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4106{
4107 register STRLEN delta;
a0d0e21e 4108 if (!ptr || !SvPOKp(sv))
79072805 4109 return;
3f7c398e 4110 delta = ptr - SvPVX_const(sv);
2213622d 4111 SV_CHECK_THINKFIRST(sv);
79072805
LW
4112 if (SvTYPE(sv) < SVt_PVIV)
4113 sv_upgrade(sv,SVt_PVIV);
4114
4115 if (!SvOOK(sv)) {
50483b2c 4116 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4117 const char *pvx = SvPVX_const(sv);
a28509cc 4118 const STRLEN len = SvCUR(sv);
50483b2c 4119 SvGROW(sv, len + 1);
706aa1c9 4120 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4121 *SvEND(sv) = '\0';
4122 }
45977657 4123 SvIV_set(sv, 0);
a4bfb290
AB
4124 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4125 and we do that anyway inside the SvNIOK_off
4126 */
7a5fa8a2 4127 SvFLAGS(sv) |= SVf_OOK;
79072805 4128 }
a4bfb290 4129 SvNIOK_off(sv);
b162af07
SP
4130 SvLEN_set(sv, SvLEN(sv) - delta);
4131 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4132 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4133 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4134}
4135
954c1994
GS
4136/*
4137=for apidoc sv_catpvn
4138
4139Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4140C<len> indicates number of bytes to copy. If the SV has the UTF-8
4141status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4142Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4143
8d6d96c1
HS
4144=for apidoc sv_catpvn_flags
4145
4146Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4147C<len> indicates number of bytes to copy. If the SV has the UTF-8
4148status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4149If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4150appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4151in terms of this function.
4152
4153=cut
4154*/
4155
4156void
4157Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4158{
97aff369 4159 dVAR;
8d6d96c1 4160 STRLEN dlen;
fabdb6c0 4161 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4162
8d6d96c1
HS
4163 SvGROW(dsv, dlen + slen + 1);
4164 if (sstr == dstr)
3f7c398e 4165 sstr = SvPVX_const(dsv);
8d6d96c1 4166 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4167 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4168 *SvEND(dsv) = '\0';
4169 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4170 SvTAINT(dsv);
bddd5118
NC
4171 if (flags & SV_SMAGIC)
4172 SvSETMAGIC(dsv);
79072805
LW
4173}
4174
954c1994 4175/*
954c1994
GS
4176=for apidoc sv_catsv
4177
13e8c8e3
JH
4178Concatenates the string from SV C<ssv> onto the end of the string in
4179SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4180not 'set' magic. See C<sv_catsv_mg>.
954c1994 4181
8d6d96c1
HS
4182=for apidoc sv_catsv_flags
4183
4184Concatenates the string from SV C<ssv> onto the end of the string in
4185SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4186bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4187and C<sv_catsv_nomg> are implemented in terms of this function.
4188
4189=cut */
4190
ef50df4b 4191void
8d6d96c1 4192Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4193{
97aff369 4194 dVAR;
bddd5118 4195 if (ssv) {
00b6aa41
AL
4196 STRLEN slen;
4197 const char *spv = SvPV_const(ssv, slen);
4198 if (spv) {
bddd5118
NC
4199 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4200 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4201 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4202 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4203 dsv->sv_flags doesn't have that bit set.
4fd84b44 4204 Andy Dougherty 12 Oct 2001
bddd5118
NC
4205 */
4206 const I32 sutf8 = DO_UTF8(ssv);
4207 I32 dutf8;
13e8c8e3 4208
bddd5118
NC
4209 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4210 mg_get(dsv);
4211 dutf8 = DO_UTF8(dsv);
8d6d96c1 4212
bddd5118
NC
4213 if (dutf8 != sutf8) {
4214 if (dutf8) {
4215 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4216 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4217
bddd5118
NC
4218 sv_utf8_upgrade(csv);
4219 spv = SvPV_const(csv, slen);
4220 }
4221 else
4222 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4223 }
bddd5118 4224 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4225 }
560a288e 4226 }
bddd5118
NC
4227 if (flags & SV_SMAGIC)
4228 SvSETMAGIC(dsv);
79072805
LW
4229}
4230
954c1994 4231/*
954c1994
GS
4232=for apidoc sv_catpv
4233
4234Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4235If the SV has the UTF-8 status set, then the bytes appended should be
4236valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4237
d5ce4a7c 4238=cut */
954c1994 4239
ef50df4b 4240void
0c981600 4241Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4242{
97aff369 4243 dVAR;
79072805 4244 register STRLEN len;
463ee0b2 4245 STRLEN tlen;
748a9306 4246 char *junk;
79072805 4247
0c981600 4248 if (!ptr)
79072805 4249 return;
748a9306 4250 junk = SvPV_force(sv, tlen);
0c981600 4251 len = strlen(ptr);
463ee0b2 4252 SvGROW(sv, tlen + len + 1);
0c981600 4253 if (ptr == junk)
3f7c398e 4254 ptr = SvPVX_const(sv);
0c981600 4255 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4256 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4257 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4258 SvTAINT(sv);
79072805
LW
4259}
4260
954c1994
GS
4261/*
4262=for apidoc sv_catpv_mg
4263
4264Like C<sv_catpv>, but also handles 'set' magic.
4265
4266=cut
4267*/
4268
ef50df4b 4269void
0c981600 4270Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4271{
0c981600 4272 sv_catpv(sv,ptr);
ef50df4b
GS
4273 SvSETMAGIC(sv);
4274}
4275
645c22ef
DM
4276/*
4277=for apidoc newSV
4278
561b68a9
SH
4279Creates a new SV. A non-zero C<len> parameter indicates the number of
4280bytes of preallocated string space the SV should have. An extra byte for a
4281trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4282space is allocated.) The reference count for the new SV is set to 1.
4283
4284In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4285parameter, I<x>, a debug aid which allowed callers to identify themselves.
4286This aid has been superseded by a new build option, PERL_MEM_LOG (see
4287L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4288modules supporting older perls.
645c22ef
DM
4289
4290=cut
4291*/
4292
79072805 4293SV *
864dbfa3 4294Perl_newSV(pTHX_ STRLEN len)
79072805 4295{
97aff369 4296 dVAR;
79072805 4297 register SV *sv;
1c846c1f 4298
4561caa4 4299 new_SV(sv);
79072805
LW
4300 if (len) {
4301 sv_upgrade(sv, SVt_PV);
4302 SvGROW(sv, len + 1);
4303 }
4304 return sv;
4305}
954c1994 4306/*
92110913 4307=for apidoc sv_magicext
954c1994 4308
68795e93 4309Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4310supplied vtable and returns a pointer to the magic added.
92110913 4311
2d8d5d5a
SH
4312Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4313In particular, you can add magic to SvREADONLY SVs, and add more than
4314one instance of the same 'how'.
645c22ef 4315
2d8d5d5a
SH
4316If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4317stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4318special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4319to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4320
2d8d5d5a 4321(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4322
4323=cut
4324*/
92110913 4325MAGIC *
92e67595 4326Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4327 const char* name, I32 namlen)
79072805 4328{
97aff369 4329 dVAR;
79072805 4330 MAGIC* mg;
68795e93 4331
92110913 4332 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4333 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4334 }
a02a5408 4335 Newxz(mg, 1, MAGIC);
79072805 4336 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4337 SvMAGIC_set(sv, mg);
75f9d97a 4338
05f95b08
SB
4339 /* Sometimes a magic contains a reference loop, where the sv and
4340 object refer to each other. To prevent a reference loop that
4341 would prevent such objects being freed, we look for such loops
4342 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4343
4344 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4345 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4346
4347 */
14befaf4
DM
4348 if (!obj || obj == sv ||
4349 how == PERL_MAGIC_arylen ||
4350 how == PERL_MAGIC_qr ||
8d2f4536 4351 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4352 (SvTYPE(obj) == SVt_PVGV &&
4353 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4354 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4355 GvFORM(obj) == (CV*)sv)))
75f9d97a 4356 {
8990e307 4357 mg->mg_obj = obj;
75f9d97a 4358 }
85e6fe83 4359 else {
b37c2d43 4360 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4361 mg->mg_flags |= MGf_REFCOUNTED;
4362 }
b5ccf5f2
YST
4363
4364 /* Normal self-ties simply pass a null object, and instead of
4365 using mg_obj directly, use the SvTIED_obj macro to produce a
4366 new RV as needed. For glob "self-ties", we are tieing the PVIO
4367 with an RV obj pointing to the glob containing the PVIO. In
4368 this case, to avoid a reference loop, we need to weaken the
4369 reference.
4370 */
4371
4372 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4373 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4374 {
4375 sv_rvweaken(obj);
4376 }
4377
79072805 4378 mg->mg_type = how;
565764a8 4379 mg->mg_len = namlen;
9cbac4c7 4380 if (name) {
92110913 4381 if (namlen > 0)
1edc1566 4382 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4383 else if (namlen == HEf_SVKEY)
b37c2d43 4384 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4385 else
92110913 4386 mg->mg_ptr = (char *) name;
9cbac4c7 4387 }
92110913 4388 mg->mg_virtual = vtable;
68795e93 4389
92110913
NIS
4390 mg_magical(sv);
4391 if (SvGMAGICAL(sv))
4392 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4393 return mg;
4394}
4395
4396/*
4397=for apidoc sv_magic
1c846c1f 4398
92110913
NIS
4399Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4400then adds a new magic item of type C<how> to the head of the magic list.
4401
2d8d5d5a
SH
4402See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4403handling of the C<name> and C<namlen> arguments.
4404
4509d3fb
SB
4405You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4406to add more than one instance of the same 'how'.
4407
92110913
NIS
4408=cut
4409*/
4410
4411void
4412Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4413{
97aff369 4414 dVAR;
92e67595 4415 MGVTBL *vtable;
92110913 4416 MAGIC* mg;
92110913 4417
f8c7b90f 4418#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4419 if (SvIsCOW(sv))
4420 sv_force_normal_flags(sv, 0);
4421#endif
92110913 4422 if (SvREADONLY(sv)) {
d8084ca5
DM
4423 if (
4424 /* its okay to attach magic to shared strings; the subsequent
4425 * upgrade to PVMG will unshare the string */
4426 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4427
4428 && IN_PERL_RUNTIME
92110913
NIS
4429 && how != PERL_MAGIC_regex_global
4430 && how != PERL_MAGIC_bm
4431 && how != PERL_MAGIC_fm
4432 && how != PERL_MAGIC_sv
e6469971 4433 && how != PERL_MAGIC_backref
92110913
NIS
4434 )
4435 {
4436 Perl_croak(aTHX_ PL_no_modify);
4437 }
4438 }
4439 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4440 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4441 /* sv_magic() refuses to add a magic of the same 'how' as an
4442 existing one
92110913 4443 */
2a509ed3 4444 if (how == PERL_MAGIC_taint) {
92110913 4445 mg->mg_len |= 1;
2a509ed3
NC
4446 /* Any scalar which already had taint magic on which someone
4447 (erroneously?) did SvIOK_on() or similar will now be
4448 incorrectly sporting public "OK" flags. */
4449 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4450 }
92110913
NIS
4451 return;
4452 }
4453 }
68795e93 4454
79072805 4455 switch (how) {
14befaf4 4456 case PERL_MAGIC_sv:
92110913 4457 vtable = &PL_vtbl_sv;
79072805 4458 break;
14befaf4 4459 case PERL_MAGIC_overload:
92110913 4460 vtable = &PL_vtbl_amagic;
a0d0e21e 4461 break;
14befaf4 4462 case PERL_MAGIC_overload_elem:
92110913 4463 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4464 break;
14befaf4 4465 case PERL_MAGIC_overload_table:
92110913 4466 vtable = &PL_vtbl_ovrld;
a0d0e21e 4467 break;
14befaf4 4468 case PERL_MAGIC_bm:
92110913 4469 vtable = &PL_vtbl_bm;
79072805 4470 break;
14befaf4 4471 case PERL_MAGIC_regdata:
92110913 4472 vtable = &PL_vtbl_regdata;
6cef1e77 4473 break;
14befaf4 4474 case PERL_MAGIC_regdatum:
92110913 4475 vtable = &PL_vtbl_regdatum;
6cef1e77 4476 break;
14befaf4 4477 case PERL_MAGIC_env:
92110913 4478 vtable = &PL_vtbl_env;
79072805 4479 break;
14befaf4 4480 case PERL_MAGIC_fm:
92110913 4481 vtable = &PL_vtbl_fm;
55497cff 4482 break;
14befaf4 4483 case PERL_MAGIC_envelem:
92110913 4484 vtable = &PL_vtbl_envelem;
79072805 4485 break;
14befaf4 4486 case PERL_MAGIC_regex_global:
92110913 4487 vtable = &PL_vtbl_mglob;
93a17b20 4488 break;
14befaf4 4489 case PERL_MAGIC_isa:
92110913 4490 vtable = &PL_vtbl_isa;
463ee0b2 4491 break;
14befaf4 4492 case PERL_MAGIC_isaelem:
92110913 4493 vtable = &PL_vtbl_isaelem;
463ee0b2 4494 break;
14befaf4 4495 case PERL_MAGIC_nkeys:
92110913 4496 vtable = &PL_vtbl_nkeys;
16660edb 4497 break;
14befaf4 4498 case PERL_MAGIC_dbfile:
aec46f14 4499 vtable = NULL;
93a17b20 4500 break;
14befaf4 4501 case PERL_MAGIC_dbline:
92110913 4502 vtable = &PL_vtbl_dbline;
79072805 4503 break;
36477c24 4504#ifdef USE_LOCALE_COLLATE
14befaf4 4505 case PERL_MAGIC_collxfrm:
92110913 4506 vtable = &PL_vtbl_collxfrm;
bbce6d69 4507 break;
36477c24 4508#endif /* USE_LOCALE_COLLATE */
14befaf4 4509 case PERL_MAGIC_tied:
92110913 4510 vtable = &PL_vtbl_pack;
463ee0b2 4511 break;
14befaf4
DM
4512 case PERL_MAGIC_tiedelem:
4513 case PERL_MAGIC_tiedscalar:
92110913 4514 vtable = &PL_vtbl_packelem;
463ee0b2 4515 break;
14befaf4 4516 case PERL_MAGIC_qr:
92110913 4517 vtable = &PL_vtbl_regexp;
c277df42 4518 break;
b3ca2e83
NC
4519 case PERL_MAGIC_hints:
4520 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4521 case PERL_MAGIC_sig:
92110913 4522 vtable = &PL_vtbl_sig;
79072805 4523 break;
14befaf4 4524 case PERL_MAGIC_sigelem:
92110913 4525 vtable = &PL_vtbl_sigelem;
79072805 4526 break;
14befaf4 4527 case PERL_MAGIC_taint:
92110913 4528 vtable = &PL_vtbl_taint;
463ee0b2 4529 break;
14befaf4 4530 case PERL_MAGIC_uvar:
92110913 4531 vtable = &PL_vtbl_uvar;
79072805 4532 break;
14befaf4 4533 case PERL_MAGIC_vec:
92110913 4534 vtable = &PL_vtbl_vec;
79072805 4535 break;
a3874608 4536 case PERL_MAGIC_arylen_p:
bfcb3514 4537 case PERL_MAGIC_rhash:
8d2f4536 4538 case PERL_MAGIC_symtab:
ece467f9 4539 case PERL_MAGIC_vstring:
aec46f14 4540 vtable = NULL;
ece467f9 4541 break;
7e8c5dac
HS
4542 case PERL_MAGIC_utf8:
4543 vtable = &PL_vtbl_utf8;
4544 break;
14befaf4 4545 case PERL_MAGIC_substr:
92110913 4546 vtable = &PL_vtbl_substr;
79072805 4547 break;
14befaf4 4548 case PERL_MAGIC_defelem:
92110913 4549 vtable = &PL_vtbl_defelem;
5f05dabc 4550 break;
14befaf4 4551 case PERL_MAGIC_arylen:
92110913 4552 vtable = &PL_vtbl_arylen;
79072805 4553 break;
14befaf4 4554 case PERL_MAGIC_pos:
92110913 4555 vtable = &PL_vtbl_pos;
a0d0e21e 4556 break;
14befaf4 4557 case PERL_MAGIC_backref:
92110913 4558 vtable = &PL_vtbl_backref;
810b8aa5 4559 break;
b3ca2e83
NC
4560 case PERL_MAGIC_hintselem:
4561 vtable = &PL_vtbl_hintselem;
4562 break;
14befaf4
DM
4563 case PERL_MAGIC_ext:
4564 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4565 /* Useful for attaching extension internal data to perl vars. */
4566 /* Note that multiple extensions may clash if magical scalars */
4567 /* etc holding private data from one are passed to another. */
aec46f14 4568 vtable = NULL;
a0d0e21e 4569 break;
79072805 4570 default:
14befaf4 4571 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4572 }
68795e93 4573
92110913 4574 /* Rest of work is done else where */
aec46f14 4575 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4576
92110913
NIS
4577 switch (how) {
4578 case PERL_MAGIC_taint:
4579 mg->mg_len = 1;
4580 break;
4581 case PERL_MAGIC_ext:
4582 case PERL_MAGIC_dbfile:
4583 SvRMAGICAL_on(sv);
4584 break;
4585 }
463ee0b2
LW
4586}
4587
c461cf8f
JH
4588/*
4589=for apidoc sv_unmagic
4590
645c22ef 4591Removes all magic of type C<type> from an SV.
c461cf8f
JH
4592
4593=cut
4594*/
4595
463ee0b2 4596int
864dbfa3 4597Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4598{
4599 MAGIC* mg;
4600 MAGIC** mgp;
91bba347 4601 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4602 return 0;
064cf529 4603 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4604 for (mg = *mgp; mg; mg = *mgp) {
4605 if (mg->mg_type == type) {
e1ec3a88 4606 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4607 *mgp = mg->mg_moremagic;
1d7c1841 4608 if (vtbl && vtbl->svt_free)
fc0dc3b3 4609 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4610 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4611 if (mg->mg_len > 0)
1edc1566 4612 Safefree(mg->mg_ptr);
565764a8 4613 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4614 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4615 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4616 Safefree(mg->mg_ptr);
9cbac4c7 4617 }
a0d0e21e
LW
4618 if (mg->mg_flags & MGf_REFCOUNTED)
4619 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4620 Safefree(mg);
4621 }
4622 else
4623 mgp = &mg->mg_moremagic;
79072805 4624 }
91bba347 4625 if (!SvMAGIC(sv)) {
463ee0b2 4626 SvMAGICAL_off(sv);
c268c2a6 4627 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4628 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4629 }
4630
4631 return 0;
79072805
LW
4632}
4633
c461cf8f
JH
4634/*
4635=for apidoc sv_rvweaken
4636
645c22ef
DM
4637Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4638referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4639push a back-reference to this RV onto the array of backreferences
4640associated with that magic.
c461cf8f
JH
4641
4642=cut
4643*/
4644
810b8aa5 4645SV *
864dbfa3 4646Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4647{
4648 SV *tsv;
4649 if (!SvOK(sv)) /* let undefs pass */
4650 return sv;
4651 if (!SvROK(sv))
cea2e8a9 4652 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4653 else if (SvWEAKREF(sv)) {
810b8aa5 4654 if (ckWARN(WARN_MISC))
9014280d 4655 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4656 return sv;
4657 }
4658 tsv = SvRV(sv);
e15faf7d 4659 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4660 SvWEAKREF_on(sv);
1c846c1f 4661 SvREFCNT_dec(tsv);
810b8aa5
GS
4662 return sv;
4663}
4664
645c22ef
DM
4665/* Give tsv backref magic if it hasn't already got it, then push a
4666 * back-reference to sv onto the array associated with the backref magic.
4667 */
4668
e15faf7d
NC
4669void
4670Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4671{
97aff369 4672 dVAR;
810b8aa5 4673 AV *av;
86f55936
NC
4674
4675 if (SvTYPE(tsv) == SVt_PVHV) {
4676 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4677
4678 av = *avp;
4679 if (!av) {
4680 /* There is no AV in the offical place - try a fixup. */
4681 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4682
4683 if (mg) {
4684 /* Aha. They've got it stowed in magic. Bring it back. */
4685 av = (AV*)mg->mg_obj;
4686 /* Stop mg_free decreasing the refernce count. */
4687 mg->mg_obj = NULL;
4688 /* Stop mg_free even calling the destructor, given that
4689 there's no AV to free up. */
4690 mg->mg_virtual = 0;
4691 sv_unmagic(tsv, PERL_MAGIC_backref);
4692 } else {
4693 av = newAV();
4694 AvREAL_off(av);
b37c2d43 4695 SvREFCNT_inc_simple_void(av);
86f55936
NC
4696 }
4697 *avp = av;
4698 }
4699 } else {
4700 const MAGIC *const mg
4701 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4702 if (mg)
4703 av = (AV*)mg->mg_obj;
4704 else {
4705 av = newAV();
4706 AvREAL_off(av);
4707 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4708 /* av now has a refcnt of 2, which avoids it getting freed
4709 * before us during global cleanup. The extra ref is removed
4710 * by magic_killbackrefs() when tsv is being freed */
4711 }
810b8aa5 4712 }
d91d49e8 4713 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4714 av_extend(av, AvFILLp(av)+1);
4715 }
4716 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4717}
4718
645c22ef
DM
4719/* delete a back-reference to ourselves from the backref magic associated
4720 * with the SV we point to.
4721 */
4722
1c846c1f 4723STATIC void
e15faf7d 4724S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4725{
97aff369 4726 dVAR;
86f55936 4727 AV *av = NULL;
810b8aa5
GS
4728 SV **svp;
4729 I32 i;
86f55936
NC
4730
4731 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4732 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4733 /* We mustn't attempt to "fix up" the hash here by moving the
4734 backreference array back to the hv_aux structure, as that is stored
4735 in the main HvARRAY(), and hfreentries assumes that no-one
4736 reallocates HvARRAY() while it is running. */
86f55936
NC
4737 }
4738 if (!av) {
4739 const MAGIC *const mg
4740 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4741 if (mg)
4742 av = (AV *)mg->mg_obj;
4743 }
4744 if (!av) {
e15faf7d
NC
4745 if (PL_in_clean_all)
4746 return;
cea2e8a9 4747 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4748 }
4749
4750 if (SvIS_FREED(av))
4751 return;
4752
810b8aa5 4753 svp = AvARRAY(av);
6a76db8b
NC
4754 /* We shouldn't be in here more than once, but for paranoia reasons lets
4755 not assume this. */
4756 for (i = AvFILLp(av); i >= 0; i--) {
4757 if (svp[i] == sv) {
4758 const SSize_t fill = AvFILLp(av);
4759 if (i != fill) {
4760 /* We weren't the last entry.
4761 An unordered list has this property that you can take the
4762 last element off the end to fill the hole, and it's still
4763 an unordered list :-)
4764 */
4765 svp[i] = svp[fill];
4766 }
a0714e2c 4767 svp[fill] = NULL;
6a76db8b
NC
4768 AvFILLp(av) = fill - 1;
4769 }
4770 }
810b8aa5
GS
4771}
4772
86f55936
NC
4773int
4774Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4775{
4776 SV **svp = AvARRAY(av);
4777
4778 PERL_UNUSED_ARG(sv);
4779
4780 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4781 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4782 if (svp && !SvIS_FREED(av)) {
4783 SV *const *const last = svp + AvFILLp(av);
4784
4785 while (svp <= last) {
4786 if (*svp) {
4787 SV *const referrer = *svp;
4788 if (SvWEAKREF(referrer)) {
4789 /* XXX Should we check that it hasn't changed? */
4790 SvRV_set(referrer, 0);
4791 SvOK_off(referrer);
4792 SvWEAKREF_off(referrer);
4793 } else if (SvTYPE(referrer) == SVt_PVGV ||
4794 SvTYPE(referrer) == SVt_PVLV) {
4795 /* You lookin' at me? */
4796 assert(GvSTASH(referrer));
4797 assert(GvSTASH(referrer) == (HV*)sv);
4798 GvSTASH(referrer) = 0;
4799 } else {
4800 Perl_croak(aTHX_
4801 "panic: magic_killbackrefs (flags=%"UVxf")",
4802 (UV)SvFLAGS(referrer));
4803 }
4804
a0714e2c 4805 *svp = NULL;
86f55936
NC
4806 }
4807 svp++;
4808 }
4809 }
4810 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4811 return 0;
4812}
4813
954c1994
GS
4814/*
4815=for apidoc sv_insert
4816
4817Inserts a string at the specified offset/length within the SV. Similar to
4818the Perl substr() function.
4819
4820=cut
4821*/
4822
79072805 4823void
e1ec3a88 4824Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4825{
97aff369 4826 dVAR;
79072805
LW
4827 register char *big;
4828 register char *mid;
4829 register char *midend;
4830 register char *bigend;
4831 register I32 i;
6ff81951 4832 STRLEN curlen;
1c846c1f 4833
79072805 4834
8990e307 4835 if (!bigstr)
cea2e8a9 4836 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4837 SvPV_force(bigstr, curlen);
60fa28ff 4838 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4839 if (offset + len > curlen) {
4840 SvGROW(bigstr, offset+len+1);
93524f2b 4841 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4842 SvCUR_set(bigstr, offset+len);
4843 }
79072805 4844
69b47968 4845 SvTAINT(bigstr);
79072805
LW
4846 i = littlelen - len;
4847 if (i > 0) { /* string might grow */
a0d0e21e 4848 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4849 mid = big + offset + len;
4850 midend = bigend = big + SvCUR(bigstr);
4851 bigend += i;
4852 *bigend = '\0';
4853 while (midend > mid) /* shove everything down */
4854 *--bigend = *--midend;
4855 Move(little,big+offset,littlelen,char);
b162af07 4856 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4857 SvSETMAGIC(bigstr);
4858 return;
4859 }
4860 else if (i == 0) {
463ee0b2 4861 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4862 SvSETMAGIC(bigstr);
4863 return;
4864 }
4865
463ee0b2 4866 big = SvPVX(bigstr);
79072805
LW
4867 mid = big + offset;
4868 midend = mid + len;
4869 bigend = big + SvCUR(bigstr);
4870
4871 if (midend > bigend)
cea2e8a9 4872 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4873
4874 if (mid - big > bigend - midend) { /* faster to shorten from end */
4875 if (littlelen) {
4876 Move(little, mid, littlelen,char);
4877 mid += littlelen;
4878 }
4879 i = bigend - midend;
4880 if (i > 0) {
4881 Move(midend, mid, i,char);
4882 mid += i;
4883 }
4884 *mid = '\0';
4885 SvCUR_set(bigstr, mid - big);
4886 }
155aba94 4887 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4888 midend -= littlelen;
4889 mid = midend;
4890 sv_chop(bigstr,midend-i);
4891 big += i;
4892 while (i--)
4893 *--midend = *--big;
4894 if (littlelen)
4895 Move(little, mid, littlelen,char);
4896 }
4897 else if (littlelen) {
4898 midend -= littlelen;
4899 sv_chop(bigstr,midend);
4900 Move(little,midend,littlelen,char);
4901 }
4902 else {
4903 sv_chop(bigstr,midend);
4904 }
4905 SvSETMAGIC(bigstr);
4906}
4907
c461cf8f
JH
4908/*
4909=for apidoc sv_replace
4910
4911Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4912The target SV physically takes over ownership of the body of the source SV
4913and inherits its flags; however, the target keeps any magic it owns,
4914and any magic in the source is discarded.
ff276b08 4915Note that this is a rather specialist SV copying operation; most of the
645c22ef 4916time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4917
4918=cut
4919*/
79072805
LW
4920
4921void
864dbfa3 4922Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4923{
97aff369 4924 dVAR;
a3b680e6 4925 const U32 refcnt = SvREFCNT(sv);
765f542d 4926 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4927 if (SvREFCNT(nsv) != 1) {
7437becc 4928 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4929 UVuf " != 1)", (UV) SvREFCNT(nsv));
4930 }
93a17b20 4931 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4932 if (SvMAGICAL(nsv))
4933 mg_free(nsv);
4934 else
4935 sv_upgrade(nsv, SVt_PVMG);
b162af07 4936 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4937 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4938 SvMAGICAL_off(sv);
b162af07 4939 SvMAGIC_set(sv, NULL);
93a17b20 4940 }
79072805
LW
4941 SvREFCNT(sv) = 0;
4942 sv_clear(sv);
477f5d66 4943 assert(!SvREFCNT(sv));
fd0854ff
DM
4944#ifdef DEBUG_LEAKING_SCALARS
4945 sv->sv_flags = nsv->sv_flags;
4946 sv->sv_any = nsv->sv_any;
4947 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4948 sv->sv_u = nsv->sv_u;
fd0854ff 4949#else
79072805 4950 StructCopy(nsv,sv,SV);
fd0854ff 4951#endif
7b2c381c
NC
4952 /* Currently could join these into one piece of pointer arithmetic, but
4953 it would be unclear. */
4954 if(SvTYPE(sv) == SVt_IV)
4955 SvANY(sv)
339049b0 4956 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4957 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4958 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4959 }
4960
fd0854ff 4961
f8c7b90f 4962#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4963 if (SvIsCOW_normal(nsv)) {
4964 /* We need to follow the pointers around the loop to make the
4965 previous SV point to sv, rather than nsv. */
4966 SV *next;
4967 SV *current = nsv;
4968 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4969 assert(next);
4970 current = next;
3f7c398e 4971 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4972 }
4973 /* Make the SV before us point to the SV after us. */
4974 if (DEBUG_C_TEST) {
4975 PerlIO_printf(Perl_debug_log, "previous is\n");
4976 sv_dump(current);
a29f6d03
NC
4977 PerlIO_printf(Perl_debug_log,
4978 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4979 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4980 }
a29f6d03 4981 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4982 }
4983#endif
79072805 4984 SvREFCNT(sv) = refcnt;
1edc1566 4985 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4986 SvREFCNT(nsv) = 0;
463ee0b2 4987 del_SV(nsv);
79072805
LW
4988}
4989
c461cf8f
JH
4990/*
4991=for apidoc sv_clear
4992
645c22ef
DM
4993Clear an SV: call any destructors, free up any memory used by the body,
4994and free the body itself. The SV's head is I<not> freed, although
4995its type is set to all 1's so that it won't inadvertently be assumed
4996to be live during global destruction etc.
4997This function should only be called when REFCNT is zero. Most of the time
4998you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4999instead.
c461cf8f
JH
5000
5001=cut
5002*/
5003
79072805 5004void
864dbfa3 5005Perl_sv_clear(pTHX_ register SV *sv)
79072805 5006{
27da23d5 5007 dVAR;
82bb6deb 5008 const U32 type = SvTYPE(sv);
8edfc514
NC
5009 const struct body_details *const sv_type_details
5010 = bodies_by_type + type;
82bb6deb 5011
79072805
LW
5012 assert(sv);
5013 assert(SvREFCNT(sv) == 0);
5014
d2a0f284
JC
5015 if (type <= SVt_IV) {
5016 /* See the comment in sv.h about the collusion between this early
5017 return and the overloading of the NULL and IV slots in the size
5018 table. */
82bb6deb 5019 return;
d2a0f284 5020 }
82bb6deb 5021
ed6116ce 5022 if (SvOBJECT(sv)) {
3280af22 5023 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5024 dSP;
893645bd 5025 HV* stash;
d460ef45 5026 do {
b464bac0 5027 CV* destructor;
4e8e7886 5028 stash = SvSTASH(sv);
32251b26 5029 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5030 if (destructor) {
1b6737cc 5031 SV* const tmpref = newRV(sv);
5cc433a6 5032 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5033 ENTER;
e788e7d3 5034 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5035 EXTEND(SP, 2);
5036 PUSHMARK(SP);
5cc433a6 5037 PUSHs(tmpref);
4e8e7886 5038 PUTBACK;
44389ee9 5039 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5040
5041
d3acc0f7 5042 POPSTACK;
3095d977 5043 SPAGAIN;
4e8e7886 5044 LEAVE;
5cc433a6
AB
5045 if(SvREFCNT(tmpref) < 2) {
5046 /* tmpref is not kept alive! */
5047 SvREFCNT(sv)--;
b162af07 5048 SvRV_set(tmpref, NULL);
5cc433a6
AB
5049 SvROK_off(tmpref);
5050 }
5051 SvREFCNT_dec(tmpref);
4e8e7886
GS
5052 }
5053 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5054
6f44e0a4
JP
5055
5056 if (SvREFCNT(sv)) {
5057 if (PL_in_clean_objs)
cea2e8a9 5058 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5059 HvNAME_get(stash));
6f44e0a4
JP
5060 /* DESTROY gave object new lease on life */
5061 return;
5062 }
a0d0e21e 5063 }
4e8e7886 5064
a0d0e21e 5065 if (SvOBJECT(sv)) {
4e8e7886 5066 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5067 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5068 if (type != SVt_PVIO)
3280af22 5069 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5070 }
463ee0b2 5071 }
82bb6deb 5072 if (type >= SVt_PVMG) {
e736a858
NC
5073 HV *ourstash;
5074 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5075 (ourstash = OURSTASH(sv))) {
5076 SvREFCNT_dec(ourstash);
5077 } else if (SvMAGIC(sv))
524189f1 5078 mg_free(sv);
00b1698f 5079 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5080 SvREFCNT_dec(SvSTASH(sv));
5081 }
82bb6deb 5082 switch (type) {
8990e307 5083 case SVt_PVIO:
df0bd2f4
GS
5084 if (IoIFP(sv) &&
5085 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5086 IoIFP(sv) != PerlIO_stdout() &&
5087 IoIFP(sv) != PerlIO_stderr())
93578b34 5088 {
f2b5be74 5089 io_close((IO*)sv, FALSE);
93578b34 5090 }
1d7c1841 5091 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5092 PerlDir_close(IoDIRP(sv));
1d7c1841 5093 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5094 Safefree(IoTOP_NAME(sv));
5095 Safefree(IoFMT_NAME(sv));
5096 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5097 goto freescalar;
79072805 5098 case SVt_PVBM:
a0d0e21e 5099 goto freescalar;
79072805 5100 case SVt_PVCV:
748a9306 5101 case SVt_PVFM:
85e6fe83 5102 cv_undef((CV*)sv);
a0d0e21e 5103 goto freescalar;
79072805 5104 case SVt_PVHV:
86f55936 5105 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5106 hv_undef((HV*)sv);
a0d0e21e 5107 break;
79072805 5108 case SVt_PVAV:
85e6fe83 5109 av_undef((AV*)sv);
a0d0e21e 5110 break;
02270b4e 5111 case SVt_PVLV:
dd28f7bb
DM
5112 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5113 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5114 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5115 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5116 }
5117 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5118 SvREFCNT_dec(LvTARG(sv));
02270b4e 5119 goto freescalar;
a0d0e21e 5120 case SVt_PVGV:
1edc1566 5121 gp_free((GV*)sv);
acda4c6a
NC
5122 if (GvNAME_HEK(sv)) {
5123 unshare_hek(GvNAME_HEK(sv));
5124 }
893645bd
NC
5125 /* If we're in a stash, we don't own a reference to it. However it does
5126 have a back reference to us, which needs to be cleared. */
5127 if (GvSTASH(sv))
5128 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5129 case SVt_PVMG:
79072805
LW
5130 case SVt_PVNV:
5131 case SVt_PVIV:
a0d0e21e 5132 freescalar:
5228ca4e
NC
5133 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5134 if (SvOOK(sv)) {
93524f2b 5135 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5136 /* Don't even bother with turning off the OOK flag. */
5137 }
79072805 5138 case SVt_PV:
a0d0e21e 5139 case SVt_RV:
810b8aa5 5140 if (SvROK(sv)) {
b37c2d43 5141 SV * const target = SvRV(sv);
810b8aa5 5142 if (SvWEAKREF(sv))
e15faf7d 5143 sv_del_backref(target, sv);
810b8aa5 5144 else
e15faf7d 5145 SvREFCNT_dec(target);
810b8aa5 5146 }
f8c7b90f 5147#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5148 else if (SvPVX_const(sv)) {
765f542d
NC
5149 if (SvIsCOW(sv)) {
5150 /* I believe I need to grab the global SV mutex here and
5151 then recheck the COW status. */
46187eeb
NC
5152 if (DEBUG_C_TEST) {
5153 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5154 sv_dump(sv);
46187eeb 5155 }
bdd68bc3
NC
5156 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5157 SV_COW_NEXT_SV(sv));
765f542d
NC
5158 /* And drop it here. */
5159 SvFAKE_off(sv);
5160 } else if (SvLEN(sv)) {
3f7c398e 5161 Safefree(SvPVX_const(sv));
765f542d
NC
5162 }
5163 }
5164#else
3f7c398e 5165 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5166 Safefree(SvPVX_mutable(sv));
3f7c398e 5167 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5168 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5169 SvFAKE_off(sv);
5170 }
765f542d 5171#endif
79072805
LW
5172 break;
5173 case SVt_NV:
79072805
LW
5174 break;
5175 }
5176
893645bd
NC
5177 SvFLAGS(sv) &= SVf_BREAK;
5178 SvFLAGS(sv) |= SVTYPEMASK;
5179
8edfc514 5180 if (sv_type_details->arena) {
b9502f15 5181 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5182 &PL_body_roots[type]);
5183 }
d2a0f284 5184 else if (sv_type_details->body_size) {
8edfc514
NC
5185 my_safefree(SvANY(sv));
5186 }
79072805
LW
5187}
5188
645c22ef
DM
5189/*
5190=for apidoc sv_newref
5191
5192Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5193instead.
5194
5195=cut
5196*/
5197
79072805 5198SV *
864dbfa3 5199Perl_sv_newref(pTHX_ SV *sv)
79072805 5200{
96a5add6 5201 PERL_UNUSED_CONTEXT;
463ee0b2 5202 if (sv)
4db098f4 5203 (SvREFCNT(sv))++;
79072805
LW
5204 return sv;
5205}
5206
c461cf8f
JH
5207/*
5208=for apidoc sv_free
5209
645c22ef
DM
5210Decrement an SV's reference count, and if it drops to zero, call
5211C<sv_clear> to invoke destructors and free up any memory used by
5212the body; finally, deallocate the SV's head itself.
5213Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5214
5215=cut
5216*/
5217
79072805 5218void
864dbfa3 5219Perl_sv_free(pTHX_ SV *sv)
79072805 5220{
27da23d5 5221 dVAR;
79072805
LW
5222 if (!sv)
5223 return;
a0d0e21e
LW
5224 if (SvREFCNT(sv) == 0) {
5225 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5226 /* this SV's refcnt has been artificially decremented to
5227 * trigger cleanup */
a0d0e21e 5228 return;
3280af22 5229 if (PL_in_clean_all) /* All is fair */
1edc1566 5230 return;
d689ffdd
JP
5231 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5232 /* make sure SvREFCNT(sv)==0 happens very seldom */
5233 SvREFCNT(sv) = (~(U32)0)/2;
5234 return;
5235 }
41e4abd8 5236 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5237 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5238 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5239 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5240#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5241 Perl_dump_sv_child(aTHX_ sv);
5242#endif
5243 }
79072805
LW
5244 return;
5245 }
4db098f4 5246 if (--(SvREFCNT(sv)) > 0)
8990e307 5247 return;
8c4d3c90
NC
5248 Perl_sv_free2(aTHX_ sv);
5249}
5250
5251void
5252Perl_sv_free2(pTHX_ SV *sv)
5253{
27da23d5 5254 dVAR;
463ee0b2
LW
5255#ifdef DEBUGGING
5256 if (SvTEMP(sv)) {
0453d815 5257 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5258 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5259 "Attempt to free temp prematurely: SV 0x%"UVxf
5260 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5261 return;
79072805 5262 }
463ee0b2 5263#endif
d689ffdd
JP
5264 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5265 /* make sure SvREFCNT(sv)==0 happens very seldom */
5266 SvREFCNT(sv) = (~(U32)0)/2;
5267 return;
5268 }
79072805 5269 sv_clear(sv);
477f5d66
CS
5270 if (! SvREFCNT(sv))
5271 del_SV(sv);
79072805
LW
5272}
5273
954c1994
GS
5274/*
5275=for apidoc sv_len
5276
645c22ef
DM
5277Returns the length of the string in the SV. Handles magic and type
5278coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5279
5280=cut
5281*/
5282
79072805 5283STRLEN
864dbfa3 5284Perl_sv_len(pTHX_ register SV *sv)
79072805 5285{
463ee0b2 5286 STRLEN len;
79072805
LW
5287
5288 if (!sv)
5289 return 0;
5290
8990e307 5291 if (SvGMAGICAL(sv))
565764a8 5292 len = mg_length(sv);
8990e307 5293 else
4d84ee25 5294 (void)SvPV_const(sv, len);
463ee0b2 5295 return len;
79072805
LW
5296}
5297
c461cf8f
JH
5298/*
5299=for apidoc sv_len_utf8
5300
5301Returns the number of characters in the string in an SV, counting wide
1e54db1a 5302UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5303
5304=cut
5305*/
5306
7e8c5dac
HS
5307/*
5308 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5309 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5310 * (Note that the mg_len is not the length of the mg_ptr field.
5311 * This allows the cache to store the character length of the string without
5312 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5313 *
7e8c5dac
HS
5314 */
5315
a0ed51b3 5316STRLEN
864dbfa3 5317Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5318{
a0ed51b3
LW
5319 if (!sv)
5320 return 0;
5321
a0ed51b3 5322 if (SvGMAGICAL(sv))
b76347f2 5323 return mg_length(sv);
a0ed51b3 5324 else
b76347f2 5325 {
26346457 5326 STRLEN len;
e62f0680 5327 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5328
26346457
NC
5329 if (PL_utf8cache) {
5330 STRLEN ulen;
5331 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5332
5333 if (mg && mg->mg_len != -1) {
5334 ulen = mg->mg_len;
5335 if (PL_utf8cache < 0) {
5336 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5337 if (real != ulen) {
5338 /* Need to turn the assertions off otherwise we may
5339 recurse infinitely while printing error messages.
5340 */
5341 SAVEI8(PL_utf8cache);
5342 PL_utf8cache = 0;
5343 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5344 " real %"UVf" for %"SVf,
26346457
NC
5345 (UV) ulen, (UV) real, sv);
5346 }
5347 }
5348 }
5349 else {
5350 ulen = Perl_utf8_length(aTHX_ s, s + len);
5351 if (!SvREADONLY(sv)) {
5352 if (!mg) {
5353 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5354 &PL_vtbl_utf8, 0, 0);
5355 }
cb9e20bb 5356 assert(mg);
26346457 5357 mg->mg_len = ulen;
cb9e20bb 5358 }
cb9e20bb 5359 }
26346457 5360 return ulen;
7e8c5dac 5361 }
26346457 5362 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5363 }
5364}
5365
9564a3bd
NC
5366/* Walk forwards to find the byte corresponding to the passed in UTF-8
5367 offset. */
bdf30dd6
NC
5368static STRLEN
5369S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send,
5370 STRLEN uoffset)
5371{
5372 const U8 *s = start;
5373
d4c19fe8
AL
5374 PERL_UNUSED_CONTEXT;
5375
bdf30dd6
NC
5376 while (s < send && uoffset--)
5377 s += UTF8SKIP(s);
5378 if (s > send) {
5379 /* This is the existing behaviour. Possibly it should be a croak, as
5380 it's actually a bounds error */
5381 s = send;
5382 }
5383 return s - start;
5384}
5385
9564a3bd
NC
5386/* Given the length of the string in both bytes and UTF-8 characters, decide
5387 whether to walk forwards or backwards to find the byte corresponding to
5388 the passed in UTF-8 offset. */
c336ad0b
NC
5389static STRLEN
5390S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send,
5391 STRLEN uoffset, STRLEN uend)
5392{
5393 STRLEN backw = uend - uoffset;
5394 if (uoffset < 2 * backw) {
25a8a4ef 5395 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5396 forward (that's where the 2 * backw comes from).
5397 (The real figure of course depends on the UTF-8 data.) */
5398 return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset);
5399 }
5400
5401 while (backw--) {
5402 send--;
5403 while (UTF8_IS_CONTINUATION(*send))
5404 send--;
5405 }
5406 return send - start;
5407}
5408
9564a3bd
NC
5409/* For the string representation of the given scalar, find the byte
5410 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5411 give another position in the string, *before* the sought offset, which
5412 (which is always true, as 0, 0 is a valid pair of positions), which should
5413 help reduce the amount of linear searching.
5414 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5415 will be used to reduce the amount of linear searching. The cache will be
5416 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5417static STRLEN
5418S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5419 const U8 *const send, STRLEN uoffset,
5420 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5421 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5422 bool found = FALSE;
5423
75c33c12
NC
5424 assert (uoffset >= uoffset0);
5425
c336ad0b 5426 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5427 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5428 if ((*mgp)->mg_ptr) {
5429 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5430 if (cache[0] == uoffset) {
5431 /* An exact match. */
5432 return cache[1];
5433 }
ab455f60
NC
5434 if (cache[2] == uoffset) {
5435 /* An exact match. */
5436 return cache[3];
5437 }
668af93f
NC
5438
5439 if (cache[0] < uoffset) {
d8b2e1f9
NC
5440 /* The cache already knows part of the way. */
5441 if (cache[0] > uoffset0) {
5442 /* The cache knows more than the passed in pair */
5443 uoffset0 = cache[0];
5444 boffset0 = cache[1];
5445 }
5446 if ((*mgp)->mg_len != -1) {
5447 /* And we know the end too. */
5448 boffset = boffset0
5449 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5450 uoffset - uoffset0,
5451 (*mgp)->mg_len - uoffset0);
5452 } else {
5453 boffset = boffset0
5454 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5455 send, uoffset - uoffset0);
5456 }
dd7c5fd3
NC
5457 }
5458 else if (cache[2] < uoffset) {
5459 /* We're between the two cache entries. */
5460 if (cache[2] > uoffset0) {
5461 /* and the cache knows more than the passed in pair */
5462 uoffset0 = cache[2];
5463 boffset0 = cache[3];
5464 }
5465
668af93f
NC
5466 boffset = boffset0
5467 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5468 start + cache[1],
5469 uoffset - uoffset0,
5470 cache[0] - uoffset0);
dd7c5fd3
NC
5471 } else {
5472 boffset = boffset0
5473 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5474 start + cache[3],
5475 uoffset - uoffset0,
5476 cache[2] - uoffset0);
d8b2e1f9 5477 }
668af93f 5478 found = TRUE;
d8b2e1f9
NC
5479 }
5480 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5481 /* If we can take advantage of a passed in offset, do so. */
5482 /* In fact, offset0 is either 0, or less than offset, so don't
5483 need to worry about the other possibility. */
5484 boffset = boffset0
5485 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5486 uoffset - uoffset0,
5487 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5488 found = TRUE;
5489 }
28ccbf94 5490 }
c336ad0b
NC
5491
5492 if (!found || PL_utf8cache < 0) {
75c33c12
NC
5493 const STRLEN real_boffset
5494 = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5495 send, uoffset - uoffset0);
5496
c336ad0b
NC
5497 if (found && PL_utf8cache < 0) {
5498 if (real_boffset != boffset) {
5499 /* Need to turn the assertions off otherwise we may recurse
5500 infinitely while printing error messages. */
5501 SAVEI8(PL_utf8cache);
5502 PL_utf8cache = 0;
5503 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5504 " real %"UVf" for %"SVf,
5505 (UV) boffset, (UV) real_boffset, sv);
5506 }
5507 }
5508 boffset = real_boffset;
28ccbf94 5509 }
0905937d 5510
ab455f60 5511 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5512 return boffset;
5513}
5514
9564a3bd
NC
5515
5516/*
5517=for apidoc sv_pos_u2b
5518
5519Converts the value pointed to by offsetp from a count of UTF-8 chars from
5520the start of the string, to a count of the equivalent number of bytes; if
5521lenp is non-zero, it does the same to lenp, but this time starting from
5522the offset, rather than from the start of the string. Handles magic and
5523type coercion.
5524
5525=cut
5526*/
5527
5528/*
5529 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5530 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5531 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5532 *
5533 */
5534
a0ed51b3 5535void
864dbfa3 5536Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5537{
245d4a47 5538 const U8 *start;
a0ed51b3
LW
5539 STRLEN len;
5540
5541 if (!sv)
5542 return;
5543
245d4a47 5544 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5545 if (len) {
bdf30dd6
NC
5546 STRLEN uoffset = (STRLEN) *offsetp;
5547 const U8 * const send = start + len;
0905937d 5548 MAGIC *mg = NULL;
28ccbf94
NC
5549 STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send,
5550 uoffset, 0, 0);
bdf30dd6
NC
5551
5552 *offsetp = (I32) boffset;
5553
5554 if (lenp) {
28ccbf94
NC
5555 /* Convert the relative offset to absolute. */
5556 STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5557 STRLEN boffset2
5558 = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2,
5559 uoffset, boffset) - boffset;
bdf30dd6 5560
28ccbf94 5561 *lenp = boffset2;
bdf30dd6 5562 }
7e8c5dac
HS
5563 }
5564 else {
5565 *offsetp = 0;
5566 if (lenp)
5567 *lenp = 0;
a0ed51b3 5568 }
e23c8137 5569
a0ed51b3
LW
5570 return;
5571}
5572
9564a3bd
NC
5573/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5574 byte length pairing. The (byte) length of the total SV is passed in too,
5575 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5576 may not have updated SvCUR, so we can't rely on reading it directly.
5577
5578 The proffered utf8/byte length pairing isn't used if the cache already has
5579 two pairs, and swapping either for the proffered pair would increase the
5580 RMS of the intervals between known byte offsets.
5581
5582 The cache itself consists of 4 STRLEN values
5583 0: larger UTF-8 offset
5584 1: corresponding byte offset
5585 2: smaller UTF-8 offset
5586 3: corresponding byte offset
5587
5588 Unused cache pairs have the value 0, 0.
5589 Keeping the cache "backwards" means that the invariant of
5590 cache[0] >= cache[2] is maintained even with empty slots, which means that
5591 the code that uses it doesn't need to worry if only 1 entry has actually
5592 been set to non-zero. It also makes the "position beyond the end of the
5593 cache" logic much simpler, as the first slot is always the one to start
5594 from.
645c22ef 5595*/
ec07b5e0 5596static void
ab455f60
NC
5597S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5598 STRLEN blen)
ec07b5e0
NC
5599{
5600 STRLEN *cache;
5601 if (SvREADONLY(sv))
5602 return;
5603
5604 if (!*mgp) {
5605 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5606 0);
5607 (*mgp)->mg_len = -1;
5608 }
5609 assert(*mgp);
5610
5611 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5612 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5613 (*mgp)->mg_ptr = (char *) cache;
5614 }
5615 assert(cache);
5616
5617 if (PL_utf8cache < 0) {
ef816a78 5618 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5619 const U8 *const end = start + byte;
5620 STRLEN realutf8 = 0;
5621
5622 while (start < end) {
5623 start += UTF8SKIP(start);
5624 realutf8++;
5625 }
5626
5627 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5628 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5629 doesn't? I don't know whether this difference was introduced with
5630 the caching code in 5.8.1. */
ec07b5e0
NC
5631
5632 if (realutf8 != utf8) {
5633 /* Need to turn the assertions off otherwise we may recurse
5634 infinitely while printing error messages. */
5635 SAVEI8(PL_utf8cache);
5636 PL_utf8cache = 0;
5637 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5638 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5639 }
5640 }
ab455f60
NC
5641
5642 /* Cache is held with the later position first, to simplify the code
5643 that deals with unbounded ends. */
5644
5645 ASSERT_UTF8_CACHE(cache);
5646 if (cache[1] == 0) {
5647 /* Cache is totally empty */
5648 cache[0] = utf8;
5649 cache[1] = byte;
5650 } else if (cache[3] == 0) {
5651 if (byte > cache[1]) {
5652 /* New one is larger, so goes first. */
5653 cache[2] = cache[0];
5654 cache[3] = cache[1];
5655 cache[0] = utf8;
5656 cache[1] = byte;
5657 } else {
5658 cache[2] = utf8;
5659 cache[3] = byte;
5660 }
5661 } else {
5662#define THREEWAY_SQUARE(a,b,c,d) \
5663 ((float)((d) - (c))) * ((float)((d) - (c))) \
5664 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5665 + ((float)((b) - (a))) * ((float)((b) - (a)))
5666
5667 /* Cache has 2 slots in use, and we know three potential pairs.
5668 Keep the two that give the lowest RMS distance. Do the
5669 calcualation in bytes simply because we always know the byte
5670 length. squareroot has the same ordering as the positive value,
5671 so don't bother with the actual square root. */
5672 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5673 if (byte > cache[1]) {
5674 /* New position is after the existing pair of pairs. */
5675 const float keep_earlier
5676 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5677 const float keep_later
5678 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5679
5680 if (keep_later < keep_earlier) {
5681 if (keep_later < existing) {
5682 cache[2] = cache[0];
5683 cache[3] = cache[1];
5684 cache[0] = utf8;
5685 cache[1] = byte;
5686 }
5687 }
5688 else {
5689 if (keep_earlier < existing) {
5690 cache[0] = utf8;
5691 cache[1] = byte;
5692 }
5693 }
5694 }
57d7fbf1
NC
5695 else if (byte > cache[3]) {
5696 /* New position is between the existing pair of pairs. */
5697 const float keep_earlier
5698 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5699 const float keep_later
5700 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5701
5702 if (keep_later < keep_earlier) {
5703 if (keep_later < existing) {
5704 cache[2] = utf8;
5705 cache[3] = byte;
5706 }
5707 }
5708 else {
5709 if (keep_earlier < existing) {
5710 cache[0] = utf8;
5711 cache[1] = byte;
5712 }
5713 }
5714 }
5715 else {
5716 /* New position is before the existing pair of pairs. */
5717 const float keep_earlier
5718 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5719 const float keep_later
5720 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5721
5722 if (keep_later < keep_earlier) {
5723 if (keep_later < existing) {
5724 cache[2] = utf8;
5725 cache[3] = byte;
5726 }
5727 }
5728 else {
5729 if (keep_earlier < existing) {
5730 cache[0] = cache[2];
5731 cache[1] = cache[3];
5732 cache[2] = utf8;
5733 cache[3] = byte;
5734 }
5735 }
5736 }
ab455f60 5737 }
0905937d 5738 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5739}
5740
5741/* If we don't know the character offset of the end of a region, our only
5742 option is to walk forwards to the target byte offset. */
5743static STRLEN
5744S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5745{
5746 STRLEN len = 0;
5747 while (s < target) {
5748 STRLEN n = 1;
5749
5750 /* Call utf8n_to_uvchr() to validate the sequence
5751 * (unless a simple non-UTF character) */
5752 if (!UTF8_IS_INVARIANT(*s))
5753 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5754 if (n > 0) {
5755 s += n;
5756 len++;
5757 }
5758 else
5759 break;
5760 }
5761 return len;
5762}
5763
5764/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5765 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5766 backward is half the speed of walking forward. */
ec07b5e0
NC
5767static STRLEN
5768S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5769 STRLEN endu)
5770{
5771 const STRLEN forw = target - s;
5772 STRLEN backw = end - target;
5773
5774 if (forw < 2 * backw) {
5775 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5776 }
5777
5778 while (end > target) {
5779 end--;
5780 while (UTF8_IS_CONTINUATION(*end)) {
5781 end--;
5782 }
5783 endu--;
5784 }
5785 return endu;
5786}
5787
9564a3bd
NC
5788/*
5789=for apidoc sv_pos_b2u
5790
5791Converts the value pointed to by offsetp from a count of bytes from the
5792start of the string, to a count of the equivalent number of UTF-8 chars.
5793Handles magic and type coercion.
5794
5795=cut
5796*/
5797
5798/*
5799 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5800 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5801 * byte offsets.
5802 *
5803 */
a0ed51b3 5804void
7e8c5dac 5805Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5806{
83003860 5807 const U8* s;
ec07b5e0 5808 const STRLEN byte = *offsetp;
7087a21c 5809 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5810 STRLEN blen;
ec07b5e0
NC
5811 MAGIC* mg = NULL;
5812 const U8* send;
a922f900 5813 bool found = FALSE;
a0ed51b3
LW
5814
5815 if (!sv)
5816 return;
5817
ab455f60 5818 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5819
ab455f60 5820 if (blen < byte)
ec07b5e0 5821 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5822
ec07b5e0 5823 send = s + byte;
a67d7df9 5824
ffca234a
NC
5825 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5826 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5827 if (mg->mg_ptr) {
d4c19fe8 5828 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5829 if (cache[1] == byte) {
ec07b5e0
NC
5830 /* An exact match. */
5831 *offsetp = cache[0];
ec07b5e0 5832 return;
7e8c5dac 5833 }
ab455f60
NC
5834 if (cache[3] == byte) {
5835 /* An exact match. */
5836 *offsetp = cache[2];
5837 return;
5838 }
668af93f
NC
5839
5840 if (cache[1] < byte) {
ec07b5e0 5841 /* We already know part of the way. */
b9f984a5
NC
5842 if (mg->mg_len != -1) {
5843 /* Actually, we know the end too. */
5844 len = cache[0]
5845 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5846 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5847 } else {
5848 len = cache[0]
5849 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5850 }
7e8c5dac 5851 }
9f985e4c
NC
5852 else if (cache[3] < byte) {
5853 /* We're between the two cached pairs, so we do the calculation
5854 offset by the byte/utf-8 positions for the earlier pair,
5855 then add the utf-8 characters from the string start to
5856 there. */
5857 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5858 s + cache[1], cache[0] - cache[2])
5859 + cache[2];
5860
5861 }
5862 else { /* cache[3] > byte */
5863 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5864 cache[2]);
7e8c5dac 5865
7e8c5dac 5866 }
ec07b5e0 5867 ASSERT_UTF8_CACHE(cache);
a922f900 5868 found = TRUE;
ffca234a 5869 } else if (mg->mg_len != -1) {
ab455f60 5870 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5871 found = TRUE;
7e8c5dac 5872 }
a0ed51b3 5873 }
a922f900
NC
5874 if (!found || PL_utf8cache < 0) {
5875 const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5876
5877 if (found && PL_utf8cache < 0) {
5878 if (len != real_len) {
5879 /* Need to turn the assertions off otherwise we may recurse
5880 infinitely while printing error messages. */
5881 SAVEI8(PL_utf8cache);
5882 PL_utf8cache = 0;
5883 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5884 " real %"UVf" for %"SVf,
5885 (UV) len, (UV) real_len, sv);
5886 }
5887 }
5888 len = real_len;
ec07b5e0
NC
5889 }
5890 *offsetp = len;
5891
ab455f60 5892 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5893}
5894
954c1994
GS
5895/*
5896=for apidoc sv_eq
5897
5898Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5899identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5900coerce its args to strings if necessary.
954c1994
GS
5901
5902=cut
5903*/
5904
79072805 5905I32
e01b9e88 5906Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5907{
97aff369 5908 dVAR;
e1ec3a88 5909 const char *pv1;
463ee0b2 5910 STRLEN cur1;
e1ec3a88 5911 const char *pv2;
463ee0b2 5912 STRLEN cur2;
e01b9e88 5913 I32 eq = 0;
bd61b366 5914 char *tpv = NULL;
a0714e2c 5915 SV* svrecode = NULL;
79072805 5916
e01b9e88 5917 if (!sv1) {
79072805
LW
5918 pv1 = "";
5919 cur1 = 0;
5920 }
463ee0b2 5921 else
4d84ee25 5922 pv1 = SvPV_const(sv1, cur1);
79072805 5923
e01b9e88
SC
5924 if (!sv2){
5925 pv2 = "";
5926 cur2 = 0;
92d29cee 5927 }
e01b9e88 5928 else
4d84ee25 5929 pv2 = SvPV_const(sv2, cur2);
79072805 5930
cf48d248 5931 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5932 /* Differing utf8ness.
5933 * Do not UTF8size the comparands as a side-effect. */
5934 if (PL_encoding) {
5935 if (SvUTF8(sv1)) {
553e1bcc
AT
5936 svrecode = newSVpvn(pv2, cur2);
5937 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5938 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5939 }
5940 else {
553e1bcc
AT
5941 svrecode = newSVpvn(pv1, cur1);
5942 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5943 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5944 }
5945 /* Now both are in UTF-8. */
0a1bd7ac
DM
5946 if (cur1 != cur2) {
5947 SvREFCNT_dec(svrecode);
799ef3cb 5948 return FALSE;
0a1bd7ac 5949 }
799ef3cb
JH
5950 }
5951 else {
5952 bool is_utf8 = TRUE;
5953
5954 if (SvUTF8(sv1)) {
5955 /* sv1 is the UTF-8 one,
5956 * if is equal it must be downgrade-able */
9d4ba2ae 5957 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5958 &cur1, &is_utf8);
5959 if (pv != pv1)
553e1bcc 5960 pv1 = tpv = pv;
799ef3cb
JH
5961 }
5962 else {
5963 /* sv2 is the UTF-8 one,
5964 * if is equal it must be downgrade-able */
9d4ba2ae 5965 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5966 &cur2, &is_utf8);
5967 if (pv != pv2)
553e1bcc 5968 pv2 = tpv = pv;
799ef3cb
JH
5969 }
5970 if (is_utf8) {
5971 /* Downgrade not possible - cannot be eq */
bf694877 5972 assert (tpv == 0);
799ef3cb
JH
5973 return FALSE;
5974 }
5975 }
cf48d248
JH
5976 }
5977
5978 if (cur1 == cur2)
765f542d 5979 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5980
b37c2d43 5981 SvREFCNT_dec(svrecode);
553e1bcc
AT
5982 if (tpv)
5983 Safefree(tpv);
cf48d248 5984
e01b9e88 5985 return eq;
79072805
LW
5986}
5987
954c1994
GS
5988/*
5989=for apidoc sv_cmp
5990
5991Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5992string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5993C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5994coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5995
5996=cut
5997*/
5998
79072805 5999I32
e01b9e88 6000Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 6001{
97aff369 6002 dVAR;
560a288e 6003 STRLEN cur1, cur2;
e1ec3a88 6004 const char *pv1, *pv2;
bd61b366 6005 char *tpv = NULL;
cf48d248 6006 I32 cmp;
a0714e2c 6007 SV *svrecode = NULL;
560a288e 6008
e01b9e88
SC
6009 if (!sv1) {
6010 pv1 = "";
560a288e
GS
6011 cur1 = 0;
6012 }
e01b9e88 6013 else
4d84ee25 6014 pv1 = SvPV_const(sv1, cur1);
560a288e 6015
553e1bcc 6016 if (!sv2) {
e01b9e88 6017 pv2 = "";
560a288e
GS
6018 cur2 = 0;
6019 }
e01b9e88 6020 else
4d84ee25 6021 pv2 = SvPV_const(sv2, cur2);
79072805 6022
cf48d248 6023 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6024 /* Differing utf8ness.
6025 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6026 if (SvUTF8(sv1)) {
799ef3cb 6027 if (PL_encoding) {
553e1bcc
AT
6028 svrecode = newSVpvn(pv2, cur2);
6029 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6030 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6031 }
6032 else {
e1ec3a88 6033 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6034 }
cf48d248
JH
6035 }
6036 else {
799ef3cb 6037 if (PL_encoding) {
553e1bcc
AT
6038 svrecode = newSVpvn(pv1, cur1);
6039 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6040 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6041 }
6042 else {
e1ec3a88 6043 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6044 }
cf48d248
JH
6045 }
6046 }
6047
e01b9e88 6048 if (!cur1) {
cf48d248 6049 cmp = cur2 ? -1 : 0;
e01b9e88 6050 } else if (!cur2) {
cf48d248
JH
6051 cmp = 1;
6052 } else {
e1ec3a88 6053 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6054
6055 if (retval) {
cf48d248 6056 cmp = retval < 0 ? -1 : 1;
e01b9e88 6057 } else if (cur1 == cur2) {
cf48d248
JH
6058 cmp = 0;
6059 } else {
6060 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6061 }
cf48d248 6062 }
16660edb 6063
b37c2d43 6064 SvREFCNT_dec(svrecode);
553e1bcc
AT
6065 if (tpv)
6066 Safefree(tpv);
cf48d248
JH
6067
6068 return cmp;
bbce6d69 6069}
16660edb 6070
c461cf8f
JH
6071/*
6072=for apidoc sv_cmp_locale
6073
645c22ef
DM
6074Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6075'use bytes' aware, handles get magic, and will coerce its args to strings
6076if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6077
6078=cut
6079*/
6080
bbce6d69 6081I32
864dbfa3 6082Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6083{
97aff369 6084 dVAR;
36477c24 6085#ifdef USE_LOCALE_COLLATE
16660edb 6086
bbce6d69 6087 char *pv1, *pv2;
6088 STRLEN len1, len2;
6089 I32 retval;
16660edb 6090
3280af22 6091 if (PL_collation_standard)
bbce6d69 6092 goto raw_compare;
16660edb 6093
bbce6d69 6094 len1 = 0;
8ac85365 6095 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6096 len2 = 0;
8ac85365 6097 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6098
bbce6d69 6099 if (!pv1 || !len1) {
6100 if (pv2 && len2)
6101 return -1;
6102 else
6103 goto raw_compare;
6104 }
6105 else {
6106 if (!pv2 || !len2)
6107 return 1;
6108 }
16660edb 6109
bbce6d69 6110 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6111
bbce6d69 6112 if (retval)
16660edb 6113 return retval < 0 ? -1 : 1;
6114
bbce6d69 6115 /*
6116 * When the result of collation is equality, that doesn't mean
6117 * that there are no differences -- some locales exclude some
6118 * characters from consideration. So to avoid false equalities,
6119 * we use the raw string as a tiebreaker.
6120 */
16660edb 6121
bbce6d69 6122 raw_compare:
5f66b61c 6123 /*FALLTHROUGH*/
16660edb 6124
36477c24 6125#endif /* USE_LOCALE_COLLATE */
16660edb 6126
bbce6d69 6127 return sv_cmp(sv1, sv2);
6128}
79072805 6129
645c22ef 6130
36477c24 6131#ifdef USE_LOCALE_COLLATE
645c22ef 6132
7a4c00b4 6133/*
645c22ef
DM
6134=for apidoc sv_collxfrm
6135
6136Add Collate Transform magic to an SV if it doesn't already have it.
6137
6138Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6139scalar data of the variable, but transformed to such a format that a normal
6140memory comparison can be used to compare the data according to the locale
6141settings.
6142
6143=cut
6144*/
6145
bbce6d69 6146char *
864dbfa3 6147Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6148{
97aff369 6149 dVAR;
7a4c00b4 6150 MAGIC *mg;
16660edb 6151
14befaf4 6152 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6153 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6154 const char *s;
6155 char *xf;
bbce6d69 6156 STRLEN len, xlen;
6157
7a4c00b4 6158 if (mg)
6159 Safefree(mg->mg_ptr);
93524f2b 6160 s = SvPV_const(sv, len);
bbce6d69 6161 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6162 if (SvREADONLY(sv)) {
6163 SAVEFREEPV(xf);
6164 *nxp = xlen;
3280af22 6165 return xf + sizeof(PL_collation_ix);
ff0cee69 6166 }
7a4c00b4 6167 if (! mg) {
d83f0a82
NC
6168#ifdef PERL_OLD_COPY_ON_WRITE
6169 if (SvIsCOW(sv))
6170 sv_force_normal_flags(sv, 0);
6171#endif
6172 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6173 0, 0);
7a4c00b4 6174 assert(mg);
bbce6d69 6175 }
7a4c00b4 6176 mg->mg_ptr = xf;
565764a8 6177 mg->mg_len = xlen;
7a4c00b4 6178 }
6179 else {
ff0cee69 6180 if (mg) {
6181 mg->mg_ptr = NULL;
565764a8 6182 mg->mg_len = -1;
ff0cee69 6183 }
bbce6d69 6184 }
6185 }
7a4c00b4 6186 if (mg && mg->mg_ptr) {
565764a8 6187 *nxp = mg->mg_len;
3280af22 6188 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6189 }
6190 else {
6191 *nxp = 0;
6192 return NULL;
16660edb 6193 }
79072805
LW
6194}
6195
36477c24 6196#endif /* USE_LOCALE_COLLATE */
bbce6d69 6197
c461cf8f
JH
6198/*
6199=for apidoc sv_gets
6200
6201Get a line from the filehandle and store it into the SV, optionally
6202appending to the currently-stored string.
6203
6204=cut
6205*/
6206
79072805 6207char *
864dbfa3 6208Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6209{
97aff369 6210 dVAR;
e1ec3a88 6211 const char *rsptr;
c07a80fd 6212 STRLEN rslen;
6213 register STDCHAR rslast;
6214 register STDCHAR *bp;
6215 register I32 cnt;
9c5ffd7c 6216 I32 i = 0;
8bfdd7d9 6217 I32 rspara = 0;
c07a80fd 6218
bc44a8a2
NC
6219 if (SvTHINKFIRST(sv))
6220 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6221 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6222 from <>.
6223 However, perlbench says it's slower, because the existing swipe code
6224 is faster than copy on write.
6225 Swings and roundabouts. */
862a34c6 6226 SvUPGRADE(sv, SVt_PV);
99491443 6227
ff68c719 6228 SvSCREAM_off(sv);
efd8b2ba
AE
6229
6230 if (append) {
6231 if (PerlIO_isutf8(fp)) {
6232 if (!SvUTF8(sv)) {
6233 sv_utf8_upgrade_nomg(sv);
6234 sv_pos_u2b(sv,&append,0);
6235 }
6236 } else if (SvUTF8(sv)) {
561b68a9 6237 SV * const tsv = newSV(0);
efd8b2ba
AE
6238 sv_gets(tsv, fp, 0);
6239 sv_utf8_upgrade_nomg(tsv);
6240 SvCUR_set(sv,append);
6241 sv_catsv(sv,tsv);
6242 sv_free(tsv);
6243 goto return_string_or_null;
6244 }
6245 }
6246
6247 SvPOK_only(sv);
6248 if (PerlIO_isutf8(fp))
6249 SvUTF8_on(sv);
c07a80fd 6250
923e4eb5 6251 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6252 /* we always read code in line mode */
6253 rsptr = "\n";
6254 rslen = 1;
6255 }
6256 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6257 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6258 of amount we are going to read -- may result in mallocing
6259 more memory than we really need if the layers below reduce
6260 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6261 */
e311fd51 6262 Stat_t st;
e468d35b 6263 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6264 const Off_t offset = PerlIO_tell(fp);
58f1856e 6265 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6266 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6267 }
6268 }
c07a80fd 6269 rsptr = NULL;
6270 rslen = 0;
6271 }
3280af22 6272 else if (RsRECORD(PL_rs)) {
e311fd51 6273 I32 bytesread;
5b2b9c68 6274 char *buffer;
acbd132f 6275 U32 recsize;
5b2b9c68
HM
6276
6277 /* Grab the size of the record we're getting */
acbd132f 6278 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6279 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6280 /* Go yank in */
6281#ifdef VMS
6282 /* VMS wants read instead of fread, because fread doesn't respect */
6283 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6284 /* doing, but we've got no other real choice - except avoid stdio
6285 as implementation - perhaps write a :vms layer ?
6286 */
5b2b9c68
HM
6287 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6288#else
6289 bytesread = PerlIO_read(fp, buffer, recsize);
6290#endif
27e6ca2d
AE
6291 if (bytesread < 0)
6292 bytesread = 0;
e311fd51 6293 SvCUR_set(sv, bytesread += append);
e670df4e 6294 buffer[bytesread] = '\0';
efd8b2ba 6295 goto return_string_or_null;
5b2b9c68 6296 }
3280af22 6297 else if (RsPARA(PL_rs)) {
c07a80fd 6298 rsptr = "\n\n";
6299 rslen = 2;
8bfdd7d9 6300 rspara = 1;
c07a80fd 6301 }
7d59b7e4
NIS
6302 else {
6303 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6304 if (PerlIO_isutf8(fp)) {
6305 rsptr = SvPVutf8(PL_rs, rslen);
6306 }
6307 else {
6308 if (SvUTF8(PL_rs)) {
6309 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6310 Perl_croak(aTHX_ "Wide character in $/");
6311 }
6312 }
93524f2b 6313 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6314 }
6315 }
6316
c07a80fd 6317 rslast = rslen ? rsptr[rslen - 1] : '\0';
6318
8bfdd7d9 6319 if (rspara) { /* have to do this both before and after */
79072805 6320 do { /* to make sure file boundaries work right */
760ac839 6321 if (PerlIO_eof(fp))
a0d0e21e 6322 return 0;
760ac839 6323 i = PerlIO_getc(fp);
79072805 6324 if (i != '\n') {
a0d0e21e
LW
6325 if (i == -1)
6326 return 0;
760ac839 6327 PerlIO_ungetc(fp,i);
79072805
LW
6328 break;
6329 }
6330 } while (i != EOF);
6331 }
c07a80fd 6332
760ac839
LW
6333 /* See if we know enough about I/O mechanism to cheat it ! */
6334
6335 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6336 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6337 enough here - and may even be a macro allowing compile
6338 time optimization.
6339 */
6340
6341 if (PerlIO_fast_gets(fp)) {
6342
6343 /*
6344 * We're going to steal some values from the stdio struct
6345 * and put EVERYTHING in the innermost loop into registers.
6346 */
6347 register STDCHAR *ptr;
6348 STRLEN bpx;
6349 I32 shortbuffered;
6350
16660edb 6351#if defined(VMS) && defined(PERLIO_IS_STDIO)
6352 /* An ungetc()d char is handled separately from the regular
6353 * buffer, so we getc() it back out and stuff it in the buffer.
6354 */
6355 i = PerlIO_getc(fp);
6356 if (i == EOF) return 0;
6357 *(--((*fp)->_ptr)) = (unsigned char) i;
6358 (*fp)->_cnt++;
6359#endif
c07a80fd 6360
c2960299 6361 /* Here is some breathtakingly efficient cheating */
c07a80fd 6362
a20bf0c3 6363 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6364 /* make sure we have the room */
7a5fa8a2 6365 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6366 /* Not room for all of it
7a5fa8a2 6367 if we are looking for a separator and room for some
e468d35b
NIS
6368 */
6369 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6370 /* just process what we have room for */
79072805
LW
6371 shortbuffered = cnt - SvLEN(sv) + append + 1;
6372 cnt -= shortbuffered;
6373 }
6374 else {
6375 shortbuffered = 0;
bbce6d69 6376 /* remember that cnt can be negative */
eb160463 6377 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6378 }
6379 }
7a5fa8a2 6380 else
79072805 6381 shortbuffered = 0;
3f7c398e 6382 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6383 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6384 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6385 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6386 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6387 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6388 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6389 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6390 for (;;) {
6391 screamer:
93a17b20 6392 if (cnt > 0) {
c07a80fd 6393 if (rslen) {
760ac839
LW
6394 while (cnt > 0) { /* this | eat */
6395 cnt--;
c07a80fd 6396 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6397 goto thats_all_folks; /* screams | sed :-) */
6398 }
6399 }
6400 else {
1c846c1f
NIS
6401 Copy(ptr, bp, cnt, char); /* this | eat */
6402 bp += cnt; /* screams | dust */
c07a80fd 6403 ptr += cnt; /* louder | sed :-) */
a5f75d66 6404 cnt = 0;
93a17b20 6405 }
79072805
LW
6406 }
6407
748a9306 6408 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6409 cnt = shortbuffered;
6410 shortbuffered = 0;
3f7c398e 6411 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6412 SvCUR_set(sv, bpx);
6413 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6414 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6415 continue;
6416 }
6417
16660edb 6418 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6419 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6420 PTR2UV(ptr),(long)cnt));
cc00df79 6421 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6422#if 0
16660edb 6423 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6424 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6425 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6426 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6427#endif
1c846c1f 6428 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6429 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6430 another abstraction. */
760ac839 6431 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6432#if 0
16660edb 6433 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6434 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6435 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6436 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6437#endif
a20bf0c3
JH
6438 cnt = PerlIO_get_cnt(fp);
6439 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6440 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6441 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6442
748a9306
LW
6443 if (i == EOF) /* all done for ever? */
6444 goto thats_really_all_folks;
6445
3f7c398e 6446 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6447 SvCUR_set(sv, bpx);
6448 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6449 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6450
eb160463 6451 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6452
c07a80fd 6453 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6454 goto thats_all_folks;
79072805
LW
6455 }
6456
6457thats_all_folks:
3f7c398e 6458 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6459 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6460 goto screamer; /* go back to the fray */
79072805
LW
6461thats_really_all_folks:
6462 if (shortbuffered)
6463 cnt += shortbuffered;
16660edb 6464 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6465 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6466 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6467 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6468 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6469 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6470 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6471 *bp = '\0';
3f7c398e 6472 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6473 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6474 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6475 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6476 }
6477 else
79072805 6478 {
6edd2cd5 6479 /*The big, slow, and stupid way. */
27da23d5 6480#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6481 STDCHAR *buf = NULL;
a02a5408 6482 Newx(buf, 8192, STDCHAR);
6edd2cd5 6483 assert(buf);
4d2c4e07 6484#else
6edd2cd5 6485 STDCHAR buf[8192];
4d2c4e07 6486#endif
79072805 6487
760ac839 6488screamer2:
c07a80fd 6489 if (rslen) {
00b6aa41 6490 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6491 bp = buf;
eb160463 6492 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6493 ; /* keep reading */
6494 cnt = bp - buf;
c07a80fd 6495 }
6496 else {
760ac839 6497 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6498 /* Accomodate broken VAXC compiler, which applies U8 cast to
6499 * both args of ?: operator, causing EOF to change into 255
6500 */
37be0adf 6501 if (cnt > 0)
cbe9e203
JH
6502 i = (U8)buf[cnt - 1];
6503 else
37be0adf 6504 i = EOF;
c07a80fd 6505 }
79072805 6506
cbe9e203
JH
6507 if (cnt < 0)
6508 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6509 if (append)
6510 sv_catpvn(sv, (char *) buf, cnt);
6511 else
6512 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6513
6514 if (i != EOF && /* joy */
6515 (!rslen ||
6516 SvCUR(sv) < rslen ||
3f7c398e 6517 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6518 {
6519 append = -1;
63e4d877
CS
6520 /*
6521 * If we're reading from a TTY and we get a short read,
6522 * indicating that the user hit his EOF character, we need
6523 * to notice it now, because if we try to read from the TTY
6524 * again, the EOF condition will disappear.
6525 *
6526 * The comparison of cnt to sizeof(buf) is an optimization
6527 * that prevents unnecessary calls to feof().
6528 *
6529 * - jik 9/25/96
6530 */
bb7a0f54 6531 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 6532 goto screamer2;
79072805 6533 }
6edd2cd5 6534
27da23d5 6535#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6536 Safefree(buf);
6537#endif
79072805
LW
6538 }
6539
8bfdd7d9 6540 if (rspara) { /* have to do this both before and after */
c07a80fd 6541 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6542 i = PerlIO_getc(fp);
79072805 6543 if (i != '\n') {
760ac839 6544 PerlIO_ungetc(fp,i);
79072805
LW
6545 break;
6546 }
6547 }
6548 }
c07a80fd 6549
efd8b2ba 6550return_string_or_null:
bd61b366 6551 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6552}
6553
954c1994
GS
6554/*
6555=for apidoc sv_inc
6556
645c22ef
DM
6557Auto-increment of the value in the SV, doing string to numeric conversion
6558if necessary. Handles 'get' magic.
954c1994
GS
6559
6560=cut
6561*/
6562
79072805 6563void
864dbfa3 6564Perl_sv_inc(pTHX_ register SV *sv)
79072805 6565{
97aff369 6566 dVAR;
79072805 6567 register char *d;
463ee0b2 6568 int flags;
79072805
LW
6569
6570 if (!sv)
6571 return;
5b295bef 6572 SvGETMAGIC(sv);
ed6116ce 6573 if (SvTHINKFIRST(sv)) {
765f542d
NC
6574 if (SvIsCOW(sv))
6575 sv_force_normal_flags(sv, 0);
0f15f207 6576 if (SvREADONLY(sv)) {
923e4eb5 6577 if (IN_PERL_RUNTIME)
cea2e8a9 6578 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6579 }
a0d0e21e 6580 if (SvROK(sv)) {
b5be31e9 6581 IV i;
9e7bc3e8
JD
6582 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6583 return;
56431972 6584 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6585 sv_unref(sv);
6586 sv_setiv(sv, i);
a0d0e21e 6587 }
ed6116ce 6588 }
8990e307 6589 flags = SvFLAGS(sv);
28e5dec8
JH
6590 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6591 /* It's (privately or publicly) a float, but not tested as an
6592 integer, so test it to see. */
d460ef45 6593 (void) SvIV(sv);
28e5dec8
JH
6594 flags = SvFLAGS(sv);
6595 }
6596 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6597 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6598#ifdef PERL_PRESERVE_IVUV
28e5dec8 6599 oops_its_int:
59d8ce62 6600#endif
25da4f38
IZ
6601 if (SvIsUV(sv)) {
6602 if (SvUVX(sv) == UV_MAX)
a1e868e7 6603 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6604 else
6605 (void)SvIOK_only_UV(sv);
607fa7f2 6606 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6607 } else {
6608 if (SvIVX(sv) == IV_MAX)
28e5dec8 6609 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6610 else {
6611 (void)SvIOK_only(sv);
45977657 6612 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6613 }
55497cff 6614 }
79072805
LW
6615 return;
6616 }
28e5dec8
JH
6617 if (flags & SVp_NOK) {
6618 (void)SvNOK_only(sv);
9d6ce603 6619 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6620 return;
6621 }
6622
3f7c398e 6623 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6624 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6625 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6626 (void)SvIOK_only(sv);
45977657 6627 SvIV_set(sv, 1);
79072805
LW
6628 return;
6629 }
463ee0b2 6630 d = SvPVX(sv);
79072805
LW
6631 while (isALPHA(*d)) d++;
6632 while (isDIGIT(*d)) d++;
6633 if (*d) {
28e5dec8 6634#ifdef PERL_PRESERVE_IVUV
d1be9408 6635 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6636 warnings. Probably ought to make the sv_iv_please() that does
6637 the conversion if possible, and silently. */
504618e9 6638 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6639 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6640 /* Need to try really hard to see if it's an integer.
6641 9.22337203685478e+18 is an integer.
6642 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6643 so $a="9.22337203685478e+18"; $a+0; $a++
6644 needs to be the same as $a="9.22337203685478e+18"; $a++
6645 or we go insane. */
d460ef45 6646
28e5dec8
JH
6647 (void) sv_2iv(sv);
6648 if (SvIOK(sv))
6649 goto oops_its_int;
6650
6651 /* sv_2iv *should* have made this an NV */
6652 if (flags & SVp_NOK) {
6653 (void)SvNOK_only(sv);
9d6ce603 6654 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6655 return;
6656 }
6657 /* I don't think we can get here. Maybe I should assert this
6658 And if we do get here I suspect that sv_setnv will croak. NWC
6659 Fall through. */
6660#if defined(USE_LONG_DOUBLE)
6661 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 6662 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6663#else
1779d84d 6664 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 6665 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6666#endif
6667 }
6668#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6669 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6670 return;
6671 }
6672 d--;
3f7c398e 6673 while (d >= SvPVX_const(sv)) {
79072805
LW
6674 if (isDIGIT(*d)) {
6675 if (++*d <= '9')
6676 return;
6677 *(d--) = '0';
6678 }
6679 else {
9d116dd7
JH
6680#ifdef EBCDIC
6681 /* MKS: The original code here died if letters weren't consecutive.
6682 * at least it didn't have to worry about non-C locales. The
6683 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6684 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6685 * [A-Za-z] are accepted by isALPHA in the C locale.
6686 */
6687 if (*d != 'z' && *d != 'Z') {
6688 do { ++*d; } while (!isALPHA(*d));
6689 return;
6690 }
6691 *(d--) -= 'z' - 'a';
6692#else
79072805
LW
6693 ++*d;
6694 if (isALPHA(*d))
6695 return;
6696 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6697#endif
79072805
LW
6698 }
6699 }
6700 /* oh,oh, the number grew */
6701 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6702 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6703 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6704 *d = d[-1];
6705 if (isDIGIT(d[1]))
6706 *d = '1';
6707 else
6708 *d = d[1];
6709}
6710
954c1994
GS
6711/*
6712=for apidoc sv_dec
6713
645c22ef
DM
6714Auto-decrement of the value in the SV, doing string to numeric conversion
6715if necessary. Handles 'get' magic.
954c1994
GS
6716
6717=cut
6718*/
6719
79072805 6720void
864dbfa3 6721Perl_sv_dec(pTHX_ register SV *sv)
79072805 6722{
97aff369 6723 dVAR;
463ee0b2
LW
6724 int flags;
6725
79072805
LW
6726 if (!sv)
6727 return;
5b295bef 6728 SvGETMAGIC(sv);
ed6116ce 6729 if (SvTHINKFIRST(sv)) {
765f542d
NC
6730 if (SvIsCOW(sv))
6731 sv_force_normal_flags(sv, 0);
0f15f207 6732 if (SvREADONLY(sv)) {
923e4eb5 6733 if (IN_PERL_RUNTIME)
cea2e8a9 6734 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6735 }
a0d0e21e 6736 if (SvROK(sv)) {
b5be31e9 6737 IV i;
9e7bc3e8
JD
6738 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6739 return;
56431972 6740 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6741 sv_unref(sv);
6742 sv_setiv(sv, i);
a0d0e21e 6743 }
ed6116ce 6744 }
28e5dec8
JH
6745 /* Unlike sv_inc we don't have to worry about string-never-numbers
6746 and keeping them magic. But we mustn't warn on punting */
8990e307 6747 flags = SvFLAGS(sv);
28e5dec8
JH
6748 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6749 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6750#ifdef PERL_PRESERVE_IVUV
28e5dec8 6751 oops_its_int:
59d8ce62 6752#endif
25da4f38
IZ
6753 if (SvIsUV(sv)) {
6754 if (SvUVX(sv) == 0) {
6755 (void)SvIOK_only(sv);
45977657 6756 SvIV_set(sv, -1);
25da4f38
IZ
6757 }
6758 else {
6759 (void)SvIOK_only_UV(sv);
f4eee32f 6760 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6761 }
25da4f38
IZ
6762 } else {
6763 if (SvIVX(sv) == IV_MIN)
65202027 6764 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6765 else {
6766 (void)SvIOK_only(sv);
45977657 6767 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6768 }
55497cff 6769 }
6770 return;
6771 }
28e5dec8 6772 if (flags & SVp_NOK) {
9d6ce603 6773 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6774 (void)SvNOK_only(sv);
6775 return;
6776 }
8990e307 6777 if (!(flags & SVp_POK)) {
ef088171
NC
6778 if ((flags & SVTYPEMASK) < SVt_PVIV)
6779 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6780 SvIV_set(sv, -1);
6781 (void)SvIOK_only(sv);
79072805
LW
6782 return;
6783 }
28e5dec8
JH
6784#ifdef PERL_PRESERVE_IVUV
6785 {
504618e9 6786 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6787 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6788 /* Need to try really hard to see if it's an integer.
6789 9.22337203685478e+18 is an integer.
6790 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6791 so $a="9.22337203685478e+18"; $a+0; $a--
6792 needs to be the same as $a="9.22337203685478e+18"; $a--
6793 or we go insane. */
d460ef45 6794
28e5dec8
JH
6795 (void) sv_2iv(sv);
6796 if (SvIOK(sv))
6797 goto oops_its_int;
6798
6799 /* sv_2iv *should* have made this an NV */
6800 if (flags & SVp_NOK) {
6801 (void)SvNOK_only(sv);
9d6ce603 6802 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6803 return;
6804 }
6805 /* I don't think we can get here. Maybe I should assert this
6806 And if we do get here I suspect that sv_setnv will croak. NWC
6807 Fall through. */
6808#if defined(USE_LONG_DOUBLE)
6809 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 6810 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6811#else
1779d84d 6812 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 6813 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6814#endif
6815 }
6816 }
6817#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6818 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6819}
6820
954c1994
GS
6821/*
6822=for apidoc sv_mortalcopy
6823
645c22ef 6824Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6825The new SV is marked as mortal. It will be destroyed "soon", either by an
6826explicit call to FREETMPS, or by an implicit call at places such as
6827statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6828
6829=cut
6830*/
6831
79072805
LW
6832/* Make a string that will exist for the duration of the expression
6833 * evaluation. Actually, it may have to last longer than that, but
6834 * hopefully we won't free it until it has been assigned to a
6835 * permanent location. */
6836
6837SV *
864dbfa3 6838Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6839{
97aff369 6840 dVAR;
463ee0b2 6841 register SV *sv;
b881518d 6842
4561caa4 6843 new_SV(sv);
79072805 6844 sv_setsv(sv,oldstr);
677b06e3
GS
6845 EXTEND_MORTAL(1);
6846 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6847 SvTEMP_on(sv);
6848 return sv;
6849}
6850
954c1994
GS
6851/*
6852=for apidoc sv_newmortal
6853
645c22ef 6854Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6855set to 1. It will be destroyed "soon", either by an explicit call to
6856FREETMPS, or by an implicit call at places such as statement boundaries.
6857See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6858
6859=cut
6860*/
6861
8990e307 6862SV *
864dbfa3 6863Perl_sv_newmortal(pTHX)
8990e307 6864{
97aff369 6865 dVAR;
8990e307
LW
6866 register SV *sv;
6867
4561caa4 6868 new_SV(sv);
8990e307 6869 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6870 EXTEND_MORTAL(1);
6871 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6872 return sv;
6873}
6874
954c1994
GS
6875/*
6876=for apidoc sv_2mortal
6877
d4236ebc
DM
6878Marks an existing SV as mortal. The SV will be destroyed "soon", either
6879by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6880statement boundaries. SvTEMP() is turned on which means that the SV's
6881string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6882and C<sv_mortalcopy>.
954c1994
GS
6883
6884=cut
6885*/
6886
79072805 6887SV *
864dbfa3 6888Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6889{
27da23d5 6890 dVAR;
79072805 6891 if (!sv)
7a5b473e 6892 return NULL;
d689ffdd 6893 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6894 return sv;
677b06e3
GS
6895 EXTEND_MORTAL(1);
6896 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6897 SvTEMP_on(sv);
79072805
LW
6898 return sv;
6899}
6900
954c1994
GS
6901/*
6902=for apidoc newSVpv
6903
6904Creates a new SV and copies a string into it. The reference count for the
6905SV is set to 1. If C<len> is zero, Perl will compute the length using
6906strlen(). For efficiency, consider using C<newSVpvn> instead.
6907
6908=cut
6909*/
6910
79072805 6911SV *
864dbfa3 6912Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6913{
97aff369 6914 dVAR;
463ee0b2 6915 register SV *sv;
79072805 6916
4561caa4 6917 new_SV(sv);
616d8c9c 6918 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6919 return sv;
6920}
6921
954c1994
GS
6922/*
6923=for apidoc newSVpvn
6924
6925Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6926SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6927string. You are responsible for ensuring that the source string is at least
9e09f5f2 6928C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6929
6930=cut
6931*/
6932
9da1e3b5 6933SV *
864dbfa3 6934Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6935{
97aff369 6936 dVAR;
9da1e3b5
MUN
6937 register SV *sv;
6938
6939 new_SV(sv);
9da1e3b5
MUN
6940 sv_setpvn(sv,s,len);
6941 return sv;
6942}
6943
bd08039b
NC
6944
6945/*
926f8064 6946=for apidoc newSVhek
bd08039b
NC
6947
6948Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6949point to the shared string table where possible. Returns a new (undefined)
6950SV if the hek is NULL.
bd08039b
NC
6951
6952=cut
6953*/
6954
6955SV *
c1b02ed8 6956Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6957{
97aff369 6958 dVAR;
5aaec2b4
NC
6959 if (!hek) {
6960 SV *sv;
6961
6962 new_SV(sv);
6963 return sv;
6964 }
6965
bd08039b
NC
6966 if (HEK_LEN(hek) == HEf_SVKEY) {
6967 return newSVsv(*(SV**)HEK_KEY(hek));
6968 } else {
6969 const int flags = HEK_FLAGS(hek);
6970 if (flags & HVhek_WASUTF8) {
6971 /* Trouble :-)
6972 Andreas would like keys he put in as utf8 to come back as utf8
6973 */
6974 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6975 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6976 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6977
6978 SvUTF8_on (sv);
6979 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6980 return sv;
6981 } else if (flags & HVhek_REHASH) {
6982 /* We don't have a pointer to the hv, so we have to replicate the
6983 flag into every HEK. This hv is using custom a hasing
6984 algorithm. Hence we can't return a shared string scalar, as
6985 that would contain the (wrong) hash value, and might get passed
6986 into an hv routine with a regular hash */
6987
b64e5050 6988 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6989 if (HEK_UTF8(hek))
6990 SvUTF8_on (sv);
6991 return sv;
6992 }
6993 /* This will be overwhelminly the most common case. */
409dfe77
NC
6994 {
6995 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6996 more efficient than sharepvn(). */
6997 SV *sv;
6998
6999 new_SV(sv);
7000 sv_upgrade(sv, SVt_PV);
7001 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7002 SvCUR_set(sv, HEK_LEN(hek));
7003 SvLEN_set(sv, 0);
7004 SvREADONLY_on(sv);
7005 SvFAKE_on(sv);
7006 SvPOK_on(sv);
7007 if (HEK_UTF8(hek))
7008 SvUTF8_on(sv);
7009 return sv;
7010 }
bd08039b
NC
7011 }
7012}
7013
1c846c1f
NIS
7014/*
7015=for apidoc newSVpvn_share
7016
3f7c398e 7017Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7018table. If the string does not already exist in the table, it is created
7019first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7020slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7021otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7022is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7023hash lookup will avoid string compare.
1c846c1f
NIS
7024
7025=cut
7026*/
7027
7028SV *
c3654f1a 7029Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7030{
97aff369 7031 dVAR;
1c846c1f 7032 register SV *sv;
c3654f1a 7033 bool is_utf8 = FALSE;
a51caccf
NC
7034 const char *const orig_src = src;
7035
c3654f1a 7036 if (len < 0) {
77caf834 7037 STRLEN tmplen = -len;
c3654f1a 7038 is_utf8 = TRUE;
75a54232 7039 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7040 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7041 len = tmplen;
7042 }
1c846c1f 7043 if (!hash)
5afd6d42 7044 PERL_HASH(hash, src, len);
1c846c1f 7045 new_SV(sv);
bdd68bc3 7046 sv_upgrade(sv, SVt_PV);
f880fe2f 7047 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7048 SvCUR_set(sv, len);
b162af07 7049 SvLEN_set(sv, 0);
1c846c1f
NIS
7050 SvREADONLY_on(sv);
7051 SvFAKE_on(sv);
7052 SvPOK_on(sv);
c3654f1a
IH
7053 if (is_utf8)
7054 SvUTF8_on(sv);
a51caccf
NC
7055 if (src != orig_src)
7056 Safefree(src);
1c846c1f
NIS
7057 return sv;
7058}
7059
645c22ef 7060
cea2e8a9 7061#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7062
7063/* pTHX_ magic can't cope with varargs, so this is a no-context
7064 * version of the main function, (which may itself be aliased to us).
7065 * Don't access this version directly.
7066 */
7067
46fc3d4c 7068SV *
cea2e8a9 7069Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7070{
cea2e8a9 7071 dTHX;
46fc3d4c 7072 register SV *sv;
7073 va_list args;
46fc3d4c 7074 va_start(args, pat);
c5be433b 7075 sv = vnewSVpvf(pat, &args);
46fc3d4c 7076 va_end(args);
7077 return sv;
7078}
cea2e8a9 7079#endif
46fc3d4c 7080
954c1994
GS
7081/*
7082=for apidoc newSVpvf
7083
645c22ef 7084Creates a new SV and initializes it with the string formatted like
954c1994
GS
7085C<sprintf>.
7086
7087=cut
7088*/
7089
cea2e8a9
GS
7090SV *
7091Perl_newSVpvf(pTHX_ const char* pat, ...)
7092{
7093 register SV *sv;
7094 va_list args;
cea2e8a9 7095 va_start(args, pat);
c5be433b 7096 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7097 va_end(args);
7098 return sv;
7099}
46fc3d4c 7100
645c22ef
DM
7101/* backend for newSVpvf() and newSVpvf_nocontext() */
7102
79072805 7103SV *
c5be433b
GS
7104Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7105{
97aff369 7106 dVAR;
c5be433b
GS
7107 register SV *sv;
7108 new_SV(sv);
4608196e 7109 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7110 return sv;
7111}
7112
954c1994
GS
7113/*
7114=for apidoc newSVnv
7115
7116Creates a new SV and copies a floating point value into it.
7117The reference count for the SV is set to 1.
7118
7119=cut
7120*/
7121
c5be433b 7122SV *
65202027 7123Perl_newSVnv(pTHX_ NV n)
79072805 7124{
97aff369 7125 dVAR;
463ee0b2 7126 register SV *sv;
79072805 7127
4561caa4 7128 new_SV(sv);
79072805
LW
7129 sv_setnv(sv,n);
7130 return sv;
7131}
7132
954c1994
GS
7133/*
7134=for apidoc newSViv
7135
7136Creates a new SV and copies an integer into it. The reference count for the
7137SV is set to 1.
7138
7139=cut
7140*/
7141
79072805 7142SV *
864dbfa3 7143Perl_newSViv(pTHX_ IV i)
79072805 7144{
97aff369 7145 dVAR;
463ee0b2 7146 register SV *sv;
79072805 7147
4561caa4 7148 new_SV(sv);
79072805
LW
7149 sv_setiv(sv,i);
7150 return sv;
7151}
7152
954c1994 7153/*
1a3327fb
JH
7154=for apidoc newSVuv
7155
7156Creates a new SV and copies an unsigned integer into it.
7157The reference count for the SV is set to 1.
7158
7159=cut
7160*/
7161
7162SV *
7163Perl_newSVuv(pTHX_ UV u)
7164{
97aff369 7165 dVAR;
1a3327fb
JH
7166 register SV *sv;
7167
7168 new_SV(sv);
7169 sv_setuv(sv,u);
7170 return sv;
7171}
7172
7173/*
954c1994
GS
7174=for apidoc newRV_noinc
7175
7176Creates an RV wrapper for an SV. The reference count for the original
7177SV is B<not> incremented.
7178
7179=cut
7180*/
7181
2304df62 7182SV *
864dbfa3 7183Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7184{
97aff369 7185 dVAR;
2304df62
AD
7186 register SV *sv;
7187
4561caa4 7188 new_SV(sv);
2304df62 7189 sv_upgrade(sv, SVt_RV);
76e3520e 7190 SvTEMP_off(tmpRef);
b162af07 7191 SvRV_set(sv, tmpRef);
2304df62 7192 SvROK_on(sv);
2304df62
AD
7193 return sv;
7194}
7195
ff276b08 7196/* newRV_inc is the official function name to use now.
645c22ef
DM
7197 * newRV_inc is in fact #defined to newRV in sv.h
7198 */
7199
5f05dabc 7200SV *
7f466ec7 7201Perl_newRV(pTHX_ SV *sv)
5f05dabc 7202{
97aff369 7203 dVAR;
7f466ec7 7204 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7205}
5f05dabc 7206
954c1994
GS
7207/*
7208=for apidoc newSVsv
7209
7210Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7211(Uses C<sv_setsv>).
954c1994
GS
7212
7213=cut
7214*/
7215
79072805 7216SV *
864dbfa3 7217Perl_newSVsv(pTHX_ register SV *old)
79072805 7218{
97aff369 7219 dVAR;
463ee0b2 7220 register SV *sv;
79072805
LW
7221
7222 if (!old)
7a5b473e 7223 return NULL;
8990e307 7224 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7225 if (ckWARN_d(WARN_INTERNAL))
9014280d 7226 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7227 return NULL;
79072805 7228 }
4561caa4 7229 new_SV(sv);
e90aabeb
NC
7230 /* SV_GMAGIC is the default for sv_setv()
7231 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7232 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7233 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7234 return sv;
79072805
LW
7235}
7236
645c22ef
DM
7237/*
7238=for apidoc sv_reset
7239
7240Underlying implementation for the C<reset> Perl function.
7241Note that the perl-level function is vaguely deprecated.
7242
7243=cut
7244*/
7245
79072805 7246void
e1ec3a88 7247Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7248{
27da23d5 7249 dVAR;
4802d5d7 7250 char todo[PERL_UCHAR_MAX+1];
79072805 7251
49d8d3a1
MB
7252 if (!stash)
7253 return;
7254
79072805 7255 if (!*s) { /* reset ?? searches */
aec46f14 7256 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7257 if (mg) {
7258 PMOP *pm = (PMOP *) mg->mg_obj;
7259 while (pm) {
7260 pm->op_pmdynflags &= ~PMdf_USED;
7261 pm = pm->op_pmnext;
7262 }
79072805
LW
7263 }
7264 return;
7265 }
7266
7267 /* reset variables */
7268
7269 if (!HvARRAY(stash))
7270 return;
463ee0b2
LW
7271
7272 Zero(todo, 256, char);
79072805 7273 while (*s) {
b464bac0
AL
7274 I32 max;
7275 I32 i = (unsigned char)*s;
79072805
LW
7276 if (s[1] == '-') {
7277 s += 2;
7278 }
4802d5d7 7279 max = (unsigned char)*s++;
79072805 7280 for ( ; i <= max; i++) {
463ee0b2
LW
7281 todo[i] = 1;
7282 }
a0d0e21e 7283 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7284 HE *entry;
79072805 7285 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7286 entry;
7287 entry = HeNEXT(entry))
7288 {
b464bac0
AL
7289 register GV *gv;
7290 register SV *sv;
7291
1edc1566 7292 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7293 continue;
1edc1566 7294 gv = (GV*)HeVAL(entry);
79072805 7295 sv = GvSV(gv);
e203899d
NC
7296 if (sv) {
7297 if (SvTHINKFIRST(sv)) {
7298 if (!SvREADONLY(sv) && SvROK(sv))
7299 sv_unref(sv);
7300 /* XXX Is this continue a bug? Why should THINKFIRST
7301 exempt us from resetting arrays and hashes? */
7302 continue;
7303 }
7304 SvOK_off(sv);
7305 if (SvTYPE(sv) >= SVt_PV) {
7306 SvCUR_set(sv, 0);
bd61b366 7307 if (SvPVX_const(sv) != NULL)
e203899d
NC
7308 *SvPVX(sv) = '\0';
7309 SvTAINT(sv);
7310 }
79072805
LW
7311 }
7312 if (GvAV(gv)) {
7313 av_clear(GvAV(gv));
7314 }
bfcb3514 7315 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7316#if defined(VMS)
7317 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7318#else /* ! VMS */
463ee0b2 7319 hv_clear(GvHV(gv));
b0269e46
AB
7320# if defined(USE_ENVIRON_ARRAY)
7321 if (gv == PL_envgv)
7322 my_clearenv();
7323# endif /* USE_ENVIRON_ARRAY */
7324#endif /* VMS */
79072805
LW
7325 }
7326 }
7327 }
7328 }
7329}
7330
645c22ef
DM
7331/*
7332=for apidoc sv_2io
7333
7334Using various gambits, try to get an IO from an SV: the IO slot if its a
7335GV; or the recursive result if we're an RV; or the IO slot of the symbol
7336named after the PV if we're a string.
7337
7338=cut
7339*/
7340
46fc3d4c 7341IO*
864dbfa3 7342Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7343{
7344 IO* io;
7345 GV* gv;
7346
7347 switch (SvTYPE(sv)) {
7348 case SVt_PVIO:
7349 io = (IO*)sv;
7350 break;
7351 case SVt_PVGV:
7352 gv = (GV*)sv;
7353 io = GvIO(gv);
7354 if (!io)
cea2e8a9 7355 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7356 break;
7357 default:
7358 if (!SvOK(sv))
cea2e8a9 7359 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7360 if (SvROK(sv))
7361 return sv_2io(SvRV(sv));
f776e3cd 7362 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7363 if (gv)
7364 io = GvIO(gv);
7365 else
7366 io = 0;
7367 if (!io)
35c1215d 7368 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7369 break;
7370 }
7371 return io;
7372}
7373
645c22ef
DM
7374/*
7375=for apidoc sv_2cv
7376
7377Using various gambits, try to get a CV from an SV; in addition, try if
7378possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7379The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7380
7381=cut
7382*/
7383
79072805 7384CV *
864dbfa3 7385Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7386{
27da23d5 7387 dVAR;
a0714e2c 7388 GV *gv = NULL;
601f1833 7389 CV *cv = NULL;
79072805 7390
85dec29a
NC
7391 if (!sv) {
7392 *st = NULL;
7393 *gvp = NULL;
7394 return NULL;
7395 }
79072805 7396 switch (SvTYPE(sv)) {
79072805
LW
7397 case SVt_PVCV:
7398 *st = CvSTASH(sv);
a0714e2c 7399 *gvp = NULL;
79072805
LW
7400 return (CV*)sv;
7401 case SVt_PVHV:
7402 case SVt_PVAV:
ef58ba18 7403 *st = NULL;
a0714e2c 7404 *gvp = NULL;
601f1833 7405 return NULL;
8990e307
LW
7406 case SVt_PVGV:
7407 gv = (GV*)sv;
a0d0e21e 7408 *gvp = gv;
8990e307
LW
7409 *st = GvESTASH(gv);
7410 goto fix_gv;
7411
79072805 7412 default:
5b295bef 7413 SvGETMAGIC(sv);
a0d0e21e 7414 if (SvROK(sv)) {
823a54a3 7415 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7416 tryAMAGICunDEREF(to_cv);
7417
62f274bf
GS
7418 sv = SvRV(sv);
7419 if (SvTYPE(sv) == SVt_PVCV) {
7420 cv = (CV*)sv;
a0714e2c 7421 *gvp = NULL;
62f274bf
GS
7422 *st = CvSTASH(cv);
7423 return cv;
7424 }
7425 else if(isGV(sv))
7426 gv = (GV*)sv;
7427 else
cea2e8a9 7428 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7429 }
62f274bf 7430 else if (isGV(sv))
79072805
LW
7431 gv = (GV*)sv;
7432 else
7a5fd60d 7433 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7434 *gvp = gv;
ef58ba18
NC
7435 if (!gv) {
7436 *st = NULL;
601f1833 7437 return NULL;
ef58ba18 7438 }
e26df76a
NC
7439 /* Some flags to gv_fetchsv mean don't really create the GV */
7440 if (SvTYPE(gv) != SVt_PVGV) {
7441 *st = NULL;
7442 return NULL;
7443 }
79072805 7444 *st = GvESTASH(gv);
8990e307 7445 fix_gv:
8ebc5c01 7446 if (lref && !GvCVu(gv)) {
4633a7c4 7447 SV *tmpsv;
748a9306 7448 ENTER;
561b68a9 7449 tmpsv = newSV(0);
bd61b366 7450 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7451 /* XXX this is probably not what they think they're getting.
7452 * It has the same effect as "sub name;", i.e. just a forward
7453 * declaration! */
774d564b 7454 newSUB(start_subparse(FALSE, 0),
4633a7c4 7455 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7456 NULL, NULL);
748a9306 7457 LEAVE;
8ebc5c01 7458 if (!GvCVu(gv))
35c1215d
NC
7459 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7460 sv);
8990e307 7461 }
8ebc5c01 7462 return GvCVu(gv);
79072805
LW
7463 }
7464}
7465
c461cf8f
JH
7466/*
7467=for apidoc sv_true
7468
7469Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7470Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7471instead use an in-line version.
c461cf8f
JH
7472
7473=cut
7474*/
7475
79072805 7476I32
864dbfa3 7477Perl_sv_true(pTHX_ register SV *sv)
79072805 7478{
8990e307
LW
7479 if (!sv)
7480 return 0;
79072805 7481 if (SvPOK(sv)) {
823a54a3
AL
7482 register const XPV* const tXpv = (XPV*)SvANY(sv);
7483 if (tXpv &&
c2f1de04 7484 (tXpv->xpv_cur > 1 ||
339049b0 7485 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7486 return 1;
7487 else
7488 return 0;
7489 }
7490 else {
7491 if (SvIOK(sv))
463ee0b2 7492 return SvIVX(sv) != 0;
79072805
LW
7493 else {
7494 if (SvNOK(sv))
463ee0b2 7495 return SvNVX(sv) != 0.0;
79072805 7496 else
463ee0b2 7497 return sv_2bool(sv);
79072805
LW
7498 }
7499 }
7500}
79072805 7501
645c22ef 7502/*
c461cf8f
JH
7503=for apidoc sv_pvn_force
7504
7505Get a sensible string out of the SV somehow.
645c22ef
DM
7506A private implementation of the C<SvPV_force> macro for compilers which
7507can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7508
8d6d96c1
HS
7509=for apidoc sv_pvn_force_flags
7510
7511Get a sensible string out of the SV somehow.
7512If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7513appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7514implemented in terms of this function.
645c22ef
DM
7515You normally want to use the various wrapper macros instead: see
7516C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7517
7518=cut
7519*/
7520
7521char *
7522Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7523{
97aff369 7524 dVAR;
6fc92669 7525 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7526 sv_force_normal_flags(sv, 0);
1c846c1f 7527
a0d0e21e 7528 if (SvPOK(sv)) {
13c5b33c
NC
7529 if (lp)
7530 *lp = SvCUR(sv);
a0d0e21e
LW
7531 }
7532 else {
a3b680e6 7533 char *s;
13c5b33c
NC
7534 STRLEN len;
7535
4d84ee25 7536 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7537 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7538 if (PL_op)
7539 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7540 ref, OP_NAME(PL_op));
4d84ee25 7541 else
b64e5050 7542 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7543 }
b64e5050 7544 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7545 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7546 OP_NAME(PL_op));
b64e5050 7547 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7548 if (lp)
7549 *lp = len;
7550
3f7c398e 7551 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7552 if (SvROK(sv))
7553 sv_unref(sv);
862a34c6 7554 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7555 SvGROW(sv, len + 1);
706aa1c9 7556 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7557 SvCUR_set(sv, len);
7558 *SvEND(sv) = '\0';
7559 }
7560 if (!SvPOK(sv)) {
7561 SvPOK_on(sv); /* validate pointer */
7562 SvTAINT(sv);
1d7c1841 7563 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7564 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7565 }
7566 }
4d84ee25 7567 return SvPVX_mutable(sv);
a0d0e21e
LW
7568}
7569
645c22ef 7570/*
645c22ef
DM
7571=for apidoc sv_pvbyten_force
7572
0feed65a 7573The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7574
7575=cut
7576*/
7577
7340a771
GS
7578char *
7579Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7580{
46ec2f14 7581 sv_pvn_force(sv,lp);
ffebcc3e 7582 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7583 *lp = SvCUR(sv);
7584 return SvPVX(sv);
7340a771
GS
7585}
7586
645c22ef 7587/*
c461cf8f
JH
7588=for apidoc sv_pvutf8n_force
7589
0feed65a 7590The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7591
7592=cut
7593*/
7594
7340a771
GS
7595char *
7596Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7597{
46ec2f14 7598 sv_pvn_force(sv,lp);
560a288e 7599 sv_utf8_upgrade(sv);
46ec2f14
TS
7600 *lp = SvCUR(sv);
7601 return SvPVX(sv);
7340a771
GS
7602}
7603
c461cf8f
JH
7604/*
7605=for apidoc sv_reftype
7606
7607Returns a string describing what the SV is a reference to.
7608
7609=cut
7610*/
7611
1cb0ed9b 7612char *
bfed75c6 7613Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7614{
07409e01
NC
7615 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7616 inside return suggests a const propagation bug in g++. */
c86bf373 7617 if (ob && SvOBJECT(sv)) {
1b6737cc 7618 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7619 return name ? name : (char *) "__ANON__";
c86bf373 7620 }
a0d0e21e
LW
7621 else {
7622 switch (SvTYPE(sv)) {
7623 case SVt_NULL:
7624 case SVt_IV:
7625 case SVt_NV:
7626 case SVt_RV:
7627 case SVt_PV:
7628 case SVt_PVIV:
7629 case SVt_PVNV:
7630 case SVt_PVMG:
7631 case SVt_PVBM:
1cb0ed9b 7632 if (SvVOK(sv))
439cb1c4 7633 return "VSTRING";
a0d0e21e
LW
7634 if (SvROK(sv))
7635 return "REF";
7636 else
7637 return "SCALAR";
1cb0ed9b 7638
07409e01 7639 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7640 /* tied lvalues should appear to be
7641 * scalars for backwards compatitbility */
7642 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7643 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7644 case SVt_PVAV: return "ARRAY";
7645 case SVt_PVHV: return "HASH";
7646 case SVt_PVCV: return "CODE";
7647 case SVt_PVGV: return "GLOB";
1d2dff63 7648 case SVt_PVFM: return "FORMAT";
27f9d8f3 7649 case SVt_PVIO: return "IO";
a0d0e21e
LW
7650 default: return "UNKNOWN";
7651 }
7652 }
7653}
7654
954c1994
GS
7655/*
7656=for apidoc sv_isobject
7657
7658Returns a boolean indicating whether the SV is an RV pointing to a blessed
7659object. If the SV is not an RV, or if the object is not blessed, then this
7660will return false.
7661
7662=cut
7663*/
7664
463ee0b2 7665int
864dbfa3 7666Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7667{
68dc0745 7668 if (!sv)
7669 return 0;
5b295bef 7670 SvGETMAGIC(sv);
85e6fe83
LW
7671 if (!SvROK(sv))
7672 return 0;
7673 sv = (SV*)SvRV(sv);
7674 if (!SvOBJECT(sv))
7675 return 0;
7676 return 1;
7677}
7678
954c1994
GS
7679/*
7680=for apidoc sv_isa
7681
7682Returns a boolean indicating whether the SV is blessed into the specified
7683class. This does not check for subtypes; use C<sv_derived_from> to verify
7684an inheritance relationship.
7685
7686=cut
7687*/
7688
85e6fe83 7689int
864dbfa3 7690Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7691{
bfcb3514 7692 const char *hvname;
68dc0745 7693 if (!sv)
7694 return 0;
5b295bef 7695 SvGETMAGIC(sv);
ed6116ce 7696 if (!SvROK(sv))
463ee0b2 7697 return 0;
ed6116ce
LW
7698 sv = (SV*)SvRV(sv);
7699 if (!SvOBJECT(sv))
463ee0b2 7700 return 0;
bfcb3514
NC
7701 hvname = HvNAME_get(SvSTASH(sv));
7702 if (!hvname)
e27ad1f2 7703 return 0;
463ee0b2 7704
bfcb3514 7705 return strEQ(hvname, name);
463ee0b2
LW
7706}
7707
954c1994
GS
7708/*
7709=for apidoc newSVrv
7710
7711Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7712it will be upgraded to one. If C<classname> is non-null then the new SV will
7713be blessed in the specified package. The new SV is returned and its
7714reference count is 1.
7715
7716=cut
7717*/
7718
463ee0b2 7719SV*
864dbfa3 7720Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7721{
97aff369 7722 dVAR;
463ee0b2
LW
7723 SV *sv;
7724
4561caa4 7725 new_SV(sv);
51cf62d8 7726
765f542d 7727 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7728 SvAMAGIC_off(rv);
51cf62d8 7729
0199fce9 7730 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7731 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7732 SvREFCNT(rv) = 0;
7733 sv_clear(rv);
7734 SvFLAGS(rv) = 0;
7735 SvREFCNT(rv) = refcnt;
0199fce9 7736
dc5494d2
NC
7737 sv_upgrade(rv, SVt_RV);
7738 } else if (SvROK(rv)) {
7739 SvREFCNT_dec(SvRV(rv));
7740 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7741 sv_upgrade(rv, SVt_RV);
7742 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7743 SvPV_free(rv);
0199fce9
JD
7744 SvCUR_set(rv, 0);
7745 SvLEN_set(rv, 0);
7746 }
51cf62d8 7747
0c34ef67 7748 SvOK_off(rv);
b162af07 7749 SvRV_set(rv, sv);
ed6116ce 7750 SvROK_on(rv);
463ee0b2 7751
a0d0e21e 7752 if (classname) {
1b6737cc 7753 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7754 (void)sv_bless(rv, stash);
7755 }
7756 return sv;
7757}
7758
954c1994
GS
7759/*
7760=for apidoc sv_setref_pv
7761
7762Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7763argument will be upgraded to an RV. That RV will be modified to point to
7764the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7765into the SV. The C<classname> argument indicates the package for the
bd61b366 7766blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7767will have a reference count of 1, and the RV will be returned.
954c1994
GS
7768
7769Do not use with other Perl types such as HV, AV, SV, CV, because those
7770objects will become corrupted by the pointer copy process.
7771
7772Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7773
7774=cut
7775*/
7776
a0d0e21e 7777SV*
864dbfa3 7778Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7779{
97aff369 7780 dVAR;
189b2af5 7781 if (!pv) {
3280af22 7782 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7783 SvSETMAGIC(rv);
7784 }
a0d0e21e 7785 else
56431972 7786 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7787 return rv;
7788}
7789
954c1994
GS
7790/*
7791=for apidoc sv_setref_iv
7792
7793Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7794argument will be upgraded to an RV. That RV will be modified to point to
7795the new SV. The C<classname> argument indicates the package for the
bd61b366 7796blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7797will have a reference count of 1, and the RV will be returned.
954c1994
GS
7798
7799=cut
7800*/
7801
a0d0e21e 7802SV*
864dbfa3 7803Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7804{
7805 sv_setiv(newSVrv(rv,classname), iv);
7806 return rv;
7807}
7808
954c1994 7809/*
e1c57cef
JH
7810=for apidoc sv_setref_uv
7811
7812Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7813argument will be upgraded to an RV. That RV will be modified to point to
7814the new SV. The C<classname> argument indicates the package for the
bd61b366 7815blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7816will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7817
7818=cut
7819*/
7820
7821SV*
7822Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7823{
7824 sv_setuv(newSVrv(rv,classname), uv);
7825 return rv;
7826}
7827
7828/*
954c1994
GS
7829=for apidoc sv_setref_nv
7830
7831Copies a double into a new SV, optionally blessing the SV. The C<rv>
7832argument will be upgraded to an RV. That RV will be modified to point to
7833the new SV. The C<classname> argument indicates the package for the
bd61b366 7834blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7835will have a reference count of 1, and the RV will be returned.
954c1994
GS
7836
7837=cut
7838*/
7839
a0d0e21e 7840SV*
65202027 7841Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7842{
7843 sv_setnv(newSVrv(rv,classname), nv);
7844 return rv;
7845}
463ee0b2 7846
954c1994
GS
7847/*
7848=for apidoc sv_setref_pvn
7849
7850Copies a string into a new SV, optionally blessing the SV. The length of the
7851string must be specified with C<n>. The C<rv> argument will be upgraded to
7852an RV. That RV will be modified to point to the new SV. The C<classname>
7853argument indicates the package for the blessing. Set C<classname> to
bd61b366 7854C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7855of 1, and the RV will be returned.
954c1994
GS
7856
7857Note that C<sv_setref_pv> copies the pointer while this copies the string.
7858
7859=cut
7860*/
7861
a0d0e21e 7862SV*
1b6737cc 7863Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7864{
7865 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7866 return rv;
7867}
7868
954c1994
GS
7869/*
7870=for apidoc sv_bless
7871
7872Blesses an SV into a specified package. The SV must be an RV. The package
7873must be designated by its stash (see C<gv_stashpv()>). The reference count
7874of the SV is unaffected.
7875
7876=cut
7877*/
7878
a0d0e21e 7879SV*
864dbfa3 7880Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7881{
97aff369 7882 dVAR;
76e3520e 7883 SV *tmpRef;
a0d0e21e 7884 if (!SvROK(sv))
cea2e8a9 7885 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7886 tmpRef = SvRV(sv);
7887 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7888 if (SvREADONLY(tmpRef))
cea2e8a9 7889 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7890 if (SvOBJECT(tmpRef)) {
7891 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7892 --PL_sv_objcount;
76e3520e 7893 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7894 }
a0d0e21e 7895 }
76e3520e
GS
7896 SvOBJECT_on(tmpRef);
7897 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7898 ++PL_sv_objcount;
862a34c6 7899 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7900 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7901
2e3febc6
CS
7902 if (Gv_AMG(stash))
7903 SvAMAGIC_on(sv);
7904 else
7905 SvAMAGIC_off(sv);
a0d0e21e 7906
1edbfb88
AB
7907 if(SvSMAGICAL(tmpRef))
7908 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7909 mg_set(tmpRef);
7910
7911
ecdeb87c 7912
a0d0e21e
LW
7913 return sv;
7914}
7915
645c22ef 7916/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7917 */
7918
76e3520e 7919STATIC void
cea2e8a9 7920S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7921{
97aff369 7922 dVAR;
850fabdf 7923 void *xpvmg;
b37c2d43 7924 SV * const temp = sv_newmortal();
850fabdf 7925
a0d0e21e
LW
7926 assert(SvTYPE(sv) == SVt_PVGV);
7927 SvFAKE_off(sv);
180488f8
NC
7928 gv_efullname3(temp, (GV *) sv, "*");
7929
f7877b28 7930 if (GvGP(sv)) {
1edc1566 7931 gp_free((GV*)sv);
f7877b28 7932 }
e826b3c7 7933 if (GvSTASH(sv)) {
e15faf7d 7934 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7935 GvSTASH(sv) = NULL;
e826b3c7 7936 }
a5f75d66 7937 GvMULTI_off(sv);
acda4c6a
NC
7938 if (GvNAME_HEK(sv)) {
7939 unshare_hek(GvNAME_HEK(sv));
7940 }
dedf8e73 7941 SvSCREAM_off(sv);
850fabdf
GS
7942
7943 /* need to keep SvANY(sv) in the right arena */
7944 xpvmg = new_XPVMG();
7945 StructCopy(SvANY(sv), xpvmg, XPVMG);
7946 del_XPVGV(SvANY(sv));
7947 SvANY(sv) = xpvmg;
7948
a0d0e21e
LW
7949 SvFLAGS(sv) &= ~SVTYPEMASK;
7950 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7951
7952 /* Intentionally not calling any local SET magic, as this isn't so much a
7953 set operation as merely an internal storage change. */
7954 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7955}
7956
954c1994 7957/*
840a7b70 7958=for apidoc sv_unref_flags
954c1994
GS
7959
7960Unsets the RV status of the SV, and decrements the reference count of
7961whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7962as a reversal of C<newSVrv>. The C<cflags> argument can contain
7963C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7964(otherwise the decrementing is conditional on the reference count being
7965different from one or the reference being a readonly SV).
7889fe52 7966See C<SvROK_off>.
954c1994
GS
7967
7968=cut
7969*/
7970
ed6116ce 7971void
e15faf7d 7972Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7973{
b64e5050 7974 SV* const target = SvRV(ref);
810b8aa5 7975
e15faf7d
NC
7976 if (SvWEAKREF(ref)) {
7977 sv_del_backref(target, ref);
7978 SvWEAKREF_off(ref);
7979 SvRV_set(ref, NULL);
810b8aa5
GS
7980 return;
7981 }
e15faf7d
NC
7982 SvRV_set(ref, NULL);
7983 SvROK_off(ref);
7984 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7985 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7986 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7987 SvREFCNT_dec(target);
840a7b70 7988 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7989 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7990}
8990e307 7991
840a7b70 7992/*
645c22ef
DM
7993=for apidoc sv_untaint
7994
7995Untaint an SV. Use C<SvTAINTED_off> instead.
7996=cut
7997*/
7998
bbce6d69 7999void
864dbfa3 8000Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 8001{
13f57bf8 8002 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 8003 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 8004 if (mg)
565764a8 8005 mg->mg_len &= ~1;
36477c24 8006 }
bbce6d69 8007}
8008
645c22ef
DM
8009/*
8010=for apidoc sv_tainted
8011
8012Test an SV for taintedness. Use C<SvTAINTED> instead.
8013=cut
8014*/
8015
bbce6d69 8016bool
864dbfa3 8017Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8018{
13f57bf8 8019 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8020 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8021 if (mg && (mg->mg_len & 1) )
36477c24 8022 return TRUE;
8023 }
8024 return FALSE;
bbce6d69 8025}
8026
09540bc3
JH
8027/*
8028=for apidoc sv_setpviv
8029
8030Copies an integer into the given SV, also updating its string value.
8031Does not handle 'set' magic. See C<sv_setpviv_mg>.
8032
8033=cut
8034*/
8035
8036void
8037Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8038{
8039 char buf[TYPE_CHARS(UV)];
8040 char *ebuf;
b64e5050 8041 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8042
8043 sv_setpvn(sv, ptr, ebuf - ptr);
8044}
8045
8046/*
8047=for apidoc sv_setpviv_mg
8048
8049Like C<sv_setpviv>, but also handles 'set' magic.
8050
8051=cut
8052*/
8053
8054void
8055Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8056{
df7eb254 8057 sv_setpviv(sv, iv);
09540bc3
JH
8058 SvSETMAGIC(sv);
8059}
8060
cea2e8a9 8061#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8062
8063/* pTHX_ magic can't cope with varargs, so this is a no-context
8064 * version of the main function, (which may itself be aliased to us).
8065 * Don't access this version directly.
8066 */
8067
cea2e8a9
GS
8068void
8069Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8070{
8071 dTHX;
8072 va_list args;
8073 va_start(args, pat);
c5be433b 8074 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8075 va_end(args);
8076}
8077
645c22ef
DM
8078/* pTHX_ magic can't cope with varargs, so this is a no-context
8079 * version of the main function, (which may itself be aliased to us).
8080 * Don't access this version directly.
8081 */
cea2e8a9
GS
8082
8083void
8084Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8085{
8086 dTHX;
8087 va_list args;
8088 va_start(args, pat);
c5be433b 8089 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8090 va_end(args);
cea2e8a9
GS
8091}
8092#endif
8093
954c1994
GS
8094/*
8095=for apidoc sv_setpvf
8096
bffc3d17
SH
8097Works like C<sv_catpvf> but copies the text into the SV instead of
8098appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8099
8100=cut
8101*/
8102
46fc3d4c 8103void
864dbfa3 8104Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8105{
8106 va_list args;
46fc3d4c 8107 va_start(args, pat);
c5be433b 8108 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8109 va_end(args);
8110}
8111
bffc3d17
SH
8112/*
8113=for apidoc sv_vsetpvf
8114
8115Works like C<sv_vcatpvf> but copies the text into the SV instead of
8116appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8117
8118Usually used via its frontend C<sv_setpvf>.
8119
8120=cut
8121*/
645c22ef 8122
c5be433b
GS
8123void
8124Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8125{
4608196e 8126 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8127}
ef50df4b 8128
954c1994
GS
8129/*
8130=for apidoc sv_setpvf_mg
8131
8132Like C<sv_setpvf>, but also handles 'set' magic.
8133
8134=cut
8135*/
8136
ef50df4b 8137void
864dbfa3 8138Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8139{
8140 va_list args;
ef50df4b 8141 va_start(args, pat);
c5be433b 8142 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8143 va_end(args);
c5be433b
GS
8144}
8145
bffc3d17
SH
8146/*
8147=for apidoc sv_vsetpvf_mg
8148
8149Like C<sv_vsetpvf>, but also handles 'set' magic.
8150
8151Usually used via its frontend C<sv_setpvf_mg>.
8152
8153=cut
8154*/
645c22ef 8155
c5be433b
GS
8156void
8157Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8158{
4608196e 8159 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8160 SvSETMAGIC(sv);
8161}
8162
cea2e8a9 8163#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8164
8165/* pTHX_ magic can't cope with varargs, so this is a no-context
8166 * version of the main function, (which may itself be aliased to us).
8167 * Don't access this version directly.
8168 */
8169
cea2e8a9
GS
8170void
8171Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8172{
8173 dTHX;
8174 va_list args;
8175 va_start(args, pat);
c5be433b 8176 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8177 va_end(args);
8178}
8179
645c22ef
DM
8180/* pTHX_ magic can't cope with varargs, so this is a no-context
8181 * version of the main function, (which may itself be aliased to us).
8182 * Don't access this version directly.
8183 */
8184
cea2e8a9
GS
8185void
8186Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8187{
8188 dTHX;
8189 va_list args;
8190 va_start(args, pat);
c5be433b 8191 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8192 va_end(args);
cea2e8a9
GS
8193}
8194#endif
8195
954c1994
GS
8196/*
8197=for apidoc sv_catpvf
8198
d5ce4a7c
GA
8199Processes its arguments like C<sprintf> and appends the formatted
8200output to an SV. If the appended data contains "wide" characters
8201(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8202and characters >255 formatted with %c), the original SV might get
bffc3d17 8203upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8204C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8205valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8206
d5ce4a7c 8207=cut */
954c1994 8208
46fc3d4c 8209void
864dbfa3 8210Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8211{
8212 va_list args;
46fc3d4c 8213 va_start(args, pat);
c5be433b 8214 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8215 va_end(args);
8216}
8217
bffc3d17
SH
8218/*
8219=for apidoc sv_vcatpvf
8220
8221Processes its arguments like C<vsprintf> and appends the formatted output
8222to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8223
8224Usually used via its frontend C<sv_catpvf>.
8225
8226=cut
8227*/
645c22ef 8228
ef50df4b 8229void
c5be433b
GS
8230Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8231{
4608196e 8232 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8233}
8234
954c1994
GS
8235/*
8236=for apidoc sv_catpvf_mg
8237
8238Like C<sv_catpvf>, but also handles 'set' magic.
8239
8240=cut
8241*/
8242
c5be433b 8243void
864dbfa3 8244Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8245{
8246 va_list args;
ef50df4b 8247 va_start(args, pat);
c5be433b 8248 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8249 va_end(args);
c5be433b
GS
8250}
8251
bffc3d17
SH
8252/*
8253=for apidoc sv_vcatpvf_mg
8254
8255Like C<sv_vcatpvf>, but also handles 'set' magic.
8256
8257Usually used via its frontend C<sv_catpvf_mg>.
8258
8259=cut
8260*/
645c22ef 8261
c5be433b
GS
8262void
8263Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8264{
4608196e 8265 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8266 SvSETMAGIC(sv);
8267}
8268
954c1994
GS
8269/*
8270=for apidoc sv_vsetpvfn
8271
bffc3d17 8272Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8273appending it.
8274
bffc3d17 8275Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8276
954c1994
GS
8277=cut
8278*/
8279
46fc3d4c 8280void
7d5ea4e7 8281Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8282{
8283 sv_setpvn(sv, "", 0);
7d5ea4e7 8284 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8285}
8286
2d00ba3b 8287STATIC I32
9dd79c3f 8288S_expect_number(pTHX_ char** pattern)
211dfcf1 8289{
97aff369 8290 dVAR;
211dfcf1
HS
8291 I32 var = 0;
8292 switch (**pattern) {
8293 case '1': case '2': case '3':
8294 case '4': case '5': case '6':
8295 case '7': case '8': case '9':
2fba7546
GA
8296 var = *(*pattern)++ - '0';
8297 while (isDIGIT(**pattern)) {
5f66b61c 8298 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8299 if (tmp < var)
8300 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8301 var = tmp;
8302 }
211dfcf1
HS
8303 }
8304 return var;
8305}
211dfcf1 8306
c445ea15
AL
8307STATIC char *
8308S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8309{
a3b680e6 8310 const int neg = nv < 0;
4151a5fe 8311 UV uv;
4151a5fe
IZ
8312
8313 if (neg)
8314 nv = -nv;
8315 if (nv < UV_MAX) {
b464bac0 8316 char *p = endbuf;
4151a5fe 8317 nv += 0.5;
028f8eaa 8318 uv = (UV)nv;
4151a5fe
IZ
8319 if (uv & 1 && uv == nv)
8320 uv--; /* Round to even */
8321 do {
a3b680e6 8322 const unsigned dig = uv % 10;
4151a5fe
IZ
8323 *--p = '0' + dig;
8324 } while (uv /= 10);
8325 if (neg)
8326 *--p = '-';
8327 *len = endbuf - p;
8328 return p;
8329 }
bd61b366 8330 return NULL;
4151a5fe
IZ
8331}
8332
8333
954c1994
GS
8334/*
8335=for apidoc sv_vcatpvfn
8336
8337Processes its arguments like C<vsprintf> and appends the formatted output
8338to an SV. Uses an array of SVs if the C style variable argument list is
8339missing (NULL). When running with taint checks enabled, indicates via
8340C<maybe_tainted> if results are untrustworthy (often due to the use of
8341locales).
8342
bffc3d17 8343Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8344
954c1994
GS
8345=cut
8346*/
8347
8896765a
RB
8348
8349#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8350 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8351 vec_utf8 = DO_UTF8(vecsv);
8352
1ef29b0e
RGS
8353/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8354
46fc3d4c 8355void
7d5ea4e7 8356Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8357{
97aff369 8358 dVAR;
46fc3d4c 8359 char *p;
8360 char *q;
a3b680e6 8361 const char *patend;
fc36a67e 8362 STRLEN origlen;
46fc3d4c 8363 I32 svix = 0;
27da23d5 8364 static const char nullstr[] = "(null)";
a0714e2c 8365 SV *argsv = NULL;
b464bac0
AL
8366 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8367 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8368 SV *nsv = NULL;
4151a5fe
IZ
8369 /* Times 4: a decimal digit takes more than 3 binary digits.
8370 * NV_DIG: mantissa takes than many decimal digits.
8371 * Plus 32: Playing safe. */
8372 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8373 /* large enough for "%#.#f" --chip */
8374 /* what about long double NVs? --jhi */
db79b45b 8375
53c1dcc0
AL
8376 PERL_UNUSED_ARG(maybe_tainted);
8377
46fc3d4c 8378 /* no matter what, this is a string now */
fc36a67e 8379 (void)SvPV_force(sv, origlen);
46fc3d4c 8380
8896765a 8381 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8382 if (patlen == 0)
8383 return;
0dbb1585 8384 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8385 if (args) {
8386 const char * const s = va_arg(*args, char*);
8387 sv_catpv(sv, s ? s : nullstr);
8388 }
8389 else if (svix < svmax) {
8390 sv_catsv(sv, *svargs);
2d03de9c
AL
8391 }
8392 return;
0dbb1585 8393 }
8896765a
RB
8394 if (args && patlen == 3 && pat[0] == '%' &&
8395 pat[1] == '-' && pat[2] == 'p') {
8396 argsv = va_arg(*args, SV*);
8397 sv_catsv(sv, argsv);
8896765a 8398 return;
46fc3d4c 8399 }
8400
1d917b39 8401#ifndef USE_LONG_DOUBLE
4151a5fe 8402 /* special-case "%.<number>[gf]" */
7af36d83 8403 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8404 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8405 unsigned digits = 0;
8406 const char *pp;
8407
8408 pp = pat + 2;
8409 while (*pp >= '0' && *pp <= '9')
8410 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8411 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8412 NV nv;
8413
7af36d83 8414 if (svix < svmax)
4151a5fe
IZ
8415 nv = SvNV(*svargs);
8416 else
8417 return;
8418 if (*pp == 'g') {
2873255c
NC
8419 /* Add check for digits != 0 because it seems that some
8420 gconverts are buggy in this case, and we don't yet have
8421 a Configure test for this. */
8422 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8423 /* 0, point, slack */
2e59c212 8424 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8425 sv_catpv(sv, ebuf);
8426 if (*ebuf) /* May return an empty string for digits==0 */
8427 return;
8428 }
8429 } else if (!digits) {
8430 STRLEN l;
8431
8432 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8433 sv_catpvn(sv, p, l);
8434 return;
8435 }
8436 }
8437 }
8438 }
1d917b39 8439#endif /* !USE_LONG_DOUBLE */
4151a5fe 8440
2cf2cfc6 8441 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8442 has_utf8 = TRUE;
2cf2cfc6 8443
46fc3d4c 8444 patend = (char*)pat + patlen;
8445 for (p = (char*)pat; p < patend; p = q) {
8446 bool alt = FALSE;
8447 bool left = FALSE;
b22c7a20 8448 bool vectorize = FALSE;
211dfcf1 8449 bool vectorarg = FALSE;
2cf2cfc6 8450 bool vec_utf8 = FALSE;
46fc3d4c 8451 char fill = ' ';
8452 char plus = 0;
8453 char intsize = 0;
8454 STRLEN width = 0;
fc36a67e 8455 STRLEN zeros = 0;
46fc3d4c 8456 bool has_precis = FALSE;
8457 STRLEN precis = 0;
c445ea15 8458 const I32 osvix = svix;
2cf2cfc6 8459 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8460#ifdef HAS_LDBL_SPRINTF_BUG
8461 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8462 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8463 bool fix_ldbl_sprintf_bug = FALSE;
8464#endif
205f51d8 8465
46fc3d4c 8466 char esignbuf[4];
89ebb4a3 8467 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8468 STRLEN esignlen = 0;
8469
bd61b366 8470 const char *eptr = NULL;
fc36a67e 8471 STRLEN elen = 0;
a0714e2c 8472 SV *vecsv = NULL;
4608196e 8473 const U8 *vecstr = NULL;
b22c7a20 8474 STRLEN veclen = 0;
934abaf1 8475 char c = 0;
46fc3d4c 8476 int i;
9c5ffd7c 8477 unsigned base = 0;
8c8eb53c
RB
8478 IV iv = 0;
8479 UV uv = 0;
9e5b023a
JH
8480 /* we need a long double target in case HAS_LONG_DOUBLE but
8481 not USE_LONG_DOUBLE
8482 */
35fff930 8483#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8484 long double nv;
8485#else
65202027 8486 NV nv;
9e5b023a 8487#endif
46fc3d4c 8488 STRLEN have;
8489 STRLEN need;
8490 STRLEN gap;
7af36d83 8491 const char *dotstr = ".";
b22c7a20 8492 STRLEN dotstrlen = 1;
211dfcf1 8493 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8494 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8495 I32 epix = 0; /* explicit precision index */
8496 I32 evix = 0; /* explicit vector index */
eb3fce90 8497 bool asterisk = FALSE;
46fc3d4c 8498
211dfcf1 8499 /* echo everything up to the next format specification */
46fc3d4c 8500 for (q = p; q < patend && *q != '%'; ++q) ;
8501 if (q > p) {
db79b45b
JH
8502 if (has_utf8 && !pat_utf8)
8503 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8504 else
8505 sv_catpvn(sv, p, q - p);
46fc3d4c 8506 p = q;
8507 }
8508 if (q++ >= patend)
8509 break;
8510
211dfcf1
HS
8511/*
8512 We allow format specification elements in this order:
8513 \d+\$ explicit format parameter index
8514 [-+ 0#]+ flags
a472f209 8515 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8516 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8517 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8518 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8519 [hlqLV] size
8896765a
RB
8520 [%bcdefginopsuxDFOUX] format (mandatory)
8521*/
8522
8523 if (args) {
8524/*
8525 As of perl5.9.3, printf format checking is on by default.
8526 Internally, perl uses %p formats to provide an escape to
8527 some extended formatting. This block deals with those
8528 extensions: if it does not match, (char*)q is reset and
8529 the normal format processing code is used.
8530
8531 Currently defined extensions are:
8532 %p include pointer address (standard)
8533 %-p (SVf) include an SV (previously %_)
8534 %-<num>p include an SV with precision <num>
8535 %1p (VDf) include a v-string (as %vd)
8536 %<num>p reserved for future extensions
8537
8538 Robin Barker 2005-07-14
211dfcf1 8539*/
8896765a
RB
8540 char* r = q;
8541 bool sv = FALSE;
8542 STRLEN n = 0;
8543 if (*q == '-')
8544 sv = *q++;
c445ea15 8545 n = expect_number(&q);
8896765a
RB
8546 if (*q++ == 'p') {
8547 if (sv) { /* SVf */
8548 if (n) {
8549 precis = n;
8550 has_precis = TRUE;
8551 }
8552 argsv = va_arg(*args, SV*);
8553 eptr = SvPVx_const(argsv, elen);
8554 if (DO_UTF8(argsv))
8555 is_utf8 = TRUE;
8556 goto string;
8557 }
8558#if vdNUMBER
8559 else if (n == vdNUMBER) { /* VDf */
8560 vectorize = TRUE;
8561 VECTORIZE_ARGS
8562 goto format_vd;
8563 }
8564#endif
8565 else if (n) {
8566 if (ckWARN_d(WARN_INTERNAL))
8567 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8568 "internal %%<num>p might conflict with future printf extensions");
8569 }
8570 }
8571 q = r;
8572 }
8573
c445ea15 8574 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8575 if (*q == '$') {
8576 ++q;
8577 efix = width;
8578 } else {
8579 goto gotwidth;
8580 }
8581 }
8582
fc36a67e 8583 /* FLAGS */
8584
46fc3d4c 8585 while (*q) {
8586 switch (*q) {
8587 case ' ':
8588 case '+':
8589 plus = *q++;
8590 continue;
8591
8592 case '-':
8593 left = TRUE;
8594 q++;
8595 continue;
8596
8597 case '0':
8598 fill = *q++;
8599 continue;
8600
8601 case '#':
8602 alt = TRUE;
8603 q++;
8604 continue;
8605
fc36a67e 8606 default:
8607 break;
8608 }
8609 break;
8610 }
46fc3d4c 8611
211dfcf1 8612 tryasterisk:
eb3fce90 8613 if (*q == '*') {
211dfcf1 8614 q++;
c445ea15 8615 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8616 if (*q++ != '$')
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;
fc36a67e 8726 precis = (i < 0) ? 0 : i;
fc36a67e 8727 }
8728 else {
8729 precis = 0;
8730 while (isDIGIT(*q))
8731 precis = precis * 10 + (*q++ - '0');
8732 }
8733 has_precis = TRUE;
8734 }
46fc3d4c 8735
fc36a67e 8736 /* SIZE */
46fc3d4c 8737
fc36a67e 8738 switch (*q) {
c623ac67
GS
8739#ifdef WIN32
8740 case 'I': /* Ix, I32x, and I64x */
8741# ifdef WIN64
8742 if (q[1] == '6' && q[2] == '4') {
8743 q += 3;
8744 intsize = 'q';
8745 break;
8746 }
8747# endif
8748 if (q[1] == '3' && q[2] == '2') {
8749 q += 3;
8750 break;
8751 }
8752# ifdef WIN64
8753 intsize = 'q';
8754# endif
8755 q++;
8756 break;
8757#endif
9e5b023a 8758#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8759 case 'L': /* Ld */
5f66b61c 8760 /*FALLTHROUGH*/
e5c81feb 8761#ifdef HAS_QUAD
6f9bb7fd 8762 case 'q': /* qd */
9e5b023a 8763#endif
6f9bb7fd
GS
8764 intsize = 'q';
8765 q++;
8766 break;
8767#endif
fc36a67e 8768 case 'l':
9e5b023a 8769#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8770 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8771 intsize = 'q';
8772 q += 2;
46fc3d4c 8773 break;
cf2093f6 8774 }
fc36a67e 8775#endif
5f66b61c 8776 /*FALLTHROUGH*/
fc36a67e 8777 case 'h':
5f66b61c 8778 /*FALLTHROUGH*/
fc36a67e 8779 case 'V':
8780 intsize = *q++;
46fc3d4c 8781 break;
8782 }
8783
fc36a67e 8784 /* CONVERSION */
8785
211dfcf1
HS
8786 if (*q == '%') {
8787 eptr = q++;
8788 elen = 1;
26372e71
GA
8789 if (vectorize) {
8790 c = '%';
8791 goto unknown;
8792 }
211dfcf1
HS
8793 goto string;
8794 }
8795
26372e71 8796 if (!vectorize && !args) {
86c51f8b
NC
8797 if (efix) {
8798 const I32 i = efix-1;
8799 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8800 } else {
8801 argsv = (svix >= 0 && svix < svmax)
8802 ? svargs[svix++] : &PL_sv_undef;
8803 }
863811b2 8804 }
211dfcf1 8805
46fc3d4c 8806 switch (c = *q++) {
8807
8808 /* STRINGS */
8809
46fc3d4c 8810 case 'c':
26372e71
GA
8811 if (vectorize)
8812 goto unknown;
8813 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8814 if ((uv > 255 ||
8815 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8816 && !IN_BYTES) {
dfe13c55 8817 eptr = (char*)utf8buf;
9041c2e3 8818 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8819 is_utf8 = TRUE;
7e2040f0
GS
8820 }
8821 else {
8822 c = (char)uv;
8823 eptr = &c;
8824 elen = 1;
a0ed51b3 8825 }
46fc3d4c 8826 goto string;
8827
46fc3d4c 8828 case 's':
26372e71
GA
8829 if (vectorize)
8830 goto unknown;
8831 if (args) {
fc36a67e 8832 eptr = va_arg(*args, char*);
c635e13b 8833 if (eptr)
1d7c1841
GS
8834#ifdef MACOS_TRADITIONAL
8835 /* On MacOS, %#s format is used for Pascal strings */
8836 if (alt)
8837 elen = *eptr++;
8838 else
8839#endif
c635e13b 8840 elen = strlen(eptr);
8841 else {
27da23d5 8842 eptr = (char *)nullstr;
c635e13b 8843 elen = sizeof nullstr - 1;
8844 }
46fc3d4c 8845 }
211dfcf1 8846 else {
4d84ee25 8847 eptr = SvPVx_const(argsv, elen);
7e2040f0 8848 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8849 if (has_precis && precis < elen) {
8850 I32 p = precis;
7e2040f0 8851 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8852 precis = p;
8853 }
8854 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8855 width += elen - sv_len_utf8(argsv);
a0ed51b3 8856 }
2cf2cfc6 8857 is_utf8 = TRUE;
a0ed51b3
LW
8858 }
8859 }
fc36a67e 8860
46fc3d4c 8861 string:
8862 if (has_precis && elen > precis)
8863 elen = precis;
8864 break;
8865
8866 /* INTEGERS */
8867
fc36a67e 8868 case 'p':
be75b157 8869 if (alt || vectorize)
c2e66d9e 8870 goto unknown;
211dfcf1 8871 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8872 base = 16;
8873 goto integer;
8874
46fc3d4c 8875 case 'D':
29fe7a80 8876#ifdef IV_IS_QUAD
22f3ae8c 8877 intsize = 'q';
29fe7a80 8878#else
46fc3d4c 8879 intsize = 'l';
29fe7a80 8880#endif
5f66b61c 8881 /*FALLTHROUGH*/
46fc3d4c 8882 case 'd':
8883 case 'i':
8896765a
RB
8884#if vdNUMBER
8885 format_vd:
8886#endif
b22c7a20 8887 if (vectorize) {
ba210ebe 8888 STRLEN ulen;
211dfcf1
HS
8889 if (!veclen)
8890 continue;
2cf2cfc6
A
8891 if (vec_utf8)
8892 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8893 UTF8_ALLOW_ANYUV);
b22c7a20 8894 else {
e83d50c9 8895 uv = *vecstr;
b22c7a20
GS
8896 ulen = 1;
8897 }
8898 vecstr += ulen;
8899 veclen -= ulen;
e83d50c9
JP
8900 if (plus)
8901 esignbuf[esignlen++] = plus;
b22c7a20
GS
8902 }
8903 else if (args) {
46fc3d4c 8904 switch (intsize) {
8905 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8906 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8907 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8908 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8909#ifdef HAS_QUAD
8910 case 'q': iv = va_arg(*args, Quad_t); break;
8911#endif
46fc3d4c 8912 }
8913 }
8914 else {
b10c0dba 8915 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8916 switch (intsize) {
b10c0dba
MHM
8917 case 'h': iv = (short)tiv; break;
8918 case 'l': iv = (long)tiv; break;
8919 case 'V':
8920 default: iv = tiv; break;
cf2093f6 8921#ifdef HAS_QUAD
b10c0dba 8922 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8923#endif
46fc3d4c 8924 }
8925 }
e83d50c9
JP
8926 if ( !vectorize ) /* we already set uv above */
8927 {
8928 if (iv >= 0) {
8929 uv = iv;
8930 if (plus)
8931 esignbuf[esignlen++] = plus;
8932 }
8933 else {
8934 uv = -iv;
8935 esignbuf[esignlen++] = '-';
8936 }
46fc3d4c 8937 }
8938 base = 10;
8939 goto integer;
8940
fc36a67e 8941 case 'U':
29fe7a80 8942#ifdef IV_IS_QUAD
22f3ae8c 8943 intsize = 'q';
29fe7a80 8944#else
fc36a67e 8945 intsize = 'l';
29fe7a80 8946#endif
5f66b61c 8947 /*FALLTHROUGH*/
fc36a67e 8948 case 'u':
8949 base = 10;
8950 goto uns_integer;
8951
4f19785b
WSI
8952 case 'b':
8953 base = 2;
8954 goto uns_integer;
8955
46fc3d4c 8956 case 'O':
29fe7a80 8957#ifdef IV_IS_QUAD
22f3ae8c 8958 intsize = 'q';
29fe7a80 8959#else
46fc3d4c 8960 intsize = 'l';
29fe7a80 8961#endif
5f66b61c 8962 /*FALLTHROUGH*/
46fc3d4c 8963 case 'o':
8964 base = 8;
8965 goto uns_integer;
8966
8967 case 'X':
46fc3d4c 8968 case 'x':
8969 base = 16;
46fc3d4c 8970
8971 uns_integer:
b22c7a20 8972 if (vectorize) {
ba210ebe 8973 STRLEN ulen;
b22c7a20 8974 vector:
211dfcf1
HS
8975 if (!veclen)
8976 continue;
2cf2cfc6
A
8977 if (vec_utf8)
8978 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8979 UTF8_ALLOW_ANYUV);
b22c7a20 8980 else {
a05b299f 8981 uv = *vecstr;
b22c7a20
GS
8982 ulen = 1;
8983 }
8984 vecstr += ulen;
8985 veclen -= ulen;
8986 }
8987 else if (args) {
46fc3d4c 8988 switch (intsize) {
8989 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8990 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8991 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8992 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8993#ifdef HAS_QUAD
9e3321a5 8994 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8995#endif
46fc3d4c 8996 }
8997 }
8998 else {
b10c0dba 8999 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 9000 switch (intsize) {
b10c0dba
MHM
9001 case 'h': uv = (unsigned short)tuv; break;
9002 case 'l': uv = (unsigned long)tuv; break;
9003 case 'V':
9004 default: uv = tuv; break;
cf2093f6 9005#ifdef HAS_QUAD
b10c0dba 9006 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 9007#endif
46fc3d4c 9008 }
9009 }
9010
9011 integer:
4d84ee25
NC
9012 {
9013 char *ptr = ebuf + sizeof ebuf;
9014 switch (base) {
9015 unsigned dig;
9016 case 16:
9017 if (!uv)
9018 alt = FALSE;
9019 p = (char*)((c == 'X')
9020 ? "0123456789ABCDEF" : "0123456789abcdef");
9021 do {
9022 dig = uv & 15;
9023 *--ptr = p[dig];
9024 } while (uv >>= 4);
9025 if (alt) {
9026 esignbuf[esignlen++] = '0';
9027 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9028 }
9029 break;
9030 case 8:
9031 do {
9032 dig = uv & 7;
9033 *--ptr = '0' + dig;
9034 } while (uv >>= 3);
9035 if (alt && *ptr != '0')
9036 *--ptr = '0';
9037 break;
9038 case 2:
ed2b91d2
GA
9039 if (!uv)
9040 alt = FALSE;
4d84ee25
NC
9041 do {
9042 dig = uv & 1;
9043 *--ptr = '0' + dig;
9044 } while (uv >>= 1);
9045 if (alt) {
9046 esignbuf[esignlen++] = '0';
9047 esignbuf[esignlen++] = 'b';
9048 }
9049 break;
9050 default: /* it had better be ten or less */
9051 do {
9052 dig = uv % base;
9053 *--ptr = '0' + dig;
9054 } while (uv /= base);
9055 break;
46fc3d4c 9056 }
4d84ee25
NC
9057 elen = (ebuf + sizeof ebuf) - ptr;
9058 eptr = ptr;
9059 if (has_precis) {
9060 if (precis > elen)
9061 zeros = precis - elen;
9062 else if (precis == 0 && elen == 1 && *eptr == '0')
9063 elen = 0;
eda88b6d 9064 }
c10ed8b9 9065 }
46fc3d4c 9066 break;
9067
9068 /* FLOATING POINT */
9069
fc36a67e 9070 case 'F':
9071 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9072 /*FALLTHROUGH*/
46fc3d4c 9073 case 'e': case 'E':
fc36a67e 9074 case 'f':
46fc3d4c 9075 case 'g': case 'G':
26372e71
GA
9076 if (vectorize)
9077 goto unknown;
46fc3d4c 9078
9079 /* This is evil, but floating point is even more evil */
9080
9e5b023a
JH
9081 /* for SV-style calling, we can only get NV
9082 for C-style calling, we assume %f is double;
9083 for simplicity we allow any of %Lf, %llf, %qf for long double
9084 */
9085 switch (intsize) {
9086 case 'V':
9087#if defined(USE_LONG_DOUBLE)
9088 intsize = 'q';
9089#endif
9090 break;
8a2e3f14 9091/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9092 case 'l':
5f66b61c 9093 /*FALLTHROUGH*/
9e5b023a
JH
9094 default:
9095#if defined(USE_LONG_DOUBLE)
9096 intsize = args ? 0 : 'q';
9097#endif
9098 break;
9099 case 'q':
9100#if defined(HAS_LONG_DOUBLE)
9101 break;
9102#else
5f66b61c 9103 /*FALLTHROUGH*/
9e5b023a
JH
9104#endif
9105 case 'h':
9e5b023a
JH
9106 goto unknown;
9107 }
9108
9109 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9110 nv = (args) ?
35fff930
JH
9111#if LONG_DOUBLESIZE > DOUBLESIZE
9112 intsize == 'q' ?
205f51d8
AS
9113 va_arg(*args, long double) :
9114 va_arg(*args, double)
35fff930 9115#else
205f51d8 9116 va_arg(*args, double)
35fff930 9117#endif
9e5b023a 9118 : SvNVx(argsv);
fc36a67e 9119
9120 need = 0;
9121 if (c != 'e' && c != 'E') {
9122 i = PERL_INT_MIN;
9e5b023a
JH
9123 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9124 will cast our (long double) to (double) */
73b309ea 9125 (void)Perl_frexp(nv, &i);
fc36a67e 9126 if (i == PERL_INT_MIN)
cea2e8a9 9127 Perl_die(aTHX_ "panic: frexp");
c635e13b 9128 if (i > 0)
fc36a67e 9129 need = BIT_DIGITS(i);
9130 }
9131 need += has_precis ? precis : 6; /* known default */
20f6aaab 9132
fc36a67e 9133 if (need < width)
9134 need = width;
9135
20f6aaab
AS
9136#ifdef HAS_LDBL_SPRINTF_BUG
9137 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9138 with sfio - Allen <allens@cpan.org> */
9139
9140# ifdef DBL_MAX
9141# define MY_DBL_MAX DBL_MAX
9142# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9143# if DOUBLESIZE >= 8
9144# define MY_DBL_MAX 1.7976931348623157E+308L
9145# else
9146# define MY_DBL_MAX 3.40282347E+38L
9147# endif
9148# endif
9149
9150# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9151# define MY_DBL_MAX_BUG 1L
20f6aaab 9152# else
205f51d8 9153# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9154# endif
20f6aaab 9155
205f51d8
AS
9156# ifdef DBL_MIN
9157# define MY_DBL_MIN DBL_MIN
9158# else /* XXX guessing! -Allen */
9159# if DOUBLESIZE >= 8
9160# define MY_DBL_MIN 2.2250738585072014E-308L
9161# else
9162# define MY_DBL_MIN 1.17549435E-38L
9163# endif
9164# endif
20f6aaab 9165
205f51d8
AS
9166 if ((intsize == 'q') && (c == 'f') &&
9167 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9168 (need < DBL_DIG)) {
9169 /* it's going to be short enough that
9170 * long double precision is not needed */
9171
9172 if ((nv <= 0L) && (nv >= -0L))
9173 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9174 else {
9175 /* would use Perl_fp_class as a double-check but not
9176 * functional on IRIX - see perl.h comments */
9177
9178 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9179 /* It's within the range that a double can represent */
9180#if defined(DBL_MAX) && !defined(DBL_MIN)
9181 if ((nv >= ((long double)1/DBL_MAX)) ||
9182 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9183#endif
205f51d8 9184 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9185 }
205f51d8
AS
9186 }
9187 if (fix_ldbl_sprintf_bug == TRUE) {
9188 double temp;
9189
9190 intsize = 0;
9191 temp = (double)nv;
9192 nv = (NV)temp;
9193 }
20f6aaab 9194 }
205f51d8
AS
9195
9196# undef MY_DBL_MAX
9197# undef MY_DBL_MAX_BUG
9198# undef MY_DBL_MIN
9199
20f6aaab
AS
9200#endif /* HAS_LDBL_SPRINTF_BUG */
9201
46fc3d4c 9202 need += 20; /* fudge factor */
80252599
GS
9203 if (PL_efloatsize < need) {
9204 Safefree(PL_efloatbuf);
9205 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9206 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9207 PL_efloatbuf[0] = '\0';
46fc3d4c 9208 }
9209
4151a5fe
IZ
9210 if ( !(width || left || plus || alt) && fill != '0'
9211 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9212 /* See earlier comment about buggy Gconvert when digits,
9213 aka precis is 0 */
9214 if ( c == 'g' && precis) {
2e59c212 9215 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9216 /* May return an empty string for digits==0 */
9217 if (*PL_efloatbuf) {
9218 elen = strlen(PL_efloatbuf);
4151a5fe 9219 goto float_converted;
4150c189 9220 }
4151a5fe
IZ
9221 } else if ( c == 'f' && !precis) {
9222 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9223 break;
9224 }
9225 }
4d84ee25
NC
9226 {
9227 char *ptr = ebuf + sizeof ebuf;
9228 *--ptr = '\0';
9229 *--ptr = c;
9230 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9231#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9232 if (intsize == 'q') {
9233 /* Copy the one or more characters in a long double
9234 * format before the 'base' ([efgEFG]) character to
9235 * the format string. */
9236 static char const prifldbl[] = PERL_PRIfldbl;
9237 char const *p = prifldbl + sizeof(prifldbl) - 3;
9238 while (p >= prifldbl) { *--ptr = *p--; }
9239 }
65202027 9240#endif
4d84ee25
NC
9241 if (has_precis) {
9242 base = precis;
9243 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9244 *--ptr = '.';
9245 }
9246 if (width) {
9247 base = width;
9248 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9249 }
9250 if (fill == '0')
9251 *--ptr = fill;
9252 if (left)
9253 *--ptr = '-';
9254 if (plus)
9255 *--ptr = plus;
9256 if (alt)
9257 *--ptr = '#';
9258 *--ptr = '%';
9259
9260 /* No taint. Otherwise we are in the strange situation
9261 * where printf() taints but print($float) doesn't.
9262 * --jhi */
9e5b023a 9263#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9264 elen = ((intsize == 'q')
9265 ? my_sprintf(PL_efloatbuf, ptr, nv)
9266 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9267#else
4150c189 9268 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9269#endif
4d84ee25 9270 }
4151a5fe 9271 float_converted:
80252599 9272 eptr = PL_efloatbuf;
46fc3d4c 9273 break;
9274
fc36a67e 9275 /* SPECIAL */
9276
9277 case 'n':
26372e71
GA
9278 if (vectorize)
9279 goto unknown;
fc36a67e 9280 i = SvCUR(sv) - origlen;
26372e71 9281 if (args) {
c635e13b 9282 switch (intsize) {
9283 case 'h': *(va_arg(*args, short*)) = i; break;
9284 default: *(va_arg(*args, int*)) = i; break;
9285 case 'l': *(va_arg(*args, long*)) = i; break;
9286 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9287#ifdef HAS_QUAD
9288 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9289#endif
c635e13b 9290 }
fc36a67e 9291 }
9dd79c3f 9292 else
211dfcf1 9293 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9294 continue; /* not "break" */
9295
9296 /* UNKNOWN */
9297
46fc3d4c 9298 default:
fc36a67e 9299 unknown:
041457d9
DM
9300 if (!args
9301 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9302 && ckWARN(WARN_PRINTF))
9303 {
c4420975 9304 SV * const msg = sv_newmortal();
35c1215d
NC
9305 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9306 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9307 if (c) {
0f4b6630 9308 if (isPRINT(c))
1c846c1f 9309 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9310 "\"%%%c\"", c & 0xFF);
9311 else
9312 Perl_sv_catpvf(aTHX_ msg,
57def98f 9313 "\"%%\\%03"UVof"\"",
0f4b6630 9314 (UV)c & 0xFF);
0f4b6630 9315 } else
396482e1 9316 sv_catpvs(msg, "end of string");
9014280d 9317 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9318 }
fb73857a 9319
9320 /* output mangled stuff ... */
9321 if (c == '\0')
9322 --q;
46fc3d4c 9323 eptr = p;
9324 elen = q - p;
fb73857a 9325
9326 /* ... right here, because formatting flags should not apply */
9327 SvGROW(sv, SvCUR(sv) + elen + 1);
9328 p = SvEND(sv);
4459522c 9329 Copy(eptr, p, elen, char);
fb73857a 9330 p += elen;
9331 *p = '\0';
3f7c398e 9332 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9333 svix = osvix;
fb73857a 9334 continue; /* not "break" */
46fc3d4c 9335 }
9336
6c94ec8b
HS
9337 /* calculate width before utf8_upgrade changes it */
9338 have = esignlen + zeros + elen;
ed2b91d2
GA
9339 if (have < zeros)
9340 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9341
d2876be5
JH
9342 if (is_utf8 != has_utf8) {
9343 if (is_utf8) {
9344 if (SvCUR(sv))
9345 sv_utf8_upgrade(sv);
9346 }
9347 else {
53c1dcc0 9348 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9349 sv_utf8_upgrade(nsv);
93524f2b 9350 eptr = SvPVX_const(nsv);
d2876be5
JH
9351 elen = SvCUR(nsv);
9352 }
9353 SvGROW(sv, SvCUR(sv) + elen + 1);
9354 p = SvEND(sv);
9355 *p = '\0';
9356 }
6af65485 9357
46fc3d4c 9358 need = (have > width ? have : width);
9359 gap = need - have;
9360
d2641cbd
PC
9361 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9362 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9363 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9364 p = SvEND(sv);
9365 if (esignlen && fill == '0') {
53c1dcc0 9366 int i;
eb160463 9367 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9368 *p++ = esignbuf[i];
9369 }
9370 if (gap && !left) {
9371 memset(p, fill, gap);
9372 p += gap;
9373 }
9374 if (esignlen && fill != '0') {
53c1dcc0 9375 int i;
eb160463 9376 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9377 *p++ = esignbuf[i];
9378 }
fc36a67e 9379 if (zeros) {
53c1dcc0 9380 int i;
fc36a67e 9381 for (i = zeros; i; i--)
9382 *p++ = '0';
9383 }
46fc3d4c 9384 if (elen) {
4459522c 9385 Copy(eptr, p, elen, char);
46fc3d4c 9386 p += elen;
9387 }
9388 if (gap && left) {
9389 memset(p, ' ', gap);
9390 p += gap;
9391 }
b22c7a20
GS
9392 if (vectorize) {
9393 if (veclen) {
4459522c 9394 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9395 p += dotstrlen;
9396 }
9397 else
9398 vectorize = FALSE; /* done iterating over vecstr */
9399 }
2cf2cfc6
A
9400 if (is_utf8)
9401 has_utf8 = TRUE;
9402 if (has_utf8)
7e2040f0 9403 SvUTF8_on(sv);
46fc3d4c 9404 *p = '\0';
3f7c398e 9405 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9406 if (vectorize) {
9407 esignlen = 0;
9408 goto vector;
9409 }
46fc3d4c 9410 }
9411}
51371543 9412
645c22ef
DM
9413/* =========================================================================
9414
9415=head1 Cloning an interpreter
9416
9417All the macros and functions in this section are for the private use of
9418the main function, perl_clone().
9419
9420The foo_dup() functions make an exact copy of an existing foo thinngy.
9421During the course of a cloning, a hash table is used to map old addresses
9422to new addresses. The table is created and manipulated with the
9423ptr_table_* functions.
9424
9425=cut
9426
9427============================================================================*/
9428
9429
1d7c1841
GS
9430#if defined(USE_ITHREADS)
9431
d4c19fe8 9432/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9433#ifndef GpREFCNT_inc
9434# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9435#endif
9436
9437
a41cc44e
NC
9438/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
9439 that currently av_dup and hv_dup are the same as sv_dup. If this changes,
9440 please unmerge ss_dup. */
d2d73c3e 9441#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9442#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9443#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9444#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9445#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9446#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9447#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9448#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9449#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9450#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9451#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9452#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9453#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9454#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9455
d2d73c3e 9456
d2f185dc
AMS
9457/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9458 regcomp.c. AMS 20010712 */
645c22ef 9459
1d7c1841 9460REGEXP *
53c1dcc0 9461Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9462{
27da23d5 9463 dVAR;
d2f185dc
AMS
9464 REGEXP *ret;
9465 int i, len, npar;
9466 struct reg_substr_datum *s;
9467
9468 if (!r)
9469 return (REGEXP *)NULL;
9470
9471 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9472 return ret;
9473
9474 len = r->offsets[0];
9475 npar = r->nparens+1;
9476
a02a5408 9477 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9478 Copy(r->program, ret->program, len+1, regnode);
9479
a02a5408 9480 Newx(ret->startp, npar, I32);
d2f185dc 9481 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9482 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9483 Copy(r->startp, ret->startp, npar, I32);
9484
a02a5408 9485 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9486 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9487 s->min_offset = r->substrs->data[i].min_offset;
9488 s->max_offset = r->substrs->data[i].max_offset;
9489 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9490 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9491 }
9492
70612e96 9493 ret->regstclass = NULL;
d2f185dc
AMS
9494 if (r->data) {
9495 struct reg_data *d;
e1ec3a88 9496 const int count = r->data->count;
53c1dcc0 9497 int i;
d2f185dc 9498
a02a5408 9499 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9500 char, struct reg_data);
a02a5408 9501 Newx(d->what, count, U8);
d2f185dc
AMS
9502
9503 d->count = count;
9504 for (i = 0; i < count; i++) {
9505 d->what[i] = r->data->what[i];
9506 switch (d->what[i]) {
a3621e74
YO
9507 /* legal options are one of: sfpont
9508 see also regcomp.h and pregfree() */
d2f185dc
AMS
9509 case 's':
9510 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9511 break;
9512 case 'p':
9513 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9514 break;
9515 case 'f':
9516 /* This is cheating. */
a02a5408 9517 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9518 StructCopy(r->data->data[i], d->data[i],
9519 struct regnode_charclass_class);
70612e96 9520 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9521 break;
9522 case 'o':
33773810
AMS
9523 /* Compiled op trees are readonly, and can thus be
9524 shared without duplication. */
b34c0dd4 9525 OP_REFCNT_LOCK;
9b978d73 9526 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9527 OP_REFCNT_UNLOCK;
9b978d73 9528 break;
d2f185dc
AMS
9529 case 'n':
9530 d->data[i] = r->data->data[i];
9531 break;
a3621e74
YO
9532 case 't':
9533 d->data[i] = r->data->data[i];
9534 OP_REFCNT_LOCK;
9535 ((reg_trie_data*)d->data[i])->refcount++;
9536 OP_REFCNT_UNLOCK;
9537 break;
9538 default:
9539 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9540 }
9541 }
9542
9543 ret->data = d;
9544 }
9545 else
9546 ret->data = NULL;
9547
a02a5408 9548 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9549 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9550
e01c5899 9551 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9552 ret->refcnt = r->refcnt;
9553 ret->minlen = r->minlen;
9554 ret->prelen = r->prelen;
9555 ret->nparens = r->nparens;
9556 ret->lastparen = r->lastparen;
9557 ret->lastcloseparen = r->lastcloseparen;
9558 ret->reganch = r->reganch;
9559
70612e96
RG
9560 ret->sublen = r->sublen;
9561
9562 if (RX_MATCH_COPIED(ret))
e01c5899 9563 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9564 else
bd61b366 9565 ret->subbeg = NULL;
f8c7b90f 9566#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9567 ret->saved_copy = NULL;
9a26048b 9568#endif
70612e96 9569
d2f185dc
AMS
9570 ptr_table_store(PL_ptr_table, r, ret);
9571 return ret;
1d7c1841
GS
9572}
9573
d2d73c3e 9574/* duplicate a file handle */
645c22ef 9575
1d7c1841 9576PerlIO *
a8fc9800 9577Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9578{
9579 PerlIO *ret;
53c1dcc0
AL
9580
9581 PERL_UNUSED_ARG(type);
73d840c0 9582
1d7c1841
GS
9583 if (!fp)
9584 return (PerlIO*)NULL;
9585
9586 /* look for it in the table first */
9587 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9588 if (ret)
9589 return ret;
9590
9591 /* create anew and remember what it is */
ecdeb87c 9592 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9593 ptr_table_store(PL_ptr_table, fp, ret);
9594 return ret;
9595}
9596
645c22ef
DM
9597/* duplicate a directory handle */
9598
1d7c1841
GS
9599DIR *
9600Perl_dirp_dup(pTHX_ DIR *dp)
9601{
96a5add6 9602 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9603 if (!dp)
9604 return (DIR*)NULL;
9605 /* XXX TODO */
9606 return dp;
9607}
9608
ff276b08 9609/* duplicate a typeglob */
645c22ef 9610
1d7c1841 9611GP *
a8fc9800 9612Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9613{
9614 GP *ret;
b37c2d43 9615
1d7c1841
GS
9616 if (!gp)
9617 return (GP*)NULL;
9618 /* look for it in the table first */
9619 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9620 if (ret)
9621 return ret;
9622
9623 /* create anew and remember what it is */
a02a5408 9624 Newxz(ret, 1, GP);
1d7c1841
GS
9625 ptr_table_store(PL_ptr_table, gp, ret);
9626
9627 /* clone */
9628 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9629 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9630 ret->gp_io = io_dup_inc(gp->gp_io, param);
9631 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9632 ret->gp_av = av_dup_inc(gp->gp_av, param);
9633 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9634 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9635 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9636 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9637 ret->gp_line = gp->gp_line;
9638 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9639 return ret;
9640}
9641
645c22ef
DM
9642/* duplicate a chain of magic */
9643
1d7c1841 9644MAGIC *
a8fc9800 9645Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9646{
cb359b41
JH
9647 MAGIC *mgprev = (MAGIC*)NULL;
9648 MAGIC *mgret;
1d7c1841
GS
9649 if (!mg)
9650 return (MAGIC*)NULL;
9651 /* look for it in the table first */
9652 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9653 if (mgret)
9654 return mgret;
9655
9656 for (; mg; mg = mg->mg_moremagic) {
9657 MAGIC *nmg;
a02a5408 9658 Newxz(nmg, 1, MAGIC);
cb359b41 9659 if (mgprev)
1d7c1841 9660 mgprev->mg_moremagic = nmg;
cb359b41
JH
9661 else
9662 mgret = nmg;
1d7c1841
GS
9663 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9664 nmg->mg_private = mg->mg_private;
9665 nmg->mg_type = mg->mg_type;
9666 nmg->mg_flags = mg->mg_flags;
14befaf4 9667 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9668 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9669 }
05bd4103 9670 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9671 /* The backref AV has its reference count deliberately bumped by
9672 1. */
9673 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9674 }
8d2f4536
NC
9675 else if (mg->mg_type == PERL_MAGIC_symtab) {
9676 nmg->mg_obj = mg->mg_obj;
9677 }
1d7c1841
GS
9678 else {
9679 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9680 ? sv_dup_inc(mg->mg_obj, param)
9681 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9682 }
9683 nmg->mg_len = mg->mg_len;
9684 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9685 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9686 if (mg->mg_len > 0) {
1d7c1841 9687 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9688 if (mg->mg_type == PERL_MAGIC_overload_table &&
9689 AMT_AMAGIC((AMT*)mg->mg_ptr))
9690 {
c445ea15 9691 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9692 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9693 I32 i;
9694 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9695 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9696 }
9697 }
9698 }
9699 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9700 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9701 }
68795e93
NIS
9702 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9703 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9704 }
1d7c1841
GS
9705 mgprev = nmg;
9706 }
9707 return mgret;
9708}
9709
645c22ef
DM
9710/* create a new pointer-mapping table */
9711
1d7c1841
GS
9712PTR_TBL_t *
9713Perl_ptr_table_new(pTHX)
9714{
9715 PTR_TBL_t *tbl;
96a5add6
AL
9716 PERL_UNUSED_CONTEXT;
9717
a02a5408 9718 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9719 tbl->tbl_max = 511;
9720 tbl->tbl_items = 0;
a02a5408 9721 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9722 return tbl;
9723}
9724
7119fd33
NC
9725#define PTR_TABLE_HASH(ptr) \
9726 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9727
93e68bfb
JC
9728/*
9729 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9730 following define) and at call to new_body_inline made below in
9731 Perl_ptr_table_store()
9732 */
9733
9734#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9735
645c22ef
DM
9736/* map an existing pointer using a table */
9737
7bf61b54 9738STATIC PTR_TBL_ENT_t *
b0e6ae5b 9739S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9740 PTR_TBL_ENT_t *tblent;
4373e329 9741 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9742 assert(tbl);
9743 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9744 for (; tblent; tblent = tblent->next) {
9745 if (tblent->oldval == sv)
7bf61b54 9746 return tblent;
1d7c1841 9747 }
d4c19fe8 9748 return NULL;
7bf61b54
NC
9749}
9750
9751void *
9752Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9753{
b0e6ae5b 9754 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9755 PERL_UNUSED_CONTEXT;
d4c19fe8 9756 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9757}
9758
645c22ef
DM
9759/* add a new entry to a pointer-mapping table */
9760
1d7c1841 9761void
44f8325f 9762Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9763{
0c9fdfe0 9764 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9765 PERL_UNUSED_CONTEXT;
1d7c1841 9766
7bf61b54
NC
9767 if (tblent) {
9768 tblent->newval = newsv;
9769 } else {
9770 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9771
d2a0f284
JC
9772 new_body_inline(tblent, PTE_SVSLOT);
9773
7bf61b54
NC
9774 tblent->oldval = oldsv;
9775 tblent->newval = newsv;
9776 tblent->next = tbl->tbl_ary[entry];
9777 tbl->tbl_ary[entry] = tblent;
9778 tbl->tbl_items++;
9779 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9780 ptr_table_split(tbl);
1d7c1841 9781 }
1d7c1841
GS
9782}
9783
645c22ef
DM
9784/* double the hash bucket size of an existing ptr table */
9785
1d7c1841
GS
9786void
9787Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9788{
9789 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9790 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9791 UV newsize = oldsize * 2;
9792 UV i;
96a5add6 9793 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9794
9795 Renew(ary, newsize, PTR_TBL_ENT_t*);
9796 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9797 tbl->tbl_max = --newsize;
9798 tbl->tbl_ary = ary;
9799 for (i=0; i < oldsize; i++, ary++) {
9800 PTR_TBL_ENT_t **curentp, **entp, *ent;
9801 if (!*ary)
9802 continue;
9803 curentp = ary + oldsize;
9804 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9805 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9806 *entp = ent->next;
9807 ent->next = *curentp;
9808 *curentp = ent;
9809 continue;
9810 }
9811 else
9812 entp = &ent->next;
9813 }
9814 }
9815}
9816
645c22ef
DM
9817/* remove all the entries from a ptr table */
9818
a0739874
DM
9819void
9820Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9821{
d5cefff9 9822 if (tbl && tbl->tbl_items) {
c445ea15 9823 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9824 UV riter = tbl->tbl_max;
a0739874 9825
d5cefff9
NC
9826 do {
9827 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9828
d5cefff9 9829 while (entry) {
00b6aa41 9830 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9831 entry = entry->next;
9832 del_pte(oentry);
9833 }
9834 } while (riter--);
a0739874 9835
d5cefff9
NC
9836 tbl->tbl_items = 0;
9837 }
a0739874
DM
9838}
9839
645c22ef
DM
9840/* clear and free a ptr table */
9841
a0739874
DM
9842void
9843Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9844{
9845 if (!tbl) {
9846 return;
9847 }
9848 ptr_table_clear(tbl);
9849 Safefree(tbl->tbl_ary);
9850 Safefree(tbl);
9851}
9852
5bd07a3d 9853
83841fad 9854void
eb86f8b3 9855Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9856{
9857 if (SvROK(sstr)) {
b162af07
SP
9858 SvRV_set(dstr, SvWEAKREF(sstr)
9859 ? sv_dup(SvRV(sstr), param)
9860 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9861
83841fad 9862 }
3f7c398e 9863 else if (SvPVX_const(sstr)) {
83841fad
NIS
9864 /* Has something there */
9865 if (SvLEN(sstr)) {
68795e93 9866 /* Normal PV - clone whole allocated space */
3f7c398e 9867 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9868 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9869 /* Not that normal - actually sstr is copy on write.
9870 But we are a true, independant SV, so: */
9871 SvREADONLY_off(dstr);
9872 SvFAKE_off(dstr);
9873 }
68795e93 9874 }
83841fad
NIS
9875 else {
9876 /* Special case - not normally malloced for some reason */
f7877b28
NC
9877 if (isGV_with_GP(sstr)) {
9878 /* Don't need to do anything here. */
9879 }
9880 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9881 /* A "shared" PV - clone it as "shared" PV */
9882 SvPV_set(dstr,
9883 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9884 param)));
83841fad
NIS
9885 }
9886 else {
9887 /* Some other special case - random pointer */
f880fe2f 9888 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9889 }
83841fad
NIS
9890 }
9891 }
9892 else {
4608196e 9893 /* Copy the NULL */
f880fe2f 9894 if (SvTYPE(dstr) == SVt_RV)
b162af07 9895 SvRV_set(dstr, NULL);
f880fe2f 9896 else
6136c704 9897 SvPV_set(dstr, NULL);
83841fad
NIS
9898 }
9899}
9900
662fb8b2
NC
9901/* duplicate an SV of any type (including AV, HV etc) */
9902
1d7c1841 9903SV *
eb86f8b3 9904Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9905{
27da23d5 9906 dVAR;
1d7c1841
GS
9907 SV *dstr;
9908
9909 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9910 return NULL;
1d7c1841
GS
9911 /* look for it in the table first */
9912 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9913 if (dstr)
9914 return dstr;
9915
0405e91e
AB
9916 if(param->flags & CLONEf_JOIN_IN) {
9917 /** We are joining here so we don't want do clone
9918 something that is bad **/
eb86f8b3
AL
9919 if (SvTYPE(sstr) == SVt_PVHV) {
9920 const char * const hvname = HvNAME_get(sstr);
9921 if (hvname)
9922 /** don't clone stashes if they already exist **/
9923 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9924 }
9925 }
9926
1d7c1841
GS
9927 /* create anew and remember what it is */
9928 new_SV(dstr);
fd0854ff
DM
9929
9930#ifdef DEBUG_LEAKING_SCALARS
9931 dstr->sv_debug_optype = sstr->sv_debug_optype;
9932 dstr->sv_debug_line = sstr->sv_debug_line;
9933 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9934 dstr->sv_debug_cloned = 1;
fd0854ff 9935 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9936#endif
9937
1d7c1841
GS
9938 ptr_table_store(PL_ptr_table, sstr, dstr);
9939
9940 /* clone */
9941 SvFLAGS(dstr) = SvFLAGS(sstr);
9942 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9943 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9944
9945#ifdef DEBUGGING
3f7c398e 9946 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9947 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9948 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9949#endif
9950
9660f481
DM
9951 /* don't clone objects whose class has asked us not to */
9952 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9953 SvFLAGS(dstr) &= ~SVTYPEMASK;
9954 SvOBJECT_off(dstr);
9955 return dstr;
9956 }
9957
1d7c1841
GS
9958 switch (SvTYPE(sstr)) {
9959 case SVt_NULL:
9960 SvANY(dstr) = NULL;
9961 break;
9962 case SVt_IV:
339049b0 9963 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9964 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9965 break;
9966 case SVt_NV:
9967 SvANY(dstr) = new_XNV();
9d6ce603 9968 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9969 break;
9970 case SVt_RV:
339049b0 9971 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9972 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9973 break;
662fb8b2
NC
9974 default:
9975 {
9976 /* These are all the types that need complex bodies allocating. */
662fb8b2 9977 void *new_body;
2bcc16b3
NC
9978 const svtype sv_type = SvTYPE(sstr);
9979 const struct body_details *const sv_type_details
9980 = bodies_by_type + sv_type;
662fb8b2 9981
93e68bfb 9982 switch (sv_type) {
662fb8b2 9983 default:
bb263b4e 9984 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9985 break;
9986
662fb8b2
NC
9987 case SVt_PVGV:
9988 if (GvUNIQUE((GV*)sstr)) {
6f207bd3 9989 NOOP; /* Do sharing here, and fall through */
662fb8b2 9990 }
c22188b4
NC
9991 case SVt_PVIO:
9992 case SVt_PVFM:
9993 case SVt_PVHV:
9994 case SVt_PVAV:
93e68bfb 9995 case SVt_PVBM:
662fb8b2 9996 case SVt_PVCV:
662fb8b2 9997 case SVt_PVLV:
662fb8b2 9998 case SVt_PVMG:
662fb8b2 9999 case SVt_PVNV:
662fb8b2 10000 case SVt_PVIV:
662fb8b2 10001 case SVt_PV:
d2a0f284 10002 assert(sv_type_details->body_size);
c22188b4 10003 if (sv_type_details->arena) {
d2a0f284 10004 new_body_inline(new_body, sv_type);
c22188b4 10005 new_body
b9502f15 10006 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
10007 } else {
10008 new_body = new_NOARENA(sv_type_details);
10009 }
1d7c1841 10010 }
662fb8b2
NC
10011 assert(new_body);
10012 SvANY(dstr) = new_body;
10013
2bcc16b3 10014#ifndef PURIFY
b9502f15
NC
10015 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10016 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 10017 sv_type_details->copy, char);
2bcc16b3
NC
10018#else
10019 Copy(((char*)SvANY(sstr)),
10020 ((char*)SvANY(dstr)),
d2a0f284 10021 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10022#endif
662fb8b2 10023
f7877b28
NC
10024 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10025 && !isGV_with_GP(dstr))
662fb8b2
NC
10026 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10027
10028 /* The Copy above means that all the source (unduplicated) pointers
10029 are now in the destination. We can check the flags and the
10030 pointers in either, but it's possible that there's less cache
10031 missing by always going for the destination.
10032 FIXME - instrument and check that assumption */
f32993d6 10033 if (sv_type >= SVt_PVMG) {
e736a858 10034 HV *ourstash;
2892acdb 10035 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
10036 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
10037 } else if (SvMAGIC(dstr))
662fb8b2
NC
10038 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10039 if (SvSTASH(dstr))
10040 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10041 }
662fb8b2 10042
f32993d6
NC
10043 /* The cast silences a GCC warning about unhandled types. */
10044 switch ((int)sv_type) {
662fb8b2
NC
10045 case SVt_PV:
10046 break;
10047 case SVt_PVIV:
10048 break;
10049 case SVt_PVNV:
10050 break;
10051 case SVt_PVMG:
10052 break;
10053 case SVt_PVBM:
10054 break;
10055 case SVt_PVLV:
10056 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10057 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10058 LvTARG(dstr) = dstr;
10059 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10060 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10061 else
10062 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10063 break;
10064 case SVt_PVGV:
acda4c6a
NC
10065 if (GvNAME_HEK(dstr))
10066 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 10067
e15faf7d
NC
10068 /* Don't call sv_add_backref here as it's going to be created
10069 as part of the magic cloning of the symbol table. */
f7877b28
NC
10070 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10071 if(isGV_with_GP(sstr)) {
10072 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10073 at the point of this comment. */
10074 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10075 (void)GpREFCNT_inc(GvGP(dstr));
10076 } else
10077 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10078 break;
10079 case SVt_PVIO:
10080 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10081 if (IoOFP(dstr) == IoIFP(sstr))
10082 IoOFP(dstr) = IoIFP(dstr);
10083 else
10084 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10085 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10086 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10087 /* I have no idea why fake dirp (rsfps)
10088 should be treated differently but otherwise
10089 we end up with leaks -- sky*/
10090 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10091 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10092 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10093 } else {
10094 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10095 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10096 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10097 if (IoDIRP(dstr)) {
10098 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10099 } else {
6f207bd3 10100 NOOP;
100ce7e1
NC
10101 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10102 }
662fb8b2
NC
10103 }
10104 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10105 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10106 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10107 break;
10108 case SVt_PVAV:
10109 if (AvARRAY((AV*)sstr)) {
10110 SV **dst_ary, **src_ary;
10111 SSize_t items = AvFILLp((AV*)sstr) + 1;
10112
10113 src_ary = AvARRAY((AV*)sstr);
a02a5408 10114 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10115 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10116 SvPV_set(dstr, (char*)dst_ary);
10117 AvALLOC((AV*)dstr) = dst_ary;
10118 if (AvREAL((AV*)sstr)) {
10119 while (items-- > 0)
10120 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10121 }
10122 else {
10123 while (items-- > 0)
10124 *dst_ary++ = sv_dup(*src_ary++, param);
10125 }
10126 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10127 while (items-- > 0) {
10128 *dst_ary++ = &PL_sv_undef;
10129 }
bfcb3514 10130 }
662fb8b2 10131 else {
bd61b366 10132 SvPV_set(dstr, NULL);
662fb8b2 10133 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10134 }
662fb8b2
NC
10135 break;
10136 case SVt_PVHV:
10137 {
cbbf8932 10138 HEK *hvname = NULL;
662fb8b2
NC
10139
10140 if (HvARRAY((HV*)sstr)) {
10141 STRLEN i = 0;
10142 const bool sharekeys = !!HvSHAREKEYS(sstr);
10143 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10144 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10145 char *darray;
a02a5408 10146 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10147 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10148 char);
10149 HvARRAY(dstr) = (HE**)darray;
10150 while (i <= sxhv->xhv_max) {
5c4138a0 10151 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10152 HvARRAY(dstr)[i] = source
10153 ? he_dup(source, sharekeys, param) : 0;
10154 ++i;
10155 }
10156 if (SvOOK(sstr)) {
00b6aa41
AL
10157 struct xpvhv_aux * const saux = HvAUX(sstr);
10158 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10159 /* This flag isn't copied. */
10160 /* SvOOK_on(hv) attacks the IV flags. */
10161 SvFLAGS(dstr) |= SVf_OOK;
10162
10163 hvname = saux->xhv_name;
dd690478
NC
10164 daux->xhv_name
10165 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10166
10167 daux->xhv_riter = saux->xhv_riter;
10168 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10169 ? he_dup(saux->xhv_eiter,
10170 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10171 daux->xhv_backreferences = saux->xhv_backreferences
10172 ? (AV*) SvREFCNT_inc(
10173 sv_dup((SV*)saux->
10174 xhv_backreferences,
10175 param))
10176 : 0;
662fb8b2
NC
10177 }
10178 }
10179 else {
bd61b366 10180 SvPV_set(dstr, NULL);
662fb8b2
NC
10181 }
10182 /* Record stashes for possible cloning in Perl_clone(). */
10183 if(hvname)
10184 av_push(param->stashes, dstr);
10185 }
10186 break;
662fb8b2 10187 case SVt_PVCV:
bb172083
NC
10188 if (!(param->flags & CLONEf_COPY_STACKS)) {
10189 CvDEPTH(dstr) = 0;
10190 }
10191 case SVt_PVFM:
662fb8b2
NC
10192 /* NOTE: not refcounted */
10193 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10194 OP_REFCNT_LOCK;
d04ba589
NC
10195 if (!CvISXSUB(dstr))
10196 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10197 OP_REFCNT_UNLOCK;
cfae286e 10198 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10199 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10200 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10201 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10202 }
10203 /* don't dup if copying back - CvGV isn't refcounted, so the
10204 * duped GV may never be freed. A bit of a hack! DAPM */
10205 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10206 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10207 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10208 CvOUTSIDE(dstr) =
10209 CvWEAKOUTSIDE(sstr)
10210 ? cv_dup( CvOUTSIDE(dstr), param)
10211 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10212 if (!CvISXSUB(dstr))
662fb8b2
NC
10213 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10214 break;
bfcb3514 10215 }
1d7c1841 10216 }
1d7c1841
GS
10217 }
10218
10219 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10220 ++PL_sv_objcount;
10221
10222 return dstr;
d2d73c3e 10223 }
1d7c1841 10224
645c22ef
DM
10225/* duplicate a context */
10226
1d7c1841 10227PERL_CONTEXT *
a8fc9800 10228Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10229{
10230 PERL_CONTEXT *ncxs;
10231
10232 if (!cxs)
10233 return (PERL_CONTEXT*)NULL;
10234
10235 /* look for it in the table first */
10236 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10237 if (ncxs)
10238 return ncxs;
10239
10240 /* create anew and remember what it is */
a02a5408 10241 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10242 ptr_table_store(PL_ptr_table, cxs, ncxs);
10243
10244 while (ix >= 0) {
c445ea15
AL
10245 PERL_CONTEXT * const cx = &cxs[ix];
10246 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10247 ncx->cx_type = cx->cx_type;
10248 if (CxTYPE(cx) == CXt_SUBST) {
10249 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10250 }
10251 else {
10252 ncx->blk_oldsp = cx->blk_oldsp;
10253 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10254 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10255 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10256 ncx->blk_oldpm = cx->blk_oldpm;
10257 ncx->blk_gimme = cx->blk_gimme;
10258 switch (CxTYPE(cx)) {
10259 case CXt_SUB:
10260 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10261 ? cv_dup_inc(cx->blk_sub.cv, param)
10262 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10263 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10264 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10265 : NULL);
d2d73c3e 10266 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10267 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10268 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10269 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10270 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10271 break;
10272 case CXt_EVAL:
10273 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10274 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10275 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10276 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10277 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10278 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10279 break;
10280 case CXt_LOOP:
10281 ncx->blk_loop.label = cx->blk_loop.label;
10282 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10283 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10284 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10285 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10286 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10287 ? cx->blk_loop.iterdata
d2d73c3e 10288 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10289 ncx->blk_loop.oldcomppad
10290 = (PAD*)ptr_table_fetch(PL_ptr_table,
10291 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10292 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10293 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10294 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10295 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10296 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10297 break;
10298 case CXt_FORMAT:
d2d73c3e
AB
10299 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10300 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10301 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10302 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10303 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10304 break;
10305 case CXt_BLOCK:
10306 case CXt_NULL:
10307 break;
10308 }
10309 }
10310 --ix;
10311 }
10312 return ncxs;
10313}
10314
645c22ef
DM
10315/* duplicate a stack info structure */
10316
1d7c1841 10317PERL_SI *
a8fc9800 10318Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10319{
10320 PERL_SI *nsi;
10321
10322 if (!si)
10323 return (PERL_SI*)NULL;
10324
10325 /* look for it in the table first */
10326 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10327 if (nsi)
10328 return nsi;
10329
10330 /* create anew and remember what it is */
a02a5408 10331 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10332 ptr_table_store(PL_ptr_table, si, nsi);
10333
d2d73c3e 10334 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10335 nsi->si_cxix = si->si_cxix;
10336 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10337 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10338 nsi->si_type = si->si_type;
d2d73c3e
AB
10339 nsi->si_prev = si_dup(si->si_prev, param);
10340 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10341 nsi->si_markoff = si->si_markoff;
10342
10343 return nsi;
10344}
10345
10346#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10347#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10348#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10349#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10350#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10351#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10352#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10353#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10354#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10355#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10356#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10357#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10358#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10359#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10360
10361/* XXXXX todo */
10362#define pv_dup_inc(p) SAVEPV(p)
10363#define pv_dup(p) SAVEPV(p)
10364#define svp_dup_inc(p,pp) any_dup(p,pp)
10365
645c22ef
DM
10366/* map any object to the new equivent - either something in the
10367 * ptr table, or something in the interpreter structure
10368 */
10369
1d7c1841 10370void *
53c1dcc0 10371Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10372{
10373 void *ret;
10374
10375 if (!v)
10376 return (void*)NULL;
10377
10378 /* look for it in the table first */
10379 ret = ptr_table_fetch(PL_ptr_table, v);
10380 if (ret)
10381 return ret;
10382
10383 /* see if it is part of the interpreter structure */
10384 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10385 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10386 else {
1d7c1841 10387 ret = v;
05ec9bb3 10388 }
1d7c1841
GS
10389
10390 return ret;
10391}
10392
645c22ef
DM
10393/* duplicate the save stack */
10394
1d7c1841 10395ANY *
a8fc9800 10396Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10397{
53c1dcc0
AL
10398 ANY * const ss = proto_perl->Tsavestack;
10399 const I32 max = proto_perl->Tsavestack_max;
10400 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10401 ANY *nss;
10402 SV *sv;
10403 GV *gv;
10404 AV *av;
10405 HV *hv;
10406 void* ptr;
10407 int intval;
10408 long longval;
10409 GP *gp;
10410 IV iv;
c4e33207 10411 char *c = NULL;
1d7c1841 10412 void (*dptr) (void*);
acfe0abc 10413 void (*dxptr) (pTHX_ void*);
1d7c1841 10414
a02a5408 10415 Newxz(nss, max, ANY);
1d7c1841
GS
10416
10417 while (ix > 0) {
b464bac0 10418 I32 i = POPINT(ss,ix);
1d7c1841
GS
10419 TOPINT(nss,ix) = i;
10420 switch (i) {
10421 case SAVEt_ITEM: /* normal string */
a41cc44e 10422 case SAVEt_SV: /* scalar reference */
1d7c1841 10423 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10424 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10425 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10426 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10427 break;
05ec9bb3
NIS
10428 case SAVEt_SHARED_PVREF: /* char* in shared space */
10429 c = (char*)POPPTR(ss,ix);
10430 TOPPTR(nss,ix) = savesharedpv(c);
10431 ptr = POPPTR(ss,ix);
10432 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10433 break;
1d7c1841
GS
10434 case SAVEt_GENERIC_SVREF: /* generic sv */
10435 case SAVEt_SVREF: /* scalar reference */
10436 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10437 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10438 ptr = POPPTR(ss,ix);
10439 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10440 break;
a41cc44e 10441 case SAVEt_HV: /* hash reference */
1d7c1841 10442 case SAVEt_AV: /* array reference */
337d28f5
NC
10443 sv = POPPTR(ss,ix);
10444 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10445 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10446 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10447 break;
10448 case SAVEt_INT: /* int reference */
10449 ptr = POPPTR(ss,ix);
10450 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10451 intval = (int)POPINT(ss,ix);
10452 TOPINT(nss,ix) = intval;
10453 break;
10454 case SAVEt_LONG: /* long reference */
10455 ptr = POPPTR(ss,ix);
10456 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10457 longval = (long)POPLONG(ss,ix);
10458 TOPLONG(nss,ix) = longval;
10459 break;
10460 case SAVEt_I32: /* I32 reference */
10461 case SAVEt_I16: /* I16 reference */
10462 case SAVEt_I8: /* I8 reference */
88effcc9 10463 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10464 ptr = POPPTR(ss,ix);
10465 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10466 i = POPINT(ss,ix);
10467 TOPINT(nss,ix) = i;
10468 break;
10469 case SAVEt_IV: /* IV reference */
10470 ptr = POPPTR(ss,ix);
10471 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10472 iv = POPIV(ss,ix);
10473 TOPIV(nss,ix) = iv;
10474 break;
a41cc44e
NC
10475 case SAVEt_HPTR: /* HV* reference */
10476 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10477 case SAVEt_SPTR: /* SV* reference */
10478 ptr = POPPTR(ss,ix);
10479 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10480 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10481 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10482 break;
10483 case SAVEt_VPTR: /* random* reference */
10484 ptr = POPPTR(ss,ix);
10485 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10486 ptr = POPPTR(ss,ix);
10487 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10488 break;
b03d03b0 10489 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10490 case SAVEt_PPTR: /* char* reference */
10491 ptr = POPPTR(ss,ix);
10492 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10493 c = (char*)POPPTR(ss,ix);
10494 TOPPTR(nss,ix) = pv_dup(c);
10495 break;
1d7c1841
GS
10496 case SAVEt_NSTAB:
10497 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10498 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10499 break;
10500 case SAVEt_GP: /* scalar reference */
10501 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10502 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10503 (void)GpREFCNT_inc(gp);
10504 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10505 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10506 c = (char*)POPPTR(ss,ix);
10507 TOPPTR(nss,ix) = pv_dup(c);
10508 iv = POPIV(ss,ix);
10509 TOPIV(nss,ix) = iv;
10510 iv = POPIV(ss,ix);
10511 TOPIV(nss,ix) = iv;
10512 break;
10513 case SAVEt_FREESV:
26d9b02f 10514 case SAVEt_MORTALIZESV:
1d7c1841 10515 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10516 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10517 break;
10518 case SAVEt_FREEOP:
10519 ptr = POPPTR(ss,ix);
10520 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10521 /* these are assumed to be refcounted properly */
53c1dcc0 10522 OP *o;
1d7c1841
GS
10523 switch (((OP*)ptr)->op_type) {
10524 case OP_LEAVESUB:
10525 case OP_LEAVESUBLV:
10526 case OP_LEAVEEVAL:
10527 case OP_LEAVE:
10528 case OP_SCOPE:
10529 case OP_LEAVEWRITE:
e977893f
GS
10530 TOPPTR(nss,ix) = ptr;
10531 o = (OP*)ptr;
10532 OpREFCNT_inc(o);
1d7c1841
GS
10533 break;
10534 default:
5f66b61c 10535 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10536 break;
10537 }
10538 }
10539 else
5f66b61c 10540 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10541 break;
10542 case SAVEt_FREEPV:
10543 c = (char*)POPPTR(ss,ix);
10544 TOPPTR(nss,ix) = pv_dup_inc(c);
10545 break;
10546 case SAVEt_CLEARSV:
10547 longval = POPLONG(ss,ix);
10548 TOPLONG(nss,ix) = longval;
10549 break;
10550 case SAVEt_DELETE:
10551 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10552 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10553 c = (char*)POPPTR(ss,ix);
10554 TOPPTR(nss,ix) = pv_dup_inc(c);
10555 i = POPINT(ss,ix);
10556 TOPINT(nss,ix) = i;
10557 break;
10558 case SAVEt_DESTRUCTOR:
10559 ptr = POPPTR(ss,ix);
10560 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10561 dptr = POPDPTR(ss,ix);
8141890a
JH
10562 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10563 any_dup(FPTR2DPTR(void *, dptr),
10564 proto_perl));
1d7c1841
GS
10565 break;
10566 case SAVEt_DESTRUCTOR_X:
10567 ptr = POPPTR(ss,ix);
10568 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10569 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10570 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10571 any_dup(FPTR2DPTR(void *, dxptr),
10572 proto_perl));
1d7c1841
GS
10573 break;
10574 case SAVEt_REGCONTEXT:
10575 case SAVEt_ALLOC:
10576 i = POPINT(ss,ix);
10577 TOPINT(nss,ix) = i;
10578 ix -= i;
10579 break;
10580 case SAVEt_STACK_POS: /* Position on Perl stack */
10581 i = POPINT(ss,ix);
10582 TOPINT(nss,ix) = i;
10583 break;
10584 case SAVEt_AELEM: /* array element */
10585 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10586 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10587 i = POPINT(ss,ix);
10588 TOPINT(nss,ix) = i;
10589 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10590 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10591 break;
10592 case SAVEt_HELEM: /* hash element */
10593 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10594 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10595 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10596 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10597 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10598 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10599 break;
10600 case SAVEt_OP:
10601 ptr = POPPTR(ss,ix);
10602 TOPPTR(nss,ix) = ptr;
10603 break;
10604 case SAVEt_HINTS:
10605 i = POPINT(ss,ix);
10606 TOPINT(nss,ix) = i;
b3ca2e83 10607 ptr = POPPTR(ss,ix);
080ac856 10608 if (ptr) {
7b6dd8c3 10609 HINTS_REFCNT_LOCK;
080ac856 10610 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10611 HINTS_REFCNT_UNLOCK;
10612 }
cbb1fbea 10613 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10614 if (i & HINT_LOCALIZE_HH) {
10615 hv = (HV*)POPPTR(ss,ix);
10616 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10617 }
1d7c1841 10618 break;
c4410b1b
GS
10619 case SAVEt_COMPPAD:
10620 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10621 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10622 break;
c3564e5c
GS
10623 case SAVEt_PADSV:
10624 longval = (long)POPLONG(ss,ix);
10625 TOPLONG(nss,ix) = longval;
10626 ptr = POPPTR(ss,ix);
10627 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10628 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10629 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10630 break;
a1bb4754 10631 case SAVEt_BOOL:
38d8b13e 10632 ptr = POPPTR(ss,ix);
b9609c01 10633 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10634 longval = (long)POPBOOL(ss,ix);
b9609c01 10635 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10636 break;
8bd2680e
MHM
10637 case SAVEt_SET_SVFLAGS:
10638 i = POPINT(ss,ix);
10639 TOPINT(nss,ix) = i;
10640 i = POPINT(ss,ix);
10641 TOPINT(nss,ix) = i;
10642 sv = (SV*)POPPTR(ss,ix);
10643 TOPPTR(nss,ix) = sv_dup(sv, param);
10644 break;
5bfb7d0e
NC
10645 case SAVEt_RE_STATE:
10646 {
10647 const struct re_save_state *const old_state
10648 = (struct re_save_state *)
10649 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10650 struct re_save_state *const new_state
10651 = (struct re_save_state *)
10652 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10653
10654 Copy(old_state, new_state, 1, struct re_save_state);
10655 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10656
10657 new_state->re_state_bostr
10658 = pv_dup(old_state->re_state_bostr);
10659 new_state->re_state_reginput
10660 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10661 new_state->re_state_regeol
10662 = pv_dup(old_state->re_state_regeol);
10663 new_state->re_state_regstartp
10664 = any_dup(old_state->re_state_regstartp, proto_perl);
10665 new_state->re_state_regendp
10666 = any_dup(old_state->re_state_regendp, proto_perl);
10667 new_state->re_state_reglastparen
10668 = any_dup(old_state->re_state_reglastparen, proto_perl);
10669 new_state->re_state_reglastcloseparen
10670 = any_dup(old_state->re_state_reglastcloseparen,
10671 proto_perl);
5bfb7d0e
NC
10672 /* XXX This just has to be broken. The old save_re_context
10673 code did SAVEGENERICPV(PL_reg_start_tmp);
10674 PL_reg_start_tmp is char **.
10675 Look above to what the dup code does for
10676 SAVEt_GENERIC_PVREF
10677 It can never have worked.
10678 So this is merely a faithful copy of the exiting bug: */
10679 new_state->re_state_reg_start_tmp
10680 = (char **) pv_dup((char *)
10681 old_state->re_state_reg_start_tmp);
10682 /* I assume that it only ever "worked" because no-one called
10683 (pseudo)fork while the regexp engine had re-entered itself.
10684 */
5bfb7d0e
NC
10685#ifdef PERL_OLD_COPY_ON_WRITE
10686 new_state->re_state_nrs
10687 = sv_dup(old_state->re_state_nrs, param);
10688#endif
10689 new_state->re_state_reg_magic
10690 = any_dup(old_state->re_state_reg_magic, proto_perl);
10691 new_state->re_state_reg_oldcurpm
10692 = any_dup(old_state->re_state_reg_oldcurpm, proto_perl);
10693 new_state->re_state_reg_curpm
10694 = any_dup(old_state->re_state_reg_curpm, proto_perl);
10695 new_state->re_state_reg_oldsaved
10696 = pv_dup(old_state->re_state_reg_oldsaved);
10697 new_state->re_state_reg_poscache
10698 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
10699 new_state->re_state_reg_starttry
10700 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
10701 break;
10702 }
68da3b2f
NC
10703 case SAVEt_COMPILE_WARNINGS:
10704 ptr = POPPTR(ss,ix);
10705 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10706 break;
1d7c1841 10707 default:
ca05af4a 10708 Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i);
1d7c1841
GS
10709 }
10710 }
10711
bd81e77b
NC
10712 return nss;
10713}
10714
10715
10716/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10717 * flag to the result. This is done for each stash before cloning starts,
10718 * so we know which stashes want their objects cloned */
10719
10720static void
10721do_mark_cloneable_stash(pTHX_ SV *sv)
10722{
10723 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10724 if (hvname) {
10725 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10726 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10727 if (cloner && GvCV(cloner)) {
10728 dSP;
10729 UV status;
10730
10731 ENTER;
10732 SAVETMPS;
10733 PUSHMARK(SP);
10734 XPUSHs(sv_2mortal(newSVhek(hvname)));
10735 PUTBACK;
10736 call_sv((SV*)GvCV(cloner), G_SCALAR);
10737 SPAGAIN;
10738 status = POPu;
10739 PUTBACK;
10740 FREETMPS;
10741 LEAVE;
10742 if (status)
10743 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10744 }
10745 }
10746}
10747
10748
10749
10750/*
10751=for apidoc perl_clone
10752
10753Create and return a new interpreter by cloning the current one.
10754
10755perl_clone takes these flags as parameters:
10756
10757CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10758without it we only clone the data and zero the stacks,
10759with it we copy the stacks and the new perl interpreter is
10760ready to run at the exact same point as the previous one.
10761The pseudo-fork code uses COPY_STACKS while the
10762threads->new doesn't.
10763
10764CLONEf_KEEP_PTR_TABLE
10765perl_clone keeps a ptr_table with the pointer of the old
10766variable as a key and the new variable as a value,
10767this allows it to check if something has been cloned and not
10768clone it again but rather just use the value and increase the
10769refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10770the ptr_table using the function
10771C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10772reason to keep it around is if you want to dup some of your own
10773variable who are outside the graph perl scans, example of this
10774code is in threads.xs create
10775
10776CLONEf_CLONE_HOST
10777This is a win32 thing, it is ignored on unix, it tells perls
10778win32host code (which is c++) to clone itself, this is needed on
10779win32 if you want to run two threads at the same time,
10780if you just want to do some stuff in a separate perl interpreter
10781and then throw it away and return to the original one,
10782you don't need to do anything.
10783
10784=cut
10785*/
10786
10787/* XXX the above needs expanding by someone who actually understands it ! */
10788EXTERN_C PerlInterpreter *
10789perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10790
10791PerlInterpreter *
10792perl_clone(PerlInterpreter *proto_perl, UV flags)
10793{
10794 dVAR;
10795#ifdef PERL_IMPLICIT_SYS
10796
10797 /* perlhost.h so we need to call into it
10798 to clone the host, CPerlHost should have a c interface, sky */
10799
10800 if (flags & CLONEf_CLONE_HOST) {
10801 return perl_clone_host(proto_perl,flags);
10802 }
10803 return perl_clone_using(proto_perl, flags,
10804 proto_perl->IMem,
10805 proto_perl->IMemShared,
10806 proto_perl->IMemParse,
10807 proto_perl->IEnv,
10808 proto_perl->IStdIO,
10809 proto_perl->ILIO,
10810 proto_perl->IDir,
10811 proto_perl->ISock,
10812 proto_perl->IProc);
10813}
10814
10815PerlInterpreter *
10816perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10817 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10818 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10819 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10820 struct IPerlDir* ipD, struct IPerlSock* ipS,
10821 struct IPerlProc* ipP)
10822{
10823 /* XXX many of the string copies here can be optimized if they're
10824 * constants; they need to be allocated as common memory and just
10825 * their pointers copied. */
10826
10827 IV i;
10828 CLONE_PARAMS clone_params;
5f66b61c 10829 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10830
5f66b61c 10831 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10832 /* for each stash, determine whether its objects should be cloned */
10833 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10834 PERL_SET_THX(my_perl);
10835
10836# ifdef DEBUGGING
7e337ee0 10837 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10838 PL_op = NULL;
10839 PL_curcop = NULL;
bd81e77b
NC
10840 PL_markstack = 0;
10841 PL_scopestack = 0;
10842 PL_savestack = 0;
10843 PL_savestack_ix = 0;
10844 PL_savestack_max = -1;
10845 PL_sig_pending = 0;
10846 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10847# else /* !DEBUGGING */
10848 Zero(my_perl, 1, PerlInterpreter);
10849# endif /* DEBUGGING */
10850
10851 /* host pointers */
10852 PL_Mem = ipM;
10853 PL_MemShared = ipMS;
10854 PL_MemParse = ipMP;
10855 PL_Env = ipE;
10856 PL_StdIO = ipStd;
10857 PL_LIO = ipLIO;
10858 PL_Dir = ipD;
10859 PL_Sock = ipS;
10860 PL_Proc = ipP;
10861#else /* !PERL_IMPLICIT_SYS */
10862 IV i;
10863 CLONE_PARAMS clone_params;
10864 CLONE_PARAMS* param = &clone_params;
5f66b61c 10865 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10866 /* for each stash, determine whether its objects should be cloned */
10867 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10868 PERL_SET_THX(my_perl);
10869
10870# ifdef DEBUGGING
7e337ee0 10871 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10872 PL_op = NULL;
10873 PL_curcop = NULL;
bd81e77b
NC
10874 PL_markstack = 0;
10875 PL_scopestack = 0;
10876 PL_savestack = 0;
10877 PL_savestack_ix = 0;
10878 PL_savestack_max = -1;
10879 PL_sig_pending = 0;
10880 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10881# else /* !DEBUGGING */
10882 Zero(my_perl, 1, PerlInterpreter);
10883# endif /* DEBUGGING */
10884#endif /* PERL_IMPLICIT_SYS */
10885 param->flags = flags;
10886 param->proto_perl = proto_perl;
10887
7cb608b5
NC
10888 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10889
fdda85ca 10890 PL_body_arenas = NULL;
bd81e77b
NC
10891 Zero(&PL_body_roots, 1, PL_body_roots);
10892
10893 PL_nice_chunk = NULL;
10894 PL_nice_chunk_size = 0;
10895 PL_sv_count = 0;
10896 PL_sv_objcount = 0;
a0714e2c
SS
10897 PL_sv_root = NULL;
10898 PL_sv_arenaroot = NULL;
bd81e77b
NC
10899
10900 PL_debug = proto_perl->Idebug;
10901
10902 PL_hash_seed = proto_perl->Ihash_seed;
10903 PL_rehash_seed = proto_perl->Irehash_seed;
10904
10905#ifdef USE_REENTRANT_API
10906 /* XXX: things like -Dm will segfault here in perlio, but doing
10907 * PERL_SET_CONTEXT(proto_perl);
10908 * breaks too many other things
10909 */
10910 Perl_reentrant_init(aTHX);
10911#endif
10912
10913 /* create SV map for pointer relocation */
10914 PL_ptr_table = ptr_table_new();
10915
10916 /* initialize these special pointers as early as possible */
10917 SvANY(&PL_sv_undef) = NULL;
10918 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10919 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10920 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10921
10922 SvANY(&PL_sv_no) = new_XPVNV();
10923 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10924 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10925 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10926 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
10927 SvCUR_set(&PL_sv_no, 0);
10928 SvLEN_set(&PL_sv_no, 1);
10929 SvIV_set(&PL_sv_no, 0);
10930 SvNV_set(&PL_sv_no, 0);
10931 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10932
10933 SvANY(&PL_sv_yes) = new_XPVNV();
10934 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10935 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10936 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 10937 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
10938 SvCUR_set(&PL_sv_yes, 1);
10939 SvLEN_set(&PL_sv_yes, 2);
10940 SvIV_set(&PL_sv_yes, 1);
10941 SvNV_set(&PL_sv_yes, 1);
10942 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10943
10944 /* create (a non-shared!) shared string table */
10945 PL_strtab = newHV();
10946 HvSHAREKEYS_off(PL_strtab);
10947 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10948 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10949
10950 PL_compiling = proto_perl->Icompiling;
10951
10952 /* These two PVs will be free'd special way so must set them same way op.c does */
10953 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10954 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10955
10956 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10957 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10958
10959 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10960 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
bd81e77b
NC
10961 if (!specialCopIO(PL_compiling.cop_io))
10962 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
cbb1fbea
NC
10963 if (PL_compiling.cop_hints) {
10964 HINTS_REFCNT_LOCK;
10965 PL_compiling.cop_hints->refcounted_he_refcnt++;
10966 HINTS_REFCNT_UNLOCK;
10967 }
bd81e77b
NC
10968 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10969
10970 /* pseudo environmental stuff */
10971 PL_origargc = proto_perl->Iorigargc;
10972 PL_origargv = proto_perl->Iorigargv;
10973
10974 param->stashes = newAV(); /* Setup array of objects to call clone on */
10975
10976 /* Set tainting stuff before PerlIO_debug can possibly get called */
10977 PL_tainting = proto_perl->Itainting;
10978 PL_taint_warn = proto_perl->Itaint_warn;
10979
10980#ifdef PERLIO_LAYERS
10981 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10982 PerlIO_clone(aTHX_ proto_perl, param);
10983#endif
10984
10985 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10986 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10987 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10988 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10989 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10990 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10991
10992 /* switches */
10993 PL_minus_c = proto_perl->Iminus_c;
10994 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10995 PL_localpatches = proto_perl->Ilocalpatches;
10996 PL_splitstr = proto_perl->Isplitstr;
10997 PL_preprocess = proto_perl->Ipreprocess;
10998 PL_minus_n = proto_perl->Iminus_n;
10999 PL_minus_p = proto_perl->Iminus_p;
11000 PL_minus_l = proto_perl->Iminus_l;
11001 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 11002 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
11003 PL_minus_F = proto_perl->Iminus_F;
11004 PL_doswitches = proto_perl->Idoswitches;
11005 PL_dowarn = proto_perl->Idowarn;
11006 PL_doextract = proto_perl->Idoextract;
11007 PL_sawampersand = proto_perl->Isawampersand;
11008 PL_unsafe = proto_perl->Iunsafe;
11009 PL_inplace = SAVEPV(proto_perl->Iinplace);
11010 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11011 PL_perldb = proto_perl->Iperldb;
11012 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11013 PL_exit_flags = proto_perl->Iexit_flags;
11014
11015 /* magical thingies */
11016 /* XXX time(&PL_basetime) when asked for? */
11017 PL_basetime = proto_perl->Ibasetime;
11018 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11019
11020 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
11021 PL_statusvalue = proto_perl->Istatusvalue;
11022#ifdef VMS
11023 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11024#else
11025 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11026#endif
11027 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11028
11029 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
11030 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
11031 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
11032
11033 /* Clone the regex array */
11034 PL_regex_padav = newAV();
11035 {
11036 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 11037 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 11038 IV i;
7f466ec7 11039 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 11040 for(i = 1; i <= len; i++) {
7a5b473e
AL
11041 const SV * const regex = regexen[i];
11042 SV * const sv =
11043 SvREPADTMP(regex)
11044 ? sv_dup_inc(regex, param)
11045 : SvREFCNT_inc(
11046 newSViv(PTR2IV(re_dup(
11047 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11048 ;
11049 av_push(PL_regex_padav, sv);
bd81e77b
NC
11050 }
11051 }
11052 PL_regex_pad = AvARRAY(PL_regex_padav);
11053
11054 /* shortcuts to various I/O objects */
11055 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11056 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11057 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11058 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11059 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11060 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11061
bd81e77b
NC
11062 /* shortcuts to regexp stuff */
11063 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11064
bd81e77b
NC
11065 /* shortcuts to misc objects */
11066 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11067
bd81e77b
NC
11068 /* shortcuts to debugging objects */
11069 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11070 PL_DBline = gv_dup(proto_perl->IDBline, param);
11071 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11072 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11073 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11074 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11075 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11076 PL_lineary = av_dup(proto_perl->Ilineary, param);
11077 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11078
bd81e77b
NC
11079 /* symbol tables */
11080 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11081 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11082 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11083 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11084 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11085
11086 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11087 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11088 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11089 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11090 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11091 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11092
11093 PL_sub_generation = proto_perl->Isub_generation;
11094
11095 /* funky return mechanisms */
11096 PL_forkprocess = proto_perl->Iforkprocess;
11097
11098 /* subprocess state */
11099 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11100
11101 /* internal state */
11102 PL_maxo = proto_perl->Imaxo;
11103 if (proto_perl->Iop_mask)
11104 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11105 else
bd61b366 11106 PL_op_mask = NULL;
bd81e77b
NC
11107 /* PL_asserting = proto_perl->Iasserting; */
11108
11109 /* current interpreter roots */
11110 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
11111 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
11112 PL_main_start = proto_perl->Imain_start;
11113 PL_eval_root = proto_perl->Ieval_root;
11114 PL_eval_start = proto_perl->Ieval_start;
11115
11116 /* runtime control stuff */
11117 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11118 PL_copline = proto_perl->Icopline;
11119
11120 PL_filemode = proto_perl->Ifilemode;
11121 PL_lastfd = proto_perl->Ilastfd;
11122 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11123 PL_Argv = NULL;
bd61b366 11124 PL_Cmd = NULL;
bd81e77b
NC
11125 PL_gensym = proto_perl->Igensym;
11126 PL_preambled = proto_perl->Ipreambled;
11127 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11128 PL_laststatval = proto_perl->Ilaststatval;
11129 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11130 PL_mess_sv = NULL;
bd81e77b
NC
11131
11132 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11133
11134 /* interpreter atexit processing */
11135 PL_exitlistlen = proto_perl->Iexitlistlen;
11136 if (PL_exitlistlen) {
11137 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11138 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11139 }
bd81e77b
NC
11140 else
11141 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11142
11143 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11144 if (PL_my_cxt_size) {
f16dd614
DM
11145 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11146 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11147 }
11148 else
11149 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11150 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11151 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11152 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11153
11154 PL_profiledata = NULL;
11155 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11156 /* PL_rsfp_filters entries have fake IoDIRP() */
11157 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11158
bd81e77b 11159 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11160
bd81e77b 11161 PAD_CLONE_VARS(proto_perl, param);
9660f481 11162
bd81e77b
NC
11163#ifdef HAVE_INTERP_INTERN
11164 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11165#endif
645c22ef 11166
bd81e77b
NC
11167 /* more statics moved here */
11168 PL_generation = proto_perl->Igeneration;
11169 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11170
bd81e77b
NC
11171 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11172 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11173
bd81e77b
NC
11174 PL_uid = proto_perl->Iuid;
11175 PL_euid = proto_perl->Ieuid;
11176 PL_gid = proto_perl->Igid;
11177 PL_egid = proto_perl->Iegid;
11178 PL_nomemok = proto_perl->Inomemok;
11179 PL_an = proto_perl->Ian;
11180 PL_evalseq = proto_perl->Ievalseq;
11181 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11182 PL_origalen = proto_perl->Iorigalen;
11183#ifdef PERL_USES_PL_PIDSTATUS
11184 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11185#endif
11186 PL_osname = SAVEPV(proto_perl->Iosname);
11187 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11188
bd81e77b 11189 PL_runops = proto_perl->Irunops;
6a78b4db 11190
bd81e77b 11191 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11192
bd81e77b
NC
11193#ifdef CSH
11194 PL_cshlen = proto_perl->Icshlen;
11195 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11196#endif
645c22ef 11197
bd81e77b
NC
11198 PL_lex_state = proto_perl->Ilex_state;
11199 PL_lex_defer = proto_perl->Ilex_defer;
11200 PL_lex_expect = proto_perl->Ilex_expect;
11201 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11202 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11203 PL_lex_starts = proto_perl->Ilex_starts;
11204 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11205 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11206 PL_lex_op = proto_perl->Ilex_op;
11207 PL_lex_inpat = proto_perl->Ilex_inpat;
11208 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11209 PL_lex_brackets = proto_perl->Ilex_brackets;
11210 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11211 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11212 PL_lex_casemods = proto_perl->Ilex_casemods;
11213 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11214 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11215
5db06880
NC
11216#ifdef PERL_MAD
11217 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11218 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11219 PL_realtokenstart = proto_perl->Irealtokenstart;
11220 PL_faketokens = proto_perl->Ifaketokens;
11221 PL_thismad = proto_perl->Ithismad;
11222 PL_thistoken = proto_perl->Ithistoken;
11223 PL_thisopen = proto_perl->Ithisopen;
11224 PL_thisstuff = proto_perl->Ithisstuff;
11225 PL_thisclose = proto_perl->Ithisclose;
11226 PL_thiswhite = proto_perl->Ithiswhite;
11227 PL_nextwhite = proto_perl->Inextwhite;
11228 PL_skipwhite = proto_perl->Iskipwhite;
11229 PL_endwhite = proto_perl->Iendwhite;
11230 PL_curforce = proto_perl->Icurforce;
5db06880 11231#else
bd81e77b
NC
11232 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11233 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11234 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11235#endif
c43294b8 11236
bd81e77b
NC
11237 /* XXX This is probably masking the deeper issue of why
11238 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11239 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11240 * (A little debugging with a watchpoint on it may help.)
11241 */
11242 if (SvANY(proto_perl->Ilinestr)) {
11243 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11244 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11245 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11246 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11247 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11248 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11249 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11250 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11251 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11252 }
11253 else {
561b68a9 11254 PL_linestr = newSV(79);
bd81e77b
NC
11255 sv_upgrade(PL_linestr,SVt_PVIV);
11256 sv_setpvn(PL_linestr,"",0);
11257 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11258 }
11259 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11260 PL_pending_ident = proto_perl->Ipending_ident;
11261 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11262
11263 PL_expect = proto_perl->Iexpect;
11264
11265 PL_multi_start = proto_perl->Imulti_start;
11266 PL_multi_end = proto_perl->Imulti_end;
11267 PL_multi_open = proto_perl->Imulti_open;
11268 PL_multi_close = proto_perl->Imulti_close;
11269
11270 PL_error_count = proto_perl->Ierror_count;
11271 PL_subline = proto_perl->Isubline;
11272 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11273
bd81e77b
NC
11274 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11275 if (SvANY(proto_perl->Ilinestr)) {
11276 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11277 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11278 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11279 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11280 PL_last_lop_op = proto_perl->Ilast_lop_op;
11281 }
11282 else {
11283 PL_last_uni = SvPVX(PL_linestr);
11284 PL_last_lop = SvPVX(PL_linestr);
11285 PL_last_lop_op = 0;
11286 }
11287 PL_in_my = proto_perl->Iin_my;
11288 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11289#ifdef FCRYPT
11290 PL_cryptseen = proto_perl->Icryptseen;
11291#endif
1d7c1841 11292
bd81e77b 11293 PL_hints = proto_perl->Ihints;
1d7c1841 11294
bd81e77b 11295 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11296
bd81e77b
NC
11297#ifdef USE_LOCALE_COLLATE
11298 PL_collation_ix = proto_perl->Icollation_ix;
11299 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11300 PL_collation_standard = proto_perl->Icollation_standard;
11301 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11302 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11303#endif /* USE_LOCALE_COLLATE */
1d7c1841 11304
bd81e77b
NC
11305#ifdef USE_LOCALE_NUMERIC
11306 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11307 PL_numeric_standard = proto_perl->Inumeric_standard;
11308 PL_numeric_local = proto_perl->Inumeric_local;
11309 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11310#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11311
bd81e77b
NC
11312 /* utf8 character classes */
11313 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11314 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11315 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11316 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11317 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11318 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11319 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11320 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11321 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11322 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11323 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11324 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11325 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11326 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11327 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11328 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11329 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11330 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11331 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11332 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11333
bd81e77b
NC
11334 /* Did the locale setup indicate UTF-8? */
11335 PL_utf8locale = proto_perl->Iutf8locale;
11336 /* Unicode features (see perlrun/-C) */
11337 PL_unicode = proto_perl->Iunicode;
1d7c1841 11338
bd81e77b
NC
11339 /* Pre-5.8 signals control */
11340 PL_signals = proto_perl->Isignals;
1d7c1841 11341
bd81e77b
NC
11342 /* times() ticks per second */
11343 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11344
bd81e77b
NC
11345 /* Recursion stopper for PerlIO_find_layer */
11346 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11347
bd81e77b
NC
11348 /* sort() routine */
11349 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11350
bd81e77b
NC
11351 /* Not really needed/useful since the reenrant_retint is "volatile",
11352 * but do it for consistency's sake. */
11353 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11354
bd81e77b
NC
11355 /* Hooks to shared SVs and locks. */
11356 PL_sharehook = proto_perl->Isharehook;
11357 PL_lockhook = proto_perl->Ilockhook;
11358 PL_unlockhook = proto_perl->Iunlockhook;
11359 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11360
bd81e77b
NC
11361 PL_runops_std = proto_perl->Irunops_std;
11362 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11363
bd81e77b
NC
11364#ifdef THREADS_HAVE_PIDS
11365 PL_ppid = proto_perl->Ippid;
11366#endif
1d7c1841 11367
bd81e77b 11368 /* swatch cache */
5c284bb0 11369 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11370 PL_last_swash_klen = 0;
11371 PL_last_swash_key[0]= '\0';
11372 PL_last_swash_tmps = (U8*)NULL;
11373 PL_last_swash_slen = 0;
1d7c1841 11374
bd81e77b
NC
11375 PL_glob_index = proto_perl->Iglob_index;
11376 PL_srand_called = proto_perl->Isrand_called;
11377 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11378 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11379
bd81e77b
NC
11380 if (proto_perl->Ipsig_pend) {
11381 Newxz(PL_psig_pend, SIG_SIZE, int);
11382 }
11383 else {
11384 PL_psig_pend = (int*)NULL;
11385 }
05ec9bb3 11386
bd81e77b
NC
11387 if (proto_perl->Ipsig_ptr) {
11388 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11389 Newxz(PL_psig_name, SIG_SIZE, SV*);
11390 for (i = 1; i < SIG_SIZE; i++) {
11391 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11392 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11393 }
11394 }
11395 else {
11396 PL_psig_ptr = (SV**)NULL;
11397 PL_psig_name = (SV**)NULL;
11398 }
05ec9bb3 11399
bd81e77b 11400 /* thrdvar.h stuff */
1d7c1841 11401
bd81e77b
NC
11402 if (flags & CLONEf_COPY_STACKS) {
11403 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11404 PL_tmps_ix = proto_perl->Ttmps_ix;
11405 PL_tmps_max = proto_perl->Ttmps_max;
11406 PL_tmps_floor = proto_perl->Ttmps_floor;
11407 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11408 i = 0;
11409 while (i <= PL_tmps_ix) {
11410 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11411 ++i;
11412 }
d2d73c3e 11413
bd81e77b
NC
11414 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11415 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11416 Newxz(PL_markstack, i, I32);
11417 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11418 - proto_perl->Tmarkstack);
11419 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11420 - proto_perl->Tmarkstack);
11421 Copy(proto_perl->Tmarkstack, PL_markstack,
11422 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11423
bd81e77b
NC
11424 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11425 * NOTE: unlike the others! */
11426 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11427 PL_scopestack_max = proto_perl->Tscopestack_max;
11428 Newxz(PL_scopestack, PL_scopestack_max, I32);
11429 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11430
bd81e77b
NC
11431 /* NOTE: si_dup() looks at PL_markstack */
11432 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11433
bd81e77b
NC
11434 /* PL_curstack = PL_curstackinfo->si_stack; */
11435 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11436 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11437
bd81e77b
NC
11438 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11439 PL_stack_base = AvARRAY(PL_curstack);
11440 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11441 - proto_perl->Tstack_base);
11442 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11443
bd81e77b
NC
11444 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11445 * NOTE: unlike the others! */
11446 PL_savestack_ix = proto_perl->Tsavestack_ix;
11447 PL_savestack_max = proto_perl->Tsavestack_max;
11448 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11449 PL_savestack = ss_dup(proto_perl, param);
11450 }
11451 else {
11452 init_stacks();
11453 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11454
11455 /* although we're not duplicating the tmps stack, we should still
11456 * add entries for any SVs on the tmps stack that got cloned by a
11457 * non-refcount means (eg a temp in @_); otherwise they will be
11458 * orphaned
11459 */
11460 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11461 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11462 proto_perl->Ttmps_stack[i]);
11463 if (nsv && !SvREFCNT(nsv)) {
11464 EXTEND_MORTAL(1);
b37c2d43 11465 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11466 }
11467 }
bd81e77b 11468 }
1d7c1841 11469
bd81e77b
NC
11470 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11471 PL_top_env = &PL_start_env;
1d7c1841 11472
bd81e77b 11473 PL_op = proto_perl->Top;
4a4c6fe3 11474
a0714e2c 11475 PL_Sv = NULL;
bd81e77b
NC
11476 PL_Xpv = (XPV*)NULL;
11477 PL_na = proto_perl->Tna;
1fcf4c12 11478
bd81e77b
NC
11479 PL_statbuf = proto_perl->Tstatbuf;
11480 PL_statcache = proto_perl->Tstatcache;
11481 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11482 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11483#ifdef HAS_TIMES
11484 PL_timesbuf = proto_perl->Ttimesbuf;
11485#endif
1d7c1841 11486
bd81e77b
NC
11487 PL_tainted = proto_perl->Ttainted;
11488 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11489 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11490 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11491 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11492 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11493 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11494 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11495 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11496 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11497
bd81e77b
NC
11498 PL_restartop = proto_perl->Trestartop;
11499 PL_in_eval = proto_perl->Tin_eval;
11500 PL_delaymagic = proto_perl->Tdelaymagic;
11501 PL_dirty = proto_perl->Tdirty;
11502 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11503
bd81e77b 11504 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11505 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11506 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11507 PL_lastgotoprobe = NULL;
bd81e77b 11508 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11509
bd81e77b
NC
11510 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11511 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11512 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11513 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11514 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11515 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11516
bd81e77b 11517 /* regex stuff */
1d7c1841 11518
bd81e77b
NC
11519 PL_screamfirst = NULL;
11520 PL_screamnext = NULL;
11521 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11522 PL_lastscream = NULL;
1d7c1841 11523
bd81e77b 11524 PL_watchaddr = NULL;
bd61b366 11525 PL_watchok = NULL;
1d7c1841 11526
bd81e77b 11527 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11528 PL_colorset = 0; /* reinits PL_colors[] */
11529 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11530
bd81e77b
NC
11531 /* RE engine - function pointers */
11532 PL_regcompp = proto_perl->Tregcompp;
11533 PL_regexecp = proto_perl->Tregexecp;
11534 PL_regint_start = proto_perl->Tregint_start;
11535 PL_regint_string = proto_perl->Tregint_string;
11536 PL_regfree = proto_perl->Tregfree;
46ab3289 11537 Zero(&PL_reg_state, 1, struct re_save_state);
bd81e77b 11538 PL_reginterp_cnt = 0;
5d9a96ca 11539 PL_regmatch_slab = NULL;
1d7c1841 11540
bd81e77b
NC
11541 /* Pluggable optimizer */
11542 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11543
bd81e77b 11544 PL_stashcache = newHV();
1d7c1841 11545
bd81e77b
NC
11546 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11547 ptr_table_free(PL_ptr_table);
11548 PL_ptr_table = NULL;
11549 }
1d7c1841 11550
bd81e77b
NC
11551 /* Call the ->CLONE method, if it exists, for each of the stashes
11552 identified by sv_dup() above.
11553 */
11554 while(av_len(param->stashes) != -1) {
11555 HV* const stash = (HV*) av_shift(param->stashes);
11556 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11557 if (cloner && GvCV(cloner)) {
11558 dSP;
11559 ENTER;
11560 SAVETMPS;
11561 PUSHMARK(SP);
11562 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11563 PUTBACK;
11564 call_sv((SV*)GvCV(cloner), G_DISCARD);
11565 FREETMPS;
11566 LEAVE;
11567 }
1d7c1841 11568 }
1d7c1841 11569
bd81e77b 11570 SvREFCNT_dec(param->stashes);
1d7c1841 11571
bd81e77b
NC
11572 /* orphaned? eg threads->new inside BEGIN or use */
11573 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11574 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11575 SAVEFREESV(PL_compcv);
11576 }
dd2155a4 11577
bd81e77b
NC
11578 return my_perl;
11579}
1d7c1841 11580
bd81e77b 11581#endif /* USE_ITHREADS */
1d7c1841 11582
bd81e77b
NC
11583/*
11584=head1 Unicode Support
1d7c1841 11585
bd81e77b 11586=for apidoc sv_recode_to_utf8
1d7c1841 11587
bd81e77b
NC
11588The encoding is assumed to be an Encode object, on entry the PV
11589of the sv is assumed to be octets in that encoding, and the sv
11590will be converted into Unicode (and UTF-8).
1d7c1841 11591
bd81e77b
NC
11592If the sv already is UTF-8 (or if it is not POK), or if the encoding
11593is not a reference, nothing is done to the sv. If the encoding is not
11594an C<Encode::XS> Encoding object, bad things will happen.
11595(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11596
bd81e77b 11597The PV of the sv is returned.
1d7c1841 11598
bd81e77b 11599=cut */
1d7c1841 11600
bd81e77b
NC
11601char *
11602Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11603{
11604 dVAR;
11605 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11606 SV *uni;
11607 STRLEN len;
11608 const char *s;
11609 dSP;
11610 ENTER;
11611 SAVETMPS;
11612 save_re_context();
11613 PUSHMARK(sp);
11614 EXTEND(SP, 3);
11615 XPUSHs(encoding);
11616 XPUSHs(sv);
11617/*
11618 NI-S 2002/07/09
11619 Passing sv_yes is wrong - it needs to be or'ed set of constants
11620 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11621 remove converted chars from source.
1d7c1841 11622
bd81e77b 11623 Both will default the value - let them.
1d7c1841 11624
bd81e77b
NC
11625 XPUSHs(&PL_sv_yes);
11626*/
11627 PUTBACK;
11628 call_method("decode", G_SCALAR);
11629 SPAGAIN;
11630 uni = POPs;
11631 PUTBACK;
11632 s = SvPV_const(uni, len);
11633 if (s != SvPVX_const(sv)) {
11634 SvGROW(sv, len + 1);
11635 Move(s, SvPVX(sv), len + 1, char);
11636 SvCUR_set(sv, len);
11637 }
11638 FREETMPS;
11639 LEAVE;
11640 SvUTF8_on(sv);
11641 return SvPVX(sv);
389edf32 11642 }
bd81e77b
NC
11643 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11644}
1d7c1841 11645
bd81e77b
NC
11646/*
11647=for apidoc sv_cat_decode
1d7c1841 11648
bd81e77b
NC
11649The encoding is assumed to be an Encode object, the PV of the ssv is
11650assumed to be octets in that encoding and decoding the input starts
11651from the position which (PV + *offset) pointed to. The dsv will be
11652concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11653when the string tstr appears in decoding output or the input ends on
11654the PV of the ssv. The value which the offset points will be modified
11655to the last input position on the ssv.
1d7c1841 11656
bd81e77b 11657Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11658
bd81e77b
NC
11659=cut */
11660
11661bool
11662Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11663 SV *ssv, int *offset, char *tstr, int tlen)
11664{
11665 dVAR;
11666 bool ret = FALSE;
11667 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11668 SV *offsv;
11669 dSP;
11670 ENTER;
11671 SAVETMPS;
11672 save_re_context();
11673 PUSHMARK(sp);
11674 EXTEND(SP, 6);
11675 XPUSHs(encoding);
11676 XPUSHs(dsv);
11677 XPUSHs(ssv);
11678 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11679 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11680 PUTBACK;
11681 call_method("cat_decode", G_SCALAR);
11682 SPAGAIN;
11683 ret = SvTRUE(TOPs);
11684 *offset = SvIV(offsv);
11685 PUTBACK;
11686 FREETMPS;
11687 LEAVE;
389edf32 11688 }
bd81e77b
NC
11689 else
11690 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11691 return ret;
1d7c1841 11692
bd81e77b 11693}
1d7c1841 11694
bd81e77b
NC
11695/* ---------------------------------------------------------------------
11696 *
11697 * support functions for report_uninit()
11698 */
1d7c1841 11699
bd81e77b
NC
11700/* the maxiumum size of array or hash where we will scan looking
11701 * for the undefined element that triggered the warning */
1d7c1841 11702
bd81e77b 11703#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11704
bd81e77b
NC
11705/* Look for an entry in the hash whose value has the same SV as val;
11706 * If so, return a mortal copy of the key. */
1d7c1841 11707
bd81e77b
NC
11708STATIC SV*
11709S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11710{
11711 dVAR;
11712 register HE **array;
11713 I32 i;
6c3182a5 11714
bd81e77b
NC
11715 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11716 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11717 return NULL;
6c3182a5 11718
bd81e77b 11719 array = HvARRAY(hv);
6c3182a5 11720
bd81e77b
NC
11721 for (i=HvMAX(hv); i>0; i--) {
11722 register HE *entry;
11723 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11724 if (HeVAL(entry) != val)
11725 continue;
11726 if ( HeVAL(entry) == &PL_sv_undef ||
11727 HeVAL(entry) == &PL_sv_placeholder)
11728 continue;
11729 if (!HeKEY(entry))
a0714e2c 11730 return NULL;
bd81e77b
NC
11731 if (HeKLEN(entry) == HEf_SVKEY)
11732 return sv_mortalcopy(HeKEY_sv(entry));
11733 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11734 }
11735 }
a0714e2c 11736 return NULL;
bd81e77b 11737}
6c3182a5 11738
bd81e77b
NC
11739/* Look for an entry in the array whose value has the same SV as val;
11740 * If so, return the index, otherwise return -1. */
6c3182a5 11741
bd81e77b
NC
11742STATIC I32
11743S_find_array_subscript(pTHX_ AV *av, SV* val)
11744{
97aff369 11745 dVAR;
bd81e77b
NC
11746 SV** svp;
11747 I32 i;
11748 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11749 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11750 return -1;
57c6e6d2 11751
bd81e77b
NC
11752 svp = AvARRAY(av);
11753 for (i=AvFILLp(av); i>=0; i--) {
11754 if (svp[i] == val && svp[i] != &PL_sv_undef)
11755 return i;
11756 }
11757 return -1;
11758}
15a5279a 11759
bd81e77b
NC
11760/* S_varname(): return the name of a variable, optionally with a subscript.
11761 * If gv is non-zero, use the name of that global, along with gvtype (one
11762 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11763 * targ. Depending on the value of the subscript_type flag, return:
11764 */
bce260cd 11765
bd81e77b
NC
11766#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11767#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11768#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11769#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11770
bd81e77b
NC
11771STATIC SV*
11772S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11773 SV* keyname, I32 aindex, int subscript_type)
11774{
1d7c1841 11775
bd81e77b
NC
11776 SV * const name = sv_newmortal();
11777 if (gv) {
11778 char buffer[2];
11779 buffer[0] = gvtype;
11780 buffer[1] = 0;
1d7c1841 11781
bd81e77b 11782 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11783
bd81e77b 11784 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11785
bd81e77b
NC
11786 if ((unsigned int)SvPVX(name)[1] <= 26) {
11787 buffer[0] = '^';
11788 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11789
bd81e77b
NC
11790 /* Swap the 1 unprintable control character for the 2 byte pretty
11791 version - ie substr($name, 1, 1) = $buffer; */
11792 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11793 }
bd81e77b
NC
11794 }
11795 else {
11796 U32 unused;
11797 CV * const cv = find_runcv(&unused);
11798 SV *sv;
11799 AV *av;
1d7c1841 11800
bd81e77b 11801 if (!cv || !CvPADLIST(cv))
a0714e2c 11802 return NULL;
bd81e77b
NC
11803 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11804 sv = *av_fetch(av, targ, FALSE);
11805 /* SvLEN in a pad name is not to be trusted */
11806 sv_setpv(name, SvPV_nolen_const(sv));
11807 }
1d7c1841 11808
bd81e77b 11809 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11810 SV * const sv = newSV(0);
bd81e77b
NC
11811 *SvPVX(name) = '$';
11812 Perl_sv_catpvf(aTHX_ name, "{%s}",
11813 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11814 SvREFCNT_dec(sv);
11815 }
11816 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11817 *SvPVX(name) = '$';
11818 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11819 }
11820 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11821 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11822
bd81e77b
NC
11823 return name;
11824}
1d7c1841 11825
1d7c1841 11826
bd81e77b
NC
11827/*
11828=for apidoc find_uninit_var
1d7c1841 11829
bd81e77b
NC
11830Find the name of the undefined variable (if any) that caused the operator o
11831to issue a "Use of uninitialized value" warning.
11832If match is true, only return a name if it's value matches uninit_sv.
11833So roughly speaking, if a unary operator (such as OP_COS) generates a
11834warning, then following the direct child of the op may yield an
11835OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11836other hand, with OP_ADD there are two branches to follow, so we only print
11837the variable name if we get an exact match.
1d7c1841 11838
bd81e77b 11839The name is returned as a mortal SV.
1d7c1841 11840
bd81e77b
NC
11841Assumes that PL_op is the op that originally triggered the error, and that
11842PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11843
bd81e77b
NC
11844=cut
11845*/
1d7c1841 11846
bd81e77b
NC
11847STATIC SV *
11848S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11849{
11850 dVAR;
11851 SV *sv;
11852 AV *av;
11853 GV *gv;
11854 OP *o, *o2, *kid;
1d7c1841 11855
bd81e77b
NC
11856 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11857 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11858 return NULL;
1d7c1841 11859
bd81e77b 11860 switch (obase->op_type) {
1d7c1841 11861
bd81e77b
NC
11862 case OP_RV2AV:
11863 case OP_RV2HV:
11864 case OP_PADAV:
11865 case OP_PADHV:
11866 {
11867 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11868 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11869 I32 index = 0;
a0714e2c 11870 SV *keysv = NULL;
bd81e77b 11871 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11872
bd81e77b
NC
11873 if (pad) { /* @lex, %lex */
11874 sv = PAD_SVl(obase->op_targ);
a0714e2c 11875 gv = NULL;
bd81e77b
NC
11876 }
11877 else {
11878 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11879 /* @global, %global */
11880 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11881 if (!gv)
11882 break;
11883 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11884 }
11885 else /* @{expr}, %{expr} */
11886 return find_uninit_var(cUNOPx(obase)->op_first,
11887 uninit_sv, match);
11888 }
1d7c1841 11889
bd81e77b
NC
11890 /* attempt to find a match within the aggregate */
11891 if (hash) {
d4c19fe8 11892 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11893 if (keysv)
11894 subscript_type = FUV_SUBSCRIPT_HASH;
11895 }
11896 else {
e15d5972 11897 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
11898 if (index >= 0)
11899 subscript_type = FUV_SUBSCRIPT_ARRAY;
11900 }
1d7c1841 11901
bd81e77b
NC
11902 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11903 break;
1d7c1841 11904
bd81e77b
NC
11905 return varname(gv, hash ? '%' : '@', obase->op_targ,
11906 keysv, index, subscript_type);
11907 }
1d7c1841 11908
bd81e77b
NC
11909 case OP_PADSV:
11910 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11911 break;
a0714e2c
SS
11912 return varname(NULL, '$', obase->op_targ,
11913 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11914
bd81e77b
NC
11915 case OP_GVSV:
11916 gv = cGVOPx_gv(obase);
11917 if (!gv || (match && GvSV(gv) != uninit_sv))
11918 break;
a0714e2c 11919 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11920
bd81e77b
NC
11921 case OP_AELEMFAST:
11922 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11923 if (match) {
11924 SV **svp;
11925 av = (AV*)PAD_SV(obase->op_targ);
11926 if (!av || SvRMAGICAL(av))
11927 break;
11928 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11929 if (!svp || *svp != uninit_sv)
11930 break;
11931 }
a0714e2c
SS
11932 return varname(NULL, '$', obase->op_targ,
11933 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11934 }
11935 else {
11936 gv = cGVOPx_gv(obase);
11937 if (!gv)
11938 break;
11939 if (match) {
11940 SV **svp;
11941 av = GvAV(gv);
11942 if (!av || SvRMAGICAL(av))
11943 break;
11944 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11945 if (!svp || *svp != uninit_sv)
11946 break;
11947 }
11948 return varname(gv, '$', 0,
a0714e2c 11949 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11950 }
11951 break;
1d7c1841 11952
bd81e77b
NC
11953 case OP_EXISTS:
11954 o = cUNOPx(obase)->op_first;
11955 if (!o || o->op_type != OP_NULL ||
11956 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11957 break;
11958 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11959
bd81e77b
NC
11960 case OP_AELEM:
11961 case OP_HELEM:
11962 if (PL_op == obase)
11963 /* $a[uninit_expr] or $h{uninit_expr} */
11964 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11965
a0714e2c 11966 gv = NULL;
bd81e77b
NC
11967 o = cBINOPx(obase)->op_first;
11968 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11969
bd81e77b 11970 /* get the av or hv, and optionally the gv */
a0714e2c 11971 sv = NULL;
bd81e77b
NC
11972 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11973 sv = PAD_SV(o->op_targ);
11974 }
11975 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11976 && cUNOPo->op_first->op_type == OP_GV)
11977 {
11978 gv = cGVOPx_gv(cUNOPo->op_first);
11979 if (!gv)
11980 break;
11981 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11982 }
11983 if (!sv)
11984 break;
11985
11986 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11987 /* index is constant */
11988 if (match) {
11989 if (SvMAGICAL(sv))
11990 break;
11991 if (obase->op_type == OP_HELEM) {
11992 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11993 if (!he || HeVAL(he) != uninit_sv)
11994 break;
11995 }
11996 else {
00b6aa41 11997 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11998 if (!svp || *svp != uninit_sv)
11999 break;
12000 }
12001 }
12002 if (obase->op_type == OP_HELEM)
12003 return varname(gv, '%', o->op_targ,
12004 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12005 else
a0714e2c 12006 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 12007 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12008 }
12009 else {
12010 /* index is an expression;
12011 * attempt to find a match within the aggregate */
12012 if (obase->op_type == OP_HELEM) {
d4c19fe8 12013 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
12014 if (keysv)
12015 return varname(gv, '%', o->op_targ,
12016 keysv, 0, FUV_SUBSCRIPT_HASH);
12017 }
12018 else {
d4c19fe8 12019 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
12020 if (index >= 0)
12021 return varname(gv, '@', o->op_targ,
a0714e2c 12022 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12023 }
12024 if (match)
12025 break;
12026 return varname(gv,
12027 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12028 ? '@' : '%',
a0714e2c 12029 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 12030 }
bd81e77b 12031 break;
dc507217 12032
bd81e77b
NC
12033 case OP_AASSIGN:
12034 /* only examine RHS */
12035 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 12036
bd81e77b
NC
12037 case OP_OPEN:
12038 o = cUNOPx(obase)->op_first;
12039 if (o->op_type == OP_PUSHMARK)
12040 o = o->op_sibling;
1d7c1841 12041
bd81e77b
NC
12042 if (!o->op_sibling) {
12043 /* one-arg version of open is highly magical */
a0ae6670 12044
bd81e77b
NC
12045 if (o->op_type == OP_GV) { /* open FOO; */
12046 gv = cGVOPx_gv(o);
12047 if (match && GvSV(gv) != uninit_sv)
12048 break;
12049 return varname(gv, '$', 0,
a0714e2c 12050 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12051 }
12052 /* other possibilities not handled are:
12053 * open $x; or open my $x; should return '${*$x}'
12054 * open expr; should return '$'.expr ideally
12055 */
12056 break;
12057 }
12058 goto do_op;
ccfc67b7 12059
bd81e77b
NC
12060 /* ops where $_ may be an implicit arg */
12061 case OP_TRANS:
12062 case OP_SUBST:
12063 case OP_MATCH:
12064 if ( !(obase->op_flags & OPf_STACKED)) {
12065 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12066 ? PAD_SVl(obase->op_targ)
12067 : DEFSV))
12068 {
12069 sv = sv_newmortal();
12070 sv_setpvn(sv, "$_", 2);
12071 return sv;
12072 }
12073 }
12074 goto do_op;
9f4817db 12075
bd81e77b
NC
12076 case OP_PRTF:
12077 case OP_PRINT:
12078 /* skip filehandle as it can't produce 'undef' warning */
12079 o = cUNOPx(obase)->op_first;
12080 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12081 o = o->op_sibling->op_sibling;
12082 goto do_op2;
9f4817db 12083
9f4817db 12084
bd81e77b
NC
12085 case OP_RV2SV:
12086 case OP_CUSTOM:
12087 case OP_ENTERSUB:
12088 match = 1; /* XS or custom code could trigger random warnings */
12089 goto do_op;
9f4817db 12090
bd81e77b
NC
12091 case OP_SCHOMP:
12092 case OP_CHOMP:
12093 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12094 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12095 /*FALLTHROUGH*/
5d170f3a 12096
bd81e77b
NC
12097 default:
12098 do_op:
12099 if (!(obase->op_flags & OPf_KIDS))
12100 break;
12101 o = cUNOPx(obase)->op_first;
12102
12103 do_op2:
12104 if (!o)
12105 break;
f9893866 12106
bd81e77b
NC
12107 /* if all except one arg are constant, or have no side-effects,
12108 * or are optimized away, then it's unambiguous */
5f66b61c 12109 o2 = NULL;
bd81e77b 12110 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
12111 if (kid) {
12112 const OPCODE type = kid->op_type;
12113 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
12114 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12115 || (type == OP_PUSHMARK)
bd81e77b 12116 )
bd81e77b 12117 continue;
e15d5972 12118 }
bd81e77b 12119 if (o2) { /* more than one found */
5f66b61c 12120 o2 = NULL;
bd81e77b
NC
12121 break;
12122 }
12123 o2 = kid;
12124 }
12125 if (o2)
12126 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12127
bd81e77b
NC
12128 /* scan all args */
12129 while (o) {
12130 sv = find_uninit_var(o, uninit_sv, 1);
12131 if (sv)
12132 return sv;
12133 o = o->op_sibling;
d0063567 12134 }
bd81e77b 12135 break;
f9893866 12136 }
a0714e2c 12137 return NULL;
9f4817db
JH
12138}
12139
220e2d4e 12140
bd81e77b
NC
12141/*
12142=for apidoc report_uninit
68795e93 12143
bd81e77b 12144Print appropriate "Use of uninitialized variable" warning
220e2d4e 12145
bd81e77b
NC
12146=cut
12147*/
220e2d4e 12148
bd81e77b
NC
12149void
12150Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12151{
97aff369 12152 dVAR;
bd81e77b 12153 if (PL_op) {
a0714e2c 12154 SV* varname = NULL;
bd81e77b
NC
12155 if (uninit_sv) {
12156 varname = find_uninit_var(PL_op, uninit_sv,0);
12157 if (varname)
12158 sv_insert(varname, 0, 0, " ", 1);
12159 }
12160 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12161 varname ? SvPV_nolen_const(varname) : "",
12162 " in ", OP_DESC(PL_op));
220e2d4e 12163 }
a73e8557 12164 else
bd81e77b
NC
12165 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12166 "", "", "");
220e2d4e 12167}
f9893866 12168
241d1a3b
NC
12169/*
12170 * Local variables:
12171 * c-indentation-style: bsd
12172 * c-basic-offset: 4
12173 * indent-tabs-mode: t
12174 * End:
12175 *
37442d52
RGS
12176 * ex: set ts=8 sts=4 sw=4 noet:
12177 */