This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Deep in the bowels of creating new and exciting bugs, I managed to
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b94e2f88 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137 32#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 33/* if adding more checks watch out for the following tests:
e23c8137
JH
34 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
35 * lib/utf8.t lib/Unicode/Collate/t/index.t
36 * --jhi
37 */
38#define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
39 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
40 assert((cache)[2] <= (cache)[3]); \
41 assert((cache)[3] <= (cache)[1]);} \
42 } STMT_END
e23c8137
JH
43#else
44#define ASSERT_UTF8_CACHE(cache) NOOP
45#endif
46
f8c7b90f 47#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 48#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 49#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 50/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 51 on-write. */
765f542d 52#endif
645c22ef
DM
53
54/* ============================================================================
55
56=head1 Allocation and deallocation of SVs.
57
d2a0f284
JC
58An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
59sv, av, hv...) contains type and reference count information, and for
60many types, a pointer to the body (struct xrv, xpv, xpviv...), which
61contains fields specific to each type. Some types store all they need
62in the head, so don't have a body.
63
64In all but the most memory-paranoid configuations (ex: PURIFY), heads
65and bodies are allocated out of arenas, which by default are
66approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
67Sv-bodies are allocated by their sv-type, guaranteeing size
68consistency needed to allocate safely from arrays.
69
d2a0f284
JC
70For SV-heads, the first slot in each arena is reserved, and holds a
71link to the next arena, some flags, and a note of the number of slots.
72Snaked through each arena chain is a linked list of free items; when
73this becomes empty, an extra arena is allocated and divided up into N
74items which are threaded into the free list.
75
76SV-bodies are similar, but they use arena-sets by default, which
77separate the link and info from the arena itself, and reclaim the 1st
78slot in the arena. SV-bodies are further described later.
645c22ef
DM
79
80The following global variables are associated with arenas:
81
82 PL_sv_arenaroot pointer to list of SV arenas
83 PL_sv_root pointer to list of free SV structures
84
d2a0f284
JC
85 PL_body_arenas head of linked-list of body arenas
86 PL_body_roots[] array of pointers to list of free bodies of svtype
87 arrays are indexed by the svtype needed
93e68bfb 88
d2a0f284
JC
89A few special SV heads are not allocated from an arena, but are
90instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
91The size of arenas can be changed from the default by setting
92PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
93
94The SV arena serves the secondary purpose of allowing still-live SVs
95to be located and destroyed during final cleanup.
96
97At the lowest level, the macros new_SV() and del_SV() grab and free
98an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
99to return the SV to the free list with error checking.) new_SV() calls
100more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
101SVs in the free list have their SvTYPE field set to all ones.
102
ff276b08 103At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 104perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 105start of the interpreter.
645c22ef
DM
106
107Manipulation of any of the PL_*root pointers is protected by enclosing
108LOCK_SV_MUTEX; ... UNLOCK_SV_MUTEX calls which should Do the Right Thing
109if threads are enabled.
110
111The function visit() scans the SV arenas list, and calls a specified
112function for each SV it finds which is still live - ie which has an SvTYPE
113other than all 1's, and a non-zero SvREFCNT. visit() is used by the
114following functions (specified as [function that calls visit()] / [function
115called by visit() for each SV]):
116
117 sv_report_used() / do_report_used()
f2524eef 118 dump all remaining SVs (debugging aid)
645c22ef
DM
119
120 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
121 Attempt to free all objects pointed to by RVs,
122 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
123 try to do the same for all objects indirectly
124 referenced by typeglobs too. Called once from
125 perl_destruct(), prior to calling sv_clean_all()
126 below.
127
128 sv_clean_all() / do_clean_all()
129 SvREFCNT_dec(sv) each remaining SV, possibly
130 triggering an sv_free(). It also sets the
131 SVf_BREAK flag on the SV to indicate that the
132 refcnt has been artificially lowered, and thus
133 stopping sv_free() from giving spurious warnings
134 about SVs which unexpectedly have a refcnt
135 of zero. called repeatedly from perl_destruct()
136 until there are no SVs left.
137
93e68bfb 138=head2 Arena allocator API Summary
645c22ef
DM
139
140Private API to rest of sv.c
141
142 new_SV(), del_SV(),
143
144 new_XIV(), del_XIV(),
145 new_XNV(), del_XNV(),
146 etc
147
148Public API:
149
8cf8f3d1 150 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 151
645c22ef
DM
152=cut
153
154============================================================================ */
155
4561caa4
CS
156/*
157 * "A time to plant, and a time to uproot what was planted..."
158 */
159
77354fb4
NC
160/*
161 * nice_chunk and nice_chunk size need to be set
162 * and queried under the protection of sv_mutex
163 */
164void
165Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
166{
97aff369 167 dVAR;
77354fb4
NC
168 void *new_chunk;
169 U32 new_chunk_size;
170 LOCK_SV_MUTEX;
171 new_chunk = (void *)(chunk);
172 new_chunk_size = (chunk_size);
173 if (new_chunk_size > PL_nice_chunk_size) {
174 Safefree(PL_nice_chunk);
175 PL_nice_chunk = (char *) new_chunk;
176 PL_nice_chunk_size = new_chunk_size;
177 } else {
178 Safefree(chunk);
179 }
180 UNLOCK_SV_MUTEX;
181}
cac9b346 182
fd0854ff 183#ifdef DEBUG_LEAKING_SCALARS
22162ca8 184# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
185#else
186# define FREE_SV_DEBUG_FILE(sv)
187#endif
188
48614a46
NC
189#ifdef PERL_POISON
190# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
191/* Whilst I'd love to do this, it seems that things like to check on
192 unreferenced scalars
7e337ee0 193# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 194*/
7e337ee0
JH
195# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
196 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
197#else
198# define SvARENA_CHAIN(sv) SvANY(sv)
199# define POSION_SV_HEAD(sv)
200#endif
201
053fc874
GS
202#define plant_SV(p) \
203 STMT_START { \
fd0854ff 204 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
205 POSION_SV_HEAD(p); \
206 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
207 SvFLAGS(p) = SVTYPEMASK; \
208 PL_sv_root = (p); \
209 --PL_sv_count; \
210 } STMT_END
a0d0e21e 211
fba3b22e 212/* sv_mutex must be held while calling uproot_SV() */
053fc874
GS
213#define uproot_SV(p) \
214 STMT_START { \
215 (p) = PL_sv_root; \
bb7bbd9c 216 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
217 ++PL_sv_count; \
218 } STMT_END
219
645c22ef 220
cac9b346
NC
221/* make some more SVs by adding another arena */
222
223/* sv_mutex must be held while calling more_sv() */
224STATIC SV*
225S_more_sv(pTHX)
226{
97aff369 227 dVAR;
cac9b346
NC
228 SV* sv;
229
230 if (PL_nice_chunk) {
231 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 232 PL_nice_chunk = NULL;
cac9b346
NC
233 PL_nice_chunk_size = 0;
234 }
235 else {
236 char *chunk; /* must use New here to match call to */
d2a0f284 237 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 238 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
239 }
240 uproot_SV(sv);
241 return sv;
242}
243
645c22ef
DM
244/* new_SV(): return a new, empty SV head */
245
eba0f806
DM
246#ifdef DEBUG_LEAKING_SCALARS
247/* provide a real function for a debugger to play with */
248STATIC SV*
249S_new_SV(pTHX)
250{
251 SV* sv;
252
253 LOCK_SV_MUTEX;
254 if (PL_sv_root)
255 uproot_SV(sv);
256 else
cac9b346 257 sv = S_more_sv(aTHX);
eba0f806
DM
258 UNLOCK_SV_MUTEX;
259 SvANY(sv) = 0;
260 SvREFCNT(sv) = 1;
261 SvFLAGS(sv) = 0;
fd0854ff
DM
262 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
263 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
264 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
265 sv->sv_debug_inpad = 0;
266 sv->sv_debug_cloned = 0;
fd0854ff 267 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 268
eba0f806
DM
269 return sv;
270}
271# define new_SV(p) (p)=S_new_SV(aTHX)
272
273#else
274# define new_SV(p) \
053fc874
GS
275 STMT_START { \
276 LOCK_SV_MUTEX; \
277 if (PL_sv_root) \
278 uproot_SV(p); \
279 else \
cac9b346 280 (p) = S_more_sv(aTHX); \
053fc874
GS
281 UNLOCK_SV_MUTEX; \
282 SvANY(p) = 0; \
283 SvREFCNT(p) = 1; \
284 SvFLAGS(p) = 0; \
285 } STMT_END
eba0f806 286#endif
463ee0b2 287
645c22ef
DM
288
289/* del_SV(): return an empty SV head to the free list */
290
a0d0e21e 291#ifdef DEBUGGING
4561caa4 292
053fc874
GS
293#define del_SV(p) \
294 STMT_START { \
295 LOCK_SV_MUTEX; \
aea4f609 296 if (DEBUG_D_TEST) \
053fc874
GS
297 del_sv(p); \
298 else \
299 plant_SV(p); \
300 UNLOCK_SV_MUTEX; \
301 } STMT_END
a0d0e21e 302
76e3520e 303STATIC void
cea2e8a9 304S_del_sv(pTHX_ SV *p)
463ee0b2 305{
97aff369 306 dVAR;
aea4f609 307 if (DEBUG_D_TEST) {
4633a7c4 308 SV* sva;
a3b680e6 309 bool ok = 0;
3280af22 310 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
311 const SV * const sv = sva + 1;
312 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 313 if (p >= sv && p < svend) {
a0d0e21e 314 ok = 1;
c0ff570e
NC
315 break;
316 }
a0d0e21e
LW
317 }
318 if (!ok) {
0453d815 319 if (ckWARN_d(WARN_INTERNAL))
9014280d 320 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
321 "Attempt to free non-arena SV: 0x%"UVxf
322 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
323 return;
324 }
325 }
4561caa4 326 plant_SV(p);
463ee0b2 327}
a0d0e21e 328
4561caa4
CS
329#else /* ! DEBUGGING */
330
331#define del_SV(p) plant_SV(p)
332
333#endif /* DEBUGGING */
463ee0b2 334
645c22ef
DM
335
336/*
ccfc67b7
JH
337=head1 SV Manipulation Functions
338
645c22ef
DM
339=for apidoc sv_add_arena
340
341Given a chunk of memory, link it to the head of the list of arenas,
342and split it into a list of free SVs.
343
344=cut
345*/
346
4633a7c4 347void
864dbfa3 348Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 349{
97aff369 350 dVAR;
0bd48802 351 SV* const sva = (SV*)ptr;
463ee0b2
LW
352 register SV* sv;
353 register SV* svend;
4633a7c4
LW
354
355 /* The first SV in an arena isn't an SV. */
3280af22 356 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
357 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
358 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
359
3280af22
NIS
360 PL_sv_arenaroot = sva;
361 PL_sv_root = sva + 1;
4633a7c4
LW
362
363 svend = &sva[SvREFCNT(sva) - 1];
364 sv = sva + 1;
463ee0b2 365 while (sv < svend) {
48614a46 366 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 367#ifdef DEBUGGING
978b032e 368 SvREFCNT(sv) = 0;
03e36789
NC
369#endif
370 /* Must always set typemask because it's awlays checked in on cleanup
371 when the arenas are walked looking for objects. */
8990e307 372 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
373 sv++;
374 }
48614a46 375 SvARENA_CHAIN(sv) = 0;
03e36789
NC
376#ifdef DEBUGGING
377 SvREFCNT(sv) = 0;
378#endif
4633a7c4
LW
379 SvFLAGS(sv) = SVTYPEMASK;
380}
381
055972dc
DM
382/* visit(): call the named function for each non-free SV in the arenas
383 * whose flags field matches the flags/mask args. */
645c22ef 384
5226ed68 385STATIC I32
055972dc 386S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 387{
97aff369 388 dVAR;
4633a7c4 389 SV* sva;
5226ed68 390 I32 visited = 0;
8990e307 391
3280af22 392 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 393 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 394 register SV* sv;
4561caa4 395 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
396 if (SvTYPE(sv) != SVTYPEMASK
397 && (sv->sv_flags & mask) == flags
398 && SvREFCNT(sv))
399 {
acfe0abc 400 (FCALL)(aTHX_ sv);
5226ed68
JH
401 ++visited;
402 }
8990e307
LW
403 }
404 }
5226ed68 405 return visited;
8990e307
LW
406}
407
758a08c3
JH
408#ifdef DEBUGGING
409
645c22ef
DM
410/* called by sv_report_used() for each live SV */
411
412static void
acfe0abc 413do_report_used(pTHX_ SV *sv)
645c22ef
DM
414{
415 if (SvTYPE(sv) != SVTYPEMASK) {
416 PerlIO_printf(Perl_debug_log, "****\n");
417 sv_dump(sv);
418 }
419}
758a08c3 420#endif
645c22ef
DM
421
422/*
423=for apidoc sv_report_used
424
425Dump the contents of all SVs not yet freed. (Debugging aid).
426
427=cut
428*/
429
8990e307 430void
864dbfa3 431Perl_sv_report_used(pTHX)
4561caa4 432{
ff270d3a 433#ifdef DEBUGGING
055972dc 434 visit(do_report_used, 0, 0);
96a5add6
AL
435#else
436 PERL_UNUSED_CONTEXT;
ff270d3a 437#endif
4561caa4
CS
438}
439
645c22ef
DM
440/* called by sv_clean_objs() for each live SV */
441
442static void
e15faf7d 443do_clean_objs(pTHX_ SV *ref)
645c22ef 444{
97aff369 445 dVAR;
823a54a3
AL
446 if (SvROK(ref)) {
447 SV * const target = SvRV(ref);
448 if (SvOBJECT(target)) {
449 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
450 if (SvWEAKREF(ref)) {
451 sv_del_backref(target, ref);
452 SvWEAKREF_off(ref);
453 SvRV_set(ref, NULL);
454 } else {
455 SvROK_off(ref);
456 SvRV_set(ref, NULL);
457 SvREFCNT_dec(target);
458 }
645c22ef
DM
459 }
460 }
461
462 /* XXX Might want to check arrays, etc. */
463}
464
465/* called by sv_clean_objs() for each live SV */
466
467#ifndef DISABLE_DESTRUCTOR_KLUDGE
468static void
acfe0abc 469do_clean_named_objs(pTHX_ SV *sv)
645c22ef 470{
97aff369 471 dVAR;
f7877b28 472 if (SvTYPE(sv) == SVt_PVGV && isGV_with_GP(sv) && GvGP(sv)) {
c69033f2
NC
473 if ((
474#ifdef PERL_DONT_CREATE_GVSV
475 GvSV(sv) &&
476#endif
477 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
478 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
479 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
480 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
481 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
482 {
483 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 484 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
485 SvREFCNT_dec(sv);
486 }
487 }
488}
489#endif
490
491/*
492=for apidoc sv_clean_objs
493
494Attempt to destroy all objects not yet freed
495
496=cut
497*/
498
4561caa4 499void
864dbfa3 500Perl_sv_clean_objs(pTHX)
4561caa4 501{
97aff369 502 dVAR;
3280af22 503 PL_in_clean_objs = TRUE;
055972dc 504 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 505#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 506 /* some barnacles may yet remain, clinging to typeglobs */
055972dc 507 visit(do_clean_named_objs, SVt_PVGV, SVTYPEMASK);
4561caa4 508#endif
3280af22 509 PL_in_clean_objs = FALSE;
4561caa4
CS
510}
511
645c22ef
DM
512/* called by sv_clean_all() for each live SV */
513
514static void
acfe0abc 515do_clean_all(pTHX_ SV *sv)
645c22ef 516{
97aff369 517 dVAR;
645c22ef
DM
518 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
519 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b 520 if (PL_comppad == (AV*)sv) {
7d49f689 521 PL_comppad = NULL;
4608196e 522 PL_curpad = NULL;
0e705b3b 523 }
645c22ef
DM
524 SvREFCNT_dec(sv);
525}
526
527/*
528=for apidoc sv_clean_all
529
530Decrement the refcnt of each remaining SV, possibly triggering a
531cleanup. This function may have to be called multiple times to free
ff276b08 532SVs which are in complex self-referential hierarchies.
645c22ef
DM
533
534=cut
535*/
536
5226ed68 537I32
864dbfa3 538Perl_sv_clean_all(pTHX)
8990e307 539{
97aff369 540 dVAR;
5226ed68 541 I32 cleaned;
3280af22 542 PL_in_clean_all = TRUE;
055972dc 543 cleaned = visit(do_clean_all, 0,0);
3280af22 544 PL_in_clean_all = FALSE;
5226ed68 545 return cleaned;
8990e307 546}
463ee0b2 547
5e258f8c
JC
548/*
549 ARENASETS: a meta-arena implementation which separates arena-info
550 into struct arena_set, which contains an array of struct
551 arena_descs, each holding info for a single arena. By separating
552 the meta-info from the arena, we recover the 1st slot, formerly
553 borrowed for list management. The arena_set is about the size of an
554 arena, avoiding the needless malloc overhead of a naive linked-list
555
556 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
557 memory in the last arena-set (1/2 on average). In trade, we get
558 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284
JC
559 smaller types). The recovery of the wasted space allows use of
560 small arenas for large, rare body types,
5e258f8c 561*/
5e258f8c 562struct arena_desc {
398c677b
NC
563 char *arena; /* the raw storage, allocated aligned */
564 size_t size; /* its size ~4k typ */
565 int unit_type; /* useful for arena audits */
5e258f8c
JC
566 /* info for sv-heads (eventually)
567 int count, flags;
568 */
569};
570
e6148039
NC
571struct arena_set;
572
573/* Get the maximum number of elements in set[] such that struct arena_set
574 will fit within PERL_ARENA_SIZE, which is probabably just under 4K, and
575 therefore likely to be 1 aligned memory page. */
576
577#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
578 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
579
580struct arena_set {
581 struct arena_set* next;
582 int set_size; /* ie ARENAS_PER_SET */
583 int curr; /* index of next available arena-desc */
584 struct arena_desc set[ARENAS_PER_SET];
585};
586
645c22ef
DM
587/*
588=for apidoc sv_free_arenas
589
590Deallocate the memory used by all arenas. Note that all the individual SV
591heads and bodies within the arenas must already have been freed.
592
593=cut
594*/
4633a7c4 595void
864dbfa3 596Perl_sv_free_arenas(pTHX)
4633a7c4 597{
97aff369 598 dVAR;
4633a7c4
LW
599 SV* sva;
600 SV* svanext;
93e68bfb 601 int i;
4633a7c4
LW
602
603 /* Free arenas here, but be careful about fake ones. (We assume
604 contiguity of the fake ones with the corresponding real ones.) */
605
3280af22 606 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
607 svanext = (SV*) SvANY(sva);
608 while (svanext && SvFAKE(svanext))
609 svanext = (SV*) SvANY(svanext);
610
611 if (!SvFAKE(sva))
1df70142 612 Safefree(sva);
4633a7c4 613 }
93e68bfb 614
5e258f8c
JC
615 {
616 struct arena_set *next, *aroot = (struct arena_set*) PL_body_arenas;
617
618 for (; aroot; aroot = next) {
96a5add6 619 const int max = aroot->curr;
5e258f8c
JC
620 for (i=0; i<max; i++) {
621 assert(aroot->set[i].arena);
622 Safefree(aroot->set[i].arena);
623 }
624 next = aroot->next;
625 Safefree(aroot);
626 }
627 }
dc8220bf 628 PL_body_arenas = 0;
fdda85ca 629
232d1c15 630 for (i=0; i<PERL_ARENA_ROOTS_SIZE; i++)
93e68bfb 631 PL_body_roots[i] = 0;
93e68bfb 632
43c5f42d 633 Safefree(PL_nice_chunk);
bd61b366 634 PL_nice_chunk = NULL;
3280af22
NIS
635 PL_nice_chunk_size = 0;
636 PL_sv_arenaroot = 0;
637 PL_sv_root = 0;
4633a7c4
LW
638}
639
bd81e77b
NC
640/*
641 Here are mid-level routines that manage the allocation of bodies out
642 of the various arenas. There are 5 kinds of arenas:
29489e7c 643
bd81e77b
NC
644 1. SV-head arenas, which are discussed and handled above
645 2. regular body arenas
646 3. arenas for reduced-size bodies
647 4. Hash-Entry arenas
648 5. pte arenas (thread related)
29489e7c 649
bd81e77b
NC
650 Arena types 2 & 3 are chained by body-type off an array of
651 arena-root pointers, which is indexed by svtype. Some of the
652 larger/less used body types are malloced singly, since a large
653 unused block of them is wasteful. Also, several svtypes dont have
654 bodies; the data fits into the sv-head itself. The arena-root
655 pointer thus has a few unused root-pointers (which may be hijacked
656 later for arena types 4,5)
29489e7c 657
bd81e77b
NC
658 3 differs from 2 as an optimization; some body types have several
659 unused fields in the front of the structure (which are kept in-place
660 for consistency). These bodies can be allocated in smaller chunks,
661 because the leading fields arent accessed. Pointers to such bodies
662 are decremented to point at the unused 'ghost' memory, knowing that
663 the pointers are used with offsets to the real memory.
29489e7c 664
bd81e77b
NC
665 HE, HEK arenas are managed separately, with separate code, but may
666 be merge-able later..
667
668 PTE arenas are not sv-bodies, but they share these mid-level
669 mechanics, so are considered here. The new mid-level mechanics rely
670 on the sv_type of the body being allocated, so we just reserve one
671 of the unused body-slots for PTEs, then use it in those (2) PTE
672 contexts below (line ~10k)
673*/
674
bd26d9a3 675/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
676 TBD: export properly for hv.c: S_more_he().
677*/
678void*
679Perl_get_arena(pTHX_ int arena_size)
680{
5e258f8c 681 struct arena_desc* adesc;
476a1e16 682 struct arena_set *newroot, **aroot = (struct arena_set**) &PL_body_arenas;
5e258f8c
JC
683 int curr;
684
476a1e16
JC
685 /* shouldnt need this
686 if (!arena_size) arena_size = PERL_ARENA_SIZE;
687 */
5e258f8c
JC
688
689 /* may need new arena-set to hold new arena */
476a1e16 690 if (!*aroot || (*aroot)->curr >= (*aroot)->set_size) {
5e258f8c
JC
691 Newxz(newroot, 1, struct arena_set);
692 newroot->set_size = ARENAS_PER_SET;
476a1e16
JC
693 newroot->next = *aroot;
694 *aroot = newroot;
695 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", *aroot));
5e258f8c
JC
696 }
697
698 /* ok, now have arena-set with at least 1 empty/available arena-desc */
476a1e16
JC
699 curr = (*aroot)->curr++;
700 adesc = &((*aroot)->set[curr]);
5e258f8c
JC
701 assert(!adesc->arena);
702
5e258f8c
JC
703 Newxz(adesc->arena, arena_size, char);
704 adesc->size = arena_size;
d2a0f284
JC
705 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %d\n",
706 curr, adesc->arena, arena_size));
5e258f8c
JC
707
708 return adesc->arena;
5e258f8c
JC
709}
710
53c1dcc0 711
bd81e77b 712/* return a thing to the free list */
29489e7c 713
bd81e77b
NC
714#define del_body(thing, root) \
715 STMT_START { \
00b6aa41 716 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
717 LOCK_SV_MUTEX; \
718 *thing_copy = *root; \
719 *root = (void*)thing_copy; \
720 UNLOCK_SV_MUTEX; \
721 } STMT_END
29489e7c 722
bd81e77b 723/*
d2a0f284
JC
724
725=head1 SV-Body Allocation
726
727Allocation of SV-bodies is similar to SV-heads, differing as follows;
728the allocation mechanism is used for many body types, so is somewhat
729more complicated, it uses arena-sets, and has no need for still-live
730SV detection.
731
732At the outermost level, (new|del)_X*V macros return bodies of the
733appropriate type. These macros call either (new|del)_body_type or
734(new|del)_body_allocated macro pairs, depending on specifics of the
735type. Most body types use the former pair, the latter pair is used to
736allocate body types with "ghost fields".
737
738"ghost fields" are fields that are unused in certain types, and
739consequently dont need to actually exist. They are declared because
740they're part of a "base type", which allows use of functions as
741methods. The simplest examples are AVs and HVs, 2 aggregate types
742which don't use the fields which support SCALAR semantics.
743
744For these types, the arenas are carved up into *_allocated size
745chunks, we thus avoid wasted memory for those unaccessed members.
746When bodies are allocated, we adjust the pointer back in memory by the
747size of the bit not allocated, so it's as if we allocated the full
748structure. (But things will all go boom if you write to the part that
749is "not there", because you'll be overwriting the last members of the
750preceding structure in memory.)
751
752We calculate the correction using the STRUCT_OFFSET macro. For
753example, if xpv_allocated is the same structure as XPV then the two
754OFFSETs sum to zero, and the pointer is unchanged. If the allocated
755structure is smaller (no initial NV actually allocated) then the net
756effect is to subtract the size of the NV from the pointer, to return a
757new pointer as if an initial NV were actually allocated.
758
759This is the same trick as was used for NV and IV bodies. Ironically it
760doesn't need to be used for NV bodies any more, because NV is now at
761the start of the structure. IV bodies don't need it either, because
762they are no longer allocated.
763
764In turn, the new_body_* allocators call S_new_body(), which invokes
765new_body_inline macro, which takes a lock, and takes a body off the
766linked list at PL_body_roots[sv_type], calling S_more_bodies() if
767necessary to refresh an empty list. Then the lock is released, and
768the body is returned.
769
770S_more_bodies calls get_arena(), and carves it up into an array of N
771bodies, which it strings into a linked list. It looks up arena-size
772and body-size from the body_details table described below, thus
773supporting the multiple body-types.
774
775If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
776the (new|del)_X*V macros are mapped directly to malloc/free.
777
778*/
779
780/*
781
782For each sv-type, struct body_details bodies_by_type[] carries
783parameters which control these aspects of SV handling:
784
785Arena_size determines whether arenas are used for this body type, and if
786so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
787zero, forcing individual mallocs and frees.
788
789Body_size determines how big a body is, and therefore how many fit into
790each arena. Offset carries the body-pointer adjustment needed for
791*_allocated body types, and is used in *_allocated macros.
792
793But its main purpose is to parameterize info needed in
794Perl_sv_upgrade(). The info here dramatically simplifies the function
795vs the implementation in 5.8.7, making it table-driven. All fields
796are used for this, except for arena_size.
797
798For the sv-types that have no bodies, arenas are not used, so those
799PL_body_roots[sv_type] are unused, and can be overloaded. In
800something of a special case, SVt_NULL is borrowed for HE arenas;
801PL_body_roots[SVt_NULL] is filled by S_more_he, but the
802bodies_by_type[SVt_NULL] slot is not used, as the table is not
803available in hv.c,
804
805PTEs also use arenas, but are never seen in Perl_sv_upgrade.
806Nonetheless, they get their own slot in bodies_by_type[SVt_NULL], so
807they can just use the same allocation semantics. At first, PTEs were
808also overloaded to a non-body sv-type, but this yielded hard-to-find
809malloc bugs, so was simplified by claiming a new slot. This choice
810has no consequence at this time.
811
29489e7c
DM
812*/
813
bd81e77b 814struct body_details {
0fb58b32 815 U8 body_size; /* Size to allocate */
10666ae3 816 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 817 U8 offset;
10666ae3
NC
818 unsigned int type : 4; /* We have space for a sanity check. */
819 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
820 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
821 unsigned int arena : 1; /* Allocated from an arena */
822 size_t arena_size; /* Size of arena to allocate */
bd81e77b 823};
29489e7c 824
bd81e77b
NC
825#define HADNV FALSE
826#define NONV TRUE
29489e7c 827
d2a0f284 828
bd81e77b
NC
829#ifdef PURIFY
830/* With -DPURFIY we allocate everything directly, and don't use arenas.
831 This seems a rather elegant way to simplify some of the code below. */
832#define HASARENA FALSE
833#else
834#define HASARENA TRUE
835#endif
836#define NOARENA FALSE
29489e7c 837
d2a0f284
JC
838/* Size the arenas to exactly fit a given number of bodies. A count
839 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
840 simplifying the default. If count > 0, the arena is sized to fit
841 only that many bodies, allowing arenas to be used for large, rare
842 bodies (XPVFM, XPVIO) without undue waste. The arena size is
843 limited by PERL_ARENA_SIZE, so we can safely oversize the
844 declarations.
845 */
95db5f15
MB
846#define FIT_ARENA0(body_size) \
847 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
848#define FIT_ARENAn(count,body_size) \
849 ( count * body_size <= PERL_ARENA_SIZE) \
850 ? count * body_size \
851 : FIT_ARENA0 (body_size)
852#define FIT_ARENA(count,body_size) \
853 count \
854 ? FIT_ARENAn (count, body_size) \
855 : FIT_ARENA0 (body_size)
d2a0f284 856
bd81e77b 857/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 858
bd81e77b
NC
859typedef struct {
860 STRLEN xpv_cur;
861 STRLEN xpv_len;
862} xpv_allocated;
29489e7c 863
bd81e77b 864to make its members accessible via a pointer to (say)
29489e7c 865
bd81e77b
NC
866struct xpv {
867 NV xnv_nv;
868 STRLEN xpv_cur;
869 STRLEN xpv_len;
870};
29489e7c 871
bd81e77b 872*/
29489e7c 873
bd81e77b
NC
874#define relative_STRUCT_OFFSET(longer, shorter, member) \
875 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 876
bd81e77b
NC
877/* Calculate the length to copy. Specifically work out the length less any
878 final padding the compiler needed to add. See the comment in sv_upgrade
879 for why copying the padding proved to be a bug. */
29489e7c 880
bd81e77b
NC
881#define copy_length(type, last_member) \
882 STRUCT_OFFSET(type, last_member) \
883 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 884
bd81e77b 885static const struct body_details bodies_by_type[] = {
10666ae3
NC
886 { sizeof(HE), 0, 0, SVt_NULL,
887 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284
JC
888
889 /* IVs are in the head, so the allocation size is 0.
890 However, the slot is overloaded for PTEs. */
891 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
892 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 893 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
894 NOARENA /* IVS don't need an arena */,
895 /* But PTEs need to know the size of their arena */
896 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
897 },
898
bd81e77b 899 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 900 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
901 FIT_ARENA(0, sizeof(NV)) },
902
903 /* RVs are in the head now. */
10666ae3 904 { 0, 0, 0, SVt_RV, FALSE, NONV, NOARENA, 0 },
d2a0f284 905
bd81e77b 906 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
907 { sizeof(xpv_allocated),
908 copy_length(XPV, xpv_len)
909 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
910 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 911 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 912
bd81e77b 913 /* 12 */
d2a0f284
JC
914 { sizeof(xpviv_allocated),
915 copy_length(XPVIV, xiv_u)
916 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
917 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 918 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 919
bd81e77b 920 /* 20 */
10666ae3 921 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
922 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
923
bd81e77b 924 /* 28 */
10666ae3 925 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284
JC
926 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
927
bd81e77b 928 /* 36 */
10666ae3 929 { sizeof(XPVBM), sizeof(XPVBM), 0, SVt_PVBM, TRUE, HADNV,
d2a0f284
JC
930 HASARENA, FIT_ARENA(0, sizeof(XPVBM)) },
931
bd81e77b 932 /* 48 */
10666ae3 933 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
934 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
935
bd81e77b 936 /* 64 */
10666ae3 937 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
938 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
939
940 { sizeof(xpvav_allocated),
941 copy_length(XPVAV, xmg_stash)
942 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
943 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
10666ae3 944 SVt_PVAV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
945
946 { sizeof(xpvhv_allocated),
947 copy_length(XPVHV, xmg_stash)
948 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
949 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
10666ae3 950 SVt_PVHV, TRUE, HADNV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 951
c84c4652 952 /* 56 */
4115f141 953 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 954 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 955 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 956
4115f141 957 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 958 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 959 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
960
961 /* XPVIO is 84 bytes, fits 48x */
10666ae3 962 { sizeof(XPVIO), sizeof(XPVIO), 0, SVt_PVIO, TRUE, HADNV,
d2a0f284 963 HASARENA, FIT_ARENA(24, sizeof(XPVIO)) },
bd81e77b 964};
29489e7c 965
d2a0f284
JC
966#define new_body_type(sv_type) \
967 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 968
bd81e77b
NC
969#define del_body_type(p, sv_type) \
970 del_body(p, &PL_body_roots[sv_type])
29489e7c 971
29489e7c 972
bd81e77b 973#define new_body_allocated(sv_type) \
d2a0f284 974 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 975 - bodies_by_type[sv_type].offset)
29489e7c 976
bd81e77b
NC
977#define del_body_allocated(p, sv_type) \
978 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 979
29489e7c 980
bd81e77b
NC
981#define my_safemalloc(s) (void*)safemalloc(s)
982#define my_safecalloc(s) (void*)safecalloc(s, 1)
983#define my_safefree(p) safefree((char*)p)
29489e7c 984
bd81e77b 985#ifdef PURIFY
29489e7c 986
bd81e77b
NC
987#define new_XNV() my_safemalloc(sizeof(XPVNV))
988#define del_XNV(p) my_safefree(p)
29489e7c 989
bd81e77b
NC
990#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
991#define del_XPVNV(p) my_safefree(p)
29489e7c 992
bd81e77b
NC
993#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
994#define del_XPVAV(p) my_safefree(p)
29489e7c 995
bd81e77b
NC
996#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
997#define del_XPVHV(p) my_safefree(p)
29489e7c 998
bd81e77b
NC
999#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1000#define del_XPVMG(p) my_safefree(p)
29489e7c 1001
bd81e77b
NC
1002#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1003#define del_XPVGV(p) my_safefree(p)
29489e7c 1004
bd81e77b 1005#else /* !PURIFY */
29489e7c 1006
bd81e77b
NC
1007#define new_XNV() new_body_type(SVt_NV)
1008#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1009
bd81e77b
NC
1010#define new_XPVNV() new_body_type(SVt_PVNV)
1011#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1012
bd81e77b
NC
1013#define new_XPVAV() new_body_allocated(SVt_PVAV)
1014#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1015
bd81e77b
NC
1016#define new_XPVHV() new_body_allocated(SVt_PVHV)
1017#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1018
bd81e77b
NC
1019#define new_XPVMG() new_body_type(SVt_PVMG)
1020#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1021
bd81e77b
NC
1022#define new_XPVGV() new_body_type(SVt_PVGV)
1023#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1024
bd81e77b 1025#endif /* PURIFY */
93e68bfb 1026
bd81e77b 1027/* no arena for you! */
93e68bfb 1028
bd81e77b 1029#define new_NOARENA(details) \
d2a0f284 1030 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1031#define new_NOARENAZ(details) \
d2a0f284
JC
1032 my_safecalloc((details)->body_size + (details)->offset)
1033
10666ae3
NC
1034#ifdef DEBUGGING
1035static bool done_sanity_check;
1036#endif
1037
d2a0f284
JC
1038STATIC void *
1039S_more_bodies (pTHX_ svtype sv_type)
1040{
1041 dVAR;
1042 void ** const root = &PL_body_roots[sv_type];
96a5add6 1043 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1044 const size_t body_size = bdp->body_size;
1045 char *start;
1046 const char *end;
1047
1048 assert(bdp->arena_size);
10666ae3
NC
1049
1050#ifdef DEBUGGING
1051 if (!done_sanity_check) {
ea471437 1052 unsigned int i = SVt_LAST;
10666ae3
NC
1053
1054 done_sanity_check = TRUE;
1055
1056 while (i--)
1057 assert (bodies_by_type[i].type == i);
1058 }
1059#endif
1060
d2a0f284
JC
1061 start = (char*) Perl_get_arena(aTHX_ bdp->arena_size);
1062
1063 end = start + bdp->arena_size - body_size;
1064
d2a0f284
JC
1065 /* computed count doesnt reflect the 1st slot reservation */
1066 DEBUG_m(PerlIO_printf(Perl_debug_log,
1067 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1068 start, end, bdp->arena_size, sv_type, body_size,
1069 bdp->arena_size / body_size));
d2a0f284
JC
1070
1071 *root = (void *)start;
1072
1073 while (start < end) {
1074 char * const next = start + body_size;
1075 *(void**) start = (void *)next;
1076 start = next;
1077 }
1078 *(void **)start = 0;
1079
1080 return *root;
1081}
1082
1083/* grab a new thing from the free list, allocating more if necessary.
1084 The inline version is used for speed in hot routines, and the
1085 function using it serves the rest (unless PURIFY).
1086*/
1087#define new_body_inline(xpv, sv_type) \
1088 STMT_START { \
1089 void ** const r3wt = &PL_body_roots[sv_type]; \
1090 LOCK_SV_MUTEX; \
1091 xpv = *((void **)(r3wt)) \
d4c19fe8 1092 ? *((void **)(r3wt)) : more_bodies(sv_type); \
d2a0f284
JC
1093 *(r3wt) = *(void**)(xpv); \
1094 UNLOCK_SV_MUTEX; \
1095 } STMT_END
1096
1097#ifndef PURIFY
1098
1099STATIC void *
1100S_new_body(pTHX_ svtype sv_type)
1101{
1102 dVAR;
1103 void *xpv;
1104 new_body_inline(xpv, sv_type);
1105 return xpv;
1106}
1107
1108#endif
93e68bfb 1109
bd81e77b
NC
1110/*
1111=for apidoc sv_upgrade
93e68bfb 1112
bd81e77b
NC
1113Upgrade an SV to a more complex form. Generally adds a new body type to the
1114SV, then copies across as much information as possible from the old body.
1115You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1116
bd81e77b 1117=cut
93e68bfb 1118*/
93e68bfb 1119
bd81e77b
NC
1120void
1121Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type)
cac9b346 1122{
97aff369 1123 dVAR;
bd81e77b
NC
1124 void* old_body;
1125 void* new_body;
1126 const U32 old_type = SvTYPE(sv);
d2a0f284 1127 const struct body_details *new_type_details;
bd81e77b
NC
1128 const struct body_details *const old_type_details
1129 = bodies_by_type + old_type;
cac9b346 1130
bd81e77b
NC
1131 if (new_type != SVt_PV && SvIsCOW(sv)) {
1132 sv_force_normal_flags(sv, 0);
1133 }
cac9b346 1134
bd81e77b
NC
1135 if (old_type == new_type)
1136 return;
cac9b346 1137
bd81e77b
NC
1138 if (old_type > new_type)
1139 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1140 (int)old_type, (int)new_type);
cac9b346 1141
cac9b346 1142
bd81e77b 1143 old_body = SvANY(sv);
de042e1d 1144
bd81e77b
NC
1145 /* Copying structures onto other structures that have been neatly zeroed
1146 has a subtle gotcha. Consider XPVMG
cac9b346 1147
bd81e77b
NC
1148 +------+------+------+------+------+-------+-------+
1149 | NV | CUR | LEN | IV | MAGIC | STASH |
1150 +------+------+------+------+------+-------+-------+
1151 0 4 8 12 16 20 24 28
645c22ef 1152
bd81e77b
NC
1153 where NVs are aligned to 8 bytes, so that sizeof that structure is
1154 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1155
bd81e77b
NC
1156 +------+------+------+------+------+-------+-------+------+
1157 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1158 +------+------+------+------+------+-------+-------+------+
1159 0 4 8 12 16 20 24 28 32
08742458 1160
bd81e77b 1161 so what happens if you allocate memory for this structure:
30f9da9e 1162
bd81e77b
NC
1163 +------+------+------+------+------+-------+-------+------+------+...
1164 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1165 +------+------+------+------+------+-------+-------+------+------+...
1166 0 4 8 12 16 20 24 28 32 36
bfc44f79 1167
bd81e77b
NC
1168 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1169 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1170 started out as zero once, but it's quite possible that it isn't. So now,
1171 rather than a nicely zeroed GP, you have it pointing somewhere random.
1172 Bugs ensue.
bfc44f79 1173
bd81e77b
NC
1174 (In fact, GP ends up pointing at a previous GP structure, because the
1175 principle cause of the padding in XPVMG getting garbage is a copy of
1176 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 1177
bd81e77b
NC
1178 So we are careful and work out the size of used parts of all the
1179 structures. */
bfc44f79 1180
bd81e77b
NC
1181 switch (old_type) {
1182 case SVt_NULL:
1183 break;
1184 case SVt_IV:
1185 if (new_type < SVt_PVIV) {
1186 new_type = (new_type == SVt_NV)
1187 ? SVt_PVNV : SVt_PVIV;
bd81e77b
NC
1188 }
1189 break;
1190 case SVt_NV:
1191 if (new_type < SVt_PVNV) {
1192 new_type = SVt_PVNV;
bd81e77b
NC
1193 }
1194 break;
1195 case SVt_RV:
1196 break;
1197 case SVt_PV:
1198 assert(new_type > SVt_PV);
1199 assert(SVt_IV < SVt_PV);
1200 assert(SVt_NV < SVt_PV);
1201 break;
1202 case SVt_PVIV:
1203 break;
1204 case SVt_PVNV:
1205 break;
1206 case SVt_PVMG:
1207 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1208 there's no way that it can be safely upgraded, because perl.c
1209 expects to Safefree(SvANY(PL_mess_sv)) */
1210 assert(sv != PL_mess_sv);
1211 /* This flag bit is used to mean other things in other scalar types.
1212 Given that it only has meaning inside the pad, it shouldn't be set
1213 on anything that can get upgraded. */
00b1698f 1214 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1215 break;
1216 default:
1217 if (old_type_details->cant_upgrade)
c81225bc
NC
1218 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1219 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1220 }
2fa1109b 1221 new_type_details = bodies_by_type + new_type;
645c22ef 1222
bd81e77b
NC
1223 SvFLAGS(sv) &= ~SVTYPEMASK;
1224 SvFLAGS(sv) |= new_type;
932e9ff9 1225
ab4416c0
NC
1226 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1227 the return statements above will have triggered. */
1228 assert (new_type != SVt_NULL);
bd81e77b 1229 switch (new_type) {
bd81e77b
NC
1230 case SVt_IV:
1231 assert(old_type == SVt_NULL);
1232 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1233 SvIV_set(sv, 0);
1234 return;
1235 case SVt_NV:
1236 assert(old_type == SVt_NULL);
1237 SvANY(sv) = new_XNV();
1238 SvNV_set(sv, 0);
1239 return;
1240 case SVt_RV:
1241 assert(old_type == SVt_NULL);
1242 SvANY(sv) = &sv->sv_u.svu_rv;
1243 SvRV_set(sv, 0);
1244 return;
1245 case SVt_PVHV:
bd81e77b 1246 case SVt_PVAV:
d2a0f284 1247 assert(new_type_details->body_size);
c1ae03ae
NC
1248
1249#ifndef PURIFY
1250 assert(new_type_details->arena);
d2a0f284 1251 assert(new_type_details->arena_size);
c1ae03ae 1252 /* This points to the start of the allocated area. */
d2a0f284
JC
1253 new_body_inline(new_body, new_type);
1254 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1255 new_body = ((char *)new_body) - new_type_details->offset;
1256#else
1257 /* We always allocated the full length item with PURIFY. To do this
1258 we fake things so that arena is false for all 16 types.. */
1259 new_body = new_NOARENAZ(new_type_details);
1260#endif
1261 SvANY(sv) = new_body;
1262 if (new_type == SVt_PVAV) {
1263 AvMAX(sv) = -1;
1264 AvFILLp(sv) = -1;
1265 AvREAL_only(sv);
1266 }
aeb18a1e 1267
bd81e77b
NC
1268 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1269 The target created by newSVrv also is, and it can have magic.
1270 However, it never has SvPVX set.
1271 */
1272 if (old_type >= SVt_RV) {
1273 assert(SvPVX_const(sv) == 0);
1274 }
aeb18a1e 1275
bd81e77b
NC
1276 /* Could put this in the else clause below, as PVMG must have SvPVX
1277 0 already (the assertion above) */
6136c704 1278 SvPV_set(sv, NULL);
93e68bfb 1279
bd81e77b 1280 if (old_type >= SVt_PVMG) {
e736a858 1281 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1282 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
bd81e77b
NC
1283 }
1284 break;
93e68bfb 1285
93e68bfb 1286
bd81e77b
NC
1287 case SVt_PVIV:
1288 /* XXX Is this still needed? Was it ever needed? Surely as there is
1289 no route from NV to PVIV, NOK can never be true */
1290 assert(!SvNOKp(sv));
1291 assert(!SvNOK(sv));
1292 case SVt_PVIO:
1293 case SVt_PVFM:
1294 case SVt_PVBM:
1295 case SVt_PVGV:
1296 case SVt_PVCV:
1297 case SVt_PVLV:
1298 case SVt_PVMG:
1299 case SVt_PVNV:
1300 case SVt_PV:
93e68bfb 1301
d2a0f284 1302 assert(new_type_details->body_size);
bd81e77b
NC
1303 /* We always allocated the full length item with PURIFY. To do this
1304 we fake things so that arena is false for all 16 types.. */
1305 if(new_type_details->arena) {
1306 /* This points to the start of the allocated area. */
d2a0f284
JC
1307 new_body_inline(new_body, new_type);
1308 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1309 new_body = ((char *)new_body) - new_type_details->offset;
1310 } else {
1311 new_body = new_NOARENAZ(new_type_details);
1312 }
1313 SvANY(sv) = new_body;
5e2fc214 1314
bd81e77b 1315 if (old_type_details->copy) {
f9ba3d20
NC
1316 /* There is now the potential for an upgrade from something without
1317 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1318 int offset = old_type_details->offset;
1319 int length = old_type_details->copy;
1320
1321 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1322 const int difference
f9ba3d20
NC
1323 = new_type_details->offset - old_type_details->offset;
1324 offset += difference;
1325 length -= difference;
1326 }
1327 assert (length >= 0);
1328
1329 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1330 char);
bd81e77b
NC
1331 }
1332
1333#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1334 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1335 * correct 0.0 for us. Otherwise, if the old body didn't have an
1336 * NV slot, but the new one does, then we need to initialise the
1337 * freshly created NV slot with whatever the correct bit pattern is
1338 * for 0.0 */
1339 if (old_type_details->zero_nv && !new_type_details->zero_nv)
bd81e77b 1340 SvNV_set(sv, 0);
82048762 1341#endif
5e2fc214 1342
bd81e77b 1343 if (new_type == SVt_PVIO)
f2524eef 1344 IoPAGE_LEN(sv) = 60;
bd81e77b 1345 if (old_type < SVt_RV)
6136c704 1346 SvPV_set(sv, NULL);
bd81e77b
NC
1347 break;
1348 default:
afd78fd5
JH
1349 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1350 (unsigned long)new_type);
bd81e77b 1351 }
73171d91 1352
d2a0f284
JC
1353 if (old_type_details->arena) {
1354 /* If there was an old body, then we need to free it.
1355 Note that there is an assumption that all bodies of types that
1356 can be upgraded came from arenas. Only the more complex non-
1357 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1358#ifdef PURIFY
1359 my_safefree(old_body);
1360#else
1361 del_body((void*)((char*)old_body + old_type_details->offset),
1362 &PL_body_roots[old_type]);
1363#endif
1364 }
1365}
73171d91 1366
bd81e77b
NC
1367/*
1368=for apidoc sv_backoff
73171d91 1369
bd81e77b
NC
1370Remove any string offset. You should normally use the C<SvOOK_off> macro
1371wrapper instead.
73171d91 1372
bd81e77b 1373=cut
73171d91
NC
1374*/
1375
bd81e77b
NC
1376int
1377Perl_sv_backoff(pTHX_ register SV *sv)
1378{
96a5add6 1379 PERL_UNUSED_CONTEXT;
bd81e77b
NC
1380 assert(SvOOK(sv));
1381 assert(SvTYPE(sv) != SVt_PVHV);
1382 assert(SvTYPE(sv) != SVt_PVAV);
1383 if (SvIVX(sv)) {
1384 const char * const s = SvPVX_const(sv);
1385 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1386 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1387 SvIV_set(sv, 0);
1388 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1389 }
1390 SvFLAGS(sv) &= ~SVf_OOK;
1391 return 0;
1392}
73171d91 1393
bd81e77b
NC
1394/*
1395=for apidoc sv_grow
73171d91 1396
bd81e77b
NC
1397Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1398upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1399Use the C<SvGROW> wrapper instead.
93e68bfb 1400
bd81e77b
NC
1401=cut
1402*/
93e68bfb 1403
bd81e77b
NC
1404char *
1405Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1406{
1407 register char *s;
93e68bfb 1408
5db06880
NC
1409 if (PL_madskills && newlen >= 0x100000) {
1410 PerlIO_printf(Perl_debug_log,
1411 "Allocation too large: %"UVxf"\n", (UV)newlen);
1412 }
bd81e77b
NC
1413#ifdef HAS_64K_LIMIT
1414 if (newlen >= 0x10000) {
1415 PerlIO_printf(Perl_debug_log,
1416 "Allocation too large: %"UVxf"\n", (UV)newlen);
1417 my_exit(1);
1418 }
1419#endif /* HAS_64K_LIMIT */
1420 if (SvROK(sv))
1421 sv_unref(sv);
1422 if (SvTYPE(sv) < SVt_PV) {
1423 sv_upgrade(sv, SVt_PV);
1424 s = SvPVX_mutable(sv);
1425 }
1426 else if (SvOOK(sv)) { /* pv is offset? */
1427 sv_backoff(sv);
1428 s = SvPVX_mutable(sv);
1429 if (newlen > SvLEN(sv))
1430 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1431#ifdef HAS_64K_LIMIT
1432 if (newlen >= 0x10000)
1433 newlen = 0xFFFF;
1434#endif
1435 }
1436 else
1437 s = SvPVX_mutable(sv);
aeb18a1e 1438
bd81e77b
NC
1439 if (newlen > SvLEN(sv)) { /* need more room? */
1440 newlen = PERL_STRLEN_ROUNDUP(newlen);
1441 if (SvLEN(sv) && s) {
1442#ifdef MYMALLOC
1443 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1444 if (newlen <= l) {
1445 SvLEN_set(sv, l);
1446 return s;
1447 } else
1448#endif
1449 s = saferealloc(s, newlen);
1450 }
1451 else {
1452 s = safemalloc(newlen);
1453 if (SvPVX_const(sv) && SvCUR(sv)) {
1454 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1455 }
1456 }
1457 SvPV_set(sv, s);
1458 SvLEN_set(sv, newlen);
1459 }
1460 return s;
1461}
aeb18a1e 1462
bd81e77b
NC
1463/*
1464=for apidoc sv_setiv
932e9ff9 1465
bd81e77b
NC
1466Copies an integer into the given SV, upgrading first if necessary.
1467Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1468
bd81e77b
NC
1469=cut
1470*/
463ee0b2 1471
bd81e77b
NC
1472void
1473Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1474{
97aff369 1475 dVAR;
bd81e77b
NC
1476 SV_CHECK_THINKFIRST_COW_DROP(sv);
1477 switch (SvTYPE(sv)) {
1478 case SVt_NULL:
1479 sv_upgrade(sv, SVt_IV);
1480 break;
1481 case SVt_NV:
1482 sv_upgrade(sv, SVt_PVNV);
1483 break;
1484 case SVt_RV:
1485 case SVt_PV:
1486 sv_upgrade(sv, SVt_PVIV);
1487 break;
463ee0b2 1488
bd81e77b
NC
1489 case SVt_PVGV:
1490 case SVt_PVAV:
1491 case SVt_PVHV:
1492 case SVt_PVCV:
1493 case SVt_PVFM:
1494 case SVt_PVIO:
1495 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1496 OP_DESC(PL_op));
1497 }
1498 (void)SvIOK_only(sv); /* validate number */
1499 SvIV_set(sv, i);
1500 SvTAINT(sv);
1501}
932e9ff9 1502
bd81e77b
NC
1503/*
1504=for apidoc sv_setiv_mg
d33b2eba 1505
bd81e77b 1506Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1507
bd81e77b
NC
1508=cut
1509*/
d33b2eba 1510
bd81e77b
NC
1511void
1512Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1513{
1514 sv_setiv(sv,i);
1515 SvSETMAGIC(sv);
1516}
727879eb 1517
bd81e77b
NC
1518/*
1519=for apidoc sv_setuv
d33b2eba 1520
bd81e77b
NC
1521Copies an unsigned integer into the given SV, upgrading first if necessary.
1522Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1523
bd81e77b
NC
1524=cut
1525*/
d33b2eba 1526
bd81e77b
NC
1527void
1528Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1529{
1530 /* With these two if statements:
1531 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1532
bd81e77b
NC
1533 without
1534 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1535
bd81e77b
NC
1536 If you wish to remove them, please benchmark to see what the effect is
1537 */
1538 if (u <= (UV)IV_MAX) {
1539 sv_setiv(sv, (IV)u);
1540 return;
1541 }
1542 sv_setiv(sv, 0);
1543 SvIsUV_on(sv);
1544 SvUV_set(sv, u);
1545}
d33b2eba 1546
bd81e77b
NC
1547/*
1548=for apidoc sv_setuv_mg
727879eb 1549
bd81e77b 1550Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1551
bd81e77b
NC
1552=cut
1553*/
5e2fc214 1554
bd81e77b
NC
1555void
1556Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1557{
1558 sv_setiv(sv, 0);
1559 SvIsUV_on(sv);
1560 sv_setuv(sv,u);
1561 SvSETMAGIC(sv);
1562}
5e2fc214 1563
954c1994 1564/*
bd81e77b 1565=for apidoc sv_setnv
954c1994 1566
bd81e77b
NC
1567Copies a double into the given SV, upgrading first if necessary.
1568Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1569
1570=cut
1571*/
1572
63f97190 1573void
bd81e77b 1574Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1575{
97aff369 1576 dVAR;
bd81e77b
NC
1577 SV_CHECK_THINKFIRST_COW_DROP(sv);
1578 switch (SvTYPE(sv)) {
79072805 1579 case SVt_NULL:
79072805 1580 case SVt_IV:
bd81e77b 1581 sv_upgrade(sv, SVt_NV);
79072805 1582 break;
ed6116ce 1583 case SVt_RV:
79072805 1584 case SVt_PV:
79072805 1585 case SVt_PVIV:
bd81e77b 1586 sv_upgrade(sv, SVt_PVNV);
79072805 1587 break;
bd4b1eb5 1588
bd4b1eb5 1589 case SVt_PVGV:
bd81e77b
NC
1590 case SVt_PVAV:
1591 case SVt_PVHV:
79072805 1592 case SVt_PVCV:
bd81e77b
NC
1593 case SVt_PVFM:
1594 case SVt_PVIO:
1595 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1596 OP_NAME(PL_op));
2068cd4d 1597 }
bd81e77b
NC
1598 SvNV_set(sv, num);
1599 (void)SvNOK_only(sv); /* validate number */
1600 SvTAINT(sv);
79072805
LW
1601}
1602
645c22ef 1603/*
bd81e77b 1604=for apidoc sv_setnv_mg
645c22ef 1605
bd81e77b 1606Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1607
1608=cut
1609*/
1610
bd81e77b
NC
1611void
1612Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1613{
bd81e77b
NC
1614 sv_setnv(sv,num);
1615 SvSETMAGIC(sv);
79072805
LW
1616}
1617
bd81e77b
NC
1618/* Print an "isn't numeric" warning, using a cleaned-up,
1619 * printable version of the offending string
1620 */
954c1994 1621
bd81e77b
NC
1622STATIC void
1623S_not_a_number(pTHX_ SV *sv)
79072805 1624{
97aff369 1625 dVAR;
bd81e77b
NC
1626 SV *dsv;
1627 char tmpbuf[64];
1628 const char *pv;
94463019
JH
1629
1630 if (DO_UTF8(sv)) {
396482e1 1631 dsv = sv_2mortal(newSVpvs(""));
94463019
JH
1632 pv = sv_uni_display(dsv, sv, 10, 0);
1633 } else {
1634 char *d = tmpbuf;
551405c4 1635 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1636 /* each *s can expand to 4 chars + "...\0",
1637 i.e. need room for 8 chars */
ecdeb87c 1638
00b6aa41
AL
1639 const char *s = SvPVX_const(sv);
1640 const char * const end = s + SvCUR(sv);
1641 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1642 int ch = *s & 0xFF;
1643 if (ch & 128 && !isPRINT_LC(ch)) {
1644 *d++ = 'M';
1645 *d++ = '-';
1646 ch &= 127;
1647 }
1648 if (ch == '\n') {
1649 *d++ = '\\';
1650 *d++ = 'n';
1651 }
1652 else if (ch == '\r') {
1653 *d++ = '\\';
1654 *d++ = 'r';
1655 }
1656 else if (ch == '\f') {
1657 *d++ = '\\';
1658 *d++ = 'f';
1659 }
1660 else if (ch == '\\') {
1661 *d++ = '\\';
1662 *d++ = '\\';
1663 }
1664 else if (ch == '\0') {
1665 *d++ = '\\';
1666 *d++ = '0';
1667 }
1668 else if (isPRINT_LC(ch))
1669 *d++ = ch;
1670 else {
1671 *d++ = '^';
1672 *d++ = toCTRL(ch);
1673 }
1674 }
1675 if (s < end) {
1676 *d++ = '.';
1677 *d++ = '.';
1678 *d++ = '.';
1679 }
1680 *d = '\0';
1681 pv = tmpbuf;
a0d0e21e 1682 }
a0d0e21e 1683
533c011a 1684 if (PL_op)
9014280d 1685 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1686 "Argument \"%s\" isn't numeric in %s", pv,
1687 OP_DESC(PL_op));
a0d0e21e 1688 else
9014280d 1689 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1690 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1691}
1692
c2988b20
NC
1693/*
1694=for apidoc looks_like_number
1695
645c22ef
DM
1696Test if the content of an SV looks like a number (or is a number).
1697C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1698non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1699
1700=cut
1701*/
1702
1703I32
1704Perl_looks_like_number(pTHX_ SV *sv)
1705{
a3b680e6 1706 register const char *sbegin;
c2988b20
NC
1707 STRLEN len;
1708
1709 if (SvPOK(sv)) {
3f7c398e 1710 sbegin = SvPVX_const(sv);
c2988b20
NC
1711 len = SvCUR(sv);
1712 }
1713 else if (SvPOKp(sv))
83003860 1714 sbegin = SvPV_const(sv, len);
c2988b20 1715 else
e0ab1c0e 1716 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1717 return grok_number(sbegin, len, NULL);
1718}
25da4f38 1719
180488f8
NC
1720STATIC char *
1721S_glob_2inpuv(pTHX_ GV *gv, STRLEN *len, bool want_number)
1722{
1723 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1724 SV *const buffer = sv_newmortal();
1725
1726 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1727 is on. */
1728 SvFAKE_off(gv);
1729 gv_efullname3(buffer, gv, "*");
1730 SvFLAGS(gv) |= wasfake;
1731
1732 if (want_number) {
1733 /* We know that all GVs stringify to something that is not-a-number,
1734 so no need to test that. */
1735 if (ckWARN(WARN_NUMERIC))
1736 not_a_number(buffer);
1737 /* We just want something true to return, so that S_sv_2iuv_common
1738 can tail call us and return true. */
1739 return (char *) 1;
1740 } else {
6a9cfe87
NC
1741 assert(SvPOK(buffer));
1742 if (len) {
1743 *len = SvCUR(buffer);
1744 }
1745 return SvPVX(buffer);
180488f8
NC
1746 }
1747}
1748
25da4f38
IZ
1749/* Actually, ISO C leaves conversion of UV to IV undefined, but
1750 until proven guilty, assume that things are not that bad... */
1751
645c22ef
DM
1752/*
1753 NV_PRESERVES_UV:
1754
1755 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1756 an IV (an assumption perl has been based on to date) it becomes necessary
1757 to remove the assumption that the NV always carries enough precision to
1758 recreate the IV whenever needed, and that the NV is the canonical form.
1759 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1760 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1761 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1762 1) to distinguish between IV/UV/NV slots that have cached a valid
1763 conversion where precision was lost and IV/UV/NV slots that have a
1764 valid conversion which has lost no precision
645c22ef 1765 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1766 would lose precision, the precise conversion (or differently
1767 imprecise conversion) is also performed and cached, to prevent
1768 requests for different numeric formats on the same SV causing
1769 lossy conversion chains. (lossless conversion chains are perfectly
1770 acceptable (still))
1771
1772
1773 flags are used:
1774 SvIOKp is true if the IV slot contains a valid value
1775 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1776 SvNOKp is true if the NV slot contains a valid value
1777 SvNOK is true only if the NV value is accurate
1778
1779 so
645c22ef 1780 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1781 IV(or UV) would lose accuracy over a direct conversion from PV to
1782 IV(or UV). If it would, cache both conversions, return NV, but mark
1783 SV as IOK NOKp (ie not NOK).
1784
645c22ef 1785 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1786 NV would lose accuracy over a direct conversion from PV to NV. If it
1787 would, cache both conversions, flag similarly.
1788
1789 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1790 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1791 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1792 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1793 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1794
645c22ef
DM
1795 The benefit of this is that operations such as pp_add know that if
1796 SvIOK is true for both left and right operands, then integer addition
1797 can be used instead of floating point (for cases where the result won't
1798 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1799 loss of precision compared with integer addition.
1800
1801 * making IV and NV equal status should make maths accurate on 64 bit
1802 platforms
1803 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1804 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1805 looking for SvIOK and checking for overflow will not outweigh the
1806 fp to integer speedup)
1807 * will slow down integer operations (callers of SvIV) on "inaccurate"
1808 values, as the change from SvIOK to SvIOKp will cause a call into
1809 sv_2iv each time rather than a macro access direct to the IV slot
1810 * should speed up number->string conversion on integers as IV is
645c22ef 1811 favoured when IV and NV are equally accurate
28e5dec8
JH
1812
1813 ####################################################################
645c22ef
DM
1814 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1815 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1816 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1817 ####################################################################
1818
645c22ef 1819 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1820 performance ratio.
1821*/
1822
1823#ifndef NV_PRESERVES_UV
645c22ef
DM
1824# define IS_NUMBER_UNDERFLOW_IV 1
1825# define IS_NUMBER_UNDERFLOW_UV 2
1826# define IS_NUMBER_IV_AND_UV 2
1827# define IS_NUMBER_OVERFLOW_IV 4
1828# define IS_NUMBER_OVERFLOW_UV 5
1829
1830/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1831
1832/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1833STATIC int
645c22ef 1834S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1835{
97aff369 1836 dVAR;
3f7c398e 1837 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
1838 if (SvNVX(sv) < (NV)IV_MIN) {
1839 (void)SvIOKp_on(sv);
1840 (void)SvNOK_on(sv);
45977657 1841 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1842 return IS_NUMBER_UNDERFLOW_IV;
1843 }
1844 if (SvNVX(sv) > (NV)UV_MAX) {
1845 (void)SvIOKp_on(sv);
1846 (void)SvNOK_on(sv);
1847 SvIsUV_on(sv);
607fa7f2 1848 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1849 return IS_NUMBER_OVERFLOW_UV;
1850 }
c2988b20
NC
1851 (void)SvIOKp_on(sv);
1852 (void)SvNOK_on(sv);
1853 /* Can't use strtol etc to convert this string. (See truth table in
1854 sv_2iv */
1855 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1856 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1857 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1858 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1859 } else {
1860 /* Integer is imprecise. NOK, IOKp */
1861 }
1862 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1863 }
1864 SvIsUV_on(sv);
607fa7f2 1865 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1866 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1867 if (SvUVX(sv) == UV_MAX) {
1868 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1869 possibly be preserved by NV. Hence, it must be overflow.
1870 NOK, IOKp */
1871 return IS_NUMBER_OVERFLOW_UV;
1872 }
1873 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1874 } else {
1875 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1876 }
c2988b20 1877 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1878}
645c22ef
DM
1879#endif /* !NV_PRESERVES_UV*/
1880
af359546
NC
1881STATIC bool
1882S_sv_2iuv_common(pTHX_ SV *sv) {
97aff369 1883 dVAR;
af359546 1884 if (SvNOKp(sv)) {
28e5dec8
JH
1885 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1886 * without also getting a cached IV/UV from it at the same time
1887 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1888 * IV or UV at same time to avoid this. */
1889 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1890
1891 if (SvTYPE(sv) == SVt_NV)
1892 sv_upgrade(sv, SVt_PVNV);
1893
28e5dec8
JH
1894 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1895 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1896 certainly cast into the IV range at IV_MAX, whereas the correct
1897 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1898 cases go to UV */
cab190d4
JD
1899#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1900 if (Perl_isnan(SvNVX(sv))) {
1901 SvUV_set(sv, 0);
1902 SvIsUV_on(sv);
fdbe6d7c 1903 return FALSE;
cab190d4 1904 }
cab190d4 1905#endif
28e5dec8 1906 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1907 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1908 if (SvNVX(sv) == (NV) SvIVX(sv)
1909#ifndef NV_PRESERVES_UV
1910 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1911 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1912 /* Don't flag it as "accurately an integer" if the number
1913 came from a (by definition imprecise) NV operation, and
1914 we're outside the range of NV integer precision */
1915#endif
1916 ) {
1917 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1918 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1919 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1920 PTR2UV(sv),
1921 SvNVX(sv),
1922 SvIVX(sv)));
1923
1924 } else {
1925 /* IV not precise. No need to convert from PV, as NV
1926 conversion would already have cached IV if it detected
1927 that PV->IV would be better than PV->NV->IV
1928 flags already correct - don't set public IOK. */
1929 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1930 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1931 PTR2UV(sv),
1932 SvNVX(sv),
1933 SvIVX(sv)));
1934 }
1935 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1936 but the cast (NV)IV_MIN rounds to a the value less (more
1937 negative) than IV_MIN which happens to be equal to SvNVX ??
1938 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1939 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1940 (NV)UVX == NVX are both true, but the values differ. :-(
1941 Hopefully for 2s complement IV_MIN is something like
1942 0x8000000000000000 which will be exact. NWC */
d460ef45 1943 }
25da4f38 1944 else {
607fa7f2 1945 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1946 if (
1947 (SvNVX(sv) == (NV) SvUVX(sv))
1948#ifndef NV_PRESERVES_UV
1949 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1950 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1951 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1952 /* Don't flag it as "accurately an integer" if the number
1953 came from a (by definition imprecise) NV operation, and
1954 we're outside the range of NV integer precision */
1955#endif
1956 )
1957 SvIOK_on(sv);
25da4f38 1958 SvIsUV_on(sv);
1c846c1f 1959 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1960 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1961 PTR2UV(sv),
57def98f
JH
1962 SvUVX(sv),
1963 SvUVX(sv)));
25da4f38 1964 }
748a9306
LW
1965 }
1966 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1967 UV value;
504618e9 1968 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1969 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1970 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1971 the same as the direct translation of the initial string
1972 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1973 be careful to ensure that the value with the .456 is around if the
1974 NV value is requested in the future).
1c846c1f 1975
af359546 1976 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1977 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1978 cache the NV if we are sure it's not needed.
25da4f38 1979 */
16b7a9a4 1980
c2988b20
NC
1981 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1982 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1983 == IS_NUMBER_IN_UV) {
5e045b90 1984 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1985 if (SvTYPE(sv) < SVt_PVIV)
1986 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1987 (void)SvIOK_on(sv);
c2988b20
NC
1988 } else if (SvTYPE(sv) < SVt_PVNV)
1989 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1990
f2524eef 1991 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
1992 we aren't going to call atof() below. If NVs don't preserve UVs
1993 then the value returned may have more precision than atof() will
1994 return, even though value isn't perfectly accurate. */
1995 if ((numtype & (IS_NUMBER_IN_UV
1996#ifdef NV_PRESERVES_UV
1997 | IS_NUMBER_NOT_INT
1998#endif
1999 )) == IS_NUMBER_IN_UV) {
2000 /* This won't turn off the public IOK flag if it was set above */
2001 (void)SvIOKp_on(sv);
2002
2003 if (!(numtype & IS_NUMBER_NEG)) {
2004 /* positive */;
2005 if (value <= (UV)IV_MAX) {
45977657 2006 SvIV_set(sv, (IV)value);
c2988b20 2007 } else {
af359546 2008 /* it didn't overflow, and it was positive. */
607fa7f2 2009 SvUV_set(sv, value);
c2988b20
NC
2010 SvIsUV_on(sv);
2011 }
2012 } else {
2013 /* 2s complement assumption */
2014 if (value <= (UV)IV_MIN) {
45977657 2015 SvIV_set(sv, -(IV)value);
c2988b20
NC
2016 } else {
2017 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2018 I'm assuming it will be rare. */
c2988b20
NC
2019 if (SvTYPE(sv) < SVt_PVNV)
2020 sv_upgrade(sv, SVt_PVNV);
2021 SvNOK_on(sv);
2022 SvIOK_off(sv);
2023 SvIOKp_on(sv);
9d6ce603 2024 SvNV_set(sv, -(NV)value);
45977657 2025 SvIV_set(sv, IV_MIN);
c2988b20
NC
2026 }
2027 }
2028 }
2029 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2030 will be in the previous block to set the IV slot, and the next
2031 block to set the NV slot. So no else here. */
2032
2033 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2034 != IS_NUMBER_IN_UV) {
2035 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2036 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2037
c2988b20
NC
2038 if (! numtype && ckWARN(WARN_NUMERIC))
2039 not_a_number(sv);
28e5dec8 2040
65202027 2041#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2042 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2043 PTR2UV(sv), SvNVX(sv)));
65202027 2044#else
1779d84d 2045 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2046 PTR2UV(sv), SvNVX(sv)));
65202027 2047#endif
28e5dec8 2048
28e5dec8 2049#ifdef NV_PRESERVES_UV
af359546
NC
2050 (void)SvIOKp_on(sv);
2051 (void)SvNOK_on(sv);
2052 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2053 SvIV_set(sv, I_V(SvNVX(sv)));
2054 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2055 SvIOK_on(sv);
2056 } else {
bb263b4e 2057 /*EMPTY*/; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2058 }
2059 /* UV will not work better than IV */
2060 } else {
2061 if (SvNVX(sv) > (NV)UV_MAX) {
2062 SvIsUV_on(sv);
2063 /* Integer is inaccurate. NOK, IOKp, is UV */
2064 SvUV_set(sv, UV_MAX);
af359546
NC
2065 } else {
2066 SvUV_set(sv, U_V(SvNVX(sv)));
2067 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2068 NV preservse UV so can do correct comparison. */
2069 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2070 SvIOK_on(sv);
af359546 2071 } else {
bb263b4e 2072 /*EMPTY*/; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2073 }
2074 }
4b0c9573 2075 SvIsUV_on(sv);
af359546 2076 }
28e5dec8 2077#else /* NV_PRESERVES_UV */
c2988b20
NC
2078 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2079 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2080 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2081 grok_number above. The NV slot has just been set using
2082 Atof. */
560b0c46 2083 SvNOK_on(sv);
c2988b20
NC
2084 assert (SvIOKp(sv));
2085 } else {
2086 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2087 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2088 /* Small enough to preserve all bits. */
2089 (void)SvIOKp_on(sv);
2090 SvNOK_on(sv);
45977657 2091 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2092 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2093 SvIOK_on(sv);
2094 /* Assumption: first non-preserved integer is < IV_MAX,
2095 this NV is in the preserved range, therefore: */
2096 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2097 < (UV)IV_MAX)) {
32fdb065 2098 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
2099 }
2100 } else {
2101 /* IN_UV NOT_INT
2102 0 0 already failed to read UV.
2103 0 1 already failed to read UV.
2104 1 0 you won't get here in this case. IV/UV
2105 slot set, public IOK, Atof() unneeded.
2106 1 1 already read UV.
2107 so there's no point in sv_2iuv_non_preserve() attempting
2108 to use atol, strtol, strtoul etc. */
40a17c4c 2109 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
2110 }
2111 }
28e5dec8 2112#endif /* NV_PRESERVES_UV */
25da4f38 2113 }
af359546
NC
2114 }
2115 else {
f7877b28 2116 if (isGV_with_GP(sv)) {
b21a80ce 2117 return (bool)PTR2IV(glob_2inpuv((GV *)sv, NULL, TRUE));
180488f8 2118 }
180488f8 2119
af359546
NC
2120 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2121 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2122 report_uninit(sv);
2123 }
25da4f38
IZ
2124 if (SvTYPE(sv) < SVt_IV)
2125 /* Typically the caller expects that sv_any is not NULL now. */
2126 sv_upgrade(sv, SVt_IV);
af359546
NC
2127 /* Return 0 from the caller. */
2128 return TRUE;
2129 }
2130 return FALSE;
2131}
2132
2133/*
2134=for apidoc sv_2iv_flags
2135
2136Return the integer value of an SV, doing any necessary string
2137conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2138Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2139
2140=cut
2141*/
2142
2143IV
2144Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
2145{
97aff369 2146 dVAR;
af359546 2147 if (!sv)
a0d0e21e 2148 return 0;
af359546
NC
2149 if (SvGMAGICAL(sv)) {
2150 if (flags & SV_GMAGIC)
2151 mg_get(sv);
2152 if (SvIOKp(sv))
2153 return SvIVX(sv);
2154 if (SvNOKp(sv)) {
2155 return I_V(SvNVX(sv));
2156 }
71c558c3
NC
2157 if (SvPOKp(sv) && SvLEN(sv)) {
2158 UV value;
2159 const int numtype
2160 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2161
2162 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2163 == IS_NUMBER_IN_UV) {
2164 /* It's definitely an integer */
2165 if (numtype & IS_NUMBER_NEG) {
2166 if (value < (UV)IV_MIN)
2167 return -(IV)value;
2168 } else {
2169 if (value < (UV)IV_MAX)
2170 return (IV)value;
2171 }
2172 }
2173 if (!numtype) {
2174 if (ckWARN(WARN_NUMERIC))
2175 not_a_number(sv);
2176 }
2177 return I_V(Atof(SvPVX_const(sv)));
2178 }
1c7ff15e
NC
2179 if (SvROK(sv)) {
2180 goto return_rok;
af359546 2181 }
1c7ff15e
NC
2182 assert(SvTYPE(sv) >= SVt_PVMG);
2183 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2184 } else if (SvTHINKFIRST(sv)) {
af359546 2185 if (SvROK(sv)) {
1c7ff15e 2186 return_rok:
af359546
NC
2187 if (SvAMAGIC(sv)) {
2188 SV * const tmpstr=AMG_CALLun(sv,numer);
2189 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2190 return SvIV(tmpstr);
2191 }
2192 }
2193 return PTR2IV(SvRV(sv));
2194 }
2195 if (SvIsCOW(sv)) {
2196 sv_force_normal_flags(sv, 0);
2197 }
2198 if (SvREADONLY(sv) && !SvOK(sv)) {
2199 if (ckWARN(WARN_UNINITIALIZED))
2200 report_uninit(sv);
2201 return 0;
2202 }
2203 }
2204 if (!SvIOKp(sv)) {
2205 if (S_sv_2iuv_common(aTHX_ sv))
2206 return 0;
79072805 2207 }
1d7c1841
GS
2208 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2209 PTR2UV(sv),SvIVX(sv)));
25da4f38 2210 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2211}
2212
645c22ef 2213/*
891f9566 2214=for apidoc sv_2uv_flags
645c22ef
DM
2215
2216Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2217conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2218Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2219
2220=cut
2221*/
2222
ff68c719 2223UV
891f9566 2224Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 2225{
97aff369 2226 dVAR;
ff68c719 2227 if (!sv)
2228 return 0;
2229 if (SvGMAGICAL(sv)) {
891f9566
YST
2230 if (flags & SV_GMAGIC)
2231 mg_get(sv);
ff68c719 2232 if (SvIOKp(sv))
2233 return SvUVX(sv);
2234 if (SvNOKp(sv))
2235 return U_V(SvNVX(sv));
71c558c3
NC
2236 if (SvPOKp(sv) && SvLEN(sv)) {
2237 UV value;
2238 const int numtype
2239 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2240
2241 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2242 == IS_NUMBER_IN_UV) {
2243 /* It's definitely an integer */
2244 if (!(numtype & IS_NUMBER_NEG))
2245 return value;
2246 }
2247 if (!numtype) {
2248 if (ckWARN(WARN_NUMERIC))
2249 not_a_number(sv);
2250 }
2251 return U_V(Atof(SvPVX_const(sv)));
2252 }
1c7ff15e
NC
2253 if (SvROK(sv)) {
2254 goto return_rok;
3fe9a6f1 2255 }
1c7ff15e
NC
2256 assert(SvTYPE(sv) >= SVt_PVMG);
2257 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2258 } else if (SvTHINKFIRST(sv)) {
ff68c719 2259 if (SvROK(sv)) {
1c7ff15e 2260 return_rok:
deb46114
NC
2261 if (SvAMAGIC(sv)) {
2262 SV *const tmpstr = AMG_CALLun(sv,numer);
2263 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2264 return SvUV(tmpstr);
2265 }
2266 }
2267 return PTR2UV(SvRV(sv));
ff68c719 2268 }
765f542d
NC
2269 if (SvIsCOW(sv)) {
2270 sv_force_normal_flags(sv, 0);
8a818333 2271 }
0336b60e 2272 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2273 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2274 report_uninit(sv);
ff68c719 2275 return 0;
2276 }
2277 }
af359546
NC
2278 if (!SvIOKp(sv)) {
2279 if (S_sv_2iuv_common(aTHX_ sv))
2280 return 0;
ff68c719 2281 }
25da4f38 2282
1d7c1841
GS
2283 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2284 PTR2UV(sv),SvUVX(sv)));
25da4f38 2285 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2286}
2287
645c22ef
DM
2288/*
2289=for apidoc sv_2nv
2290
2291Return the num value of an SV, doing any necessary string or integer
2292conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2293macros.
2294
2295=cut
2296*/
2297
65202027 2298NV
864dbfa3 2299Perl_sv_2nv(pTHX_ register SV *sv)
79072805 2300{
97aff369 2301 dVAR;
79072805
LW
2302 if (!sv)
2303 return 0.0;
8990e307 2304 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2305 mg_get(sv);
2306 if (SvNOKp(sv))
2307 return SvNVX(sv);
0aa395f8 2308 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2309 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2310 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2311 not_a_number(sv);
3f7c398e 2312 return Atof(SvPVX_const(sv));
a0d0e21e 2313 }
25da4f38 2314 if (SvIOKp(sv)) {
1c846c1f 2315 if (SvIsUV(sv))
65202027 2316 return (NV)SvUVX(sv);
25da4f38 2317 else
65202027 2318 return (NV)SvIVX(sv);
47a72cb8
NC
2319 }
2320 if (SvROK(sv)) {
2321 goto return_rok;
2322 }
2323 assert(SvTYPE(sv) >= SVt_PVMG);
2324 /* This falls through to the report_uninit near the end of the
2325 function. */
2326 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2327 if (SvROK(sv)) {
47a72cb8 2328 return_rok:
deb46114
NC
2329 if (SvAMAGIC(sv)) {
2330 SV *const tmpstr = AMG_CALLun(sv,numer);
2331 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2332 return SvNV(tmpstr);
2333 }
2334 }
2335 return PTR2NV(SvRV(sv));
a0d0e21e 2336 }
765f542d
NC
2337 if (SvIsCOW(sv)) {
2338 sv_force_normal_flags(sv, 0);
8a818333 2339 }
0336b60e 2340 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2341 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2342 report_uninit(sv);
ed6116ce
LW
2343 return 0.0;
2344 }
79072805
LW
2345 }
2346 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2347 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2348 sv_upgrade(sv, SVt_NV);
906f284f 2349#ifdef USE_LONG_DOUBLE
097ee67d 2350 DEBUG_c({
f93f4e46 2351 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2352 PerlIO_printf(Perl_debug_log,
2353 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2354 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2355 RESTORE_NUMERIC_LOCAL();
2356 });
65202027 2357#else
572bbb43 2358 DEBUG_c({
f93f4e46 2359 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2360 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2361 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2362 RESTORE_NUMERIC_LOCAL();
2363 });
572bbb43 2364#endif
79072805
LW
2365 }
2366 else if (SvTYPE(sv) < SVt_PVNV)
2367 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2368 if (SvNOKp(sv)) {
2369 return SvNVX(sv);
61604483 2370 }
59d8ce62 2371 if (SvIOKp(sv)) {
9d6ce603 2372 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2373#ifdef NV_PRESERVES_UV
2374 SvNOK_on(sv);
2375#else
2376 /* Only set the public NV OK flag if this NV preserves the IV */
2377 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2378 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2379 : (SvIVX(sv) == I_V(SvNVX(sv))))
2380 SvNOK_on(sv);
2381 else
2382 SvNOKp_on(sv);
2383#endif
93a17b20 2384 }
748a9306 2385 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2386 UV value;
3f7c398e 2387 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2388 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2389 not_a_number(sv);
28e5dec8 2390#ifdef NV_PRESERVES_UV
c2988b20
NC
2391 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2392 == IS_NUMBER_IN_UV) {
5e045b90 2393 /* It's definitely an integer */
9d6ce603 2394 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2395 } else
3f7c398e 2396 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2397 SvNOK_on(sv);
2398#else
3f7c398e 2399 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2400 /* Only set the public NV OK flag if this NV preserves the value in
2401 the PV at least as well as an IV/UV would.
2402 Not sure how to do this 100% reliably. */
2403 /* if that shift count is out of range then Configure's test is
2404 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2405 UV_BITS */
2406 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2407 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2408 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2409 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2410 /* Can't use strtol etc to convert this string, so don't try.
2411 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2412 SvNOK_on(sv);
2413 } else {
2414 /* value has been set. It may not be precise. */
2415 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2416 /* 2s complement assumption for (UV)IV_MIN */
2417 SvNOK_on(sv); /* Integer is too negative. */
2418 } else {
2419 SvNOKp_on(sv);
2420 SvIOKp_on(sv);
6fa402ec 2421
c2988b20 2422 if (numtype & IS_NUMBER_NEG) {
45977657 2423 SvIV_set(sv, -(IV)value);
c2988b20 2424 } else if (value <= (UV)IV_MAX) {
45977657 2425 SvIV_set(sv, (IV)value);
c2988b20 2426 } else {
607fa7f2 2427 SvUV_set(sv, value);
c2988b20
NC
2428 SvIsUV_on(sv);
2429 }
2430
2431 if (numtype & IS_NUMBER_NOT_INT) {
2432 /* I believe that even if the original PV had decimals,
2433 they are lost beyond the limit of the FP precision.
2434 However, neither is canonical, so both only get p
2435 flags. NWC, 2000/11/25 */
2436 /* Both already have p flags, so do nothing */
2437 } else {
66a1b24b 2438 const NV nv = SvNVX(sv);
c2988b20
NC
2439 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2440 if (SvIVX(sv) == I_V(nv)) {
2441 SvNOK_on(sv);
c2988b20 2442 } else {
c2988b20
NC
2443 /* It had no "." so it must be integer. */
2444 }
00b6aa41 2445 SvIOK_on(sv);
c2988b20
NC
2446 } else {
2447 /* between IV_MAX and NV(UV_MAX).
2448 Could be slightly > UV_MAX */
6fa402ec 2449
c2988b20
NC
2450 if (numtype & IS_NUMBER_NOT_INT) {
2451 /* UV and NV both imprecise. */
2452 } else {
66a1b24b 2453 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2454
2455 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2456 SvNOK_on(sv);
c2988b20 2457 }
00b6aa41 2458 SvIOK_on(sv);
c2988b20
NC
2459 }
2460 }
2461 }
2462 }
2463 }
28e5dec8 2464#endif /* NV_PRESERVES_UV */
93a17b20 2465 }
79072805 2466 else {
f7877b28 2467 if (isGV_with_GP(sv)) {
180488f8
NC
2468 glob_2inpuv((GV *)sv, NULL, TRUE);
2469 return 0.0;
2470 }
2471
041457d9 2472 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2473 report_uninit(sv);
7e25a7e9
NC
2474 assert (SvTYPE(sv) >= SVt_NV);
2475 /* Typically the caller expects that sv_any is not NULL now. */
2476 /* XXX Ilya implies that this is a bug in callers that assume this
2477 and ideally should be fixed. */
a0d0e21e 2478 return 0.0;
79072805 2479 }
572bbb43 2480#if defined(USE_LONG_DOUBLE)
097ee67d 2481 DEBUG_c({
f93f4e46 2482 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2483 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2484 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2485 RESTORE_NUMERIC_LOCAL();
2486 });
65202027 2487#else
572bbb43 2488 DEBUG_c({
f93f4e46 2489 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2490 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2491 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2492 RESTORE_NUMERIC_LOCAL();
2493 });
572bbb43 2494#endif
463ee0b2 2495 return SvNVX(sv);
79072805
LW
2496}
2497
645c22ef
DM
2498/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2499 * UV as a string towards the end of buf, and return pointers to start and
2500 * end of it.
2501 *
2502 * We assume that buf is at least TYPE_CHARS(UV) long.
2503 */
2504
864dbfa3 2505static char *
aec46f14 2506S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2507{
25da4f38 2508 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2509 char * const ebuf = ptr;
25da4f38 2510 int sign;
25da4f38
IZ
2511
2512 if (is_uv)
2513 sign = 0;
2514 else if (iv >= 0) {
2515 uv = iv;
2516 sign = 0;
2517 } else {
2518 uv = -iv;
2519 sign = 1;
2520 }
2521 do {
eb160463 2522 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2523 } while (uv /= 10);
2524 if (sign)
2525 *--ptr = '-';
2526 *peob = ebuf;
2527 return ptr;
2528}
2529
9af30d34
NC
2530/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2531 * a regexp to its stringified form.
2532 */
2533
2534static char *
2535S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
97aff369 2536 dVAR;
00b6aa41 2537 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2538
2539 if (!mg->mg_ptr) {
2540 const char *fptr = "msix";
2541 char reflags[6];
2542 char ch;
2543 int left = 0;
2544 int right = 4;
00b6aa41 2545 bool need_newline = 0;
9af30d34
NC
2546 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2547
2548 while((ch = *fptr++)) {
2549 if(reganch & 1) {
2550 reflags[left++] = ch;
2551 }
2552 else {
2553 reflags[right--] = ch;
2554 }
2555 reganch >>= 1;
2556 }
2557 if(left != 4) {
2558 reflags[left] = '-';
2559 left = 5;
2560 }
2561
2562 mg->mg_len = re->prelen + 4 + left;
2563 /*
2564 * If /x was used, we have to worry about a regex ending with a
2565 * comment later being embedded within another regex. If so, we don't
2566 * want this regex's "commentization" to leak out to the right part of
2567 * the enclosing regex, we must cap it with a newline.
2568 *
2569 * So, if /x was used, we scan backwards from the end of the regex. If
2570 * we find a '#' before we find a newline, we need to add a newline
2571 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2572 * we don't need to add anything. -jfriedl
2573 */
2574 if (PMf_EXTENDED & re->reganch) {
2575 const char *endptr = re->precomp + re->prelen;
2576 while (endptr >= re->precomp) {
2577 const char c = *(endptr--);
2578 if (c == '\n')
2579 break; /* don't need another */
2580 if (c == '#') {
2581 /* we end while in a comment, so we need a newline */
2582 mg->mg_len++; /* save space for it */
2583 need_newline = 1; /* note to add it */
2584 break;
2585 }
2586 }
2587 }
2588
2589 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2590 mg->mg_ptr[0] = '(';
2591 mg->mg_ptr[1] = '?';
2592 Copy(reflags, mg->mg_ptr+2, left, char);
2593 *(mg->mg_ptr+left+2) = ':';
2594 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2595 if (need_newline)
2596 mg->mg_ptr[mg->mg_len - 2] = '\n';
2597 mg->mg_ptr[mg->mg_len - 1] = ')';
2598 mg->mg_ptr[mg->mg_len] = 0;
2599 }
2600 PL_reginterp_cnt += re->program[0].next_off;
2601
2602 if (re->reganch & ROPT_UTF8)
2603 SvUTF8_on(sv);
2604 else
2605 SvUTF8_off(sv);
2606 if (lp)
2607 *lp = mg->mg_len;
2608 return mg->mg_ptr;
2609}
2610
645c22ef
DM
2611/*
2612=for apidoc sv_2pv_flags
2613
ff276b08 2614Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2615If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2616if necessary.
2617Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2618usually end up here too.
2619
2620=cut
2621*/
2622
8d6d96c1
HS
2623char *
2624Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2625{
97aff369 2626 dVAR;
79072805 2627 register char *s;
79072805 2628
463ee0b2 2629 if (!sv) {
cdb061a3
NC
2630 if (lp)
2631 *lp = 0;
73d840c0 2632 return (char *)"";
463ee0b2 2633 }
8990e307 2634 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2635 if (flags & SV_GMAGIC)
2636 mg_get(sv);
463ee0b2 2637 if (SvPOKp(sv)) {
cdb061a3
NC
2638 if (lp)
2639 *lp = SvCUR(sv);
10516c54
NC
2640 if (flags & SV_MUTABLE_RETURN)
2641 return SvPVX_mutable(sv);
4d84ee25
NC
2642 if (flags & SV_CONST_RETURN)
2643 return (char *)SvPVX_const(sv);
463ee0b2
LW
2644 return SvPVX(sv);
2645 }
75dfc8ec
NC
2646 if (SvIOKp(sv) || SvNOKp(sv)) {
2647 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2648 STRLEN len;
2649
2650 if (SvIOKp(sv)) {
e8ada2d0
NC
2651 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2652 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2653 } else {
e8ada2d0
NC
2654 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2655 len = strlen(tbuf);
75dfc8ec 2656 }
b5b886f0
NC
2657 assert(!SvROK(sv));
2658 {
75dfc8ec
NC
2659 dVAR;
2660
2661#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2662 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2663 tbuf[0] = '0';
2664 tbuf[1] = 0;
75dfc8ec
NC
2665 len = 1;
2666 }
2667#endif
2668 SvUPGRADE(sv, SVt_PV);
2669 if (lp)
2670 *lp = len;
2671 s = SvGROW_mutable(sv, len + 1);
2672 SvCUR_set(sv, len);
2673 SvPOKp_on(sv);
e8ada2d0 2674 return memcpy(s, tbuf, len + 1);
75dfc8ec 2675 }
463ee0b2 2676 }
1c7ff15e
NC
2677 if (SvROK(sv)) {
2678 goto return_rok;
2679 }
2680 assert(SvTYPE(sv) >= SVt_PVMG);
2681 /* This falls through to the report_uninit near the end of the
2682 function. */
2683 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2684 if (SvROK(sv)) {
1c7ff15e 2685 return_rok:
deb46114
NC
2686 if (SvAMAGIC(sv)) {
2687 SV *const tmpstr = AMG_CALLun(sv,string);
2688 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2689 /* Unwrap this: */
2690 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2691 */
2692
2693 char *pv;
2694 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2695 if (flags & SV_CONST_RETURN) {
2696 pv = (char *) SvPVX_const(tmpstr);
2697 } else {
2698 pv = (flags & SV_MUTABLE_RETURN)
2699 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2700 }
2701 if (lp)
2702 *lp = SvCUR(tmpstr);
50adf7d2 2703 } else {
deb46114 2704 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2705 }
deb46114
NC
2706 if (SvUTF8(tmpstr))
2707 SvUTF8_on(sv);
2708 else
2709 SvUTF8_off(sv);
2710 return pv;
50adf7d2 2711 }
deb46114
NC
2712 }
2713 {
75dfc8ec 2714 SV *tsv;
f9277f47 2715 MAGIC *mg;
d8eae41e
NC
2716 const SV *const referent = (SV*)SvRV(sv);
2717
2718 if (!referent) {
396482e1 2719 tsv = sv_2mortal(newSVpvs("NULLREF"));
042dae7a
NC
2720 } else if (SvTYPE(referent) == SVt_PVMG
2721 && ((SvFLAGS(referent) &
2722 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2723 == (SVs_OBJECT|SVs_SMG))
2724 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
c445ea15 2725 return stringify_regexp(sv, mg, lp);
d8eae41e
NC
2726 } else {
2727 const char *const typestr = sv_reftype(referent, 0);
2728
2729 tsv = sv_newmortal();
2730 if (SvOBJECT(referent)) {
2731 const char *const name = HvNAME_get(SvSTASH(referent));
2732 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2733 name ? name : "__ANON__" , typestr,
2734 PTR2UV(referent));
2735 }
2736 else
2737 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2738 PTR2UV(referent));
c080367d 2739 }
042dae7a
NC
2740 if (lp)
2741 *lp = SvCUR(tsv);
2742 return SvPVX(tsv);
463ee0b2 2743 }
79072805 2744 }
0336b60e 2745 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2746 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2747 report_uninit(sv);
cdb061a3
NC
2748 if (lp)
2749 *lp = 0;
73d840c0 2750 return (char *)"";
79072805 2751 }
79072805 2752 }
28e5dec8
JH
2753 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2754 /* I'm assuming that if both IV and NV are equally valid then
2755 converting the IV is going to be more efficient */
e1ec3a88
AL
2756 const U32 isIOK = SvIOK(sv);
2757 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2758 char buf[TYPE_CHARS(UV)];
2759 char *ebuf, *ptr;
2760
2761 if (SvTYPE(sv) < SVt_PVIV)
2762 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2763 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
5902b6a9
NC
2764 /* inlined from sv_setpvn */
2765 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2766 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2767 SvCUR_set(sv, ebuf - ptr);
2768 s = SvEND(sv);
2769 *s = '\0';
2770 if (isIOK)
2771 SvIOK_on(sv);
2772 else
2773 SvIOKp_on(sv);
2774 if (isUIOK)
2775 SvIsUV_on(sv);
2776 }
2777 else if (SvNOKp(sv)) {
c81271c3 2778 const int olderrno = errno;
79072805
LW
2779 if (SvTYPE(sv) < SVt_PVNV)
2780 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2781 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2782 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2783 /* some Xenix systems wipe out errno here */
79072805 2784#ifdef apollo
463ee0b2 2785 if (SvNVX(sv) == 0.0)
79072805
LW
2786 (void)strcpy(s,"0");
2787 else
2788#endif /*apollo*/
bbce6d69 2789 {
2d4389e4 2790 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2791 }
79072805 2792 errno = olderrno;
a0d0e21e
LW
2793#ifdef FIXNEGATIVEZERO
2794 if (*s == '-' && s[1] == '0' && !s[2])
2795 strcpy(s,"0");
2796#endif
79072805
LW
2797 while (*s) s++;
2798#ifdef hcx
2799 if (s[-1] == '.')
46fc3d4c 2800 *--s = '\0';
79072805
LW
2801#endif
2802 }
79072805 2803 else {
f7877b28 2804 if (isGV_with_GP(sv)) {
180488f8
NC
2805 return glob_2inpuv((GV *)sv, lp, FALSE);
2806 }
2807
041457d9 2808 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2809 report_uninit(sv);
cdb061a3 2810 if (lp)
00b6aa41 2811 *lp = 0;
25da4f38
IZ
2812 if (SvTYPE(sv) < SVt_PV)
2813 /* Typically the caller expects that sv_any is not NULL now. */
2814 sv_upgrade(sv, SVt_PV);
73d840c0 2815 return (char *)"";
79072805 2816 }
cdb061a3 2817 {
823a54a3 2818 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2819 if (lp)
2820 *lp = len;
2821 SvCUR_set(sv, len);
2822 }
79072805 2823 SvPOK_on(sv);
1d7c1841 2824 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2825 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2826 if (flags & SV_CONST_RETURN)
2827 return (char *)SvPVX_const(sv);
10516c54
NC
2828 if (flags & SV_MUTABLE_RETURN)
2829 return SvPVX_mutable(sv);
463ee0b2
LW
2830 return SvPVX(sv);
2831}
2832
645c22ef 2833/*
6050d10e
JP
2834=for apidoc sv_copypv
2835
2836Copies a stringified representation of the source SV into the
2837destination SV. Automatically performs any necessary mg_get and
54f0641b 2838coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2839UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2840sv_2pv[_flags] but operates directly on an SV instead of just the
2841string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2842would lose the UTF-8'ness of the PV.
2843
2844=cut
2845*/
2846
2847void
2848Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2849{
446eaa42 2850 STRLEN len;
53c1dcc0 2851 const char * const s = SvPV_const(ssv,len);
cb50f42d 2852 sv_setpvn(dsv,s,len);
446eaa42 2853 if (SvUTF8(ssv))
cb50f42d 2854 SvUTF8_on(dsv);
446eaa42 2855 else
cb50f42d 2856 SvUTF8_off(dsv);
6050d10e
JP
2857}
2858
2859/*
645c22ef
DM
2860=for apidoc sv_2pvbyte
2861
2862Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2863to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2864side-effect.
2865
2866Usually accessed via the C<SvPVbyte> macro.
2867
2868=cut
2869*/
2870
7340a771
GS
2871char *
2872Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2873{
0875d2fe 2874 sv_utf8_downgrade(sv,0);
97972285 2875 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2876}
2877
645c22ef 2878/*
035cbb0e
RGS
2879=for apidoc sv_2pvutf8
2880
2881Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2882to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2883
2884Usually accessed via the C<SvPVutf8> macro.
2885
2886=cut
2887*/
645c22ef 2888
7340a771
GS
2889char *
2890Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2891{
035cbb0e
RGS
2892 sv_utf8_upgrade(sv);
2893 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2894}
1c846c1f 2895
7ee2227d 2896
645c22ef
DM
2897/*
2898=for apidoc sv_2bool
2899
2900This function is only called on magical items, and is only used by
8cf8f3d1 2901sv_true() or its macro equivalent.
645c22ef
DM
2902
2903=cut
2904*/
2905
463ee0b2 2906bool
864dbfa3 2907Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2908{
97aff369 2909 dVAR;
5b295bef 2910 SvGETMAGIC(sv);
463ee0b2 2911
a0d0e21e
LW
2912 if (!SvOK(sv))
2913 return 0;
2914 if (SvROK(sv)) {
fabdb6c0
AL
2915 if (SvAMAGIC(sv)) {
2916 SV * const tmpsv = AMG_CALLun(sv,bool_);
2917 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2918 return (bool)SvTRUE(tmpsv);
2919 }
2920 return SvRV(sv) != 0;
a0d0e21e 2921 }
463ee0b2 2922 if (SvPOKp(sv)) {
53c1dcc0
AL
2923 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2924 if (Xpvtmp &&
339049b0 2925 (*sv->sv_u.svu_pv > '0' ||
11343788 2926 Xpvtmp->xpv_cur > 1 ||
339049b0 2927 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2928 return 1;
2929 else
2930 return 0;
2931 }
2932 else {
2933 if (SvIOKp(sv))
2934 return SvIVX(sv) != 0;
2935 else {
2936 if (SvNOKp(sv))
2937 return SvNVX(sv) != 0.0;
180488f8 2938 else {
f7877b28 2939 if (isGV_with_GP(sv))
180488f8
NC
2940 return TRUE;
2941 else
2942 return FALSE;
2943 }
463ee0b2
LW
2944 }
2945 }
79072805
LW
2946}
2947
c461cf8f
JH
2948/*
2949=for apidoc sv_utf8_upgrade
2950
78ea37eb 2951Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2952Forces the SV to string form if it is not already.
4411f3b6
NIS
2953Always sets the SvUTF8 flag to avoid future validity checks even
2954if all the bytes have hibit clear.
c461cf8f 2955
13a6c0e0
JH
2956This is not as a general purpose byte encoding to Unicode interface:
2957use the Encode extension for that.
2958
8d6d96c1
HS
2959=for apidoc sv_utf8_upgrade_flags
2960
78ea37eb 2961Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2962Forces the SV to string form if it is not already.
8d6d96c1
HS
2963Always sets the SvUTF8 flag to avoid future validity checks even
2964if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2965will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2966C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2967
13a6c0e0
JH
2968This is not as a general purpose byte encoding to Unicode interface:
2969use the Encode extension for that.
2970
8d6d96c1
HS
2971=cut
2972*/
2973
2974STRLEN
2975Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2976{
97aff369 2977 dVAR;
808c356f
RGS
2978 if (sv == &PL_sv_undef)
2979 return 0;
e0e62c2a
NIS
2980 if (!SvPOK(sv)) {
2981 STRLEN len = 0;
d52b7888
NC
2982 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2983 (void) sv_2pv_flags(sv,&len, flags);
2984 if (SvUTF8(sv))
2985 return len;
2986 } else {
2987 (void) SvPV_force(sv,len);
2988 }
e0e62c2a 2989 }
4411f3b6 2990
f5cee72b 2991 if (SvUTF8(sv)) {
5fec3b1d 2992 return SvCUR(sv);
f5cee72b 2993 }
5fec3b1d 2994
765f542d
NC
2995 if (SvIsCOW(sv)) {
2996 sv_force_normal_flags(sv, 0);
db42d148
NIS
2997 }
2998
88632417 2999 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 3000 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 3001 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
3002 /* This function could be much more efficient if we
3003 * had a FLAG in SVs to signal if there are any hibit
3004 * chars in the PV. Given that there isn't such a flag
3005 * make the loop as fast as possible. */
00b6aa41 3006 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3007 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3008 const U8 *t = s;
c4e7c712
NC
3009
3010 while (t < e) {
53c1dcc0 3011 const U8 ch = *t++;
00b6aa41
AL
3012 /* Check for hi bit */
3013 if (!NATIVE_IS_INVARIANT(ch)) {
3014 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
3015 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3016
3017 SvPV_free(sv); /* No longer using what was there before. */
3018 SvPV_set(sv, (char*)recoded);
3019 SvCUR_set(sv, len - 1);
3020 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 3021 break;
00b6aa41 3022 }
c4e7c712
NC
3023 }
3024 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3025 SvUTF8_on(sv);
560a288e 3026 }
4411f3b6 3027 return SvCUR(sv);
560a288e
GS
3028}
3029
c461cf8f
JH
3030/*
3031=for apidoc sv_utf8_downgrade
3032
78ea37eb
TS
3033Attempts to convert the PV of an SV from characters to bytes.
3034If the PV contains a character beyond byte, this conversion will fail;
3035in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3036true, croaks.
3037
13a6c0e0
JH
3038This is not as a general purpose Unicode to byte encoding interface:
3039use the Encode extension for that.
3040
c461cf8f
JH
3041=cut
3042*/
3043
560a288e
GS
3044bool
3045Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
3046{
97aff369 3047 dVAR;
78ea37eb 3048 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3049 if (SvCUR(sv)) {
03cfe0ae 3050 U8 *s;
652088fc 3051 STRLEN len;
fa301091 3052
765f542d
NC
3053 if (SvIsCOW(sv)) {
3054 sv_force_normal_flags(sv, 0);
3055 }
03cfe0ae
NIS
3056 s = (U8 *) SvPV(sv, len);
3057 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3058 if (fail_ok)
3059 return FALSE;
3060 else {
3061 if (PL_op)
3062 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3063 OP_DESC(PL_op));
fa301091
JH
3064 else
3065 Perl_croak(aTHX_ "Wide character");
3066 }
4b3603a4 3067 }
b162af07 3068 SvCUR_set(sv, len);
67e989fb 3069 }
560a288e 3070 }
ffebcc3e 3071 SvUTF8_off(sv);
560a288e
GS
3072 return TRUE;
3073}
3074
c461cf8f
JH
3075/*
3076=for apidoc sv_utf8_encode
3077
78ea37eb
TS
3078Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3079flag off so that it looks like octets again.
c461cf8f
JH
3080
3081=cut
3082*/
3083
560a288e
GS
3084void
3085Perl_sv_utf8_encode(pTHX_ register SV *sv)
3086{
4411f3b6 3087 (void) sv_utf8_upgrade(sv);
4c94c214
NC
3088 if (SvIsCOW(sv)) {
3089 sv_force_normal_flags(sv, 0);
3090 }
3091 if (SvREADONLY(sv)) {
3092 Perl_croak(aTHX_ PL_no_modify);
3093 }
560a288e
GS
3094 SvUTF8_off(sv);
3095}
3096
4411f3b6
NIS
3097/*
3098=for apidoc sv_utf8_decode
3099
78ea37eb
TS
3100If the PV of the SV is an octet sequence in UTF-8
3101and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3102so that it looks like a character. If the PV contains only single-byte
3103characters, the C<SvUTF8> flag stays being off.
3104Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3105
3106=cut
3107*/
3108
560a288e
GS
3109bool
3110Perl_sv_utf8_decode(pTHX_ register SV *sv)
3111{
78ea37eb 3112 if (SvPOKp(sv)) {
93524f2b
NC
3113 const U8 *c;
3114 const U8 *e;
9cbac4c7 3115
645c22ef
DM
3116 /* The octets may have got themselves encoded - get them back as
3117 * bytes
3118 */
3119 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3120 return FALSE;
3121
3122 /* it is actually just a matter of turning the utf8 flag on, but
3123 * we want to make sure everything inside is valid utf8 first.
3124 */
93524f2b 3125 c = (const U8 *) SvPVX_const(sv);
63cd0674 3126 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3127 return FALSE;
93524f2b 3128 e = (const U8 *) SvEND(sv);
511c2ff0 3129 while (c < e) {
b64e5050 3130 const U8 ch = *c++;
c4d5f83a 3131 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3132 SvUTF8_on(sv);
3133 break;
3134 }
560a288e 3135 }
560a288e
GS
3136 }
3137 return TRUE;
3138}
3139
954c1994
GS
3140/*
3141=for apidoc sv_setsv
3142
645c22ef
DM
3143Copies the contents of the source SV C<ssv> into the destination SV
3144C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3145function if the source SV needs to be reused. Does not handle 'set' magic.
3146Loosely speaking, it performs a copy-by-value, obliterating any previous
3147content of the destination.
3148
3149You probably want to use one of the assortment of wrappers, such as
3150C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3151C<SvSetMagicSV_nosteal>.
3152
8d6d96c1
HS
3153=for apidoc sv_setsv_flags
3154
645c22ef
DM
3155Copies the contents of the source SV C<ssv> into the destination SV
3156C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3157function if the source SV needs to be reused. Does not handle 'set' magic.
3158Loosely speaking, it performs a copy-by-value, obliterating any previous
3159content of the destination.
3160If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3161C<ssv> if appropriate, else not. If the C<flags> parameter has the
3162C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3163and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3164
3165You probably want to use one of the assortment of wrappers, such as
3166C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3167C<SvSetMagicSV_nosteal>.
3168
3169This is the primary function for copying scalars, and most other
3170copy-ish functions and macros use this underneath.
8d6d96c1
HS
3171
3172=cut
3173*/
3174
5d0301b7 3175static void
2eb42952 3176S_glob_assign_glob(pTHX_ SV *dstr, SV *sstr, const int dtype)
5d0301b7
NC
3177{
3178 if (dtype != SVt_PVGV) {
3179 const char * const name = GvNAME(sstr);
3180 const STRLEN len = GvNAMELEN(sstr);
3181 /* don't upgrade SVt_PVLV: it can hold a glob */
f7877b28
NC
3182 if (dtype != SVt_PVLV) {
3183 if (dtype >= SVt_PV) {
3184 SvPV_free(dstr);
3185 SvPV_set(dstr, 0);
3186 SvLEN_set(dstr, 0);
3187 SvCUR_set(dstr, 0);
3188 }
5d0301b7 3189 sv_upgrade(dstr, SVt_PVGV);
dedf8e73
NC
3190 (void)SvOK_off(dstr);
3191 SvSCREAM_on(dstr);
f7877b28 3192 }
5d0301b7
NC
3193 GvSTASH(dstr) = GvSTASH(sstr);
3194 if (GvSTASH(dstr))
3195 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3196 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3197 SvFAKE_on(dstr); /* can coerce to non-glob */
3198 }
3199
3200#ifdef GV_UNIQUE_CHECK
3201 if (GvUNIQUE((GV*)dstr)) {
3202 Perl_croak(aTHX_ PL_no_modify);
3203 }
3204#endif
3205
f7877b28
NC
3206 gp_free((GV*)dstr);
3207 SvSCREAM_off(dstr);
5d0301b7 3208 (void)SvOK_off(dstr);
f7877b28 3209 SvSCREAM_on(dstr);
dedf8e73 3210 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3211 GvGP(dstr) = gp_ref(GvGP(sstr));
3212 if (SvTAINTED(sstr))
3213 SvTAINT(dstr);
3214 if (GvIMPORTED(dstr) != GVf_IMPORTED
3215 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3216 {
3217 GvIMPORTED_on(dstr);
3218 }
3219 GvMULTI_on(dstr);
3220 return;
3221}
3222
b8473700 3223static void
2eb42952 3224S_glob_assign_ref(pTHX_ SV *dstr, SV *sstr) {
b8473700
NC
3225 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3226 SV *dref = NULL;
3227 const int intro = GvINTRO(dstr);
2440974c 3228 SV **location;
3386d083 3229 U8 import_flag = 0;
27242d61
NC
3230 const U32 stype = SvTYPE(sref);
3231
b8473700
NC
3232
3233#ifdef GV_UNIQUE_CHECK
3234 if (GvUNIQUE((GV*)dstr)) {
3235 Perl_croak(aTHX_ PL_no_modify);
3236 }
3237#endif
3238
3239 if (intro) {
3240 GvINTRO_off(dstr); /* one-shot flag */
3241 GvLINE(dstr) = CopLINE(PL_curcop);
3242 GvEGV(dstr) = (GV*)dstr;
3243 }
3244 GvMULTI_on(dstr);
27242d61 3245 switch (stype) {
b8473700 3246 case SVt_PVCV:
27242d61
NC
3247 location = (SV **) &GvCV(dstr);
3248 import_flag = GVf_IMPORTED_CV;
3249 goto common;
3250 case SVt_PVHV:
3251 location = (SV **) &GvHV(dstr);
3252 import_flag = GVf_IMPORTED_HV;
3253 goto common;
3254 case SVt_PVAV:
3255 location = (SV **) &GvAV(dstr);
3256 import_flag = GVf_IMPORTED_AV;
3257 goto common;
3258 case SVt_PVIO:
3259 location = (SV **) &GvIOp(dstr);
3260 goto common;
3261 case SVt_PVFM:
3262 location = (SV **) &GvFORM(dstr);
3263 default:
3264 location = &GvSV(dstr);
3265 import_flag = GVf_IMPORTED_SV;
3266 common:
b8473700 3267 if (intro) {
27242d61
NC
3268 if (stype == SVt_PVCV) {
3269 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3270 SvREFCNT_dec(GvCV(dstr));
3271 GvCV(dstr) = NULL;
3272 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3273 PL_sub_generation++;
3274 }
b8473700 3275 }
27242d61 3276 SAVEGENERICSV(*location);
b8473700
NC
3277 }
3278 else
27242d61
NC
3279 dref = *location;
3280 if (stype == SVt_PVCV && *location != sref) {
3281 CV* const cv = (CV*)*location;
b8473700
NC
3282 if (cv) {
3283 if (!GvCVGEN((GV*)dstr) &&
3284 (CvROOT(cv) || CvXSUB(cv)))
3285 {
3286 /* Redefining a sub - warning is mandatory if
3287 it was a const and its value changed. */
3288 if (CvCONST(cv) && CvCONST((CV*)sref)
3289 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
bb263b4e 3290 /*EMPTY*/
b8473700
NC
3291 /* They are 2 constant subroutines generated from
3292 the same constant. This probably means that
3293 they are really the "same" proxy subroutine
3294 instantiated in 2 places. Most likely this is
3295 when a constant is exported twice. Don't warn.
3296 */
3297 }
3298 else if (ckWARN(WARN_REDEFINE)
3299 || (CvCONST(cv)
3300 && (!CvCONST((CV*)sref)
3301 || sv_cmp(cv_const_sv(cv),
3302 cv_const_sv((CV*)sref))))) {
3303 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3304 CvCONST(cv)
3305 ? "Constant subroutine %s::%s redefined"
3306 : "Subroutine %s::%s redefined",
3307 HvNAME_get(GvSTASH((GV*)dstr)),
3308 GvENAME((GV*)dstr));
3309 }
3310 }
3311 if (!intro)
3312 cv_ckproto(cv, (GV*)dstr,
bd61b366 3313 SvPOK(sref) ? SvPVX_const(sref) : NULL);
b8473700 3314 }
b8473700
NC
3315 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3316 GvASSUMECV_on(dstr);
3317 PL_sub_generation++;
3318 }
2440974c 3319 *location = sref;
3386d083
NC
3320 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3321 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3322 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3323 }
3324 break;
3325 }
b37c2d43 3326 SvREFCNT_dec(dref);
b8473700
NC
3327 if (SvTAINTED(sstr))
3328 SvTAINT(dstr);
3329 return;
3330}
3331
8d6d96c1
HS
3332void
3333Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
3334{
97aff369 3335 dVAR;
8990e307
LW
3336 register U32 sflags;
3337 register int dtype;
3338 register int stype;
463ee0b2 3339
79072805
LW
3340 if (sstr == dstr)
3341 return;
765f542d 3342 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3343 if (!sstr)
3280af22 3344 sstr = &PL_sv_undef;
8990e307
LW
3345 stype = SvTYPE(sstr);
3346 dtype = SvTYPE(dstr);
79072805 3347
a0d0e21e 3348 SvAMAGIC_off(dstr);
7a5fa8a2 3349 if ( SvVOK(dstr) )
ece467f9
JP
3350 {
3351 /* need to nuke the magic */
3352 mg_free(dstr);
3353 SvRMAGICAL_off(dstr);
3354 }
9e7bc3e8 3355
463ee0b2 3356 /* There's a lot of redundancy below but we're going for speed here */
79072805 3357
8990e307 3358 switch (stype) {
79072805 3359 case SVt_NULL:
aece5585 3360 undef_sstr:
20408e3c
GS
3361 if (dtype != SVt_PVGV) {
3362 (void)SvOK_off(dstr);
3363 return;
3364 }
3365 break;
463ee0b2 3366 case SVt_IV:
aece5585
GA
3367 if (SvIOK(sstr)) {
3368 switch (dtype) {
3369 case SVt_NULL:
8990e307 3370 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3371 break;
3372 case SVt_NV:
aece5585
GA
3373 case SVt_RV:
3374 case SVt_PV:
a0d0e21e 3375 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
3376 break;
3377 }
3378 (void)SvIOK_only(dstr);
45977657 3379 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3380 if (SvIsUV(sstr))
3381 SvIsUV_on(dstr);
37c25af0
NC
3382 /* SvTAINTED can only be true if the SV has taint magic, which in
3383 turn means that the SV type is PVMG (or greater). This is the
3384 case statement for SVt_IV, so this cannot be true (whatever gcov
3385 may say). */
3386 assert(!SvTAINTED(sstr));
aece5585 3387 return;
8990e307 3388 }
aece5585
GA
3389 goto undef_sstr;
3390
463ee0b2 3391 case SVt_NV:
aece5585
GA
3392 if (SvNOK(sstr)) {
3393 switch (dtype) {
3394 case SVt_NULL:
3395 case SVt_IV:
8990e307 3396 sv_upgrade(dstr, SVt_NV);
aece5585
GA
3397 break;
3398 case SVt_RV:
3399 case SVt_PV:
3400 case SVt_PVIV:
a0d0e21e 3401 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
3402 break;
3403 }
9d6ce603 3404 SvNV_set(dstr, SvNVX(sstr));
aece5585 3405 (void)SvNOK_only(dstr);
37c25af0
NC
3406 /* SvTAINTED can only be true if the SV has taint magic, which in
3407 turn means that the SV type is PVMG (or greater). This is the
3408 case statement for SVt_NV, so this cannot be true (whatever gcov
3409 may say). */
3410 assert(!SvTAINTED(sstr));
aece5585 3411 return;
8990e307 3412 }
aece5585
GA
3413 goto undef_sstr;
3414
ed6116ce 3415 case SVt_RV:
8990e307 3416 if (dtype < SVt_RV)
ed6116ce 3417 sv_upgrade(dstr, SVt_RV);
ed6116ce 3418 break;
fc36a67e 3419 case SVt_PVFM:
f8c7b90f 3420#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3421 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3422 if (dtype < SVt_PVIV)
3423 sv_upgrade(dstr, SVt_PVIV);
3424 break;
3425 }
3426 /* Fall through */
3427#endif
3428 case SVt_PV:
8990e307 3429 if (dtype < SVt_PV)
463ee0b2 3430 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3431 break;
3432 case SVt_PVIV:
8990e307 3433 if (dtype < SVt_PVIV)
463ee0b2 3434 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3435 break;
3436 case SVt_PVNV:
8990e307 3437 if (dtype < SVt_PVNV)
463ee0b2 3438 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3439 break;
489f7bfe 3440 default:
a3b680e6
AL
3441 {
3442 const char * const type = sv_reftype(sstr,0);
533c011a 3443 if (PL_op)
a3b680e6 3444 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3445 else
a3b680e6
AL
3446 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3447 }
4633a7c4
LW
3448 break;
3449
79072805 3450 case SVt_PVGV:
8990e307 3451 if (dtype <= SVt_PVGV) {
d4c19fe8 3452 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3453 return;
79072805 3454 }
5f66b61c 3455 /*FALLTHROUGH*/
79072805 3456
489f7bfe
NC
3457 case SVt_PVMG:
3458 case SVt_PVLV:
3459 case SVt_PVBM:
8d6d96c1 3460 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3461 mg_get(sstr);
eb160463 3462 if ((int)SvTYPE(sstr) != stype) {
973f89ab 3463 stype = SvTYPE(sstr);
b8c701c1 3464 if (stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3465 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3466 return;
3467 }
973f89ab
CS
3468 }
3469 }
ded42b9f 3470 if (stype == SVt_PVLV)
862a34c6 3471 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3472 else
862a34c6 3473 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3474 }
3475
ff920335
NC
3476 /* dstr may have been upgraded. */
3477 dtype = SvTYPE(dstr);
8990e307
LW
3478 sflags = SvFLAGS(sstr);
3479
3480 if (sflags & SVf_ROK) {
acaa9288
NC
3481 if (dtype == SVt_PVGV &&
3482 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
3483 sstr = SvRV(sstr);
3484 if (sstr == dstr) {
3485 if (GvIMPORTED(dstr) != GVf_IMPORTED
3486 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3487 {
3488 GvIMPORTED_on(dstr);
3489 }
3490 GvMULTI_on(dstr);
3491 return;
3492 }
d4c19fe8 3493 glob_assign_glob(dstr, sstr, dtype);
acaa9288
NC
3494 return;
3495 }
3496
8990e307 3497 if (dtype >= SVt_PV) {
b8c701c1 3498 if (dtype == SVt_PVGV) {
d4c19fe8 3499 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3500 return;
3501 }
3f7c398e 3502 if (SvPVX_const(dstr)) {
8bd4d4c5 3503 SvPV_free(dstr);
b162af07
SP
3504 SvLEN_set(dstr, 0);
3505 SvCUR_set(dstr, 0);
a0d0e21e 3506 }
8990e307 3507 }
a0d0e21e 3508 (void)SvOK_off(dstr);
b162af07 3509 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
dfd48732
NC
3510 SvFLAGS(dstr) |= sflags & (SVf_ROK|SVf_AMAGIC);
3511 assert(!(sflags & SVp_NOK));
3512 assert(!(sflags & SVp_IOK));
3513 assert(!(sflags & SVf_NOK));
3514 assert(!(sflags & SVf_IOK));
ed6116ce 3515 }
c0c44674
NC
3516 else if (dtype == SVt_PVGV) {
3517 if (!(sflags & SVf_OK)) {
3518 if (ckWARN(WARN_MISC))
3519 Perl_warner(aTHX_ packWARN(WARN_MISC),
3520 "Undefined value assigned to typeglob");
3521 }
3522 else {
3523 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3524 if (dstr != (SV*)gv) {
3525 if (GvGP(dstr))
3526 gp_free((GV*)dstr);
3527 GvGP(dstr) = gp_ref(GvGP(gv));
3528 }
3529 }
3530 }
8990e307 3531 else if (sflags & SVp_POK) {
765f542d 3532 bool isSwipe = 0;
79072805
LW
3533
3534 /*
3535 * Check to see if we can just swipe the string. If so, it's a
3536 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3537 * It might even be a win on short strings if SvPVX_const(dstr)
3538 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3539 */
3540
120fac95
NC
3541 /* Whichever path we take through the next code, we want this true,
3542 and doing it now facilitates the COW check. */
3543 (void)SvPOK_only(dstr);
3544
765f542d 3545 if (
b8f9541a
NC
3546 /* We're not already COW */
3547 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3548#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3549 /* or we are, but dstr isn't a suitable target. */
3550 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3551#endif
3552 )
765f542d 3553 &&
765f542d
NC
3554 !(isSwipe =
3555 (sflags & SVs_TEMP) && /* slated for free anyway? */
3556 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3557 (!(flags & SV_NOSTEAL)) &&
3558 /* and we're allowed to steal temps */
765f542d
NC
3559 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3560 SvLEN(sstr) && /* and really is a string */
645c22ef 3561 /* and won't be needed again, potentially */
765f542d 3562 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3563#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3564 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3565 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3566 && SvTYPE(sstr) >= SVt_PVIV)
3567#endif
3568 ) {
3569 /* Failed the swipe test, and it's not a shared hash key either.
3570 Have to copy the string. */
3571 STRLEN len = SvCUR(sstr);
3572 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3573 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3574 SvCUR_set(dstr, len);
3575 *SvEND(dstr) = '\0';
765f542d 3576 } else {
f8c7b90f 3577 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3578 be true in here. */
765f542d
NC
3579 /* Either it's a shared hash key, or it's suitable for
3580 copy-on-write or we can swipe the string. */
46187eeb 3581 if (DEBUG_C_TEST) {
ed252734 3582 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3583 sv_dump(sstr);
3584 sv_dump(dstr);
46187eeb 3585 }
f8c7b90f 3586#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3587 if (!isSwipe) {
3588 /* I believe I should acquire a global SV mutex if
3589 it's a COW sv (not a shared hash key) to stop
3590 it going un copy-on-write.
3591 If the source SV has gone un copy on write between up there
3592 and down here, then (assert() that) it is of the correct
3593 form to make it copy on write again */
3594 if ((sflags & (SVf_FAKE | SVf_READONLY))
3595 != (SVf_FAKE | SVf_READONLY)) {
3596 SvREADONLY_on(sstr);
3597 SvFAKE_on(sstr);
3598 /* Make the source SV into a loop of 1.
3599 (about to become 2) */
a29f6d03 3600 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3601 }
3602 }
3603#endif
3604 /* Initial code is common. */
94010e71
NC
3605 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3606 SvPV_free(dstr);
79072805 3607 }
765f542d 3608
765f542d
NC
3609 if (!isSwipe) {
3610 /* making another shared SV. */
3611 STRLEN cur = SvCUR(sstr);
3612 STRLEN len = SvLEN(sstr);
f8c7b90f 3613#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3614 if (len) {
b8f9541a 3615 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3616 /* SvIsCOW_normal */
3617 /* splice us in between source and next-after-source. */
a29f6d03
NC
3618 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3619 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3620 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3621 } else
3622#endif
3623 {
765f542d 3624 /* SvIsCOW_shared_hash */
46187eeb
NC
3625 DEBUG_C(PerlIO_printf(Perl_debug_log,
3626 "Copy on write: Sharing hash\n"));
b8f9541a 3627
bdd68bc3 3628 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3629 SvPV_set(dstr,
d1db91c6 3630 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3631 }
87a1ef3d
SP
3632 SvLEN_set(dstr, len);
3633 SvCUR_set(dstr, cur);
765f542d
NC
3634 SvREADONLY_on(dstr);
3635 SvFAKE_on(dstr);
3636 /* Relesase a global SV mutex. */
3637 }
3638 else
765f542d 3639 { /* Passes the swipe test. */
78d1e721 3640 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3641 SvLEN_set(dstr, SvLEN(sstr));
3642 SvCUR_set(dstr, SvCUR(sstr));
3643
3644 SvTEMP_off(dstr);
3645 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 3646 SvPV_set(sstr, NULL);
765f542d
NC
3647 SvLEN_set(sstr, 0);
3648 SvCUR_set(sstr, 0);
3649 SvTEMP_off(sstr);
3650 }
3651 }
8990e307 3652 if (sflags & SVp_NOK) {
9d6ce603 3653 SvNV_set(dstr, SvNVX(sstr));
79072805 3654 }
8990e307 3655 if (sflags & SVp_IOK) {
23525414
NC
3656 SvRELEASE_IVX(dstr);
3657 SvIV_set(dstr, SvIVX(sstr));
3658 /* Must do this otherwise some other overloaded use of 0x80000000
3659 gets confused. I guess SVpbm_VALID */
2b1c7e3e 3660 if (sflags & SVf_IVisUV)
25da4f38 3661 SvIsUV_on(dstr);
79072805 3662 }
dd2eae66
NC
3663 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8
3664 |SVf_AMAGIC);
4f2da183
NC
3665 {
3666 const MAGIC * const smg = SvVOK(sstr);
3667 if (smg) {
3668 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3669 smg->mg_ptr, smg->mg_len);
3670 SvRMAGICAL_on(dstr);
3671 }
7a5fa8a2 3672 }
79072805 3673 }
5d581361 3674 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 3675 (void)SvOK_off(dstr);
dd2eae66
NC
3676 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK
3677 |SVf_AMAGIC);
5d581361
NC
3678 if (sflags & SVp_IOK) {
3679 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
3680 SvIV_set(dstr, SvIVX(sstr));
3681 }
3332b3c1 3682 if (sflags & SVp_NOK) {
9d6ce603 3683 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3684 }
3685 }
79072805 3686 else {
f7877b28 3687 if (isGV_with_GP(sstr)) {
180488f8
NC
3688 /* This stringification rule for globs is spread in 3 places.
3689 This feels bad. FIXME. */
3690 const U32 wasfake = sflags & SVf_FAKE;
3691
3692 /* FAKE globs can get coerced, so need to turn this off
3693 temporarily if it is on. */
3694 SvFAKE_off(sstr);
3695 gv_efullname3(dstr, (GV *)sstr, "*");
3696 SvFLAGS(sstr) |= wasfake;
dd2eae66 3697 SvFLAGS(dstr) |= sflags & SVf_AMAGIC;
180488f8 3698 }
20408e3c
GS
3699 else
3700 (void)SvOK_off(dstr);
a0d0e21e 3701 }
27c9684d
AP
3702 if (SvTAINTED(sstr))
3703 SvTAINT(dstr);
79072805
LW
3704}
3705
954c1994
GS
3706/*
3707=for apidoc sv_setsv_mg
3708
3709Like C<sv_setsv>, but also handles 'set' magic.
3710
3711=cut
3712*/
3713
79072805 3714void
864dbfa3 3715Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3716{
3717 sv_setsv(dstr,sstr);
3718 SvSETMAGIC(dstr);
3719}
3720
f8c7b90f 3721#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3722SV *
3723Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3724{
3725 STRLEN cur = SvCUR(sstr);
3726 STRLEN len = SvLEN(sstr);
3727 register char *new_pv;
3728
3729 if (DEBUG_C_TEST) {
3730 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3731 sstr, dstr);
3732 sv_dump(sstr);
3733 if (dstr)
3734 sv_dump(dstr);
3735 }
3736
3737 if (dstr) {
3738 if (SvTHINKFIRST(dstr))
3739 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3740 else if (SvPVX_const(dstr))
3741 Safefree(SvPVX_const(dstr));
ed252734
NC
3742 }
3743 else
3744 new_SV(dstr);
862a34c6 3745 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3746
3747 assert (SvPOK(sstr));
3748 assert (SvPOKp(sstr));
3749 assert (!SvIOK(sstr));
3750 assert (!SvIOKp(sstr));
3751 assert (!SvNOK(sstr));
3752 assert (!SvNOKp(sstr));
3753
3754 if (SvIsCOW(sstr)) {
3755
3756 if (SvLEN(sstr) == 0) {
3757 /* source is a COW shared hash key. */
ed252734
NC
3758 DEBUG_C(PerlIO_printf(Perl_debug_log,
3759 "Fast copy on write: Sharing hash\n"));
d1db91c6 3760 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3761 goto common_exit;
3762 }
3763 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3764 } else {
3765 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3766 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3767 SvREADONLY_on(sstr);
3768 SvFAKE_on(sstr);
3769 DEBUG_C(PerlIO_printf(Perl_debug_log,
3770 "Fast copy on write: Converting sstr to COW\n"));
3771 SV_COW_NEXT_SV_SET(dstr, sstr);
3772 }
3773 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3774 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3775
3776 common_exit:
3777 SvPV_set(dstr, new_pv);
3778 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3779 if (SvUTF8(sstr))
3780 SvUTF8_on(dstr);
87a1ef3d
SP
3781 SvLEN_set(dstr, len);
3782 SvCUR_set(dstr, cur);
ed252734
NC
3783 if (DEBUG_C_TEST) {
3784 sv_dump(dstr);
3785 }
3786 return dstr;
3787}
3788#endif
3789
954c1994
GS
3790/*
3791=for apidoc sv_setpvn
3792
3793Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3794bytes to be copied. If the C<ptr> argument is NULL the SV will become
3795undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3796
3797=cut
3798*/
3799
ef50df4b 3800void
864dbfa3 3801Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3802{
97aff369 3803 dVAR;
c6f8c383 3804 register char *dptr;
22c522df 3805
765f542d 3806 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3807 if (!ptr) {
a0d0e21e 3808 (void)SvOK_off(sv);
463ee0b2
LW
3809 return;
3810 }
22c522df
JH
3811 else {
3812 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3813 const IV iv = len;
9c5ffd7c
JH
3814 if (iv < 0)
3815 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3816 }
862a34c6 3817 SvUPGRADE(sv, SVt_PV);
c6f8c383 3818
5902b6a9 3819 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3820 Move(ptr,dptr,len,char);
3821 dptr[len] = '\0';
79072805 3822 SvCUR_set(sv, len);
1aa99e6b 3823 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3824 SvTAINT(sv);
79072805
LW
3825}
3826
954c1994
GS
3827/*
3828=for apidoc sv_setpvn_mg
3829
3830Like C<sv_setpvn>, but also handles 'set' magic.
3831
3832=cut
3833*/
3834
79072805 3835void
864dbfa3 3836Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3837{
3838 sv_setpvn(sv,ptr,len);
3839 SvSETMAGIC(sv);
3840}
3841
954c1994
GS
3842/*
3843=for apidoc sv_setpv
3844
3845Copies a string into an SV. The string must be null-terminated. Does not
3846handle 'set' magic. See C<sv_setpv_mg>.
3847
3848=cut
3849*/
3850
ef50df4b 3851void
864dbfa3 3852Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805 3853{
97aff369 3854 dVAR;
79072805
LW
3855 register STRLEN len;
3856
765f542d 3857 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3858 if (!ptr) {
a0d0e21e 3859 (void)SvOK_off(sv);
463ee0b2
LW
3860 return;
3861 }
79072805 3862 len = strlen(ptr);
862a34c6 3863 SvUPGRADE(sv, SVt_PV);
c6f8c383 3864
79072805 3865 SvGROW(sv, len + 1);
463ee0b2 3866 Move(ptr,SvPVX(sv),len+1,char);
79072805 3867 SvCUR_set(sv, len);
1aa99e6b 3868 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3869 SvTAINT(sv);
3870}
3871
954c1994
GS
3872/*
3873=for apidoc sv_setpv_mg
3874
3875Like C<sv_setpv>, but also handles 'set' magic.
3876
3877=cut
3878*/
3879
463ee0b2 3880void
864dbfa3 3881Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3882{
3883 sv_setpv(sv,ptr);
3884 SvSETMAGIC(sv);
3885}
3886
954c1994
GS
3887/*
3888=for apidoc sv_usepvn
3889
794a0d33
JH
3890Tells an SV to use C<ptr> to find its string value. Normally the
3891string is stored inside the SV but sv_usepvn allows the SV to use an
3892outside string. The C<ptr> should point to memory that was allocated
3893by C<malloc>. The string length, C<len>, must be supplied. This
3894function will realloc (i.e. move) the memory pointed to by C<ptr>,
3895so that pointer should not be freed or used by the programmer after
3896giving it to sv_usepvn, and neither should any pointers from "behind"
3897that pointer (e.g. ptr + 1) be used. Does not handle 'set' magic.
954c1994
GS
3898See C<sv_usepvn_mg>.
3899
3900=cut
3901*/
3902
ef50df4b 3903void
864dbfa3 3904Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
463ee0b2 3905{
97aff369 3906 dVAR;
1936d2a7 3907 STRLEN allocate;
765f542d 3908 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3909 SvUPGRADE(sv, SVt_PV);
463ee0b2 3910 if (!ptr) {
a0d0e21e 3911 (void)SvOK_off(sv);
463ee0b2
LW
3912 return;
3913 }
3f7c398e 3914 if (SvPVX_const(sv))
8bd4d4c5 3915 SvPV_free(sv);
1936d2a7
NC
3916
3917 allocate = PERL_STRLEN_ROUNDUP(len + 1);
7a9b70e9 3918 ptr = saferealloc (ptr, allocate);
f880fe2f 3919 SvPV_set(sv, ptr);
463ee0b2 3920 SvCUR_set(sv, len);
1936d2a7 3921 SvLEN_set(sv, allocate);
463ee0b2 3922 *SvEND(sv) = '\0';
1aa99e6b 3923 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3924 SvTAINT(sv);
79072805
LW
3925}
3926
954c1994
GS
3927/*
3928=for apidoc sv_usepvn_mg
3929
3930Like C<sv_usepvn>, but also handles 'set' magic.
3931
3932=cut
3933*/
3934
ef50df4b 3935void
864dbfa3 3936Perl_sv_usepvn_mg(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
ef50df4b 3937{
51c1089b 3938 sv_usepvn(sv,ptr,len);
ef50df4b
GS
3939 SvSETMAGIC(sv);
3940}
3941
f8c7b90f 3942#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3943/* Need to do this *after* making the SV normal, as we need the buffer
3944 pointer to remain valid until after we've copied it. If we let go too early,
3945 another thread could invalidate it by unsharing last of the same hash key
3946 (which it can do by means other than releasing copy-on-write Svs)
3947 or by changing the other copy-on-write SVs in the loop. */
3948STATIC void
bdd68bc3 3949S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3950{
3951 if (len) { /* this SV was SvIsCOW_normal(sv) */
3952 /* we need to find the SV pointing to us. */
cf5629ad 3953 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3954
765f542d
NC
3955 if (current == sv) {
3956 /* The SV we point to points back to us (there were only two of us
3957 in the loop.)
3958 Hence other SV is no longer copy on write either. */
3959 SvFAKE_off(after);
3960 SvREADONLY_off(after);
3961 } else {
3962 /* We need to follow the pointers around the loop. */
3963 SV *next;
3964 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3965 assert (next);
3966 current = next;
3967 /* don't loop forever if the structure is bust, and we have
3968 a pointer into a closed loop. */
3969 assert (current != after);
3f7c398e 3970 assert (SvPVX_const(current) == pvx);
765f542d
NC
3971 }
3972 /* Make the SV before us point to the SV after us. */
a29f6d03 3973 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3974 }
3975 } else {
bdd68bc3 3976 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3977 }
3978}
3979
3980int
3981Perl_sv_release_IVX(pTHX_ register SV *sv)
3982{
3983 if (SvIsCOW(sv))
3984 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
3985 SvOOK_off(sv);
3986 return 0;
765f542d
NC
3987}
3988#endif
645c22ef
DM
3989/*
3990=for apidoc sv_force_normal_flags
3991
3992Undo various types of fakery on an SV: if the PV is a shared string, make
3993a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
3994an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
3995we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
3996then a copy-on-write scalar drops its PV buffer (if any) and becomes
3997SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 3998set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
3999C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4000with flags set to 0.
645c22ef
DM
4001
4002=cut
4003*/
4004
6fc92669 4005void
840a7b70 4006Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4007{
97aff369 4008 dVAR;
f8c7b90f 4009#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4010 if (SvREADONLY(sv)) {
4011 /* At this point I believe I should acquire a global SV mutex. */
4012 if (SvFAKE(sv)) {
b64e5050 4013 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4014 const STRLEN len = SvLEN(sv);
4015 const STRLEN cur = SvCUR(sv);
a28509cc 4016 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4017 if (DEBUG_C_TEST) {
4018 PerlIO_printf(Perl_debug_log,
4019 "Copy on write: Force normal %ld\n",
4020 (long) flags);
e419cbc5 4021 sv_dump(sv);
46187eeb 4022 }
765f542d
NC
4023 SvFAKE_off(sv);
4024 SvREADONLY_off(sv);
9f653bb5 4025 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4026 SvPV_set(sv, NULL);
87a1ef3d 4027 SvLEN_set(sv, 0);
765f542d
NC
4028 if (flags & SV_COW_DROP_PV) {
4029 /* OK, so we don't need to copy our buffer. */
4030 SvPOK_off(sv);
4031 } else {
4032 SvGROW(sv, cur + 1);
4033 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4034 SvCUR_set(sv, cur);
765f542d
NC
4035 *SvEND(sv) = '\0';
4036 }
bdd68bc3 4037 sv_release_COW(sv, pvx, len, next);
46187eeb 4038 if (DEBUG_C_TEST) {
e419cbc5 4039 sv_dump(sv);
46187eeb 4040 }
765f542d 4041 }
923e4eb5 4042 else if (IN_PERL_RUNTIME)
765f542d
NC
4043 Perl_croak(aTHX_ PL_no_modify);
4044 /* At this point I believe that I can drop the global SV mutex. */
4045 }
4046#else
2213622d 4047 if (SvREADONLY(sv)) {
1c846c1f 4048 if (SvFAKE(sv)) {
b64e5050 4049 const char * const pvx = SvPVX_const(sv);
66a1b24b 4050 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4051 SvFAKE_off(sv);
4052 SvREADONLY_off(sv);
bd61b366 4053 SvPV_set(sv, NULL);
66a1b24b 4054 SvLEN_set(sv, 0);
1c846c1f 4055 SvGROW(sv, len + 1);
706aa1c9 4056 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4057 *SvEND(sv) = '\0';
bdd68bc3 4058 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4059 }
923e4eb5 4060 else if (IN_PERL_RUNTIME)
cea2e8a9 4061 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4062 }
765f542d 4063#endif
2213622d 4064 if (SvROK(sv))
840a7b70 4065 sv_unref_flags(sv, flags);
6fc92669
GS
4066 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4067 sv_unglob(sv);
0f15f207 4068}
1c846c1f 4069
645c22ef 4070/*
954c1994
GS
4071=for apidoc sv_chop
4072
1c846c1f 4073Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4074SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4075the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4076string. Uses the "OOK hack".
3f7c398e 4077Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4078refer to the same chunk of data.
954c1994
GS
4079
4080=cut
4081*/
4082
79072805 4083void
f54cb97a 4084Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4085{
4086 register STRLEN delta;
a0d0e21e 4087 if (!ptr || !SvPOKp(sv))
79072805 4088 return;
3f7c398e 4089 delta = ptr - SvPVX_const(sv);
2213622d 4090 SV_CHECK_THINKFIRST(sv);
79072805
LW
4091 if (SvTYPE(sv) < SVt_PVIV)
4092 sv_upgrade(sv,SVt_PVIV);
4093
4094 if (!SvOOK(sv)) {
50483b2c 4095 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4096 const char *pvx = SvPVX_const(sv);
a28509cc 4097 const STRLEN len = SvCUR(sv);
50483b2c 4098 SvGROW(sv, len + 1);
706aa1c9 4099 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4100 *SvEND(sv) = '\0';
4101 }
45977657 4102 SvIV_set(sv, 0);
a4bfb290
AB
4103 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4104 and we do that anyway inside the SvNIOK_off
4105 */
7a5fa8a2 4106 SvFLAGS(sv) |= SVf_OOK;
79072805 4107 }
a4bfb290 4108 SvNIOK_off(sv);
b162af07
SP
4109 SvLEN_set(sv, SvLEN(sv) - delta);
4110 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4111 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4112 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4113}
4114
954c1994
GS
4115/*
4116=for apidoc sv_catpvn
4117
4118Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4119C<len> indicates number of bytes to copy. If the SV has the UTF-8
4120status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4121Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4122
8d6d96c1
HS
4123=for apidoc sv_catpvn_flags
4124
4125Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4126C<len> indicates number of bytes to copy. If the SV has the UTF-8
4127status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4128If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4129appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4130in terms of this function.
4131
4132=cut
4133*/
4134
4135void
4136Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4137{
97aff369 4138 dVAR;
8d6d96c1 4139 STRLEN dlen;
fabdb6c0 4140 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4141
8d6d96c1
HS
4142 SvGROW(dsv, dlen + slen + 1);
4143 if (sstr == dstr)
3f7c398e 4144 sstr = SvPVX_const(dsv);
8d6d96c1 4145 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4146 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4147 *SvEND(dsv) = '\0';
4148 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4149 SvTAINT(dsv);
bddd5118
NC
4150 if (flags & SV_SMAGIC)
4151 SvSETMAGIC(dsv);
79072805
LW
4152}
4153
954c1994 4154/*
954c1994
GS
4155=for apidoc sv_catsv
4156
13e8c8e3
JH
4157Concatenates the string from SV C<ssv> onto the end of the string in
4158SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4159not 'set' magic. See C<sv_catsv_mg>.
954c1994 4160
8d6d96c1
HS
4161=for apidoc sv_catsv_flags
4162
4163Concatenates the string from SV C<ssv> onto the end of the string in
4164SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4165bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4166and C<sv_catsv_nomg> are implemented in terms of this function.
4167
4168=cut */
4169
ef50df4b 4170void
8d6d96c1 4171Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4172{
97aff369 4173 dVAR;
bddd5118 4174 if (ssv) {
00b6aa41
AL
4175 STRLEN slen;
4176 const char *spv = SvPV_const(ssv, slen);
4177 if (spv) {
bddd5118
NC
4178 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4179 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4180 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4181 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4182 dsv->sv_flags doesn't have that bit set.
4fd84b44 4183 Andy Dougherty 12 Oct 2001
bddd5118
NC
4184 */
4185 const I32 sutf8 = DO_UTF8(ssv);
4186 I32 dutf8;
13e8c8e3 4187
bddd5118
NC
4188 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4189 mg_get(dsv);
4190 dutf8 = DO_UTF8(dsv);
8d6d96c1 4191
bddd5118
NC
4192 if (dutf8 != sutf8) {
4193 if (dutf8) {
4194 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4195 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4196
bddd5118
NC
4197 sv_utf8_upgrade(csv);
4198 spv = SvPV_const(csv, slen);
4199 }
4200 else
4201 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4202 }
bddd5118 4203 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4204 }
560a288e 4205 }
bddd5118
NC
4206 if (flags & SV_SMAGIC)
4207 SvSETMAGIC(dsv);
79072805
LW
4208}
4209
954c1994 4210/*
954c1994
GS
4211=for apidoc sv_catpv
4212
4213Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4214If the SV has the UTF-8 status set, then the bytes appended should be
4215valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4216
d5ce4a7c 4217=cut */
954c1994 4218
ef50df4b 4219void
0c981600 4220Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4221{
97aff369 4222 dVAR;
79072805 4223 register STRLEN len;
463ee0b2 4224 STRLEN tlen;
748a9306 4225 char *junk;
79072805 4226
0c981600 4227 if (!ptr)
79072805 4228 return;
748a9306 4229 junk = SvPV_force(sv, tlen);
0c981600 4230 len = strlen(ptr);
463ee0b2 4231 SvGROW(sv, tlen + len + 1);
0c981600 4232 if (ptr == junk)
3f7c398e 4233 ptr = SvPVX_const(sv);
0c981600 4234 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4235 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4236 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4237 SvTAINT(sv);
79072805
LW
4238}
4239
954c1994
GS
4240/*
4241=for apidoc sv_catpv_mg
4242
4243Like C<sv_catpv>, but also handles 'set' magic.
4244
4245=cut
4246*/
4247
ef50df4b 4248void
0c981600 4249Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4250{
0c981600 4251 sv_catpv(sv,ptr);
ef50df4b
GS
4252 SvSETMAGIC(sv);
4253}
4254
645c22ef
DM
4255/*
4256=for apidoc newSV
4257
561b68a9
SH
4258Creates a new SV. A non-zero C<len> parameter indicates the number of
4259bytes of preallocated string space the SV should have. An extra byte for a
4260trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4261space is allocated.) The reference count for the new SV is set to 1.
4262
4263In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4264parameter, I<x>, a debug aid which allowed callers to identify themselves.
4265This aid has been superseded by a new build option, PERL_MEM_LOG (see
4266L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4267modules supporting older perls.
645c22ef
DM
4268
4269=cut
4270*/
4271
79072805 4272SV *
864dbfa3 4273Perl_newSV(pTHX_ STRLEN len)
79072805 4274{
97aff369 4275 dVAR;
79072805 4276 register SV *sv;
1c846c1f 4277
4561caa4 4278 new_SV(sv);
79072805
LW
4279 if (len) {
4280 sv_upgrade(sv, SVt_PV);
4281 SvGROW(sv, len + 1);
4282 }
4283 return sv;
4284}
954c1994 4285/*
92110913 4286=for apidoc sv_magicext
954c1994 4287
68795e93 4288Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4289supplied vtable and returns a pointer to the magic added.
92110913 4290
2d8d5d5a
SH
4291Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4292In particular, you can add magic to SvREADONLY SVs, and add more than
4293one instance of the same 'how'.
645c22ef 4294
2d8d5d5a
SH
4295If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4296stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4297special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4298to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4299
2d8d5d5a 4300(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4301
4302=cut
4303*/
92110913 4304MAGIC *
92e67595 4305Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4306 const char* name, I32 namlen)
79072805 4307{
97aff369 4308 dVAR;
79072805 4309 MAGIC* mg;
68795e93 4310
92110913 4311 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4312 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4313 }
a02a5408 4314 Newxz(mg, 1, MAGIC);
79072805 4315 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4316 SvMAGIC_set(sv, mg);
75f9d97a 4317
05f95b08
SB
4318 /* Sometimes a magic contains a reference loop, where the sv and
4319 object refer to each other. To prevent a reference loop that
4320 would prevent such objects being freed, we look for such loops
4321 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4322
4323 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4324 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4325
4326 */
14befaf4
DM
4327 if (!obj || obj == sv ||
4328 how == PERL_MAGIC_arylen ||
4329 how == PERL_MAGIC_qr ||
8d2f4536 4330 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4331 (SvTYPE(obj) == SVt_PVGV &&
4332 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4333 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4334 GvFORM(obj) == (CV*)sv)))
75f9d97a 4335 {
8990e307 4336 mg->mg_obj = obj;
75f9d97a 4337 }
85e6fe83 4338 else {
b37c2d43 4339 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4340 mg->mg_flags |= MGf_REFCOUNTED;
4341 }
b5ccf5f2
YST
4342
4343 /* Normal self-ties simply pass a null object, and instead of
4344 using mg_obj directly, use the SvTIED_obj macro to produce a
4345 new RV as needed. For glob "self-ties", we are tieing the PVIO
4346 with an RV obj pointing to the glob containing the PVIO. In
4347 this case, to avoid a reference loop, we need to weaken the
4348 reference.
4349 */
4350
4351 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4352 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4353 {
4354 sv_rvweaken(obj);
4355 }
4356
79072805 4357 mg->mg_type = how;
565764a8 4358 mg->mg_len = namlen;
9cbac4c7 4359 if (name) {
92110913 4360 if (namlen > 0)
1edc1566 4361 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4362 else if (namlen == HEf_SVKEY)
b37c2d43 4363 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4364 else
92110913 4365 mg->mg_ptr = (char *) name;
9cbac4c7 4366 }
92110913 4367 mg->mg_virtual = vtable;
68795e93 4368
92110913
NIS
4369 mg_magical(sv);
4370 if (SvGMAGICAL(sv))
4371 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4372 return mg;
4373}
4374
4375/*
4376=for apidoc sv_magic
1c846c1f 4377
92110913
NIS
4378Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4379then adds a new magic item of type C<how> to the head of the magic list.
4380
2d8d5d5a
SH
4381See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4382handling of the C<name> and C<namlen> arguments.
4383
4509d3fb
SB
4384You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4385to add more than one instance of the same 'how'.
4386
92110913
NIS
4387=cut
4388*/
4389
4390void
4391Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4392{
97aff369 4393 dVAR;
92e67595 4394 MGVTBL *vtable;
92110913 4395 MAGIC* mg;
92110913 4396
f8c7b90f 4397#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4398 if (SvIsCOW(sv))
4399 sv_force_normal_flags(sv, 0);
4400#endif
92110913 4401 if (SvREADONLY(sv)) {
d8084ca5
DM
4402 if (
4403 /* its okay to attach magic to shared strings; the subsequent
4404 * upgrade to PVMG will unshare the string */
4405 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4406
4407 && IN_PERL_RUNTIME
92110913
NIS
4408 && how != PERL_MAGIC_regex_global
4409 && how != PERL_MAGIC_bm
4410 && how != PERL_MAGIC_fm
4411 && how != PERL_MAGIC_sv
e6469971 4412 && how != PERL_MAGIC_backref
92110913
NIS
4413 )
4414 {
4415 Perl_croak(aTHX_ PL_no_modify);
4416 }
4417 }
4418 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4419 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4420 /* sv_magic() refuses to add a magic of the same 'how' as an
4421 existing one
92110913 4422 */
2a509ed3 4423 if (how == PERL_MAGIC_taint) {
92110913 4424 mg->mg_len |= 1;
2a509ed3
NC
4425 /* Any scalar which already had taint magic on which someone
4426 (erroneously?) did SvIOK_on() or similar will now be
4427 incorrectly sporting public "OK" flags. */
4428 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4429 }
92110913
NIS
4430 return;
4431 }
4432 }
68795e93 4433
79072805 4434 switch (how) {
14befaf4 4435 case PERL_MAGIC_sv:
92110913 4436 vtable = &PL_vtbl_sv;
79072805 4437 break;
14befaf4 4438 case PERL_MAGIC_overload:
92110913 4439 vtable = &PL_vtbl_amagic;
a0d0e21e 4440 break;
14befaf4 4441 case PERL_MAGIC_overload_elem:
92110913 4442 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4443 break;
14befaf4 4444 case PERL_MAGIC_overload_table:
92110913 4445 vtable = &PL_vtbl_ovrld;
a0d0e21e 4446 break;
14befaf4 4447 case PERL_MAGIC_bm:
92110913 4448 vtable = &PL_vtbl_bm;
79072805 4449 break;
14befaf4 4450 case PERL_MAGIC_regdata:
92110913 4451 vtable = &PL_vtbl_regdata;
6cef1e77 4452 break;
14befaf4 4453 case PERL_MAGIC_regdatum:
92110913 4454 vtable = &PL_vtbl_regdatum;
6cef1e77 4455 break;
14befaf4 4456 case PERL_MAGIC_env:
92110913 4457 vtable = &PL_vtbl_env;
79072805 4458 break;
14befaf4 4459 case PERL_MAGIC_fm:
92110913 4460 vtable = &PL_vtbl_fm;
55497cff 4461 break;
14befaf4 4462 case PERL_MAGIC_envelem:
92110913 4463 vtable = &PL_vtbl_envelem;
79072805 4464 break;
14befaf4 4465 case PERL_MAGIC_regex_global:
92110913 4466 vtable = &PL_vtbl_mglob;
93a17b20 4467 break;
14befaf4 4468 case PERL_MAGIC_isa:
92110913 4469 vtable = &PL_vtbl_isa;
463ee0b2 4470 break;
14befaf4 4471 case PERL_MAGIC_isaelem:
92110913 4472 vtable = &PL_vtbl_isaelem;
463ee0b2 4473 break;
14befaf4 4474 case PERL_MAGIC_nkeys:
92110913 4475 vtable = &PL_vtbl_nkeys;
16660edb 4476 break;
14befaf4 4477 case PERL_MAGIC_dbfile:
aec46f14 4478 vtable = NULL;
93a17b20 4479 break;
14befaf4 4480 case PERL_MAGIC_dbline:
92110913 4481 vtable = &PL_vtbl_dbline;
79072805 4482 break;
36477c24 4483#ifdef USE_LOCALE_COLLATE
14befaf4 4484 case PERL_MAGIC_collxfrm:
92110913 4485 vtable = &PL_vtbl_collxfrm;
bbce6d69 4486 break;
36477c24 4487#endif /* USE_LOCALE_COLLATE */
14befaf4 4488 case PERL_MAGIC_tied:
92110913 4489 vtable = &PL_vtbl_pack;
463ee0b2 4490 break;
14befaf4
DM
4491 case PERL_MAGIC_tiedelem:
4492 case PERL_MAGIC_tiedscalar:
92110913 4493 vtable = &PL_vtbl_packelem;
463ee0b2 4494 break;
14befaf4 4495 case PERL_MAGIC_qr:
92110913 4496 vtable = &PL_vtbl_regexp;
c277df42 4497 break;
b3ca2e83
NC
4498 case PERL_MAGIC_hints:
4499 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4500 case PERL_MAGIC_sig:
92110913 4501 vtable = &PL_vtbl_sig;
79072805 4502 break;
14befaf4 4503 case PERL_MAGIC_sigelem:
92110913 4504 vtable = &PL_vtbl_sigelem;
79072805 4505 break;
14befaf4 4506 case PERL_MAGIC_taint:
92110913 4507 vtable = &PL_vtbl_taint;
463ee0b2 4508 break;
14befaf4 4509 case PERL_MAGIC_uvar:
92110913 4510 vtable = &PL_vtbl_uvar;
79072805 4511 break;
14befaf4 4512 case PERL_MAGIC_vec:
92110913 4513 vtable = &PL_vtbl_vec;
79072805 4514 break;
a3874608 4515 case PERL_MAGIC_arylen_p:
bfcb3514 4516 case PERL_MAGIC_rhash:
8d2f4536 4517 case PERL_MAGIC_symtab:
ece467f9 4518 case PERL_MAGIC_vstring:
aec46f14 4519 vtable = NULL;
ece467f9 4520 break;
7e8c5dac
HS
4521 case PERL_MAGIC_utf8:
4522 vtable = &PL_vtbl_utf8;
4523 break;
14befaf4 4524 case PERL_MAGIC_substr:
92110913 4525 vtable = &PL_vtbl_substr;
79072805 4526 break;
14befaf4 4527 case PERL_MAGIC_defelem:
92110913 4528 vtable = &PL_vtbl_defelem;
5f05dabc 4529 break;
14befaf4 4530 case PERL_MAGIC_arylen:
92110913 4531 vtable = &PL_vtbl_arylen;
79072805 4532 break;
14befaf4 4533 case PERL_MAGIC_pos:
92110913 4534 vtable = &PL_vtbl_pos;
a0d0e21e 4535 break;
14befaf4 4536 case PERL_MAGIC_backref:
92110913 4537 vtable = &PL_vtbl_backref;
810b8aa5 4538 break;
b3ca2e83
NC
4539 case PERL_MAGIC_hintselem:
4540 vtable = &PL_vtbl_hintselem;
4541 break;
14befaf4
DM
4542 case PERL_MAGIC_ext:
4543 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4544 /* Useful for attaching extension internal data to perl vars. */
4545 /* Note that multiple extensions may clash if magical scalars */
4546 /* etc holding private data from one are passed to another. */
aec46f14 4547 vtable = NULL;
a0d0e21e 4548 break;
79072805 4549 default:
14befaf4 4550 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4551 }
68795e93 4552
92110913 4553 /* Rest of work is done else where */
aec46f14 4554 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4555
92110913
NIS
4556 switch (how) {
4557 case PERL_MAGIC_taint:
4558 mg->mg_len = 1;
4559 break;
4560 case PERL_MAGIC_ext:
4561 case PERL_MAGIC_dbfile:
4562 SvRMAGICAL_on(sv);
4563 break;
4564 }
463ee0b2
LW
4565}
4566
c461cf8f
JH
4567/*
4568=for apidoc sv_unmagic
4569
645c22ef 4570Removes all magic of type C<type> from an SV.
c461cf8f
JH
4571
4572=cut
4573*/
4574
463ee0b2 4575int
864dbfa3 4576Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4577{
4578 MAGIC* mg;
4579 MAGIC** mgp;
91bba347 4580 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4581 return 0;
064cf529 4582 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4583 for (mg = *mgp; mg; mg = *mgp) {
4584 if (mg->mg_type == type) {
e1ec3a88 4585 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4586 *mgp = mg->mg_moremagic;
1d7c1841 4587 if (vtbl && vtbl->svt_free)
fc0dc3b3 4588 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4589 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4590 if (mg->mg_len > 0)
1edc1566 4591 Safefree(mg->mg_ptr);
565764a8 4592 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4593 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4594 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4595 Safefree(mg->mg_ptr);
9cbac4c7 4596 }
a0d0e21e
LW
4597 if (mg->mg_flags & MGf_REFCOUNTED)
4598 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4599 Safefree(mg);
4600 }
4601 else
4602 mgp = &mg->mg_moremagic;
79072805 4603 }
91bba347 4604 if (!SvMAGIC(sv)) {
463ee0b2 4605 SvMAGICAL_off(sv);
c268c2a6 4606 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4607 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4608 }
4609
4610 return 0;
79072805
LW
4611}
4612
c461cf8f
JH
4613/*
4614=for apidoc sv_rvweaken
4615
645c22ef
DM
4616Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4617referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4618push a back-reference to this RV onto the array of backreferences
4619associated with that magic.
c461cf8f
JH
4620
4621=cut
4622*/
4623
810b8aa5 4624SV *
864dbfa3 4625Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4626{
4627 SV *tsv;
4628 if (!SvOK(sv)) /* let undefs pass */
4629 return sv;
4630 if (!SvROK(sv))
cea2e8a9 4631 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4632 else if (SvWEAKREF(sv)) {
810b8aa5 4633 if (ckWARN(WARN_MISC))
9014280d 4634 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4635 return sv;
4636 }
4637 tsv = SvRV(sv);
e15faf7d 4638 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4639 SvWEAKREF_on(sv);
1c846c1f 4640 SvREFCNT_dec(tsv);
810b8aa5
GS
4641 return sv;
4642}
4643
645c22ef
DM
4644/* Give tsv backref magic if it hasn't already got it, then push a
4645 * back-reference to sv onto the array associated with the backref magic.
4646 */
4647
e15faf7d
NC
4648void
4649Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4650{
97aff369 4651 dVAR;
810b8aa5 4652 AV *av;
86f55936
NC
4653
4654 if (SvTYPE(tsv) == SVt_PVHV) {
4655 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4656
4657 av = *avp;
4658 if (!av) {
4659 /* There is no AV in the offical place - try a fixup. */
4660 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4661
4662 if (mg) {
4663 /* Aha. They've got it stowed in magic. Bring it back. */
4664 av = (AV*)mg->mg_obj;
4665 /* Stop mg_free decreasing the refernce count. */
4666 mg->mg_obj = NULL;
4667 /* Stop mg_free even calling the destructor, given that
4668 there's no AV to free up. */
4669 mg->mg_virtual = 0;
4670 sv_unmagic(tsv, PERL_MAGIC_backref);
4671 } else {
4672 av = newAV();
4673 AvREAL_off(av);
b37c2d43 4674 SvREFCNT_inc_simple_void(av);
86f55936
NC
4675 }
4676 *avp = av;
4677 }
4678 } else {
4679 const MAGIC *const mg
4680 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4681 if (mg)
4682 av = (AV*)mg->mg_obj;
4683 else {
4684 av = newAV();
4685 AvREAL_off(av);
4686 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4687 /* av now has a refcnt of 2, which avoids it getting freed
4688 * before us during global cleanup. The extra ref is removed
4689 * by magic_killbackrefs() when tsv is being freed */
4690 }
810b8aa5 4691 }
d91d49e8 4692 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4693 av_extend(av, AvFILLp(av)+1);
4694 }
4695 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4696}
4697
645c22ef
DM
4698/* delete a back-reference to ourselves from the backref magic associated
4699 * with the SV we point to.
4700 */
4701
1c846c1f 4702STATIC void
e15faf7d 4703S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4704{
97aff369 4705 dVAR;
86f55936 4706 AV *av = NULL;
810b8aa5
GS
4707 SV **svp;
4708 I32 i;
86f55936
NC
4709
4710 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4711 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4712 /* We mustn't attempt to "fix up" the hash here by moving the
4713 backreference array back to the hv_aux structure, as that is stored
4714 in the main HvARRAY(), and hfreentries assumes that no-one
4715 reallocates HvARRAY() while it is running. */
86f55936
NC
4716 }
4717 if (!av) {
4718 const MAGIC *const mg
4719 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4720 if (mg)
4721 av = (AV *)mg->mg_obj;
4722 }
4723 if (!av) {
e15faf7d
NC
4724 if (PL_in_clean_all)
4725 return;
cea2e8a9 4726 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4727 }
4728
4729 if (SvIS_FREED(av))
4730 return;
4731
810b8aa5 4732 svp = AvARRAY(av);
6a76db8b
NC
4733 /* We shouldn't be in here more than once, but for paranoia reasons lets
4734 not assume this. */
4735 for (i = AvFILLp(av); i >= 0; i--) {
4736 if (svp[i] == sv) {
4737 const SSize_t fill = AvFILLp(av);
4738 if (i != fill) {
4739 /* We weren't the last entry.
4740 An unordered list has this property that you can take the
4741 last element off the end to fill the hole, and it's still
4742 an unordered list :-)
4743 */
4744 svp[i] = svp[fill];
4745 }
a0714e2c 4746 svp[fill] = NULL;
6a76db8b
NC
4747 AvFILLp(av) = fill - 1;
4748 }
4749 }
810b8aa5
GS
4750}
4751
86f55936
NC
4752int
4753Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4754{
4755 SV **svp = AvARRAY(av);
4756
4757 PERL_UNUSED_ARG(sv);
4758
4759 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4760 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4761 if (svp && !SvIS_FREED(av)) {
4762 SV *const *const last = svp + AvFILLp(av);
4763
4764 while (svp <= last) {
4765 if (*svp) {
4766 SV *const referrer = *svp;
4767 if (SvWEAKREF(referrer)) {
4768 /* XXX Should we check that it hasn't changed? */
4769 SvRV_set(referrer, 0);
4770 SvOK_off(referrer);
4771 SvWEAKREF_off(referrer);
4772 } else if (SvTYPE(referrer) == SVt_PVGV ||
4773 SvTYPE(referrer) == SVt_PVLV) {
4774 /* You lookin' at me? */
4775 assert(GvSTASH(referrer));
4776 assert(GvSTASH(referrer) == (HV*)sv);
4777 GvSTASH(referrer) = 0;
4778 } else {
4779 Perl_croak(aTHX_
4780 "panic: magic_killbackrefs (flags=%"UVxf")",
4781 (UV)SvFLAGS(referrer));
4782 }
4783
a0714e2c 4784 *svp = NULL;
86f55936
NC
4785 }
4786 svp++;
4787 }
4788 }
4789 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4790 return 0;
4791}
4792
954c1994
GS
4793/*
4794=for apidoc sv_insert
4795
4796Inserts a string at the specified offset/length within the SV. Similar to
4797the Perl substr() function.
4798
4799=cut
4800*/
4801
79072805 4802void
e1ec3a88 4803Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4804{
97aff369 4805 dVAR;
79072805
LW
4806 register char *big;
4807 register char *mid;
4808 register char *midend;
4809 register char *bigend;
4810 register I32 i;
6ff81951 4811 STRLEN curlen;
1c846c1f 4812
79072805 4813
8990e307 4814 if (!bigstr)
cea2e8a9 4815 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4816 SvPV_force(bigstr, curlen);
60fa28ff 4817 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4818 if (offset + len > curlen) {
4819 SvGROW(bigstr, offset+len+1);
93524f2b 4820 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4821 SvCUR_set(bigstr, offset+len);
4822 }
79072805 4823
69b47968 4824 SvTAINT(bigstr);
79072805
LW
4825 i = littlelen - len;
4826 if (i > 0) { /* string might grow */
a0d0e21e 4827 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4828 mid = big + offset + len;
4829 midend = bigend = big + SvCUR(bigstr);
4830 bigend += i;
4831 *bigend = '\0';
4832 while (midend > mid) /* shove everything down */
4833 *--bigend = *--midend;
4834 Move(little,big+offset,littlelen,char);
b162af07 4835 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4836 SvSETMAGIC(bigstr);
4837 return;
4838 }
4839 else if (i == 0) {
463ee0b2 4840 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4841 SvSETMAGIC(bigstr);
4842 return;
4843 }
4844
463ee0b2 4845 big = SvPVX(bigstr);
79072805
LW
4846 mid = big + offset;
4847 midend = mid + len;
4848 bigend = big + SvCUR(bigstr);
4849
4850 if (midend > bigend)
cea2e8a9 4851 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4852
4853 if (mid - big > bigend - midend) { /* faster to shorten from end */
4854 if (littlelen) {
4855 Move(little, mid, littlelen,char);
4856 mid += littlelen;
4857 }
4858 i = bigend - midend;
4859 if (i > 0) {
4860 Move(midend, mid, i,char);
4861 mid += i;
4862 }
4863 *mid = '\0';
4864 SvCUR_set(bigstr, mid - big);
4865 }
155aba94 4866 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4867 midend -= littlelen;
4868 mid = midend;
4869 sv_chop(bigstr,midend-i);
4870 big += i;
4871 while (i--)
4872 *--midend = *--big;
4873 if (littlelen)
4874 Move(little, mid, littlelen,char);
4875 }
4876 else if (littlelen) {
4877 midend -= littlelen;
4878 sv_chop(bigstr,midend);
4879 Move(little,midend,littlelen,char);
4880 }
4881 else {
4882 sv_chop(bigstr,midend);
4883 }
4884 SvSETMAGIC(bigstr);
4885}
4886
c461cf8f
JH
4887/*
4888=for apidoc sv_replace
4889
4890Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4891The target SV physically takes over ownership of the body of the source SV
4892and inherits its flags; however, the target keeps any magic it owns,
4893and any magic in the source is discarded.
ff276b08 4894Note that this is a rather specialist SV copying operation; most of the
645c22ef 4895time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4896
4897=cut
4898*/
79072805
LW
4899
4900void
864dbfa3 4901Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4902{
97aff369 4903 dVAR;
a3b680e6 4904 const U32 refcnt = SvREFCNT(sv);
765f542d 4905 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4906 if (SvREFCNT(nsv) != 1) {
7437becc 4907 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4908 UVuf " != 1)", (UV) SvREFCNT(nsv));
4909 }
93a17b20 4910 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4911 if (SvMAGICAL(nsv))
4912 mg_free(nsv);
4913 else
4914 sv_upgrade(nsv, SVt_PVMG);
b162af07 4915 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4916 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4917 SvMAGICAL_off(sv);
b162af07 4918 SvMAGIC_set(sv, NULL);
93a17b20 4919 }
79072805
LW
4920 SvREFCNT(sv) = 0;
4921 sv_clear(sv);
477f5d66 4922 assert(!SvREFCNT(sv));
fd0854ff
DM
4923#ifdef DEBUG_LEAKING_SCALARS
4924 sv->sv_flags = nsv->sv_flags;
4925 sv->sv_any = nsv->sv_any;
4926 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4927 sv->sv_u = nsv->sv_u;
fd0854ff 4928#else
79072805 4929 StructCopy(nsv,sv,SV);
fd0854ff 4930#endif
7b2c381c
NC
4931 /* Currently could join these into one piece of pointer arithmetic, but
4932 it would be unclear. */
4933 if(SvTYPE(sv) == SVt_IV)
4934 SvANY(sv)
339049b0 4935 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4936 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4937 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4938 }
4939
fd0854ff 4940
f8c7b90f 4941#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4942 if (SvIsCOW_normal(nsv)) {
4943 /* We need to follow the pointers around the loop to make the
4944 previous SV point to sv, rather than nsv. */
4945 SV *next;
4946 SV *current = nsv;
4947 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4948 assert(next);
4949 current = next;
3f7c398e 4950 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4951 }
4952 /* Make the SV before us point to the SV after us. */
4953 if (DEBUG_C_TEST) {
4954 PerlIO_printf(Perl_debug_log, "previous is\n");
4955 sv_dump(current);
a29f6d03
NC
4956 PerlIO_printf(Perl_debug_log,
4957 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4958 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4959 }
a29f6d03 4960 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4961 }
4962#endif
79072805 4963 SvREFCNT(sv) = refcnt;
1edc1566 4964 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4965 SvREFCNT(nsv) = 0;
463ee0b2 4966 del_SV(nsv);
79072805
LW
4967}
4968
c461cf8f
JH
4969/*
4970=for apidoc sv_clear
4971
645c22ef
DM
4972Clear an SV: call any destructors, free up any memory used by the body,
4973and free the body itself. The SV's head is I<not> freed, although
4974its type is set to all 1's so that it won't inadvertently be assumed
4975to be live during global destruction etc.
4976This function should only be called when REFCNT is zero. Most of the time
4977you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4978instead.
c461cf8f
JH
4979
4980=cut
4981*/
4982
79072805 4983void
864dbfa3 4984Perl_sv_clear(pTHX_ register SV *sv)
79072805 4985{
27da23d5 4986 dVAR;
82bb6deb 4987 const U32 type = SvTYPE(sv);
8edfc514
NC
4988 const struct body_details *const sv_type_details
4989 = bodies_by_type + type;
82bb6deb 4990
79072805
LW
4991 assert(sv);
4992 assert(SvREFCNT(sv) == 0);
4993
d2a0f284
JC
4994 if (type <= SVt_IV) {
4995 /* See the comment in sv.h about the collusion between this early
4996 return and the overloading of the NULL and IV slots in the size
4997 table. */
82bb6deb 4998 return;
d2a0f284 4999 }
82bb6deb 5000
ed6116ce 5001 if (SvOBJECT(sv)) {
3280af22 5002 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5003 dSP;
893645bd 5004 HV* stash;
d460ef45 5005 do {
b464bac0 5006 CV* destructor;
4e8e7886 5007 stash = SvSTASH(sv);
32251b26 5008 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5009 if (destructor) {
1b6737cc 5010 SV* const tmpref = newRV(sv);
5cc433a6 5011 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5012 ENTER;
e788e7d3 5013 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5014 EXTEND(SP, 2);
5015 PUSHMARK(SP);
5cc433a6 5016 PUSHs(tmpref);
4e8e7886 5017 PUTBACK;
44389ee9 5018 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5019
5020
d3acc0f7 5021 POPSTACK;
3095d977 5022 SPAGAIN;
4e8e7886 5023 LEAVE;
5cc433a6
AB
5024 if(SvREFCNT(tmpref) < 2) {
5025 /* tmpref is not kept alive! */
5026 SvREFCNT(sv)--;
b162af07 5027 SvRV_set(tmpref, NULL);
5cc433a6
AB
5028 SvROK_off(tmpref);
5029 }
5030 SvREFCNT_dec(tmpref);
4e8e7886
GS
5031 }
5032 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5033
6f44e0a4
JP
5034
5035 if (SvREFCNT(sv)) {
5036 if (PL_in_clean_objs)
cea2e8a9 5037 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5038 HvNAME_get(stash));
6f44e0a4
JP
5039 /* DESTROY gave object new lease on life */
5040 return;
5041 }
a0d0e21e 5042 }
4e8e7886 5043
a0d0e21e 5044 if (SvOBJECT(sv)) {
4e8e7886 5045 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5046 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5047 if (type != SVt_PVIO)
3280af22 5048 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5049 }
463ee0b2 5050 }
82bb6deb 5051 if (type >= SVt_PVMG) {
e736a858
NC
5052 HV *ourstash;
5053 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5054 (ourstash = OURSTASH(sv))) {
5055 SvREFCNT_dec(ourstash);
5056 } else if (SvMAGIC(sv))
524189f1 5057 mg_free(sv);
00b1698f 5058 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5059 SvREFCNT_dec(SvSTASH(sv));
5060 }
82bb6deb 5061 switch (type) {
8990e307 5062 case SVt_PVIO:
df0bd2f4
GS
5063 if (IoIFP(sv) &&
5064 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5065 IoIFP(sv) != PerlIO_stdout() &&
5066 IoIFP(sv) != PerlIO_stderr())
93578b34 5067 {
f2b5be74 5068 io_close((IO*)sv, FALSE);
93578b34 5069 }
1d7c1841 5070 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5071 PerlDir_close(IoDIRP(sv));
1d7c1841 5072 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5073 Safefree(IoTOP_NAME(sv));
5074 Safefree(IoFMT_NAME(sv));
5075 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5076 goto freescalar;
79072805 5077 case SVt_PVBM:
a0d0e21e 5078 goto freescalar;
79072805 5079 case SVt_PVCV:
748a9306 5080 case SVt_PVFM:
85e6fe83 5081 cv_undef((CV*)sv);
a0d0e21e 5082 goto freescalar;
79072805 5083 case SVt_PVHV:
86f55936 5084 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5085 hv_undef((HV*)sv);
a0d0e21e 5086 break;
79072805 5087 case SVt_PVAV:
85e6fe83 5088 av_undef((AV*)sv);
a0d0e21e 5089 break;
02270b4e 5090 case SVt_PVLV:
dd28f7bb
DM
5091 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5092 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5093 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5094 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5095 }
5096 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5097 SvREFCNT_dec(LvTARG(sv));
02270b4e 5098 goto freescalar;
a0d0e21e 5099 case SVt_PVGV:
1edc1566 5100 gp_free((GV*)sv);
acda4c6a
NC
5101 if (GvNAME_HEK(sv)) {
5102 unshare_hek(GvNAME_HEK(sv));
5103 }
893645bd
NC
5104 /* If we're in a stash, we don't own a reference to it. However it does
5105 have a back reference to us, which needs to be cleared. */
5106 if (GvSTASH(sv))
5107 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5108 case SVt_PVMG:
79072805
LW
5109 case SVt_PVNV:
5110 case SVt_PVIV:
a0d0e21e 5111 freescalar:
5228ca4e
NC
5112 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5113 if (SvOOK(sv)) {
93524f2b 5114 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5115 /* Don't even bother with turning off the OOK flag. */
5116 }
79072805 5117 case SVt_PV:
a0d0e21e 5118 case SVt_RV:
810b8aa5 5119 if (SvROK(sv)) {
b37c2d43 5120 SV * const target = SvRV(sv);
810b8aa5 5121 if (SvWEAKREF(sv))
e15faf7d 5122 sv_del_backref(target, sv);
810b8aa5 5123 else
e15faf7d 5124 SvREFCNT_dec(target);
810b8aa5 5125 }
f8c7b90f 5126#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5127 else if (SvPVX_const(sv)) {
765f542d
NC
5128 if (SvIsCOW(sv)) {
5129 /* I believe I need to grab the global SV mutex here and
5130 then recheck the COW status. */
46187eeb
NC
5131 if (DEBUG_C_TEST) {
5132 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5133 sv_dump(sv);
46187eeb 5134 }
bdd68bc3
NC
5135 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5136 SV_COW_NEXT_SV(sv));
765f542d
NC
5137 /* And drop it here. */
5138 SvFAKE_off(sv);
5139 } else if (SvLEN(sv)) {
3f7c398e 5140 Safefree(SvPVX_const(sv));
765f542d
NC
5141 }
5142 }
5143#else
3f7c398e 5144 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5145 Safefree(SvPVX_mutable(sv));
3f7c398e 5146 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5147 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5148 SvFAKE_off(sv);
5149 }
765f542d 5150#endif
79072805
LW
5151 break;
5152 case SVt_NV:
79072805
LW
5153 break;
5154 }
5155
893645bd
NC
5156 SvFLAGS(sv) &= SVf_BREAK;
5157 SvFLAGS(sv) |= SVTYPEMASK;
5158
8edfc514 5159 if (sv_type_details->arena) {
b9502f15 5160 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5161 &PL_body_roots[type]);
5162 }
d2a0f284 5163 else if (sv_type_details->body_size) {
8edfc514
NC
5164 my_safefree(SvANY(sv));
5165 }
79072805
LW
5166}
5167
645c22ef
DM
5168/*
5169=for apidoc sv_newref
5170
5171Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5172instead.
5173
5174=cut
5175*/
5176
79072805 5177SV *
864dbfa3 5178Perl_sv_newref(pTHX_ SV *sv)
79072805 5179{
96a5add6 5180 PERL_UNUSED_CONTEXT;
463ee0b2 5181 if (sv)
4db098f4 5182 (SvREFCNT(sv))++;
79072805
LW
5183 return sv;
5184}
5185
c461cf8f
JH
5186/*
5187=for apidoc sv_free
5188
645c22ef
DM
5189Decrement an SV's reference count, and if it drops to zero, call
5190C<sv_clear> to invoke destructors and free up any memory used by
5191the body; finally, deallocate the SV's head itself.
5192Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5193
5194=cut
5195*/
5196
79072805 5197void
864dbfa3 5198Perl_sv_free(pTHX_ SV *sv)
79072805 5199{
27da23d5 5200 dVAR;
79072805
LW
5201 if (!sv)
5202 return;
a0d0e21e
LW
5203 if (SvREFCNT(sv) == 0) {
5204 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5205 /* this SV's refcnt has been artificially decremented to
5206 * trigger cleanup */
a0d0e21e 5207 return;
3280af22 5208 if (PL_in_clean_all) /* All is fair */
1edc1566 5209 return;
d689ffdd
JP
5210 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5211 /* make sure SvREFCNT(sv)==0 happens very seldom */
5212 SvREFCNT(sv) = (~(U32)0)/2;
5213 return;
5214 }
41e4abd8 5215 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5216 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5217 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5218 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5219#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5220 Perl_dump_sv_child(aTHX_ sv);
5221#endif
5222 }
79072805
LW
5223 return;
5224 }
4db098f4 5225 if (--(SvREFCNT(sv)) > 0)
8990e307 5226 return;
8c4d3c90
NC
5227 Perl_sv_free2(aTHX_ sv);
5228}
5229
5230void
5231Perl_sv_free2(pTHX_ SV *sv)
5232{
27da23d5 5233 dVAR;
463ee0b2
LW
5234#ifdef DEBUGGING
5235 if (SvTEMP(sv)) {
0453d815 5236 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5237 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5238 "Attempt to free temp prematurely: SV 0x%"UVxf
5239 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5240 return;
79072805 5241 }
463ee0b2 5242#endif
d689ffdd
JP
5243 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5244 /* make sure SvREFCNT(sv)==0 happens very seldom */
5245 SvREFCNT(sv) = (~(U32)0)/2;
5246 return;
5247 }
79072805 5248 sv_clear(sv);
477f5d66
CS
5249 if (! SvREFCNT(sv))
5250 del_SV(sv);
79072805
LW
5251}
5252
954c1994
GS
5253/*
5254=for apidoc sv_len
5255
645c22ef
DM
5256Returns the length of the string in the SV. Handles magic and type
5257coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5258
5259=cut
5260*/
5261
79072805 5262STRLEN
864dbfa3 5263Perl_sv_len(pTHX_ register SV *sv)
79072805 5264{
463ee0b2 5265 STRLEN len;
79072805
LW
5266
5267 if (!sv)
5268 return 0;
5269
8990e307 5270 if (SvGMAGICAL(sv))
565764a8 5271 len = mg_length(sv);
8990e307 5272 else
4d84ee25 5273 (void)SvPV_const(sv, len);
463ee0b2 5274 return len;
79072805
LW
5275}
5276
c461cf8f
JH
5277/*
5278=for apidoc sv_len_utf8
5279
5280Returns the number of characters in the string in an SV, counting wide
1e54db1a 5281UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5282
5283=cut
5284*/
5285
7e8c5dac
HS
5286/*
5287 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5288 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5289 * (Note that the mg_len is not the length of the mg_ptr field.
5290 * This allows the cache to store the character length of the string without
5291 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5292 *
7e8c5dac
HS
5293 */
5294
a0ed51b3 5295STRLEN
864dbfa3 5296Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5297{
a0ed51b3
LW
5298 if (!sv)
5299 return 0;
5300
a0ed51b3 5301 if (SvGMAGICAL(sv))
b76347f2 5302 return mg_length(sv);
a0ed51b3 5303 else
b76347f2 5304 {
26346457 5305 STRLEN len;
e62f0680 5306 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5307
26346457
NC
5308 if (PL_utf8cache) {
5309 STRLEN ulen;
5310 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5311
5312 if (mg && mg->mg_len != -1) {
5313 ulen = mg->mg_len;
5314 if (PL_utf8cache < 0) {
5315 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5316 if (real != ulen) {
5317 /* Need to turn the assertions off otherwise we may
5318 recurse infinitely while printing error messages.
5319 */
5320 SAVEI8(PL_utf8cache);
5321 PL_utf8cache = 0;
5322 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5323 " real %"UVf" for %"SVf,
26346457
NC
5324 (UV) ulen, (UV) real, sv);
5325 }
5326 }
5327 }
5328 else {
5329 ulen = Perl_utf8_length(aTHX_ s, s + len);
5330 if (!SvREADONLY(sv)) {
5331 if (!mg) {
5332 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5333 &PL_vtbl_utf8, 0, 0);
5334 }
cb9e20bb 5335 assert(mg);
26346457 5336 mg->mg_len = ulen;
cb9e20bb 5337 }
cb9e20bb 5338 }
26346457 5339 return ulen;
7e8c5dac 5340 }
26346457 5341 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5342 }
5343}
5344
9564a3bd
NC
5345/* Walk forwards to find the byte corresponding to the passed in UTF-8
5346 offset. */
bdf30dd6
NC
5347static STRLEN
5348S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send,
5349 STRLEN uoffset)
5350{
5351 const U8 *s = start;
5352
d4c19fe8
AL
5353 PERL_UNUSED_CONTEXT;
5354
bdf30dd6
NC
5355 while (s < send && uoffset--)
5356 s += UTF8SKIP(s);
5357 if (s > send) {
5358 /* This is the existing behaviour. Possibly it should be a croak, as
5359 it's actually a bounds error */
5360 s = send;
5361 }
5362 return s - start;
5363}
5364
9564a3bd
NC
5365/* Given the length of the string in both bytes and UTF-8 characters, decide
5366 whether to walk forwards or backwards to find the byte corresponding to
5367 the passed in UTF-8 offset. */
c336ad0b
NC
5368static STRLEN
5369S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send,
5370 STRLEN uoffset, STRLEN uend)
5371{
5372 STRLEN backw = uend - uoffset;
5373 if (uoffset < 2 * backw) {
25a8a4ef 5374 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5375 forward (that's where the 2 * backw comes from).
5376 (The real figure of course depends on the UTF-8 data.) */
5377 return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset);
5378 }
5379
5380 while (backw--) {
5381 send--;
5382 while (UTF8_IS_CONTINUATION(*send))
5383 send--;
5384 }
5385 return send - start;
5386}
5387
9564a3bd
NC
5388/* For the string representation of the given scalar, find the byte
5389 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5390 give another position in the string, *before* the sought offset, which
5391 (which is always true, as 0, 0 is a valid pair of positions), which should
5392 help reduce the amount of linear searching.
5393 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5394 will be used to reduce the amount of linear searching. The cache will be
5395 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5396static STRLEN
5397S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5398 const U8 *const send, STRLEN uoffset,
5399 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5400 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5401 bool found = FALSE;
5402
75c33c12
NC
5403 assert (uoffset >= uoffset0);
5404
c336ad0b 5405 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5406 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5407 if ((*mgp)->mg_ptr) {
5408 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5409 if (cache[0] == uoffset) {
5410 /* An exact match. */
5411 return cache[1];
5412 }
ab455f60
NC
5413 if (cache[2] == uoffset) {
5414 /* An exact match. */
5415 return cache[3];
5416 }
668af93f
NC
5417
5418 if (cache[0] < uoffset) {
d8b2e1f9
NC
5419 /* The cache already knows part of the way. */
5420 if (cache[0] > uoffset0) {
5421 /* The cache knows more than the passed in pair */
5422 uoffset0 = cache[0];
5423 boffset0 = cache[1];
5424 }
5425 if ((*mgp)->mg_len != -1) {
5426 /* And we know the end too. */
5427 boffset = boffset0
5428 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5429 uoffset - uoffset0,
5430 (*mgp)->mg_len - uoffset0);
5431 } else {
5432 boffset = boffset0
5433 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5434 send, uoffset - uoffset0);
5435 }
dd7c5fd3
NC
5436 }
5437 else if (cache[2] < uoffset) {
5438 /* We're between the two cache entries. */
5439 if (cache[2] > uoffset0) {
5440 /* and the cache knows more than the passed in pair */
5441 uoffset0 = cache[2];
5442 boffset0 = cache[3];
5443 }
5444
668af93f
NC
5445 boffset = boffset0
5446 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5447 start + cache[1],
5448 uoffset - uoffset0,
5449 cache[0] - uoffset0);
dd7c5fd3
NC
5450 } else {
5451 boffset = boffset0
5452 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5453 start + cache[3],
5454 uoffset - uoffset0,
5455 cache[2] - uoffset0);
d8b2e1f9 5456 }
668af93f 5457 found = TRUE;
d8b2e1f9
NC
5458 }
5459 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5460 /* If we can take advantage of a passed in offset, do so. */
5461 /* In fact, offset0 is either 0, or less than offset, so don't
5462 need to worry about the other possibility. */
5463 boffset = boffset0
5464 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5465 uoffset - uoffset0,
5466 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5467 found = TRUE;
5468 }
28ccbf94 5469 }
c336ad0b
NC
5470
5471 if (!found || PL_utf8cache < 0) {
75c33c12
NC
5472 const STRLEN real_boffset
5473 = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5474 send, uoffset - uoffset0);
5475
c336ad0b
NC
5476 if (found && PL_utf8cache < 0) {
5477 if (real_boffset != boffset) {
5478 /* Need to turn the assertions off otherwise we may recurse
5479 infinitely while printing error messages. */
5480 SAVEI8(PL_utf8cache);
5481 PL_utf8cache = 0;
5482 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5483 " real %"UVf" for %"SVf,
5484 (UV) boffset, (UV) real_boffset, sv);
5485 }
5486 }
5487 boffset = real_boffset;
28ccbf94 5488 }
0905937d 5489
ab455f60 5490 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5491 return boffset;
5492}
5493
9564a3bd
NC
5494
5495/*
5496=for apidoc sv_pos_u2b
5497
5498Converts the value pointed to by offsetp from a count of UTF-8 chars from
5499the start of the string, to a count of the equivalent number of bytes; if
5500lenp is non-zero, it does the same to lenp, but this time starting from
5501the offset, rather than from the start of the string. Handles magic and
5502type coercion.
5503
5504=cut
5505*/
5506
5507/*
5508 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5509 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5510 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5511 *
5512 */
5513
a0ed51b3 5514void
864dbfa3 5515Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5516{
245d4a47 5517 const U8 *start;
a0ed51b3
LW
5518 STRLEN len;
5519
5520 if (!sv)
5521 return;
5522
245d4a47 5523 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5524 if (len) {
bdf30dd6
NC
5525 STRLEN uoffset = (STRLEN) *offsetp;
5526 const U8 * const send = start + len;
0905937d 5527 MAGIC *mg = NULL;
28ccbf94
NC
5528 STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send,
5529 uoffset, 0, 0);
bdf30dd6
NC
5530
5531 *offsetp = (I32) boffset;
5532
5533 if (lenp) {
28ccbf94
NC
5534 /* Convert the relative offset to absolute. */
5535 STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5536 STRLEN boffset2
5537 = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2,
5538 uoffset, boffset) - boffset;
bdf30dd6 5539
28ccbf94 5540 *lenp = boffset2;
bdf30dd6 5541 }
7e8c5dac
HS
5542 }
5543 else {
5544 *offsetp = 0;
5545 if (lenp)
5546 *lenp = 0;
a0ed51b3 5547 }
e23c8137 5548
a0ed51b3
LW
5549 return;
5550}
5551
9564a3bd
NC
5552/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5553 byte length pairing. The (byte) length of the total SV is passed in too,
5554 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5555 may not have updated SvCUR, so we can't rely on reading it directly.
5556
5557 The proffered utf8/byte length pairing isn't used if the cache already has
5558 two pairs, and swapping either for the proffered pair would increase the
5559 RMS of the intervals between known byte offsets.
5560
5561 The cache itself consists of 4 STRLEN values
5562 0: larger UTF-8 offset
5563 1: corresponding byte offset
5564 2: smaller UTF-8 offset
5565 3: corresponding byte offset
5566
5567 Unused cache pairs have the value 0, 0.
5568 Keeping the cache "backwards" means that the invariant of
5569 cache[0] >= cache[2] is maintained even with empty slots, which means that
5570 the code that uses it doesn't need to worry if only 1 entry has actually
5571 been set to non-zero. It also makes the "position beyond the end of the
5572 cache" logic much simpler, as the first slot is always the one to start
5573 from.
645c22ef 5574*/
ec07b5e0 5575static void
ab455f60
NC
5576S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5577 STRLEN blen)
ec07b5e0
NC
5578{
5579 STRLEN *cache;
5580 if (SvREADONLY(sv))
5581 return;
5582
5583 if (!*mgp) {
5584 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5585 0);
5586 (*mgp)->mg_len = -1;
5587 }
5588 assert(*mgp);
5589
5590 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5591 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5592 (*mgp)->mg_ptr = (char *) cache;
5593 }
5594 assert(cache);
5595
5596 if (PL_utf8cache < 0) {
ef816a78 5597 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5598 const U8 *const end = start + byte;
5599 STRLEN realutf8 = 0;
5600
5601 while (start < end) {
5602 start += UTF8SKIP(start);
5603 realutf8++;
5604 }
5605
5606 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5607 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5608 doesn't? I don't know whether this difference was introduced with
5609 the caching code in 5.8.1. */
ec07b5e0
NC
5610
5611 if (realutf8 != utf8) {
5612 /* Need to turn the assertions off otherwise we may recurse
5613 infinitely while printing error messages. */
5614 SAVEI8(PL_utf8cache);
5615 PL_utf8cache = 0;
5616 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5617 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5618 }
5619 }
ab455f60
NC
5620
5621 /* Cache is held with the later position first, to simplify the code
5622 that deals with unbounded ends. */
5623
5624 ASSERT_UTF8_CACHE(cache);
5625 if (cache[1] == 0) {
5626 /* Cache is totally empty */
5627 cache[0] = utf8;
5628 cache[1] = byte;
5629 } else if (cache[3] == 0) {
5630 if (byte > cache[1]) {
5631 /* New one is larger, so goes first. */
5632 cache[2] = cache[0];
5633 cache[3] = cache[1];
5634 cache[0] = utf8;
5635 cache[1] = byte;
5636 } else {
5637 cache[2] = utf8;
5638 cache[3] = byte;
5639 }
5640 } else {
5641#define THREEWAY_SQUARE(a,b,c,d) \
5642 ((float)((d) - (c))) * ((float)((d) - (c))) \
5643 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5644 + ((float)((b) - (a))) * ((float)((b) - (a)))
5645
5646 /* Cache has 2 slots in use, and we know three potential pairs.
5647 Keep the two that give the lowest RMS distance. Do the
5648 calcualation in bytes simply because we always know the byte
5649 length. squareroot has the same ordering as the positive value,
5650 so don't bother with the actual square root. */
5651 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5652 if (byte > cache[1]) {
5653 /* New position is after the existing pair of pairs. */
5654 const float keep_earlier
5655 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5656 const float keep_later
5657 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5658
5659 if (keep_later < keep_earlier) {
5660 if (keep_later < existing) {
5661 cache[2] = cache[0];
5662 cache[3] = cache[1];
5663 cache[0] = utf8;
5664 cache[1] = byte;
5665 }
5666 }
5667 else {
5668 if (keep_earlier < existing) {
5669 cache[0] = utf8;
5670 cache[1] = byte;
5671 }
5672 }
5673 }
57d7fbf1
NC
5674 else if (byte > cache[3]) {
5675 /* New position is between the existing pair of pairs. */
5676 const float keep_earlier
5677 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5678 const float keep_later
5679 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5680
5681 if (keep_later < keep_earlier) {
5682 if (keep_later < existing) {
5683 cache[2] = utf8;
5684 cache[3] = byte;
5685 }
5686 }
5687 else {
5688 if (keep_earlier < existing) {
5689 cache[0] = utf8;
5690 cache[1] = byte;
5691 }
5692 }
5693 }
5694 else {
5695 /* New position is before the existing pair of pairs. */
5696 const float keep_earlier
5697 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5698 const float keep_later
5699 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5700
5701 if (keep_later < keep_earlier) {
5702 if (keep_later < existing) {
5703 cache[2] = utf8;
5704 cache[3] = byte;
5705 }
5706 }
5707 else {
5708 if (keep_earlier < existing) {
5709 cache[0] = cache[2];
5710 cache[1] = cache[3];
5711 cache[2] = utf8;
5712 cache[3] = byte;
5713 }
5714 }
5715 }
ab455f60 5716 }
0905937d 5717 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5718}
5719
5720/* If we don't know the character offset of the end of a region, our only
5721 option is to walk forwards to the target byte offset. */
5722static STRLEN
5723S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5724{
5725 STRLEN len = 0;
5726 while (s < target) {
5727 STRLEN n = 1;
5728
5729 /* Call utf8n_to_uvchr() to validate the sequence
5730 * (unless a simple non-UTF character) */
5731 if (!UTF8_IS_INVARIANT(*s))
5732 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5733 if (n > 0) {
5734 s += n;
5735 len++;
5736 }
5737 else
5738 break;
5739 }
5740 return len;
5741}
5742
5743/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5744 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5745 backward is half the speed of walking forward. */
ec07b5e0
NC
5746static STRLEN
5747S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5748 STRLEN endu)
5749{
5750 const STRLEN forw = target - s;
5751 STRLEN backw = end - target;
5752
5753 if (forw < 2 * backw) {
5754 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5755 }
5756
5757 while (end > target) {
5758 end--;
5759 while (UTF8_IS_CONTINUATION(*end)) {
5760 end--;
5761 }
5762 endu--;
5763 }
5764 return endu;
5765}
5766
9564a3bd
NC
5767/*
5768=for apidoc sv_pos_b2u
5769
5770Converts the value pointed to by offsetp from a count of bytes from the
5771start of the string, to a count of the equivalent number of UTF-8 chars.
5772Handles magic and type coercion.
5773
5774=cut
5775*/
5776
5777/*
5778 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5779 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5780 * byte offsets.
5781 *
5782 */
a0ed51b3 5783void
7e8c5dac 5784Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5785{
83003860 5786 const U8* s;
ec07b5e0 5787 const STRLEN byte = *offsetp;
7087a21c 5788 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5789 STRLEN blen;
ec07b5e0
NC
5790 MAGIC* mg = NULL;
5791 const U8* send;
a922f900 5792 bool found = FALSE;
a0ed51b3
LW
5793
5794 if (!sv)
5795 return;
5796
ab455f60 5797 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5798
ab455f60 5799 if (blen < byte)
ec07b5e0 5800 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5801
ec07b5e0 5802 send = s + byte;
a67d7df9 5803
ffca234a
NC
5804 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5805 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5806 if (mg->mg_ptr) {
d4c19fe8 5807 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5808 if (cache[1] == byte) {
ec07b5e0
NC
5809 /* An exact match. */
5810 *offsetp = cache[0];
ec07b5e0 5811 return;
7e8c5dac 5812 }
ab455f60
NC
5813 if (cache[3] == byte) {
5814 /* An exact match. */
5815 *offsetp = cache[2];
5816 return;
5817 }
668af93f
NC
5818
5819 if (cache[1] < byte) {
ec07b5e0 5820 /* We already know part of the way. */
b9f984a5
NC
5821 if (mg->mg_len != -1) {
5822 /* Actually, we know the end too. */
5823 len = cache[0]
5824 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5825 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5826 } else {
5827 len = cache[0]
5828 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5829 }
7e8c5dac 5830 }
9f985e4c
NC
5831 else if (cache[3] < byte) {
5832 /* We're between the two cached pairs, so we do the calculation
5833 offset by the byte/utf-8 positions for the earlier pair,
5834 then add the utf-8 characters from the string start to
5835 there. */
5836 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5837 s + cache[1], cache[0] - cache[2])
5838 + cache[2];
5839
5840 }
5841 else { /* cache[3] > byte */
5842 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5843 cache[2]);
7e8c5dac 5844
7e8c5dac 5845 }
ec07b5e0 5846 ASSERT_UTF8_CACHE(cache);
a922f900 5847 found = TRUE;
ffca234a 5848 } else if (mg->mg_len != -1) {
ab455f60 5849 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5850 found = TRUE;
7e8c5dac 5851 }
a0ed51b3 5852 }
a922f900
NC
5853 if (!found || PL_utf8cache < 0) {
5854 const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5855
5856 if (found && PL_utf8cache < 0) {
5857 if (len != real_len) {
5858 /* Need to turn the assertions off otherwise we may recurse
5859 infinitely while printing error messages. */
5860 SAVEI8(PL_utf8cache);
5861 PL_utf8cache = 0;
5862 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5863 " real %"UVf" for %"SVf,
5864 (UV) len, (UV) real_len, sv);
5865 }
5866 }
5867 len = real_len;
ec07b5e0
NC
5868 }
5869 *offsetp = len;
5870
ab455f60 5871 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5872}
5873
954c1994
GS
5874/*
5875=for apidoc sv_eq
5876
5877Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5878identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5879coerce its args to strings if necessary.
954c1994
GS
5880
5881=cut
5882*/
5883
79072805 5884I32
e01b9e88 5885Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5886{
97aff369 5887 dVAR;
e1ec3a88 5888 const char *pv1;
463ee0b2 5889 STRLEN cur1;
e1ec3a88 5890 const char *pv2;
463ee0b2 5891 STRLEN cur2;
e01b9e88 5892 I32 eq = 0;
bd61b366 5893 char *tpv = NULL;
a0714e2c 5894 SV* svrecode = NULL;
79072805 5895
e01b9e88 5896 if (!sv1) {
79072805
LW
5897 pv1 = "";
5898 cur1 = 0;
5899 }
463ee0b2 5900 else
4d84ee25 5901 pv1 = SvPV_const(sv1, cur1);
79072805 5902
e01b9e88
SC
5903 if (!sv2){
5904 pv2 = "";
5905 cur2 = 0;
92d29cee 5906 }
e01b9e88 5907 else
4d84ee25 5908 pv2 = SvPV_const(sv2, cur2);
79072805 5909
cf48d248 5910 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5911 /* Differing utf8ness.
5912 * Do not UTF8size the comparands as a side-effect. */
5913 if (PL_encoding) {
5914 if (SvUTF8(sv1)) {
553e1bcc
AT
5915 svrecode = newSVpvn(pv2, cur2);
5916 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5917 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5918 }
5919 else {
553e1bcc
AT
5920 svrecode = newSVpvn(pv1, cur1);
5921 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5922 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5923 }
5924 /* Now both are in UTF-8. */
0a1bd7ac
DM
5925 if (cur1 != cur2) {
5926 SvREFCNT_dec(svrecode);
799ef3cb 5927 return FALSE;
0a1bd7ac 5928 }
799ef3cb
JH
5929 }
5930 else {
5931 bool is_utf8 = TRUE;
5932
5933 if (SvUTF8(sv1)) {
5934 /* sv1 is the UTF-8 one,
5935 * if is equal it must be downgrade-able */
9d4ba2ae 5936 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5937 &cur1, &is_utf8);
5938 if (pv != pv1)
553e1bcc 5939 pv1 = tpv = pv;
799ef3cb
JH
5940 }
5941 else {
5942 /* sv2 is the UTF-8 one,
5943 * if is equal it must be downgrade-able */
9d4ba2ae 5944 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5945 &cur2, &is_utf8);
5946 if (pv != pv2)
553e1bcc 5947 pv2 = tpv = pv;
799ef3cb
JH
5948 }
5949 if (is_utf8) {
5950 /* Downgrade not possible - cannot be eq */
bf694877 5951 assert (tpv == 0);
799ef3cb
JH
5952 return FALSE;
5953 }
5954 }
cf48d248
JH
5955 }
5956
5957 if (cur1 == cur2)
765f542d 5958 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5959
b37c2d43 5960 SvREFCNT_dec(svrecode);
553e1bcc
AT
5961 if (tpv)
5962 Safefree(tpv);
cf48d248 5963
e01b9e88 5964 return eq;
79072805
LW
5965}
5966
954c1994
GS
5967/*
5968=for apidoc sv_cmp
5969
5970Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5971string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5972C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5973coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5974
5975=cut
5976*/
5977
79072805 5978I32
e01b9e88 5979Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5980{
97aff369 5981 dVAR;
560a288e 5982 STRLEN cur1, cur2;
e1ec3a88 5983 const char *pv1, *pv2;
bd61b366 5984 char *tpv = NULL;
cf48d248 5985 I32 cmp;
a0714e2c 5986 SV *svrecode = NULL;
560a288e 5987
e01b9e88
SC
5988 if (!sv1) {
5989 pv1 = "";
560a288e
GS
5990 cur1 = 0;
5991 }
e01b9e88 5992 else
4d84ee25 5993 pv1 = SvPV_const(sv1, cur1);
560a288e 5994
553e1bcc 5995 if (!sv2) {
e01b9e88 5996 pv2 = "";
560a288e
GS
5997 cur2 = 0;
5998 }
e01b9e88 5999 else
4d84ee25 6000 pv2 = SvPV_const(sv2, cur2);
79072805 6001
cf48d248 6002 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6003 /* Differing utf8ness.
6004 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6005 if (SvUTF8(sv1)) {
799ef3cb 6006 if (PL_encoding) {
553e1bcc
AT
6007 svrecode = newSVpvn(pv2, cur2);
6008 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6009 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6010 }
6011 else {
e1ec3a88 6012 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6013 }
cf48d248
JH
6014 }
6015 else {
799ef3cb 6016 if (PL_encoding) {
553e1bcc
AT
6017 svrecode = newSVpvn(pv1, cur1);
6018 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6019 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6020 }
6021 else {
e1ec3a88 6022 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6023 }
cf48d248
JH
6024 }
6025 }
6026
e01b9e88 6027 if (!cur1) {
cf48d248 6028 cmp = cur2 ? -1 : 0;
e01b9e88 6029 } else if (!cur2) {
cf48d248
JH
6030 cmp = 1;
6031 } else {
e1ec3a88 6032 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6033
6034 if (retval) {
cf48d248 6035 cmp = retval < 0 ? -1 : 1;
e01b9e88 6036 } else if (cur1 == cur2) {
cf48d248
JH
6037 cmp = 0;
6038 } else {
6039 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6040 }
cf48d248 6041 }
16660edb 6042
b37c2d43 6043 SvREFCNT_dec(svrecode);
553e1bcc
AT
6044 if (tpv)
6045 Safefree(tpv);
cf48d248
JH
6046
6047 return cmp;
bbce6d69 6048}
16660edb 6049
c461cf8f
JH
6050/*
6051=for apidoc sv_cmp_locale
6052
645c22ef
DM
6053Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6054'use bytes' aware, handles get magic, and will coerce its args to strings
6055if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6056
6057=cut
6058*/
6059
bbce6d69 6060I32
864dbfa3 6061Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6062{
97aff369 6063 dVAR;
36477c24 6064#ifdef USE_LOCALE_COLLATE
16660edb 6065
bbce6d69 6066 char *pv1, *pv2;
6067 STRLEN len1, len2;
6068 I32 retval;
16660edb 6069
3280af22 6070 if (PL_collation_standard)
bbce6d69 6071 goto raw_compare;
16660edb 6072
bbce6d69 6073 len1 = 0;
8ac85365 6074 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6075 len2 = 0;
8ac85365 6076 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6077
bbce6d69 6078 if (!pv1 || !len1) {
6079 if (pv2 && len2)
6080 return -1;
6081 else
6082 goto raw_compare;
6083 }
6084 else {
6085 if (!pv2 || !len2)
6086 return 1;
6087 }
16660edb 6088
bbce6d69 6089 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6090
bbce6d69 6091 if (retval)
16660edb 6092 return retval < 0 ? -1 : 1;
6093
bbce6d69 6094 /*
6095 * When the result of collation is equality, that doesn't mean
6096 * that there are no differences -- some locales exclude some
6097 * characters from consideration. So to avoid false equalities,
6098 * we use the raw string as a tiebreaker.
6099 */
16660edb 6100
bbce6d69 6101 raw_compare:
5f66b61c 6102 /*FALLTHROUGH*/
16660edb 6103
36477c24 6104#endif /* USE_LOCALE_COLLATE */
16660edb 6105
bbce6d69 6106 return sv_cmp(sv1, sv2);
6107}
79072805 6108
645c22ef 6109
36477c24 6110#ifdef USE_LOCALE_COLLATE
645c22ef 6111
7a4c00b4 6112/*
645c22ef
DM
6113=for apidoc sv_collxfrm
6114
6115Add Collate Transform magic to an SV if it doesn't already have it.
6116
6117Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6118scalar data of the variable, but transformed to such a format that a normal
6119memory comparison can be used to compare the data according to the locale
6120settings.
6121
6122=cut
6123*/
6124
bbce6d69 6125char *
864dbfa3 6126Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6127{
97aff369 6128 dVAR;
7a4c00b4 6129 MAGIC *mg;
16660edb 6130
14befaf4 6131 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6132 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6133 const char *s;
6134 char *xf;
bbce6d69 6135 STRLEN len, xlen;
6136
7a4c00b4 6137 if (mg)
6138 Safefree(mg->mg_ptr);
93524f2b 6139 s = SvPV_const(sv, len);
bbce6d69 6140 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6141 if (SvREADONLY(sv)) {
6142 SAVEFREEPV(xf);
6143 *nxp = xlen;
3280af22 6144 return xf + sizeof(PL_collation_ix);
ff0cee69 6145 }
7a4c00b4 6146 if (! mg) {
d83f0a82
NC
6147#ifdef PERL_OLD_COPY_ON_WRITE
6148 if (SvIsCOW(sv))
6149 sv_force_normal_flags(sv, 0);
6150#endif
6151 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6152 0, 0);
7a4c00b4 6153 assert(mg);
bbce6d69 6154 }
7a4c00b4 6155 mg->mg_ptr = xf;
565764a8 6156 mg->mg_len = xlen;
7a4c00b4 6157 }
6158 else {
ff0cee69 6159 if (mg) {
6160 mg->mg_ptr = NULL;
565764a8 6161 mg->mg_len = -1;
ff0cee69 6162 }
bbce6d69 6163 }
6164 }
7a4c00b4 6165 if (mg && mg->mg_ptr) {
565764a8 6166 *nxp = mg->mg_len;
3280af22 6167 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6168 }
6169 else {
6170 *nxp = 0;
6171 return NULL;
16660edb 6172 }
79072805
LW
6173}
6174
36477c24 6175#endif /* USE_LOCALE_COLLATE */
bbce6d69 6176
c461cf8f
JH
6177/*
6178=for apidoc sv_gets
6179
6180Get a line from the filehandle and store it into the SV, optionally
6181appending to the currently-stored string.
6182
6183=cut
6184*/
6185
79072805 6186char *
864dbfa3 6187Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6188{
97aff369 6189 dVAR;
e1ec3a88 6190 const char *rsptr;
c07a80fd 6191 STRLEN rslen;
6192 register STDCHAR rslast;
6193 register STDCHAR *bp;
6194 register I32 cnt;
9c5ffd7c 6195 I32 i = 0;
8bfdd7d9 6196 I32 rspara = 0;
c07a80fd 6197
bc44a8a2
NC
6198 if (SvTHINKFIRST(sv))
6199 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6200 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6201 from <>.
6202 However, perlbench says it's slower, because the existing swipe code
6203 is faster than copy on write.
6204 Swings and roundabouts. */
862a34c6 6205 SvUPGRADE(sv, SVt_PV);
99491443 6206
ff68c719 6207 SvSCREAM_off(sv);
efd8b2ba
AE
6208
6209 if (append) {
6210 if (PerlIO_isutf8(fp)) {
6211 if (!SvUTF8(sv)) {
6212 sv_utf8_upgrade_nomg(sv);
6213 sv_pos_u2b(sv,&append,0);
6214 }
6215 } else if (SvUTF8(sv)) {
561b68a9 6216 SV * const tsv = newSV(0);
efd8b2ba
AE
6217 sv_gets(tsv, fp, 0);
6218 sv_utf8_upgrade_nomg(tsv);
6219 SvCUR_set(sv,append);
6220 sv_catsv(sv,tsv);
6221 sv_free(tsv);
6222 goto return_string_or_null;
6223 }
6224 }
6225
6226 SvPOK_only(sv);
6227 if (PerlIO_isutf8(fp))
6228 SvUTF8_on(sv);
c07a80fd 6229
923e4eb5 6230 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6231 /* we always read code in line mode */
6232 rsptr = "\n";
6233 rslen = 1;
6234 }
6235 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6236 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6237 of amount we are going to read -- may result in mallocing
6238 more memory than we really need if the layers below reduce
6239 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6240 */
e311fd51 6241 Stat_t st;
e468d35b 6242 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6243 const Off_t offset = PerlIO_tell(fp);
58f1856e 6244 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6245 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6246 }
6247 }
c07a80fd 6248 rsptr = NULL;
6249 rslen = 0;
6250 }
3280af22 6251 else if (RsRECORD(PL_rs)) {
e311fd51 6252 I32 bytesread;
5b2b9c68 6253 char *buffer;
acbd132f 6254 U32 recsize;
5b2b9c68
HM
6255
6256 /* Grab the size of the record we're getting */
acbd132f 6257 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6258 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6259 /* Go yank in */
6260#ifdef VMS
6261 /* VMS wants read instead of fread, because fread doesn't respect */
6262 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6263 /* doing, but we've got no other real choice - except avoid stdio
6264 as implementation - perhaps write a :vms layer ?
6265 */
5b2b9c68
HM
6266 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6267#else
6268 bytesread = PerlIO_read(fp, buffer, recsize);
6269#endif
27e6ca2d
AE
6270 if (bytesread < 0)
6271 bytesread = 0;
e311fd51 6272 SvCUR_set(sv, bytesread += append);
e670df4e 6273 buffer[bytesread] = '\0';
efd8b2ba 6274 goto return_string_or_null;
5b2b9c68 6275 }
3280af22 6276 else if (RsPARA(PL_rs)) {
c07a80fd 6277 rsptr = "\n\n";
6278 rslen = 2;
8bfdd7d9 6279 rspara = 1;
c07a80fd 6280 }
7d59b7e4
NIS
6281 else {
6282 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6283 if (PerlIO_isutf8(fp)) {
6284 rsptr = SvPVutf8(PL_rs, rslen);
6285 }
6286 else {
6287 if (SvUTF8(PL_rs)) {
6288 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6289 Perl_croak(aTHX_ "Wide character in $/");
6290 }
6291 }
93524f2b 6292 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6293 }
6294 }
6295
c07a80fd 6296 rslast = rslen ? rsptr[rslen - 1] : '\0';
6297
8bfdd7d9 6298 if (rspara) { /* have to do this both before and after */
79072805 6299 do { /* to make sure file boundaries work right */
760ac839 6300 if (PerlIO_eof(fp))
a0d0e21e 6301 return 0;
760ac839 6302 i = PerlIO_getc(fp);
79072805 6303 if (i != '\n') {
a0d0e21e
LW
6304 if (i == -1)
6305 return 0;
760ac839 6306 PerlIO_ungetc(fp,i);
79072805
LW
6307 break;
6308 }
6309 } while (i != EOF);
6310 }
c07a80fd 6311
760ac839
LW
6312 /* See if we know enough about I/O mechanism to cheat it ! */
6313
6314 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6315 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6316 enough here - and may even be a macro allowing compile
6317 time optimization.
6318 */
6319
6320 if (PerlIO_fast_gets(fp)) {
6321
6322 /*
6323 * We're going to steal some values from the stdio struct
6324 * and put EVERYTHING in the innermost loop into registers.
6325 */
6326 register STDCHAR *ptr;
6327 STRLEN bpx;
6328 I32 shortbuffered;
6329
16660edb 6330#if defined(VMS) && defined(PERLIO_IS_STDIO)
6331 /* An ungetc()d char is handled separately from the regular
6332 * buffer, so we getc() it back out and stuff it in the buffer.
6333 */
6334 i = PerlIO_getc(fp);
6335 if (i == EOF) return 0;
6336 *(--((*fp)->_ptr)) = (unsigned char) i;
6337 (*fp)->_cnt++;
6338#endif
c07a80fd 6339
c2960299 6340 /* Here is some breathtakingly efficient cheating */
c07a80fd 6341
a20bf0c3 6342 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6343 /* make sure we have the room */
7a5fa8a2 6344 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6345 /* Not room for all of it
7a5fa8a2 6346 if we are looking for a separator and room for some
e468d35b
NIS
6347 */
6348 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6349 /* just process what we have room for */
79072805
LW
6350 shortbuffered = cnt - SvLEN(sv) + append + 1;
6351 cnt -= shortbuffered;
6352 }
6353 else {
6354 shortbuffered = 0;
bbce6d69 6355 /* remember that cnt can be negative */
eb160463 6356 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6357 }
6358 }
7a5fa8a2 6359 else
79072805 6360 shortbuffered = 0;
3f7c398e 6361 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6362 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6363 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6364 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6365 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6366 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6367 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6368 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6369 for (;;) {
6370 screamer:
93a17b20 6371 if (cnt > 0) {
c07a80fd 6372 if (rslen) {
760ac839
LW
6373 while (cnt > 0) { /* this | eat */
6374 cnt--;
c07a80fd 6375 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6376 goto thats_all_folks; /* screams | sed :-) */
6377 }
6378 }
6379 else {
1c846c1f
NIS
6380 Copy(ptr, bp, cnt, char); /* this | eat */
6381 bp += cnt; /* screams | dust */
c07a80fd 6382 ptr += cnt; /* louder | sed :-) */
a5f75d66 6383 cnt = 0;
93a17b20 6384 }
79072805
LW
6385 }
6386
748a9306 6387 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6388 cnt = shortbuffered;
6389 shortbuffered = 0;
3f7c398e 6390 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6391 SvCUR_set(sv, bpx);
6392 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6393 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6394 continue;
6395 }
6396
16660edb 6397 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6398 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6399 PTR2UV(ptr),(long)cnt));
cc00df79 6400 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6401#if 0
16660edb 6402 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6403 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6404 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6405 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6406#endif
1c846c1f 6407 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6408 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6409 another abstraction. */
760ac839 6410 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6411#if 0
16660edb 6412 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6413 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6414 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6415 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6416#endif
a20bf0c3
JH
6417 cnt = PerlIO_get_cnt(fp);
6418 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6419 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6420 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6421
748a9306
LW
6422 if (i == EOF) /* all done for ever? */
6423 goto thats_really_all_folks;
6424
3f7c398e 6425 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6426 SvCUR_set(sv, bpx);
6427 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6428 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6429
eb160463 6430 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6431
c07a80fd 6432 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6433 goto thats_all_folks;
79072805
LW
6434 }
6435
6436thats_all_folks:
3f7c398e 6437 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6438 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6439 goto screamer; /* go back to the fray */
79072805
LW
6440thats_really_all_folks:
6441 if (shortbuffered)
6442 cnt += shortbuffered;
16660edb 6443 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6444 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6445 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6446 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6447 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6448 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6449 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6450 *bp = '\0';
3f7c398e 6451 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6452 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6453 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6454 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6455 }
6456 else
79072805 6457 {
6edd2cd5 6458 /*The big, slow, and stupid way. */
27da23d5 6459#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6460 STDCHAR *buf = NULL;
a02a5408 6461 Newx(buf, 8192, STDCHAR);
6edd2cd5 6462 assert(buf);
4d2c4e07 6463#else
6edd2cd5 6464 STDCHAR buf[8192];
4d2c4e07 6465#endif
79072805 6466
760ac839 6467screamer2:
c07a80fd 6468 if (rslen) {
00b6aa41 6469 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6470 bp = buf;
eb160463 6471 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6472 ; /* keep reading */
6473 cnt = bp - buf;
c07a80fd 6474 }
6475 else {
760ac839 6476 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6477 /* Accomodate broken VAXC compiler, which applies U8 cast to
6478 * both args of ?: operator, causing EOF to change into 255
6479 */
37be0adf 6480 if (cnt > 0)
cbe9e203
JH
6481 i = (U8)buf[cnt - 1];
6482 else
37be0adf 6483 i = EOF;
c07a80fd 6484 }
79072805 6485
cbe9e203
JH
6486 if (cnt < 0)
6487 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6488 if (append)
6489 sv_catpvn(sv, (char *) buf, cnt);
6490 else
6491 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6492
6493 if (i != EOF && /* joy */
6494 (!rslen ||
6495 SvCUR(sv) < rslen ||
3f7c398e 6496 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6497 {
6498 append = -1;
63e4d877
CS
6499 /*
6500 * If we're reading from a TTY and we get a short read,
6501 * indicating that the user hit his EOF character, we need
6502 * to notice it now, because if we try to read from the TTY
6503 * again, the EOF condition will disappear.
6504 *
6505 * The comparison of cnt to sizeof(buf) is an optimization
6506 * that prevents unnecessary calls to feof().
6507 *
6508 * - jik 9/25/96
6509 */
6510 if (!(cnt < sizeof(buf) && PerlIO_eof(fp)))
6511 goto screamer2;
79072805 6512 }
6edd2cd5 6513
27da23d5 6514#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6515 Safefree(buf);
6516#endif
79072805
LW
6517 }
6518
8bfdd7d9 6519 if (rspara) { /* have to do this both before and after */
c07a80fd 6520 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6521 i = PerlIO_getc(fp);
79072805 6522 if (i != '\n') {
760ac839 6523 PerlIO_ungetc(fp,i);
79072805
LW
6524 break;
6525 }
6526 }
6527 }
c07a80fd 6528
efd8b2ba 6529return_string_or_null:
bd61b366 6530 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6531}
6532
954c1994
GS
6533/*
6534=for apidoc sv_inc
6535
645c22ef
DM
6536Auto-increment of the value in the SV, doing string to numeric conversion
6537if necessary. Handles 'get' magic.
954c1994
GS
6538
6539=cut
6540*/
6541
79072805 6542void
864dbfa3 6543Perl_sv_inc(pTHX_ register SV *sv)
79072805 6544{
97aff369 6545 dVAR;
79072805 6546 register char *d;
463ee0b2 6547 int flags;
79072805
LW
6548
6549 if (!sv)
6550 return;
5b295bef 6551 SvGETMAGIC(sv);
ed6116ce 6552 if (SvTHINKFIRST(sv)) {
765f542d
NC
6553 if (SvIsCOW(sv))
6554 sv_force_normal_flags(sv, 0);
0f15f207 6555 if (SvREADONLY(sv)) {
923e4eb5 6556 if (IN_PERL_RUNTIME)
cea2e8a9 6557 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6558 }
a0d0e21e 6559 if (SvROK(sv)) {
b5be31e9 6560 IV i;
9e7bc3e8
JD
6561 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6562 return;
56431972 6563 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6564 sv_unref(sv);
6565 sv_setiv(sv, i);
a0d0e21e 6566 }
ed6116ce 6567 }
8990e307 6568 flags = SvFLAGS(sv);
28e5dec8
JH
6569 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6570 /* It's (privately or publicly) a float, but not tested as an
6571 integer, so test it to see. */
d460ef45 6572 (void) SvIV(sv);
28e5dec8
JH
6573 flags = SvFLAGS(sv);
6574 }
6575 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6576 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6577#ifdef PERL_PRESERVE_IVUV
28e5dec8 6578 oops_its_int:
59d8ce62 6579#endif
25da4f38
IZ
6580 if (SvIsUV(sv)) {
6581 if (SvUVX(sv) == UV_MAX)
a1e868e7 6582 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6583 else
6584 (void)SvIOK_only_UV(sv);
607fa7f2 6585 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6586 } else {
6587 if (SvIVX(sv) == IV_MAX)
28e5dec8 6588 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6589 else {
6590 (void)SvIOK_only(sv);
45977657 6591 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6592 }
55497cff 6593 }
79072805
LW
6594 return;
6595 }
28e5dec8
JH
6596 if (flags & SVp_NOK) {
6597 (void)SvNOK_only(sv);
9d6ce603 6598 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6599 return;
6600 }
6601
3f7c398e 6602 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6603 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6604 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6605 (void)SvIOK_only(sv);
45977657 6606 SvIV_set(sv, 1);
79072805
LW
6607 return;
6608 }
463ee0b2 6609 d = SvPVX(sv);
79072805
LW
6610 while (isALPHA(*d)) d++;
6611 while (isDIGIT(*d)) d++;
6612 if (*d) {
28e5dec8 6613#ifdef PERL_PRESERVE_IVUV
d1be9408 6614 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6615 warnings. Probably ought to make the sv_iv_please() that does
6616 the conversion if possible, and silently. */
504618e9 6617 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6618 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6619 /* Need to try really hard to see if it's an integer.
6620 9.22337203685478e+18 is an integer.
6621 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6622 so $a="9.22337203685478e+18"; $a+0; $a++
6623 needs to be the same as $a="9.22337203685478e+18"; $a++
6624 or we go insane. */
d460ef45 6625
28e5dec8
JH
6626 (void) sv_2iv(sv);
6627 if (SvIOK(sv))
6628 goto oops_its_int;
6629
6630 /* sv_2iv *should* have made this an NV */
6631 if (flags & SVp_NOK) {
6632 (void)SvNOK_only(sv);
9d6ce603 6633 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6634 return;
6635 }
6636 /* I don't think we can get here. Maybe I should assert this
6637 And if we do get here I suspect that sv_setnv will croak. NWC
6638 Fall through. */
6639#if defined(USE_LONG_DOUBLE)
6640 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 6641 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6642#else
1779d84d 6643 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 6644 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6645#endif
6646 }
6647#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6648 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6649 return;
6650 }
6651 d--;
3f7c398e 6652 while (d >= SvPVX_const(sv)) {
79072805
LW
6653 if (isDIGIT(*d)) {
6654 if (++*d <= '9')
6655 return;
6656 *(d--) = '0';
6657 }
6658 else {
9d116dd7
JH
6659#ifdef EBCDIC
6660 /* MKS: The original code here died if letters weren't consecutive.
6661 * at least it didn't have to worry about non-C locales. The
6662 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6663 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6664 * [A-Za-z] are accepted by isALPHA in the C locale.
6665 */
6666 if (*d != 'z' && *d != 'Z') {
6667 do { ++*d; } while (!isALPHA(*d));
6668 return;
6669 }
6670 *(d--) -= 'z' - 'a';
6671#else
79072805
LW
6672 ++*d;
6673 if (isALPHA(*d))
6674 return;
6675 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6676#endif
79072805
LW
6677 }
6678 }
6679 /* oh,oh, the number grew */
6680 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6681 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6682 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6683 *d = d[-1];
6684 if (isDIGIT(d[1]))
6685 *d = '1';
6686 else
6687 *d = d[1];
6688}
6689
954c1994
GS
6690/*
6691=for apidoc sv_dec
6692
645c22ef
DM
6693Auto-decrement of the value in the SV, doing string to numeric conversion
6694if necessary. Handles 'get' magic.
954c1994
GS
6695
6696=cut
6697*/
6698
79072805 6699void
864dbfa3 6700Perl_sv_dec(pTHX_ register SV *sv)
79072805 6701{
97aff369 6702 dVAR;
463ee0b2
LW
6703 int flags;
6704
79072805
LW
6705 if (!sv)
6706 return;
5b295bef 6707 SvGETMAGIC(sv);
ed6116ce 6708 if (SvTHINKFIRST(sv)) {
765f542d
NC
6709 if (SvIsCOW(sv))
6710 sv_force_normal_flags(sv, 0);
0f15f207 6711 if (SvREADONLY(sv)) {
923e4eb5 6712 if (IN_PERL_RUNTIME)
cea2e8a9 6713 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6714 }
a0d0e21e 6715 if (SvROK(sv)) {
b5be31e9 6716 IV i;
9e7bc3e8
JD
6717 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6718 return;
56431972 6719 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6720 sv_unref(sv);
6721 sv_setiv(sv, i);
a0d0e21e 6722 }
ed6116ce 6723 }
28e5dec8
JH
6724 /* Unlike sv_inc we don't have to worry about string-never-numbers
6725 and keeping them magic. But we mustn't warn on punting */
8990e307 6726 flags = SvFLAGS(sv);
28e5dec8
JH
6727 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6728 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6729#ifdef PERL_PRESERVE_IVUV
28e5dec8 6730 oops_its_int:
59d8ce62 6731#endif
25da4f38
IZ
6732 if (SvIsUV(sv)) {
6733 if (SvUVX(sv) == 0) {
6734 (void)SvIOK_only(sv);
45977657 6735 SvIV_set(sv, -1);
25da4f38
IZ
6736 }
6737 else {
6738 (void)SvIOK_only_UV(sv);
f4eee32f 6739 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6740 }
25da4f38
IZ
6741 } else {
6742 if (SvIVX(sv) == IV_MIN)
65202027 6743 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6744 else {
6745 (void)SvIOK_only(sv);
45977657 6746 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6747 }
55497cff 6748 }
6749 return;
6750 }
28e5dec8 6751 if (flags & SVp_NOK) {
9d6ce603 6752 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6753 (void)SvNOK_only(sv);
6754 return;
6755 }
8990e307 6756 if (!(flags & SVp_POK)) {
ef088171
NC
6757 if ((flags & SVTYPEMASK) < SVt_PVIV)
6758 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6759 SvIV_set(sv, -1);
6760 (void)SvIOK_only(sv);
79072805
LW
6761 return;
6762 }
28e5dec8
JH
6763#ifdef PERL_PRESERVE_IVUV
6764 {
504618e9 6765 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6766 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6767 /* Need to try really hard to see if it's an integer.
6768 9.22337203685478e+18 is an integer.
6769 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6770 so $a="9.22337203685478e+18"; $a+0; $a--
6771 needs to be the same as $a="9.22337203685478e+18"; $a--
6772 or we go insane. */
d460ef45 6773
28e5dec8
JH
6774 (void) sv_2iv(sv);
6775 if (SvIOK(sv))
6776 goto oops_its_int;
6777
6778 /* sv_2iv *should* have made this an NV */
6779 if (flags & SVp_NOK) {
6780 (void)SvNOK_only(sv);
9d6ce603 6781 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6782 return;
6783 }
6784 /* I don't think we can get here. Maybe I should assert this
6785 And if we do get here I suspect that sv_setnv will croak. NWC
6786 Fall through. */
6787#if defined(USE_LONG_DOUBLE)
6788 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 6789 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6790#else
1779d84d 6791 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 6792 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6793#endif
6794 }
6795 }
6796#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6797 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6798}
6799
954c1994
GS
6800/*
6801=for apidoc sv_mortalcopy
6802
645c22ef 6803Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6804The new SV is marked as mortal. It will be destroyed "soon", either by an
6805explicit call to FREETMPS, or by an implicit call at places such as
6806statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6807
6808=cut
6809*/
6810
79072805
LW
6811/* Make a string that will exist for the duration of the expression
6812 * evaluation. Actually, it may have to last longer than that, but
6813 * hopefully we won't free it until it has been assigned to a
6814 * permanent location. */
6815
6816SV *
864dbfa3 6817Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6818{
97aff369 6819 dVAR;
463ee0b2 6820 register SV *sv;
b881518d 6821
4561caa4 6822 new_SV(sv);
79072805 6823 sv_setsv(sv,oldstr);
677b06e3
GS
6824 EXTEND_MORTAL(1);
6825 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6826 SvTEMP_on(sv);
6827 return sv;
6828}
6829
954c1994
GS
6830/*
6831=for apidoc sv_newmortal
6832
645c22ef 6833Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6834set to 1. It will be destroyed "soon", either by an explicit call to
6835FREETMPS, or by an implicit call at places such as statement boundaries.
6836See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6837
6838=cut
6839*/
6840
8990e307 6841SV *
864dbfa3 6842Perl_sv_newmortal(pTHX)
8990e307 6843{
97aff369 6844 dVAR;
8990e307
LW
6845 register SV *sv;
6846
4561caa4 6847 new_SV(sv);
8990e307 6848 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6849 EXTEND_MORTAL(1);
6850 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6851 return sv;
6852}
6853
954c1994
GS
6854/*
6855=for apidoc sv_2mortal
6856
d4236ebc
DM
6857Marks an existing SV as mortal. The SV will be destroyed "soon", either
6858by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6859statement boundaries. SvTEMP() is turned on which means that the SV's
6860string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6861and C<sv_mortalcopy>.
954c1994
GS
6862
6863=cut
6864*/
6865
79072805 6866SV *
864dbfa3 6867Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6868{
27da23d5 6869 dVAR;
79072805 6870 if (!sv)
7a5b473e 6871 return NULL;
d689ffdd 6872 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6873 return sv;
677b06e3
GS
6874 EXTEND_MORTAL(1);
6875 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6876 SvTEMP_on(sv);
79072805
LW
6877 return sv;
6878}
6879
954c1994
GS
6880/*
6881=for apidoc newSVpv
6882
6883Creates a new SV and copies a string into it. The reference count for the
6884SV is set to 1. If C<len> is zero, Perl will compute the length using
6885strlen(). For efficiency, consider using C<newSVpvn> instead.
6886
6887=cut
6888*/
6889
79072805 6890SV *
864dbfa3 6891Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6892{
97aff369 6893 dVAR;
463ee0b2 6894 register SV *sv;
79072805 6895
4561caa4 6896 new_SV(sv);
616d8c9c 6897 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6898 return sv;
6899}
6900
954c1994
GS
6901/*
6902=for apidoc newSVpvn
6903
6904Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6905SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6906string. You are responsible for ensuring that the source string is at least
9e09f5f2 6907C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6908
6909=cut
6910*/
6911
9da1e3b5 6912SV *
864dbfa3 6913Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6914{
97aff369 6915 dVAR;
9da1e3b5
MUN
6916 register SV *sv;
6917
6918 new_SV(sv);
9da1e3b5
MUN
6919 sv_setpvn(sv,s,len);
6920 return sv;
6921}
6922
bd08039b
NC
6923
6924/*
926f8064 6925=for apidoc newSVhek
bd08039b
NC
6926
6927Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6928point to the shared string table where possible. Returns a new (undefined)
6929SV if the hek is NULL.
bd08039b
NC
6930
6931=cut
6932*/
6933
6934SV *
c1b02ed8 6935Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6936{
97aff369 6937 dVAR;
5aaec2b4
NC
6938 if (!hek) {
6939 SV *sv;
6940
6941 new_SV(sv);
6942 return sv;
6943 }
6944
bd08039b
NC
6945 if (HEK_LEN(hek) == HEf_SVKEY) {
6946 return newSVsv(*(SV**)HEK_KEY(hek));
6947 } else {
6948 const int flags = HEK_FLAGS(hek);
6949 if (flags & HVhek_WASUTF8) {
6950 /* Trouble :-)
6951 Andreas would like keys he put in as utf8 to come back as utf8
6952 */
6953 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6954 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6955 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6956
6957 SvUTF8_on (sv);
6958 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6959 return sv;
6960 } else if (flags & HVhek_REHASH) {
6961 /* We don't have a pointer to the hv, so we have to replicate the
6962 flag into every HEK. This hv is using custom a hasing
6963 algorithm. Hence we can't return a shared string scalar, as
6964 that would contain the (wrong) hash value, and might get passed
6965 into an hv routine with a regular hash */
6966
b64e5050 6967 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6968 if (HEK_UTF8(hek))
6969 SvUTF8_on (sv);
6970 return sv;
6971 }
6972 /* This will be overwhelminly the most common case. */
409dfe77
NC
6973 {
6974 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6975 more efficient than sharepvn(). */
6976 SV *sv;
6977
6978 new_SV(sv);
6979 sv_upgrade(sv, SVt_PV);
6980 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
6981 SvCUR_set(sv, HEK_LEN(hek));
6982 SvLEN_set(sv, 0);
6983 SvREADONLY_on(sv);
6984 SvFAKE_on(sv);
6985 SvPOK_on(sv);
6986 if (HEK_UTF8(hek))
6987 SvUTF8_on(sv);
6988 return sv;
6989 }
bd08039b
NC
6990 }
6991}
6992
1c846c1f
NIS
6993/*
6994=for apidoc newSVpvn_share
6995
3f7c398e 6996Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
6997table. If the string does not already exist in the table, it is created
6998first. Turns on READONLY and FAKE. The string's hash is stored in the UV
6999slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7000otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7001is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7002hash lookup will avoid string compare.
1c846c1f
NIS
7003
7004=cut
7005*/
7006
7007SV *
c3654f1a 7008Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7009{
97aff369 7010 dVAR;
1c846c1f 7011 register SV *sv;
c3654f1a 7012 bool is_utf8 = FALSE;
a51caccf
NC
7013 const char *const orig_src = src;
7014
c3654f1a 7015 if (len < 0) {
77caf834 7016 STRLEN tmplen = -len;
c3654f1a 7017 is_utf8 = TRUE;
75a54232 7018 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7019 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7020 len = tmplen;
7021 }
1c846c1f 7022 if (!hash)
5afd6d42 7023 PERL_HASH(hash, src, len);
1c846c1f 7024 new_SV(sv);
bdd68bc3 7025 sv_upgrade(sv, SVt_PV);
f880fe2f 7026 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7027 SvCUR_set(sv, len);
b162af07 7028 SvLEN_set(sv, 0);
1c846c1f
NIS
7029 SvREADONLY_on(sv);
7030 SvFAKE_on(sv);
7031 SvPOK_on(sv);
c3654f1a
IH
7032 if (is_utf8)
7033 SvUTF8_on(sv);
a51caccf
NC
7034 if (src != orig_src)
7035 Safefree(src);
1c846c1f
NIS
7036 return sv;
7037}
7038
645c22ef 7039
cea2e8a9 7040#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7041
7042/* pTHX_ magic can't cope with varargs, so this is a no-context
7043 * version of the main function, (which may itself be aliased to us).
7044 * Don't access this version directly.
7045 */
7046
46fc3d4c 7047SV *
cea2e8a9 7048Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7049{
cea2e8a9 7050 dTHX;
46fc3d4c 7051 register SV *sv;
7052 va_list args;
46fc3d4c 7053 va_start(args, pat);
c5be433b 7054 sv = vnewSVpvf(pat, &args);
46fc3d4c 7055 va_end(args);
7056 return sv;
7057}
cea2e8a9 7058#endif
46fc3d4c 7059
954c1994
GS
7060/*
7061=for apidoc newSVpvf
7062
645c22ef 7063Creates a new SV and initializes it with the string formatted like
954c1994
GS
7064C<sprintf>.
7065
7066=cut
7067*/
7068
cea2e8a9
GS
7069SV *
7070Perl_newSVpvf(pTHX_ const char* pat, ...)
7071{
7072 register SV *sv;
7073 va_list args;
cea2e8a9 7074 va_start(args, pat);
c5be433b 7075 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7076 va_end(args);
7077 return sv;
7078}
46fc3d4c 7079
645c22ef
DM
7080/* backend for newSVpvf() and newSVpvf_nocontext() */
7081
79072805 7082SV *
c5be433b
GS
7083Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7084{
97aff369 7085 dVAR;
c5be433b
GS
7086 register SV *sv;
7087 new_SV(sv);
4608196e 7088 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7089 return sv;
7090}
7091
954c1994
GS
7092/*
7093=for apidoc newSVnv
7094
7095Creates a new SV and copies a floating point value into it.
7096The reference count for the SV is set to 1.
7097
7098=cut
7099*/
7100
c5be433b 7101SV *
65202027 7102Perl_newSVnv(pTHX_ NV n)
79072805 7103{
97aff369 7104 dVAR;
463ee0b2 7105 register SV *sv;
79072805 7106
4561caa4 7107 new_SV(sv);
79072805
LW
7108 sv_setnv(sv,n);
7109 return sv;
7110}
7111
954c1994
GS
7112/*
7113=for apidoc newSViv
7114
7115Creates a new SV and copies an integer into it. The reference count for the
7116SV is set to 1.
7117
7118=cut
7119*/
7120
79072805 7121SV *
864dbfa3 7122Perl_newSViv(pTHX_ IV i)
79072805 7123{
97aff369 7124 dVAR;
463ee0b2 7125 register SV *sv;
79072805 7126
4561caa4 7127 new_SV(sv);
79072805
LW
7128 sv_setiv(sv,i);
7129 return sv;
7130}
7131
954c1994 7132/*
1a3327fb
JH
7133=for apidoc newSVuv
7134
7135Creates a new SV and copies an unsigned integer into it.
7136The reference count for the SV is set to 1.
7137
7138=cut
7139*/
7140
7141SV *
7142Perl_newSVuv(pTHX_ UV u)
7143{
97aff369 7144 dVAR;
1a3327fb
JH
7145 register SV *sv;
7146
7147 new_SV(sv);
7148 sv_setuv(sv,u);
7149 return sv;
7150}
7151
7152/*
954c1994
GS
7153=for apidoc newRV_noinc
7154
7155Creates an RV wrapper for an SV. The reference count for the original
7156SV is B<not> incremented.
7157
7158=cut
7159*/
7160
2304df62 7161SV *
864dbfa3 7162Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7163{
97aff369 7164 dVAR;
2304df62
AD
7165 register SV *sv;
7166
4561caa4 7167 new_SV(sv);
2304df62 7168 sv_upgrade(sv, SVt_RV);
76e3520e 7169 SvTEMP_off(tmpRef);
b162af07 7170 SvRV_set(sv, tmpRef);
2304df62 7171 SvROK_on(sv);
2304df62
AD
7172 return sv;
7173}
7174
ff276b08 7175/* newRV_inc is the official function name to use now.
645c22ef
DM
7176 * newRV_inc is in fact #defined to newRV in sv.h
7177 */
7178
5f05dabc 7179SV *
7f466ec7 7180Perl_newRV(pTHX_ SV *sv)
5f05dabc 7181{
97aff369 7182 dVAR;
7f466ec7 7183 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7184}
5f05dabc 7185
954c1994
GS
7186/*
7187=for apidoc newSVsv
7188
7189Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7190(Uses C<sv_setsv>).
954c1994
GS
7191
7192=cut
7193*/
7194
79072805 7195SV *
864dbfa3 7196Perl_newSVsv(pTHX_ register SV *old)
79072805 7197{
97aff369 7198 dVAR;
463ee0b2 7199 register SV *sv;
79072805
LW
7200
7201 if (!old)
7a5b473e 7202 return NULL;
8990e307 7203 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7204 if (ckWARN_d(WARN_INTERNAL))
9014280d 7205 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7206 return NULL;
79072805 7207 }
4561caa4 7208 new_SV(sv);
e90aabeb
NC
7209 /* SV_GMAGIC is the default for sv_setv()
7210 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7211 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7212 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7213 return sv;
79072805
LW
7214}
7215
645c22ef
DM
7216/*
7217=for apidoc sv_reset
7218
7219Underlying implementation for the C<reset> Perl function.
7220Note that the perl-level function is vaguely deprecated.
7221
7222=cut
7223*/
7224
79072805 7225void
e1ec3a88 7226Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7227{
27da23d5 7228 dVAR;
4802d5d7 7229 char todo[PERL_UCHAR_MAX+1];
79072805 7230
49d8d3a1
MB
7231 if (!stash)
7232 return;
7233
79072805 7234 if (!*s) { /* reset ?? searches */
aec46f14 7235 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7236 if (mg) {
7237 PMOP *pm = (PMOP *) mg->mg_obj;
7238 while (pm) {
7239 pm->op_pmdynflags &= ~PMdf_USED;
7240 pm = pm->op_pmnext;
7241 }
79072805
LW
7242 }
7243 return;
7244 }
7245
7246 /* reset variables */
7247
7248 if (!HvARRAY(stash))
7249 return;
463ee0b2
LW
7250
7251 Zero(todo, 256, char);
79072805 7252 while (*s) {
b464bac0
AL
7253 I32 max;
7254 I32 i = (unsigned char)*s;
79072805
LW
7255 if (s[1] == '-') {
7256 s += 2;
7257 }
4802d5d7 7258 max = (unsigned char)*s++;
79072805 7259 for ( ; i <= max; i++) {
463ee0b2
LW
7260 todo[i] = 1;
7261 }
a0d0e21e 7262 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7263 HE *entry;
79072805 7264 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7265 entry;
7266 entry = HeNEXT(entry))
7267 {
b464bac0
AL
7268 register GV *gv;
7269 register SV *sv;
7270
1edc1566 7271 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7272 continue;
1edc1566 7273 gv = (GV*)HeVAL(entry);
79072805 7274 sv = GvSV(gv);
e203899d
NC
7275 if (sv) {
7276 if (SvTHINKFIRST(sv)) {
7277 if (!SvREADONLY(sv) && SvROK(sv))
7278 sv_unref(sv);
7279 /* XXX Is this continue a bug? Why should THINKFIRST
7280 exempt us from resetting arrays and hashes? */
7281 continue;
7282 }
7283 SvOK_off(sv);
7284 if (SvTYPE(sv) >= SVt_PV) {
7285 SvCUR_set(sv, 0);
bd61b366 7286 if (SvPVX_const(sv) != NULL)
e203899d
NC
7287 *SvPVX(sv) = '\0';
7288 SvTAINT(sv);
7289 }
79072805
LW
7290 }
7291 if (GvAV(gv)) {
7292 av_clear(GvAV(gv));
7293 }
bfcb3514 7294 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7295#if defined(VMS)
7296 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7297#else /* ! VMS */
463ee0b2 7298 hv_clear(GvHV(gv));
b0269e46
AB
7299# if defined(USE_ENVIRON_ARRAY)
7300 if (gv == PL_envgv)
7301 my_clearenv();
7302# endif /* USE_ENVIRON_ARRAY */
7303#endif /* VMS */
79072805
LW
7304 }
7305 }
7306 }
7307 }
7308}
7309
645c22ef
DM
7310/*
7311=for apidoc sv_2io
7312
7313Using various gambits, try to get an IO from an SV: the IO slot if its a
7314GV; or the recursive result if we're an RV; or the IO slot of the symbol
7315named after the PV if we're a string.
7316
7317=cut
7318*/
7319
46fc3d4c 7320IO*
864dbfa3 7321Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7322{
7323 IO* io;
7324 GV* gv;
7325
7326 switch (SvTYPE(sv)) {
7327 case SVt_PVIO:
7328 io = (IO*)sv;
7329 break;
7330 case SVt_PVGV:
7331 gv = (GV*)sv;
7332 io = GvIO(gv);
7333 if (!io)
cea2e8a9 7334 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7335 break;
7336 default:
7337 if (!SvOK(sv))
cea2e8a9 7338 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7339 if (SvROK(sv))
7340 return sv_2io(SvRV(sv));
f776e3cd 7341 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7342 if (gv)
7343 io = GvIO(gv);
7344 else
7345 io = 0;
7346 if (!io)
35c1215d 7347 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7348 break;
7349 }
7350 return io;
7351}
7352
645c22ef
DM
7353/*
7354=for apidoc sv_2cv
7355
7356Using various gambits, try to get a CV from an SV; in addition, try if
7357possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7358The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7359
7360=cut
7361*/
7362
79072805 7363CV *
864dbfa3 7364Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7365{
27da23d5 7366 dVAR;
a0714e2c 7367 GV *gv = NULL;
601f1833 7368 CV *cv = NULL;
79072805 7369
85dec29a
NC
7370 if (!sv) {
7371 *st = NULL;
7372 *gvp = NULL;
7373 return NULL;
7374 }
79072805 7375 switch (SvTYPE(sv)) {
79072805
LW
7376 case SVt_PVCV:
7377 *st = CvSTASH(sv);
a0714e2c 7378 *gvp = NULL;
79072805
LW
7379 return (CV*)sv;
7380 case SVt_PVHV:
7381 case SVt_PVAV:
ef58ba18 7382 *st = NULL;
a0714e2c 7383 *gvp = NULL;
601f1833 7384 return NULL;
8990e307
LW
7385 case SVt_PVGV:
7386 gv = (GV*)sv;
a0d0e21e 7387 *gvp = gv;
8990e307
LW
7388 *st = GvESTASH(gv);
7389 goto fix_gv;
7390
79072805 7391 default:
5b295bef 7392 SvGETMAGIC(sv);
a0d0e21e 7393 if (SvROK(sv)) {
823a54a3 7394 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7395 tryAMAGICunDEREF(to_cv);
7396
62f274bf
GS
7397 sv = SvRV(sv);
7398 if (SvTYPE(sv) == SVt_PVCV) {
7399 cv = (CV*)sv;
a0714e2c 7400 *gvp = NULL;
62f274bf
GS
7401 *st = CvSTASH(cv);
7402 return cv;
7403 }
7404 else if(isGV(sv))
7405 gv = (GV*)sv;
7406 else
cea2e8a9 7407 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7408 }
62f274bf 7409 else if (isGV(sv))
79072805
LW
7410 gv = (GV*)sv;
7411 else
7a5fd60d 7412 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7413 *gvp = gv;
ef58ba18
NC
7414 if (!gv) {
7415 *st = NULL;
601f1833 7416 return NULL;
ef58ba18 7417 }
e26df76a
NC
7418 /* Some flags to gv_fetchsv mean don't really create the GV */
7419 if (SvTYPE(gv) != SVt_PVGV) {
7420 *st = NULL;
7421 return NULL;
7422 }
79072805 7423 *st = GvESTASH(gv);
8990e307 7424 fix_gv:
8ebc5c01 7425 if (lref && !GvCVu(gv)) {
4633a7c4 7426 SV *tmpsv;
748a9306 7427 ENTER;
561b68a9 7428 tmpsv = newSV(0);
bd61b366 7429 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7430 /* XXX this is probably not what they think they're getting.
7431 * It has the same effect as "sub name;", i.e. just a forward
7432 * declaration! */
774d564b 7433 newSUB(start_subparse(FALSE, 0),
4633a7c4 7434 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7435 NULL, NULL);
748a9306 7436 LEAVE;
8ebc5c01 7437 if (!GvCVu(gv))
35c1215d
NC
7438 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7439 sv);
8990e307 7440 }
8ebc5c01 7441 return GvCVu(gv);
79072805
LW
7442 }
7443}
7444
c461cf8f
JH
7445/*
7446=for apidoc sv_true
7447
7448Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7449Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7450instead use an in-line version.
c461cf8f
JH
7451
7452=cut
7453*/
7454
79072805 7455I32
864dbfa3 7456Perl_sv_true(pTHX_ register SV *sv)
79072805 7457{
8990e307
LW
7458 if (!sv)
7459 return 0;
79072805 7460 if (SvPOK(sv)) {
823a54a3
AL
7461 register const XPV* const tXpv = (XPV*)SvANY(sv);
7462 if (tXpv &&
c2f1de04 7463 (tXpv->xpv_cur > 1 ||
339049b0 7464 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7465 return 1;
7466 else
7467 return 0;
7468 }
7469 else {
7470 if (SvIOK(sv))
463ee0b2 7471 return SvIVX(sv) != 0;
79072805
LW
7472 else {
7473 if (SvNOK(sv))
463ee0b2 7474 return SvNVX(sv) != 0.0;
79072805 7475 else
463ee0b2 7476 return sv_2bool(sv);
79072805
LW
7477 }
7478 }
7479}
79072805 7480
645c22ef 7481/*
c461cf8f
JH
7482=for apidoc sv_pvn_force
7483
7484Get a sensible string out of the SV somehow.
645c22ef
DM
7485A private implementation of the C<SvPV_force> macro for compilers which
7486can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7487
8d6d96c1
HS
7488=for apidoc sv_pvn_force_flags
7489
7490Get a sensible string out of the SV somehow.
7491If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7492appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7493implemented in terms of this function.
645c22ef
DM
7494You normally want to use the various wrapper macros instead: see
7495C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7496
7497=cut
7498*/
7499
7500char *
7501Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7502{
97aff369 7503 dVAR;
6fc92669 7504 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7505 sv_force_normal_flags(sv, 0);
1c846c1f 7506
a0d0e21e 7507 if (SvPOK(sv)) {
13c5b33c
NC
7508 if (lp)
7509 *lp = SvCUR(sv);
a0d0e21e
LW
7510 }
7511 else {
a3b680e6 7512 char *s;
13c5b33c
NC
7513 STRLEN len;
7514
4d84ee25 7515 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7516 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7517 if (PL_op)
7518 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7519 ref, OP_NAME(PL_op));
4d84ee25 7520 else
b64e5050 7521 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7522 }
b64e5050 7523 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7524 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7525 OP_NAME(PL_op));
b64e5050 7526 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7527 if (lp)
7528 *lp = len;
7529
3f7c398e 7530 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7531 if (SvROK(sv))
7532 sv_unref(sv);
862a34c6 7533 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7534 SvGROW(sv, len + 1);
706aa1c9 7535 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7536 SvCUR_set(sv, len);
7537 *SvEND(sv) = '\0';
7538 }
7539 if (!SvPOK(sv)) {
7540 SvPOK_on(sv); /* validate pointer */
7541 SvTAINT(sv);
1d7c1841 7542 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7543 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7544 }
7545 }
4d84ee25 7546 return SvPVX_mutable(sv);
a0d0e21e
LW
7547}
7548
645c22ef 7549/*
645c22ef
DM
7550=for apidoc sv_pvbyten_force
7551
0feed65a 7552The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7553
7554=cut
7555*/
7556
7340a771
GS
7557char *
7558Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7559{
46ec2f14 7560 sv_pvn_force(sv,lp);
ffebcc3e 7561 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7562 *lp = SvCUR(sv);
7563 return SvPVX(sv);
7340a771
GS
7564}
7565
645c22ef 7566/*
c461cf8f
JH
7567=for apidoc sv_pvutf8n_force
7568
0feed65a 7569The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7570
7571=cut
7572*/
7573
7340a771
GS
7574char *
7575Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7576{
46ec2f14 7577 sv_pvn_force(sv,lp);
560a288e 7578 sv_utf8_upgrade(sv);
46ec2f14
TS
7579 *lp = SvCUR(sv);
7580 return SvPVX(sv);
7340a771
GS
7581}
7582
c461cf8f
JH
7583/*
7584=for apidoc sv_reftype
7585
7586Returns a string describing what the SV is a reference to.
7587
7588=cut
7589*/
7590
1cb0ed9b 7591char *
bfed75c6 7592Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7593{
07409e01
NC
7594 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7595 inside return suggests a const propagation bug in g++. */
c86bf373 7596 if (ob && SvOBJECT(sv)) {
1b6737cc 7597 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7598 return name ? name : (char *) "__ANON__";
c86bf373 7599 }
a0d0e21e
LW
7600 else {
7601 switch (SvTYPE(sv)) {
7602 case SVt_NULL:
7603 case SVt_IV:
7604 case SVt_NV:
7605 case SVt_RV:
7606 case SVt_PV:
7607 case SVt_PVIV:
7608 case SVt_PVNV:
7609 case SVt_PVMG:
7610 case SVt_PVBM:
1cb0ed9b 7611 if (SvVOK(sv))
439cb1c4 7612 return "VSTRING";
a0d0e21e
LW
7613 if (SvROK(sv))
7614 return "REF";
7615 else
7616 return "SCALAR";
1cb0ed9b 7617
07409e01 7618 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7619 /* tied lvalues should appear to be
7620 * scalars for backwards compatitbility */
7621 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7622 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7623 case SVt_PVAV: return "ARRAY";
7624 case SVt_PVHV: return "HASH";
7625 case SVt_PVCV: return "CODE";
7626 case SVt_PVGV: return "GLOB";
1d2dff63 7627 case SVt_PVFM: return "FORMAT";
27f9d8f3 7628 case SVt_PVIO: return "IO";
a0d0e21e
LW
7629 default: return "UNKNOWN";
7630 }
7631 }
7632}
7633
954c1994
GS
7634/*
7635=for apidoc sv_isobject
7636
7637Returns a boolean indicating whether the SV is an RV pointing to a blessed
7638object. If the SV is not an RV, or if the object is not blessed, then this
7639will return false.
7640
7641=cut
7642*/
7643
463ee0b2 7644int
864dbfa3 7645Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7646{
68dc0745 7647 if (!sv)
7648 return 0;
5b295bef 7649 SvGETMAGIC(sv);
85e6fe83
LW
7650 if (!SvROK(sv))
7651 return 0;
7652 sv = (SV*)SvRV(sv);
7653 if (!SvOBJECT(sv))
7654 return 0;
7655 return 1;
7656}
7657
954c1994
GS
7658/*
7659=for apidoc sv_isa
7660
7661Returns a boolean indicating whether the SV is blessed into the specified
7662class. This does not check for subtypes; use C<sv_derived_from> to verify
7663an inheritance relationship.
7664
7665=cut
7666*/
7667
85e6fe83 7668int
864dbfa3 7669Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7670{
bfcb3514 7671 const char *hvname;
68dc0745 7672 if (!sv)
7673 return 0;
5b295bef 7674 SvGETMAGIC(sv);
ed6116ce 7675 if (!SvROK(sv))
463ee0b2 7676 return 0;
ed6116ce
LW
7677 sv = (SV*)SvRV(sv);
7678 if (!SvOBJECT(sv))
463ee0b2 7679 return 0;
bfcb3514
NC
7680 hvname = HvNAME_get(SvSTASH(sv));
7681 if (!hvname)
e27ad1f2 7682 return 0;
463ee0b2 7683
bfcb3514 7684 return strEQ(hvname, name);
463ee0b2
LW
7685}
7686
954c1994
GS
7687/*
7688=for apidoc newSVrv
7689
7690Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7691it will be upgraded to one. If C<classname> is non-null then the new SV will
7692be blessed in the specified package. The new SV is returned and its
7693reference count is 1.
7694
7695=cut
7696*/
7697
463ee0b2 7698SV*
864dbfa3 7699Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7700{
97aff369 7701 dVAR;
463ee0b2
LW
7702 SV *sv;
7703
4561caa4 7704 new_SV(sv);
51cf62d8 7705
765f542d 7706 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7707 SvAMAGIC_off(rv);
51cf62d8 7708
0199fce9 7709 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7710 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7711 SvREFCNT(rv) = 0;
7712 sv_clear(rv);
7713 SvFLAGS(rv) = 0;
7714 SvREFCNT(rv) = refcnt;
0199fce9 7715
dc5494d2
NC
7716 sv_upgrade(rv, SVt_RV);
7717 } else if (SvROK(rv)) {
7718 SvREFCNT_dec(SvRV(rv));
7719 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7720 sv_upgrade(rv, SVt_RV);
7721 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7722 SvPV_free(rv);
0199fce9
JD
7723 SvCUR_set(rv, 0);
7724 SvLEN_set(rv, 0);
7725 }
51cf62d8 7726
0c34ef67 7727 SvOK_off(rv);
b162af07 7728 SvRV_set(rv, sv);
ed6116ce 7729 SvROK_on(rv);
463ee0b2 7730
a0d0e21e 7731 if (classname) {
1b6737cc 7732 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7733 (void)sv_bless(rv, stash);
7734 }
7735 return sv;
7736}
7737
954c1994
GS
7738/*
7739=for apidoc sv_setref_pv
7740
7741Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7742argument will be upgraded to an RV. That RV will be modified to point to
7743the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7744into the SV. The C<classname> argument indicates the package for the
bd61b366 7745blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7746will have a reference count of 1, and the RV will be returned.
954c1994
GS
7747
7748Do not use with other Perl types such as HV, AV, SV, CV, because those
7749objects will become corrupted by the pointer copy process.
7750
7751Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7752
7753=cut
7754*/
7755
a0d0e21e 7756SV*
864dbfa3 7757Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7758{
97aff369 7759 dVAR;
189b2af5 7760 if (!pv) {
3280af22 7761 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7762 SvSETMAGIC(rv);
7763 }
a0d0e21e 7764 else
56431972 7765 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7766 return rv;
7767}
7768
954c1994
GS
7769/*
7770=for apidoc sv_setref_iv
7771
7772Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7773argument will be upgraded to an RV. That RV will be modified to point to
7774the new SV. The C<classname> argument indicates the package for the
bd61b366 7775blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7776will have a reference count of 1, and the RV will be returned.
954c1994
GS
7777
7778=cut
7779*/
7780
a0d0e21e 7781SV*
864dbfa3 7782Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7783{
7784 sv_setiv(newSVrv(rv,classname), iv);
7785 return rv;
7786}
7787
954c1994 7788/*
e1c57cef
JH
7789=for apidoc sv_setref_uv
7790
7791Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7792argument will be upgraded to an RV. That RV will be modified to point to
7793the new SV. The C<classname> argument indicates the package for the
bd61b366 7794blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7795will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7796
7797=cut
7798*/
7799
7800SV*
7801Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7802{
7803 sv_setuv(newSVrv(rv,classname), uv);
7804 return rv;
7805}
7806
7807/*
954c1994
GS
7808=for apidoc sv_setref_nv
7809
7810Copies a double into a new SV, optionally blessing the SV. The C<rv>
7811argument will be upgraded to an RV. That RV will be modified to point to
7812the new SV. The C<classname> argument indicates the package for the
bd61b366 7813blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7814will have a reference count of 1, and the RV will be returned.
954c1994
GS
7815
7816=cut
7817*/
7818
a0d0e21e 7819SV*
65202027 7820Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7821{
7822 sv_setnv(newSVrv(rv,classname), nv);
7823 return rv;
7824}
463ee0b2 7825
954c1994
GS
7826/*
7827=for apidoc sv_setref_pvn
7828
7829Copies a string into a new SV, optionally blessing the SV. The length of the
7830string must be specified with C<n>. The C<rv> argument will be upgraded to
7831an RV. That RV will be modified to point to the new SV. The C<classname>
7832argument indicates the package for the blessing. Set C<classname> to
bd61b366 7833C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7834of 1, and the RV will be returned.
954c1994
GS
7835
7836Note that C<sv_setref_pv> copies the pointer while this copies the string.
7837
7838=cut
7839*/
7840
a0d0e21e 7841SV*
1b6737cc 7842Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7843{
7844 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7845 return rv;
7846}
7847
954c1994
GS
7848/*
7849=for apidoc sv_bless
7850
7851Blesses an SV into a specified package. The SV must be an RV. The package
7852must be designated by its stash (see C<gv_stashpv()>). The reference count
7853of the SV is unaffected.
7854
7855=cut
7856*/
7857
a0d0e21e 7858SV*
864dbfa3 7859Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7860{
97aff369 7861 dVAR;
76e3520e 7862 SV *tmpRef;
a0d0e21e 7863 if (!SvROK(sv))
cea2e8a9 7864 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7865 tmpRef = SvRV(sv);
7866 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7867 if (SvREADONLY(tmpRef))
cea2e8a9 7868 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7869 if (SvOBJECT(tmpRef)) {
7870 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7871 --PL_sv_objcount;
76e3520e 7872 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7873 }
a0d0e21e 7874 }
76e3520e
GS
7875 SvOBJECT_on(tmpRef);
7876 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7877 ++PL_sv_objcount;
862a34c6 7878 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7879 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7880
2e3febc6
CS
7881 if (Gv_AMG(stash))
7882 SvAMAGIC_on(sv);
7883 else
7884 SvAMAGIC_off(sv);
a0d0e21e 7885
1edbfb88
AB
7886 if(SvSMAGICAL(tmpRef))
7887 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7888 mg_set(tmpRef);
7889
7890
ecdeb87c 7891
a0d0e21e
LW
7892 return sv;
7893}
7894
645c22ef 7895/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7896 */
7897
76e3520e 7898STATIC void
cea2e8a9 7899S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7900{
97aff369 7901 dVAR;
850fabdf 7902 void *xpvmg;
b37c2d43 7903 SV * const temp = sv_newmortal();
850fabdf 7904
a0d0e21e
LW
7905 assert(SvTYPE(sv) == SVt_PVGV);
7906 SvFAKE_off(sv);
180488f8
NC
7907 gv_efullname3(temp, (GV *) sv, "*");
7908
f7877b28 7909 if (GvGP(sv)) {
1edc1566 7910 gp_free((GV*)sv);
f7877b28 7911 }
e826b3c7 7912 if (GvSTASH(sv)) {
e15faf7d 7913 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7914 GvSTASH(sv) = NULL;
e826b3c7 7915 }
a5f75d66 7916 GvMULTI_off(sv);
acda4c6a
NC
7917 if (GvNAME_HEK(sv)) {
7918 unshare_hek(GvNAME_HEK(sv));
7919 }
dedf8e73 7920 SvSCREAM_off(sv);
850fabdf
GS
7921
7922 /* need to keep SvANY(sv) in the right arena */
7923 xpvmg = new_XPVMG();
7924 StructCopy(SvANY(sv), xpvmg, XPVMG);
7925 del_XPVGV(SvANY(sv));
7926 SvANY(sv) = xpvmg;
7927
a0d0e21e
LW
7928 SvFLAGS(sv) &= ~SVTYPEMASK;
7929 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7930
7931 /* Intentionally not calling any local SET magic, as this isn't so much a
7932 set operation as merely an internal storage change. */
7933 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7934}
7935
954c1994 7936/*
840a7b70 7937=for apidoc sv_unref_flags
954c1994
GS
7938
7939Unsets the RV status of the SV, and decrements the reference count of
7940whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7941as a reversal of C<newSVrv>. The C<cflags> argument can contain
7942C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7943(otherwise the decrementing is conditional on the reference count being
7944different from one or the reference being a readonly SV).
7889fe52 7945See C<SvROK_off>.
954c1994
GS
7946
7947=cut
7948*/
7949
ed6116ce 7950void
e15faf7d 7951Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7952{
b64e5050 7953 SV* const target = SvRV(ref);
810b8aa5 7954
e15faf7d
NC
7955 if (SvWEAKREF(ref)) {
7956 sv_del_backref(target, ref);
7957 SvWEAKREF_off(ref);
7958 SvRV_set(ref, NULL);
810b8aa5
GS
7959 return;
7960 }
e15faf7d
NC
7961 SvRV_set(ref, NULL);
7962 SvROK_off(ref);
7963 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7964 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7965 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7966 SvREFCNT_dec(target);
840a7b70 7967 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7968 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7969}
8990e307 7970
840a7b70 7971/*
645c22ef
DM
7972=for apidoc sv_untaint
7973
7974Untaint an SV. Use C<SvTAINTED_off> instead.
7975=cut
7976*/
7977
bbce6d69 7978void
864dbfa3 7979Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7980{
13f57bf8 7981 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 7982 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 7983 if (mg)
565764a8 7984 mg->mg_len &= ~1;
36477c24 7985 }
bbce6d69 7986}
7987
645c22ef
DM
7988/*
7989=for apidoc sv_tainted
7990
7991Test an SV for taintedness. Use C<SvTAINTED> instead.
7992=cut
7993*/
7994
bbce6d69 7995bool
864dbfa3 7996Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 7997{
13f57bf8 7998 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 7999 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8000 if (mg && (mg->mg_len & 1) )
36477c24 8001 return TRUE;
8002 }
8003 return FALSE;
bbce6d69 8004}
8005
09540bc3
JH
8006/*
8007=for apidoc sv_setpviv
8008
8009Copies an integer into the given SV, also updating its string value.
8010Does not handle 'set' magic. See C<sv_setpviv_mg>.
8011
8012=cut
8013*/
8014
8015void
8016Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8017{
8018 char buf[TYPE_CHARS(UV)];
8019 char *ebuf;
b64e5050 8020 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8021
8022 sv_setpvn(sv, ptr, ebuf - ptr);
8023}
8024
8025/*
8026=for apidoc sv_setpviv_mg
8027
8028Like C<sv_setpviv>, but also handles 'set' magic.
8029
8030=cut
8031*/
8032
8033void
8034Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8035{
df7eb254 8036 sv_setpviv(sv, iv);
09540bc3
JH
8037 SvSETMAGIC(sv);
8038}
8039
cea2e8a9 8040#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8041
8042/* pTHX_ magic can't cope with varargs, so this is a no-context
8043 * version of the main function, (which may itself be aliased to us).
8044 * Don't access this version directly.
8045 */
8046
cea2e8a9
GS
8047void
8048Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8049{
8050 dTHX;
8051 va_list args;
8052 va_start(args, pat);
c5be433b 8053 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8054 va_end(args);
8055}
8056
645c22ef
DM
8057/* pTHX_ magic can't cope with varargs, so this is a no-context
8058 * version of the main function, (which may itself be aliased to us).
8059 * Don't access this version directly.
8060 */
cea2e8a9
GS
8061
8062void
8063Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8064{
8065 dTHX;
8066 va_list args;
8067 va_start(args, pat);
c5be433b 8068 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8069 va_end(args);
cea2e8a9
GS
8070}
8071#endif
8072
954c1994
GS
8073/*
8074=for apidoc sv_setpvf
8075
bffc3d17
SH
8076Works like C<sv_catpvf> but copies the text into the SV instead of
8077appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8078
8079=cut
8080*/
8081
46fc3d4c 8082void
864dbfa3 8083Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8084{
8085 va_list args;
46fc3d4c 8086 va_start(args, pat);
c5be433b 8087 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8088 va_end(args);
8089}
8090
bffc3d17
SH
8091/*
8092=for apidoc sv_vsetpvf
8093
8094Works like C<sv_vcatpvf> but copies the text into the SV instead of
8095appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8096
8097Usually used via its frontend C<sv_setpvf>.
8098
8099=cut
8100*/
645c22ef 8101
c5be433b
GS
8102void
8103Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8104{
4608196e 8105 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8106}
ef50df4b 8107
954c1994
GS
8108/*
8109=for apidoc sv_setpvf_mg
8110
8111Like C<sv_setpvf>, but also handles 'set' magic.
8112
8113=cut
8114*/
8115
ef50df4b 8116void
864dbfa3 8117Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8118{
8119 va_list args;
ef50df4b 8120 va_start(args, pat);
c5be433b 8121 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8122 va_end(args);
c5be433b
GS
8123}
8124
bffc3d17
SH
8125/*
8126=for apidoc sv_vsetpvf_mg
8127
8128Like C<sv_vsetpvf>, but also handles 'set' magic.
8129
8130Usually used via its frontend C<sv_setpvf_mg>.
8131
8132=cut
8133*/
645c22ef 8134
c5be433b
GS
8135void
8136Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8137{
4608196e 8138 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8139 SvSETMAGIC(sv);
8140}
8141
cea2e8a9 8142#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8143
8144/* pTHX_ magic can't cope with varargs, so this is a no-context
8145 * version of the main function, (which may itself be aliased to us).
8146 * Don't access this version directly.
8147 */
8148
cea2e8a9
GS
8149void
8150Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8151{
8152 dTHX;
8153 va_list args;
8154 va_start(args, pat);
c5be433b 8155 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8156 va_end(args);
8157}
8158
645c22ef
DM
8159/* pTHX_ magic can't cope with varargs, so this is a no-context
8160 * version of the main function, (which may itself be aliased to us).
8161 * Don't access this version directly.
8162 */
8163
cea2e8a9
GS
8164void
8165Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8166{
8167 dTHX;
8168 va_list args;
8169 va_start(args, pat);
c5be433b 8170 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8171 va_end(args);
cea2e8a9
GS
8172}
8173#endif
8174
954c1994
GS
8175/*
8176=for apidoc sv_catpvf
8177
d5ce4a7c
GA
8178Processes its arguments like C<sprintf> and appends the formatted
8179output to an SV. If the appended data contains "wide" characters
8180(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8181and characters >255 formatted with %c), the original SV might get
bffc3d17 8182upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8183C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8184valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8185
d5ce4a7c 8186=cut */
954c1994 8187
46fc3d4c 8188void
864dbfa3 8189Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8190{
8191 va_list args;
46fc3d4c 8192 va_start(args, pat);
c5be433b 8193 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8194 va_end(args);
8195}
8196
bffc3d17
SH
8197/*
8198=for apidoc sv_vcatpvf
8199
8200Processes its arguments like C<vsprintf> and appends the formatted output
8201to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8202
8203Usually used via its frontend C<sv_catpvf>.
8204
8205=cut
8206*/
645c22ef 8207
ef50df4b 8208void
c5be433b
GS
8209Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8210{
4608196e 8211 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8212}
8213
954c1994
GS
8214/*
8215=for apidoc sv_catpvf_mg
8216
8217Like C<sv_catpvf>, but also handles 'set' magic.
8218
8219=cut
8220*/
8221
c5be433b 8222void
864dbfa3 8223Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8224{
8225 va_list args;
ef50df4b 8226 va_start(args, pat);
c5be433b 8227 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8228 va_end(args);
c5be433b
GS
8229}
8230
bffc3d17
SH
8231/*
8232=for apidoc sv_vcatpvf_mg
8233
8234Like C<sv_vcatpvf>, but also handles 'set' magic.
8235
8236Usually used via its frontend C<sv_catpvf_mg>.
8237
8238=cut
8239*/
645c22ef 8240
c5be433b
GS
8241void
8242Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8243{
4608196e 8244 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8245 SvSETMAGIC(sv);
8246}
8247
954c1994
GS
8248/*
8249=for apidoc sv_vsetpvfn
8250
bffc3d17 8251Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8252appending it.
8253
bffc3d17 8254Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8255
954c1994
GS
8256=cut
8257*/
8258
46fc3d4c 8259void
7d5ea4e7 8260Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8261{
8262 sv_setpvn(sv, "", 0);
7d5ea4e7 8263 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8264}
8265
2d00ba3b 8266STATIC I32
9dd79c3f 8267S_expect_number(pTHX_ char** pattern)
211dfcf1 8268{
97aff369 8269 dVAR;
211dfcf1
HS
8270 I32 var = 0;
8271 switch (**pattern) {
8272 case '1': case '2': case '3':
8273 case '4': case '5': case '6':
8274 case '7': case '8': case '9':
2fba7546
GA
8275 var = *(*pattern)++ - '0';
8276 while (isDIGIT(**pattern)) {
5f66b61c 8277 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8278 if (tmp < var)
8279 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8280 var = tmp;
8281 }
211dfcf1
HS
8282 }
8283 return var;
8284}
211dfcf1 8285
c445ea15
AL
8286STATIC char *
8287S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8288{
a3b680e6 8289 const int neg = nv < 0;
4151a5fe 8290 UV uv;
4151a5fe
IZ
8291
8292 if (neg)
8293 nv = -nv;
8294 if (nv < UV_MAX) {
b464bac0 8295 char *p = endbuf;
4151a5fe 8296 nv += 0.5;
028f8eaa 8297 uv = (UV)nv;
4151a5fe
IZ
8298 if (uv & 1 && uv == nv)
8299 uv--; /* Round to even */
8300 do {
a3b680e6 8301 const unsigned dig = uv % 10;
4151a5fe
IZ
8302 *--p = '0' + dig;
8303 } while (uv /= 10);
8304 if (neg)
8305 *--p = '-';
8306 *len = endbuf - p;
8307 return p;
8308 }
bd61b366 8309 return NULL;
4151a5fe
IZ
8310}
8311
8312
954c1994
GS
8313/*
8314=for apidoc sv_vcatpvfn
8315
8316Processes its arguments like C<vsprintf> and appends the formatted output
8317to an SV. Uses an array of SVs if the C style variable argument list is
8318missing (NULL). When running with taint checks enabled, indicates via
8319C<maybe_tainted> if results are untrustworthy (often due to the use of
8320locales).
8321
bffc3d17 8322Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8323
954c1994
GS
8324=cut
8325*/
8326
8896765a
RB
8327
8328#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8329 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8330 vec_utf8 = DO_UTF8(vecsv);
8331
1ef29b0e
RGS
8332/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8333
46fc3d4c 8334void
7d5ea4e7 8335Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8336{
97aff369 8337 dVAR;
46fc3d4c 8338 char *p;
8339 char *q;
a3b680e6 8340 const char *patend;
fc36a67e 8341 STRLEN origlen;
46fc3d4c 8342 I32 svix = 0;
27da23d5 8343 static const char nullstr[] = "(null)";
a0714e2c 8344 SV *argsv = NULL;
b464bac0
AL
8345 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8346 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8347 SV *nsv = NULL;
4151a5fe
IZ
8348 /* Times 4: a decimal digit takes more than 3 binary digits.
8349 * NV_DIG: mantissa takes than many decimal digits.
8350 * Plus 32: Playing safe. */
8351 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8352 /* large enough for "%#.#f" --chip */
8353 /* what about long double NVs? --jhi */
db79b45b 8354
53c1dcc0
AL
8355 PERL_UNUSED_ARG(maybe_tainted);
8356
46fc3d4c 8357 /* no matter what, this is a string now */
fc36a67e 8358 (void)SvPV_force(sv, origlen);
46fc3d4c 8359
8896765a 8360 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8361 if (patlen == 0)
8362 return;
0dbb1585 8363 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8364 if (args) {
8365 const char * const s = va_arg(*args, char*);
8366 sv_catpv(sv, s ? s : nullstr);
8367 }
8368 else if (svix < svmax) {
8369 sv_catsv(sv, *svargs);
2d03de9c
AL
8370 }
8371 return;
0dbb1585 8372 }
8896765a
RB
8373 if (args && patlen == 3 && pat[0] == '%' &&
8374 pat[1] == '-' && pat[2] == 'p') {
8375 argsv = va_arg(*args, SV*);
8376 sv_catsv(sv, argsv);
8896765a 8377 return;
46fc3d4c 8378 }
8379
1d917b39 8380#ifndef USE_LONG_DOUBLE
4151a5fe 8381 /* special-case "%.<number>[gf]" */
7af36d83 8382 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8383 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8384 unsigned digits = 0;
8385 const char *pp;
8386
8387 pp = pat + 2;
8388 while (*pp >= '0' && *pp <= '9')
8389 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8390 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8391 NV nv;
8392
7af36d83 8393 if (svix < svmax)
4151a5fe
IZ
8394 nv = SvNV(*svargs);
8395 else
8396 return;
8397 if (*pp == 'g') {
2873255c
NC
8398 /* Add check for digits != 0 because it seems that some
8399 gconverts are buggy in this case, and we don't yet have
8400 a Configure test for this. */
8401 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8402 /* 0, point, slack */
2e59c212 8403 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8404 sv_catpv(sv, ebuf);
8405 if (*ebuf) /* May return an empty string for digits==0 */
8406 return;
8407 }
8408 } else if (!digits) {
8409 STRLEN l;
8410
8411 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8412 sv_catpvn(sv, p, l);
8413 return;
8414 }
8415 }
8416 }
8417 }
1d917b39 8418#endif /* !USE_LONG_DOUBLE */
4151a5fe 8419
2cf2cfc6 8420 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8421 has_utf8 = TRUE;
2cf2cfc6 8422
46fc3d4c 8423 patend = (char*)pat + patlen;
8424 for (p = (char*)pat; p < patend; p = q) {
8425 bool alt = FALSE;
8426 bool left = FALSE;
b22c7a20 8427 bool vectorize = FALSE;
211dfcf1 8428 bool vectorarg = FALSE;
2cf2cfc6 8429 bool vec_utf8 = FALSE;
46fc3d4c 8430 char fill = ' ';
8431 char plus = 0;
8432 char intsize = 0;
8433 STRLEN width = 0;
fc36a67e 8434 STRLEN zeros = 0;
46fc3d4c 8435 bool has_precis = FALSE;
8436 STRLEN precis = 0;
c445ea15 8437 const I32 osvix = svix;
2cf2cfc6 8438 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8439#ifdef HAS_LDBL_SPRINTF_BUG
8440 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8441 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8442 bool fix_ldbl_sprintf_bug = FALSE;
8443#endif
205f51d8 8444
46fc3d4c 8445 char esignbuf[4];
89ebb4a3 8446 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8447 STRLEN esignlen = 0;
8448
bd61b366 8449 const char *eptr = NULL;
fc36a67e 8450 STRLEN elen = 0;
a0714e2c 8451 SV *vecsv = NULL;
4608196e 8452 const U8 *vecstr = NULL;
b22c7a20 8453 STRLEN veclen = 0;
934abaf1 8454 char c = 0;
46fc3d4c 8455 int i;
9c5ffd7c 8456 unsigned base = 0;
8c8eb53c
RB
8457 IV iv = 0;
8458 UV uv = 0;
9e5b023a
JH
8459 /* we need a long double target in case HAS_LONG_DOUBLE but
8460 not USE_LONG_DOUBLE
8461 */
35fff930 8462#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8463 long double nv;
8464#else
65202027 8465 NV nv;
9e5b023a 8466#endif
46fc3d4c 8467 STRLEN have;
8468 STRLEN need;
8469 STRLEN gap;
7af36d83 8470 const char *dotstr = ".";
b22c7a20 8471 STRLEN dotstrlen = 1;
211dfcf1 8472 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8473 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8474 I32 epix = 0; /* explicit precision index */
8475 I32 evix = 0; /* explicit vector index */
eb3fce90 8476 bool asterisk = FALSE;
46fc3d4c 8477
211dfcf1 8478 /* echo everything up to the next format specification */
46fc3d4c 8479 for (q = p; q < patend && *q != '%'; ++q) ;
8480 if (q > p) {
db79b45b
JH
8481 if (has_utf8 && !pat_utf8)
8482 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8483 else
8484 sv_catpvn(sv, p, q - p);
46fc3d4c 8485 p = q;
8486 }
8487 if (q++ >= patend)
8488 break;
8489
211dfcf1
HS
8490/*
8491 We allow format specification elements in this order:
8492 \d+\$ explicit format parameter index
8493 [-+ 0#]+ flags
a472f209 8494 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8495 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8496 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8497 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8498 [hlqLV] size
8896765a
RB
8499 [%bcdefginopsuxDFOUX] format (mandatory)
8500*/
8501
8502 if (args) {
8503/*
8504 As of perl5.9.3, printf format checking is on by default.
8505 Internally, perl uses %p formats to provide an escape to
8506 some extended formatting. This block deals with those
8507 extensions: if it does not match, (char*)q is reset and
8508 the normal format processing code is used.
8509
8510 Currently defined extensions are:
8511 %p include pointer address (standard)
8512 %-p (SVf) include an SV (previously %_)
8513 %-<num>p include an SV with precision <num>
8514 %1p (VDf) include a v-string (as %vd)
8515 %<num>p reserved for future extensions
8516
8517 Robin Barker 2005-07-14
211dfcf1 8518*/
8896765a
RB
8519 char* r = q;
8520 bool sv = FALSE;
8521 STRLEN n = 0;
8522 if (*q == '-')
8523 sv = *q++;
c445ea15 8524 n = expect_number(&q);
8896765a
RB
8525 if (*q++ == 'p') {
8526 if (sv) { /* SVf */
8527 if (n) {
8528 precis = n;
8529 has_precis = TRUE;
8530 }
8531 argsv = va_arg(*args, SV*);
8532 eptr = SvPVx_const(argsv, elen);
8533 if (DO_UTF8(argsv))
8534 is_utf8 = TRUE;
8535 goto string;
8536 }
8537#if vdNUMBER
8538 else if (n == vdNUMBER) { /* VDf */
8539 vectorize = TRUE;
8540 VECTORIZE_ARGS
8541 goto format_vd;
8542 }
8543#endif
8544 else if (n) {
8545 if (ckWARN_d(WARN_INTERNAL))
8546 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8547 "internal %%<num>p might conflict with future printf extensions");
8548 }
8549 }
8550 q = r;
8551 }
8552
c445ea15 8553 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8554 if (*q == '$') {
8555 ++q;
8556 efix = width;
8557 } else {
8558 goto gotwidth;
8559 }
8560 }
8561
fc36a67e 8562 /* FLAGS */
8563
46fc3d4c 8564 while (*q) {
8565 switch (*q) {
8566 case ' ':
8567 case '+':
8568 plus = *q++;
8569 continue;
8570
8571 case '-':
8572 left = TRUE;
8573 q++;
8574 continue;
8575
8576 case '0':
8577 fill = *q++;
8578 continue;
8579
8580 case '#':
8581 alt = TRUE;
8582 q++;
8583 continue;
8584
fc36a67e 8585 default:
8586 break;
8587 }
8588 break;
8589 }
46fc3d4c 8590
211dfcf1 8591 tryasterisk:
eb3fce90 8592 if (*q == '*') {
211dfcf1 8593 q++;
c445ea15 8594 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8595 if (*q++ != '$')
8596 goto unknown;
eb3fce90 8597 asterisk = TRUE;
211dfcf1
HS
8598 }
8599 if (*q == 'v') {
eb3fce90 8600 q++;
211dfcf1
HS
8601 if (vectorize)
8602 goto unknown;
9cbac4c7 8603 if ((vectorarg = asterisk)) {
211dfcf1
HS
8604 evix = ewix;
8605 ewix = 0;
8606 asterisk = FALSE;
8607 }
8608 vectorize = TRUE;
8609 goto tryasterisk;
eb3fce90
JH
8610 }
8611
211dfcf1 8612 if (!asterisk)
858a90f9 8613 {
7a5fa8a2 8614 if( *q == '0' )
f3583277 8615 fill = *q++;
c445ea15 8616 width = expect_number(&q);
858a90f9 8617 }
211dfcf1
HS
8618
8619 if (vectorize) {
8620 if (vectorarg) {
8621 if (args)
8622 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8623 else if (evix) {
8624 vecsv = (evix > 0 && evix <= svmax)
8625 ? svargs[evix-1] : &PL_sv_undef;
8626 } else {
8627 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8628 }
245d4a47 8629 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8630 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8631 bad with tied or overloaded values that return UTF8. */
211dfcf1 8632 if (DO_UTF8(vecsv))
2cf2cfc6 8633 is_utf8 = TRUE;
640283f5
NC
8634 else if (has_utf8) {
8635 vecsv = sv_mortalcopy(vecsv);
8636 sv_utf8_upgrade(vecsv);
8637 dotstr = SvPV_const(vecsv, dotstrlen);
8638 is_utf8 = TRUE;
8639 }
211dfcf1
HS
8640 }
8641 if (args) {
8896765a 8642 VECTORIZE_ARGS
eb3fce90 8643 }
7ad96abb 8644 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8645 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8646 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8647 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8648
8649 /* if this is a version object, we need to convert
8650 * back into v-string notation and then let the
8651 * vectorize happen normally
d7aa5382 8652 */
96b8f7ce
JP
8653 if (sv_derived_from(vecsv, "version")) {
8654 char *version = savesvpv(vecsv);
34ba6322
SP
8655 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8656 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8657 "vector argument not supported with alpha versions");
8658 goto unknown;
8659 }
96b8f7ce
JP
8660 vecsv = sv_newmortal();
8661 /* scan_vstring is expected to be called during
8662 * tokenization, so we need to fake up the end
8663 * of the buffer for it
8664 */
8665 PL_bufend = version + veclen;
8666 scan_vstring(version, vecsv);
8667 vecstr = (U8*)SvPV_const(vecsv, veclen);
8668 vec_utf8 = DO_UTF8(vecsv);
8669 Safefree(version);
d7aa5382 8670 }
211dfcf1
HS
8671 }
8672 else {
8673 vecstr = (U8*)"";
8674 veclen = 0;
8675 }
eb3fce90 8676 }
fc36a67e 8677
eb3fce90 8678 if (asterisk) {
fc36a67e 8679 if (args)
8680 i = va_arg(*args, int);
8681 else
eb3fce90
JH
8682 i = (ewix ? ewix <= svmax : svix < svmax) ?
8683 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8684 left |= (i < 0);
8685 width = (i < 0) ? -i : i;
fc36a67e 8686 }
211dfcf1 8687 gotwidth:
fc36a67e 8688
8689 /* PRECISION */
46fc3d4c 8690
fc36a67e 8691 if (*q == '.') {
8692 q++;
8693 if (*q == '*') {
211dfcf1 8694 q++;
c445ea15 8695 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8696 goto unknown;
8697 /* XXX: todo, support specified precision parameter */
8698 if (epix)
211dfcf1 8699 goto unknown;
46fc3d4c 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 precis = (i < 0) ? 0 : i;
fc36a67e 8706 }
8707 else {
8708 precis = 0;
8709 while (isDIGIT(*q))
8710 precis = precis * 10 + (*q++ - '0');
8711 }
8712 has_precis = TRUE;
8713 }
46fc3d4c 8714
fc36a67e 8715 /* SIZE */
46fc3d4c 8716
fc36a67e 8717 switch (*q) {
c623ac67
GS
8718#ifdef WIN32
8719 case 'I': /* Ix, I32x, and I64x */
8720# ifdef WIN64
8721 if (q[1] == '6' && q[2] == '4') {
8722 q += 3;
8723 intsize = 'q';
8724 break;
8725 }
8726# endif
8727 if (q[1] == '3' && q[2] == '2') {
8728 q += 3;
8729 break;
8730 }
8731# ifdef WIN64
8732 intsize = 'q';
8733# endif
8734 q++;
8735 break;
8736#endif
9e5b023a 8737#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8738 case 'L': /* Ld */
5f66b61c 8739 /*FALLTHROUGH*/
e5c81feb 8740#ifdef HAS_QUAD
6f9bb7fd 8741 case 'q': /* qd */
9e5b023a 8742#endif
6f9bb7fd
GS
8743 intsize = 'q';
8744 q++;
8745 break;
8746#endif
fc36a67e 8747 case 'l':
9e5b023a 8748#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8749 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8750 intsize = 'q';
8751 q += 2;
46fc3d4c 8752 break;
cf2093f6 8753 }
fc36a67e 8754#endif
5f66b61c 8755 /*FALLTHROUGH*/
fc36a67e 8756 case 'h':
5f66b61c 8757 /*FALLTHROUGH*/
fc36a67e 8758 case 'V':
8759 intsize = *q++;
46fc3d4c 8760 break;
8761 }
8762
fc36a67e 8763 /* CONVERSION */
8764
211dfcf1
HS
8765 if (*q == '%') {
8766 eptr = q++;
8767 elen = 1;
26372e71
GA
8768 if (vectorize) {
8769 c = '%';
8770 goto unknown;
8771 }
211dfcf1
HS
8772 goto string;
8773 }
8774
26372e71 8775 if (!vectorize && !args) {
86c51f8b
NC
8776 if (efix) {
8777 const I32 i = efix-1;
8778 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8779 } else {
8780 argsv = (svix >= 0 && svix < svmax)
8781 ? svargs[svix++] : &PL_sv_undef;
8782 }
863811b2 8783 }
211dfcf1 8784
46fc3d4c 8785 switch (c = *q++) {
8786
8787 /* STRINGS */
8788
46fc3d4c 8789 case 'c':
26372e71
GA
8790 if (vectorize)
8791 goto unknown;
8792 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8793 if ((uv > 255 ||
8794 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8795 && !IN_BYTES) {
dfe13c55 8796 eptr = (char*)utf8buf;
9041c2e3 8797 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8798 is_utf8 = TRUE;
7e2040f0
GS
8799 }
8800 else {
8801 c = (char)uv;
8802 eptr = &c;
8803 elen = 1;
a0ed51b3 8804 }
46fc3d4c 8805 goto string;
8806
46fc3d4c 8807 case 's':
26372e71
GA
8808 if (vectorize)
8809 goto unknown;
8810 if (args) {
fc36a67e 8811 eptr = va_arg(*args, char*);
c635e13b 8812 if (eptr)
1d7c1841
GS
8813#ifdef MACOS_TRADITIONAL
8814 /* On MacOS, %#s format is used for Pascal strings */
8815 if (alt)
8816 elen = *eptr++;
8817 else
8818#endif
c635e13b 8819 elen = strlen(eptr);
8820 else {
27da23d5 8821 eptr = (char *)nullstr;
c635e13b 8822 elen = sizeof nullstr - 1;
8823 }
46fc3d4c 8824 }
211dfcf1 8825 else {
4d84ee25 8826 eptr = SvPVx_const(argsv, elen);
7e2040f0 8827 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8828 if (has_precis && precis < elen) {
8829 I32 p = precis;
7e2040f0 8830 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8831 precis = p;
8832 }
8833 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8834 width += elen - sv_len_utf8(argsv);
a0ed51b3 8835 }
2cf2cfc6 8836 is_utf8 = TRUE;
a0ed51b3
LW
8837 }
8838 }
fc36a67e 8839
46fc3d4c 8840 string:
8841 if (has_precis && elen > precis)
8842 elen = precis;
8843 break;
8844
8845 /* INTEGERS */
8846
fc36a67e 8847 case 'p':
be75b157 8848 if (alt || vectorize)
c2e66d9e 8849 goto unknown;
211dfcf1 8850 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8851 base = 16;
8852 goto integer;
8853
46fc3d4c 8854 case 'D':
29fe7a80 8855#ifdef IV_IS_QUAD
22f3ae8c 8856 intsize = 'q';
29fe7a80 8857#else
46fc3d4c 8858 intsize = 'l';
29fe7a80 8859#endif
5f66b61c 8860 /*FALLTHROUGH*/
46fc3d4c 8861 case 'd':
8862 case 'i':
8896765a
RB
8863#if vdNUMBER
8864 format_vd:
8865#endif
b22c7a20 8866 if (vectorize) {
ba210ebe 8867 STRLEN ulen;
211dfcf1
HS
8868 if (!veclen)
8869 continue;
2cf2cfc6
A
8870 if (vec_utf8)
8871 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8872 UTF8_ALLOW_ANYUV);
b22c7a20 8873 else {
e83d50c9 8874 uv = *vecstr;
b22c7a20
GS
8875 ulen = 1;
8876 }
8877 vecstr += ulen;
8878 veclen -= ulen;
e83d50c9
JP
8879 if (plus)
8880 esignbuf[esignlen++] = plus;
b22c7a20
GS
8881 }
8882 else if (args) {
46fc3d4c 8883 switch (intsize) {
8884 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8885 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8886 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8887 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8888#ifdef HAS_QUAD
8889 case 'q': iv = va_arg(*args, Quad_t); break;
8890#endif
46fc3d4c 8891 }
8892 }
8893 else {
b10c0dba 8894 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8895 switch (intsize) {
b10c0dba
MHM
8896 case 'h': iv = (short)tiv; break;
8897 case 'l': iv = (long)tiv; break;
8898 case 'V':
8899 default: iv = tiv; break;
cf2093f6 8900#ifdef HAS_QUAD
b10c0dba 8901 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8902#endif
46fc3d4c 8903 }
8904 }
e83d50c9
JP
8905 if ( !vectorize ) /* we already set uv above */
8906 {
8907 if (iv >= 0) {
8908 uv = iv;
8909 if (plus)
8910 esignbuf[esignlen++] = plus;
8911 }
8912 else {
8913 uv = -iv;
8914 esignbuf[esignlen++] = '-';
8915 }
46fc3d4c 8916 }
8917 base = 10;
8918 goto integer;
8919
fc36a67e 8920 case 'U':
29fe7a80 8921#ifdef IV_IS_QUAD
22f3ae8c 8922 intsize = 'q';
29fe7a80 8923#else
fc36a67e 8924 intsize = 'l';
29fe7a80 8925#endif
5f66b61c 8926 /*FALLTHROUGH*/
fc36a67e 8927 case 'u':
8928 base = 10;
8929 goto uns_integer;
8930
4f19785b
WSI
8931 case 'b':
8932 base = 2;
8933 goto uns_integer;
8934
46fc3d4c 8935 case 'O':
29fe7a80 8936#ifdef IV_IS_QUAD
22f3ae8c 8937 intsize = 'q';
29fe7a80 8938#else
46fc3d4c 8939 intsize = 'l';
29fe7a80 8940#endif
5f66b61c 8941 /*FALLTHROUGH*/
46fc3d4c 8942 case 'o':
8943 base = 8;
8944 goto uns_integer;
8945
8946 case 'X':
46fc3d4c 8947 case 'x':
8948 base = 16;
46fc3d4c 8949
8950 uns_integer:
b22c7a20 8951 if (vectorize) {
ba210ebe 8952 STRLEN ulen;
b22c7a20 8953 vector:
211dfcf1
HS
8954 if (!veclen)
8955 continue;
2cf2cfc6
A
8956 if (vec_utf8)
8957 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8958 UTF8_ALLOW_ANYUV);
b22c7a20 8959 else {
a05b299f 8960 uv = *vecstr;
b22c7a20
GS
8961 ulen = 1;
8962 }
8963 vecstr += ulen;
8964 veclen -= ulen;
8965 }
8966 else if (args) {
46fc3d4c 8967 switch (intsize) {
8968 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8969 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8970 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8971 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8972#ifdef HAS_QUAD
9e3321a5 8973 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8974#endif
46fc3d4c 8975 }
8976 }
8977 else {
b10c0dba 8978 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8979 switch (intsize) {
b10c0dba
MHM
8980 case 'h': uv = (unsigned short)tuv; break;
8981 case 'l': uv = (unsigned long)tuv; break;
8982 case 'V':
8983 default: uv = tuv; break;
cf2093f6 8984#ifdef HAS_QUAD
b10c0dba 8985 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 8986#endif
46fc3d4c 8987 }
8988 }
8989
8990 integer:
4d84ee25
NC
8991 {
8992 char *ptr = ebuf + sizeof ebuf;
8993 switch (base) {
8994 unsigned dig;
8995 case 16:
8996 if (!uv)
8997 alt = FALSE;
8998 p = (char*)((c == 'X')
8999 ? "0123456789ABCDEF" : "0123456789abcdef");
9000 do {
9001 dig = uv & 15;
9002 *--ptr = p[dig];
9003 } while (uv >>= 4);
9004 if (alt) {
9005 esignbuf[esignlen++] = '0';
9006 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9007 }
9008 break;
9009 case 8:
9010 do {
9011 dig = uv & 7;
9012 *--ptr = '0' + dig;
9013 } while (uv >>= 3);
9014 if (alt && *ptr != '0')
9015 *--ptr = '0';
9016 break;
9017 case 2:
ed2b91d2
GA
9018 if (!uv)
9019 alt = FALSE;
4d84ee25
NC
9020 do {
9021 dig = uv & 1;
9022 *--ptr = '0' + dig;
9023 } while (uv >>= 1);
9024 if (alt) {
9025 esignbuf[esignlen++] = '0';
9026 esignbuf[esignlen++] = 'b';
9027 }
9028 break;
9029 default: /* it had better be ten or less */
9030 do {
9031 dig = uv % base;
9032 *--ptr = '0' + dig;
9033 } while (uv /= base);
9034 break;
46fc3d4c 9035 }
4d84ee25
NC
9036 elen = (ebuf + sizeof ebuf) - ptr;
9037 eptr = ptr;
9038 if (has_precis) {
9039 if (precis > elen)
9040 zeros = precis - elen;
9041 else if (precis == 0 && elen == 1 && *eptr == '0')
9042 elen = 0;
eda88b6d 9043 }
c10ed8b9 9044 }
46fc3d4c 9045 break;
9046
9047 /* FLOATING POINT */
9048
fc36a67e 9049 case 'F':
9050 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9051 /*FALLTHROUGH*/
46fc3d4c 9052 case 'e': case 'E':
fc36a67e 9053 case 'f':
46fc3d4c 9054 case 'g': case 'G':
26372e71
GA
9055 if (vectorize)
9056 goto unknown;
46fc3d4c 9057
9058 /* This is evil, but floating point is even more evil */
9059
9e5b023a
JH
9060 /* for SV-style calling, we can only get NV
9061 for C-style calling, we assume %f is double;
9062 for simplicity we allow any of %Lf, %llf, %qf for long double
9063 */
9064 switch (intsize) {
9065 case 'V':
9066#if defined(USE_LONG_DOUBLE)
9067 intsize = 'q';
9068#endif
9069 break;
8a2e3f14 9070/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9071 case 'l':
5f66b61c 9072 /*FALLTHROUGH*/
9e5b023a
JH
9073 default:
9074#if defined(USE_LONG_DOUBLE)
9075 intsize = args ? 0 : 'q';
9076#endif
9077 break;
9078 case 'q':
9079#if defined(HAS_LONG_DOUBLE)
9080 break;
9081#else
5f66b61c 9082 /*FALLTHROUGH*/
9e5b023a
JH
9083#endif
9084 case 'h':
9e5b023a
JH
9085 goto unknown;
9086 }
9087
9088 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9089 nv = (args) ?
35fff930
JH
9090#if LONG_DOUBLESIZE > DOUBLESIZE
9091 intsize == 'q' ?
205f51d8
AS
9092 va_arg(*args, long double) :
9093 va_arg(*args, double)
35fff930 9094#else
205f51d8 9095 va_arg(*args, double)
35fff930 9096#endif
9e5b023a 9097 : SvNVx(argsv);
fc36a67e 9098
9099 need = 0;
9100 if (c != 'e' && c != 'E') {
9101 i = PERL_INT_MIN;
9e5b023a
JH
9102 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9103 will cast our (long double) to (double) */
73b309ea 9104 (void)Perl_frexp(nv, &i);
fc36a67e 9105 if (i == PERL_INT_MIN)
cea2e8a9 9106 Perl_die(aTHX_ "panic: frexp");
c635e13b 9107 if (i > 0)
fc36a67e 9108 need = BIT_DIGITS(i);
9109 }
9110 need += has_precis ? precis : 6; /* known default */
20f6aaab 9111
fc36a67e 9112 if (need < width)
9113 need = width;
9114
20f6aaab
AS
9115#ifdef HAS_LDBL_SPRINTF_BUG
9116 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9117 with sfio - Allen <allens@cpan.org> */
9118
9119# ifdef DBL_MAX
9120# define MY_DBL_MAX DBL_MAX
9121# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9122# if DOUBLESIZE >= 8
9123# define MY_DBL_MAX 1.7976931348623157E+308L
9124# else
9125# define MY_DBL_MAX 3.40282347E+38L
9126# endif
9127# endif
9128
9129# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9130# define MY_DBL_MAX_BUG 1L
20f6aaab 9131# else
205f51d8 9132# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9133# endif
20f6aaab 9134
205f51d8
AS
9135# ifdef DBL_MIN
9136# define MY_DBL_MIN DBL_MIN
9137# else /* XXX guessing! -Allen */
9138# if DOUBLESIZE >= 8
9139# define MY_DBL_MIN 2.2250738585072014E-308L
9140# else
9141# define MY_DBL_MIN 1.17549435E-38L
9142# endif
9143# endif
20f6aaab 9144
205f51d8
AS
9145 if ((intsize == 'q') && (c == 'f') &&
9146 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9147 (need < DBL_DIG)) {
9148 /* it's going to be short enough that
9149 * long double precision is not needed */
9150
9151 if ((nv <= 0L) && (nv >= -0L))
9152 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9153 else {
9154 /* would use Perl_fp_class as a double-check but not
9155 * functional on IRIX - see perl.h comments */
9156
9157 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9158 /* It's within the range that a double can represent */
9159#if defined(DBL_MAX) && !defined(DBL_MIN)
9160 if ((nv >= ((long double)1/DBL_MAX)) ||
9161 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9162#endif
205f51d8 9163 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9164 }
205f51d8
AS
9165 }
9166 if (fix_ldbl_sprintf_bug == TRUE) {
9167 double temp;
9168
9169 intsize = 0;
9170 temp = (double)nv;
9171 nv = (NV)temp;
9172 }
20f6aaab 9173 }
205f51d8
AS
9174
9175# undef MY_DBL_MAX
9176# undef MY_DBL_MAX_BUG
9177# undef MY_DBL_MIN
9178
20f6aaab
AS
9179#endif /* HAS_LDBL_SPRINTF_BUG */
9180
46fc3d4c 9181 need += 20; /* fudge factor */
80252599
GS
9182 if (PL_efloatsize < need) {
9183 Safefree(PL_efloatbuf);
9184 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9185 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9186 PL_efloatbuf[0] = '\0';
46fc3d4c 9187 }
9188
4151a5fe
IZ
9189 if ( !(width || left || plus || alt) && fill != '0'
9190 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9191 /* See earlier comment about buggy Gconvert when digits,
9192 aka precis is 0 */
9193 if ( c == 'g' && precis) {
2e59c212 9194 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9195 /* May return an empty string for digits==0 */
9196 if (*PL_efloatbuf) {
9197 elen = strlen(PL_efloatbuf);
4151a5fe 9198 goto float_converted;
4150c189 9199 }
4151a5fe
IZ
9200 } else if ( c == 'f' && !precis) {
9201 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9202 break;
9203 }
9204 }
4d84ee25
NC
9205 {
9206 char *ptr = ebuf + sizeof ebuf;
9207 *--ptr = '\0';
9208 *--ptr = c;
9209 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9210#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9211 if (intsize == 'q') {
9212 /* Copy the one or more characters in a long double
9213 * format before the 'base' ([efgEFG]) character to
9214 * the format string. */
9215 static char const prifldbl[] = PERL_PRIfldbl;
9216 char const *p = prifldbl + sizeof(prifldbl) - 3;
9217 while (p >= prifldbl) { *--ptr = *p--; }
9218 }
65202027 9219#endif
4d84ee25
NC
9220 if (has_precis) {
9221 base = precis;
9222 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9223 *--ptr = '.';
9224 }
9225 if (width) {
9226 base = width;
9227 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9228 }
9229 if (fill == '0')
9230 *--ptr = fill;
9231 if (left)
9232 *--ptr = '-';
9233 if (plus)
9234 *--ptr = plus;
9235 if (alt)
9236 *--ptr = '#';
9237 *--ptr = '%';
9238
9239 /* No taint. Otherwise we are in the strange situation
9240 * where printf() taints but print($float) doesn't.
9241 * --jhi */
9e5b023a 9242#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9243 elen = ((intsize == 'q')
9244 ? my_sprintf(PL_efloatbuf, ptr, nv)
9245 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9246#else
4150c189 9247 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9248#endif
4d84ee25 9249 }
4151a5fe 9250 float_converted:
80252599 9251 eptr = PL_efloatbuf;
46fc3d4c 9252 break;
9253
fc36a67e 9254 /* SPECIAL */
9255
9256 case 'n':
26372e71
GA
9257 if (vectorize)
9258 goto unknown;
fc36a67e 9259 i = SvCUR(sv) - origlen;
26372e71 9260 if (args) {
c635e13b 9261 switch (intsize) {
9262 case 'h': *(va_arg(*args, short*)) = i; break;
9263 default: *(va_arg(*args, int*)) = i; break;
9264 case 'l': *(va_arg(*args, long*)) = i; break;
9265 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9266#ifdef HAS_QUAD
9267 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9268#endif
c635e13b 9269 }
fc36a67e 9270 }
9dd79c3f 9271 else
211dfcf1 9272 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9273 continue; /* not "break" */
9274
9275 /* UNKNOWN */
9276
46fc3d4c 9277 default:
fc36a67e 9278 unknown:
041457d9
DM
9279 if (!args
9280 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9281 && ckWARN(WARN_PRINTF))
9282 {
c4420975 9283 SV * const msg = sv_newmortal();
35c1215d
NC
9284 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9285 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9286 if (c) {
0f4b6630 9287 if (isPRINT(c))
1c846c1f 9288 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9289 "\"%%%c\"", c & 0xFF);
9290 else
9291 Perl_sv_catpvf(aTHX_ msg,
57def98f 9292 "\"%%\\%03"UVof"\"",
0f4b6630 9293 (UV)c & 0xFF);
0f4b6630 9294 } else
396482e1 9295 sv_catpvs(msg, "end of string");
9014280d 9296 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9297 }
fb73857a 9298
9299 /* output mangled stuff ... */
9300 if (c == '\0')
9301 --q;
46fc3d4c 9302 eptr = p;
9303 elen = q - p;
fb73857a 9304
9305 /* ... right here, because formatting flags should not apply */
9306 SvGROW(sv, SvCUR(sv) + elen + 1);
9307 p = SvEND(sv);
4459522c 9308 Copy(eptr, p, elen, char);
fb73857a 9309 p += elen;
9310 *p = '\0';
3f7c398e 9311 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9312 svix = osvix;
fb73857a 9313 continue; /* not "break" */
46fc3d4c 9314 }
9315
6c94ec8b
HS
9316 /* calculate width before utf8_upgrade changes it */
9317 have = esignlen + zeros + elen;
ed2b91d2
GA
9318 if (have < zeros)
9319 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9320
d2876be5
JH
9321 if (is_utf8 != has_utf8) {
9322 if (is_utf8) {
9323 if (SvCUR(sv))
9324 sv_utf8_upgrade(sv);
9325 }
9326 else {
53c1dcc0 9327 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9328 sv_utf8_upgrade(nsv);
93524f2b 9329 eptr = SvPVX_const(nsv);
d2876be5
JH
9330 elen = SvCUR(nsv);
9331 }
9332 SvGROW(sv, SvCUR(sv) + elen + 1);
9333 p = SvEND(sv);
9334 *p = '\0';
9335 }
6af65485 9336
46fc3d4c 9337 need = (have > width ? have : width);
9338 gap = need - have;
9339
d2641cbd
PC
9340 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9341 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9342 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9343 p = SvEND(sv);
9344 if (esignlen && fill == '0') {
53c1dcc0 9345 int i;
eb160463 9346 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9347 *p++ = esignbuf[i];
9348 }
9349 if (gap && !left) {
9350 memset(p, fill, gap);
9351 p += gap;
9352 }
9353 if (esignlen && fill != '0') {
53c1dcc0 9354 int i;
eb160463 9355 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9356 *p++ = esignbuf[i];
9357 }
fc36a67e 9358 if (zeros) {
53c1dcc0 9359 int i;
fc36a67e 9360 for (i = zeros; i; i--)
9361 *p++ = '0';
9362 }
46fc3d4c 9363 if (elen) {
4459522c 9364 Copy(eptr, p, elen, char);
46fc3d4c 9365 p += elen;
9366 }
9367 if (gap && left) {
9368 memset(p, ' ', gap);
9369 p += gap;
9370 }
b22c7a20
GS
9371 if (vectorize) {
9372 if (veclen) {
4459522c 9373 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9374 p += dotstrlen;
9375 }
9376 else
9377 vectorize = FALSE; /* done iterating over vecstr */
9378 }
2cf2cfc6
A
9379 if (is_utf8)
9380 has_utf8 = TRUE;
9381 if (has_utf8)
7e2040f0 9382 SvUTF8_on(sv);
46fc3d4c 9383 *p = '\0';
3f7c398e 9384 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9385 if (vectorize) {
9386 esignlen = 0;
9387 goto vector;
9388 }
46fc3d4c 9389 }
9390}
51371543 9391
645c22ef
DM
9392/* =========================================================================
9393
9394=head1 Cloning an interpreter
9395
9396All the macros and functions in this section are for the private use of
9397the main function, perl_clone().
9398
9399The foo_dup() functions make an exact copy of an existing foo thinngy.
9400During the course of a cloning, a hash table is used to map old addresses
9401to new addresses. The table is created and manipulated with the
9402ptr_table_* functions.
9403
9404=cut
9405
9406============================================================================*/
9407
9408
1d7c1841
GS
9409#if defined(USE_ITHREADS)
9410
d4c19fe8 9411/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9412#ifndef GpREFCNT_inc
9413# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9414#endif
9415
9416
a41cc44e
NC
9417/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
9418 that currently av_dup and hv_dup are the same as sv_dup. If this changes,
9419 please unmerge ss_dup. */
d2d73c3e 9420#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9421#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9422#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9423#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9424#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9425#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9426#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9427#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9428#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9429#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9430#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9431#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9432#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9433#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9434
d2d73c3e 9435
d2f185dc
AMS
9436/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9437 regcomp.c. AMS 20010712 */
645c22ef 9438
1d7c1841 9439REGEXP *
53c1dcc0 9440Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9441{
27da23d5 9442 dVAR;
d2f185dc
AMS
9443 REGEXP *ret;
9444 int i, len, npar;
9445 struct reg_substr_datum *s;
9446
9447 if (!r)
9448 return (REGEXP *)NULL;
9449
9450 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9451 return ret;
9452
9453 len = r->offsets[0];
9454 npar = r->nparens+1;
9455
a02a5408 9456 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9457 Copy(r->program, ret->program, len+1, regnode);
9458
a02a5408 9459 Newx(ret->startp, npar, I32);
d2f185dc 9460 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9461 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9462 Copy(r->startp, ret->startp, npar, I32);
9463
a02a5408 9464 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9465 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9466 s->min_offset = r->substrs->data[i].min_offset;
9467 s->max_offset = r->substrs->data[i].max_offset;
9468 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9469 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9470 }
9471
70612e96 9472 ret->regstclass = NULL;
d2f185dc
AMS
9473 if (r->data) {
9474 struct reg_data *d;
e1ec3a88 9475 const int count = r->data->count;
53c1dcc0 9476 int i;
d2f185dc 9477
a02a5408 9478 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9479 char, struct reg_data);
a02a5408 9480 Newx(d->what, count, U8);
d2f185dc
AMS
9481
9482 d->count = count;
9483 for (i = 0; i < count; i++) {
9484 d->what[i] = r->data->what[i];
9485 switch (d->what[i]) {
a3621e74
YO
9486 /* legal options are one of: sfpont
9487 see also regcomp.h and pregfree() */
d2f185dc
AMS
9488 case 's':
9489 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9490 break;
9491 case 'p':
9492 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9493 break;
9494 case 'f':
9495 /* This is cheating. */
a02a5408 9496 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9497 StructCopy(r->data->data[i], d->data[i],
9498 struct regnode_charclass_class);
70612e96 9499 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9500 break;
9501 case 'o':
33773810
AMS
9502 /* Compiled op trees are readonly, and can thus be
9503 shared without duplication. */
b34c0dd4 9504 OP_REFCNT_LOCK;
9b978d73 9505 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9506 OP_REFCNT_UNLOCK;
9b978d73 9507 break;
d2f185dc
AMS
9508 case 'n':
9509 d->data[i] = r->data->data[i];
9510 break;
a3621e74
YO
9511 case 't':
9512 d->data[i] = r->data->data[i];
9513 OP_REFCNT_LOCK;
9514 ((reg_trie_data*)d->data[i])->refcount++;
9515 OP_REFCNT_UNLOCK;
9516 break;
9517 default:
9518 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9519 }
9520 }
9521
9522 ret->data = d;
9523 }
9524 else
9525 ret->data = NULL;
9526
a02a5408 9527 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9528 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9529
e01c5899 9530 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9531 ret->refcnt = r->refcnt;
9532 ret->minlen = r->minlen;
9533 ret->prelen = r->prelen;
9534 ret->nparens = r->nparens;
9535 ret->lastparen = r->lastparen;
9536 ret->lastcloseparen = r->lastcloseparen;
9537 ret->reganch = r->reganch;
9538
70612e96
RG
9539 ret->sublen = r->sublen;
9540
9541 if (RX_MATCH_COPIED(ret))
e01c5899 9542 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9543 else
bd61b366 9544 ret->subbeg = NULL;
f8c7b90f 9545#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9546 ret->saved_copy = NULL;
9a26048b 9547#endif
70612e96 9548
d2f185dc
AMS
9549 ptr_table_store(PL_ptr_table, r, ret);
9550 return ret;
1d7c1841
GS
9551}
9552
d2d73c3e 9553/* duplicate a file handle */
645c22ef 9554
1d7c1841 9555PerlIO *
a8fc9800 9556Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9557{
9558 PerlIO *ret;
53c1dcc0
AL
9559
9560 PERL_UNUSED_ARG(type);
73d840c0 9561
1d7c1841
GS
9562 if (!fp)
9563 return (PerlIO*)NULL;
9564
9565 /* look for it in the table first */
9566 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9567 if (ret)
9568 return ret;
9569
9570 /* create anew and remember what it is */
ecdeb87c 9571 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9572 ptr_table_store(PL_ptr_table, fp, ret);
9573 return ret;
9574}
9575
645c22ef
DM
9576/* duplicate a directory handle */
9577
1d7c1841
GS
9578DIR *
9579Perl_dirp_dup(pTHX_ DIR *dp)
9580{
96a5add6 9581 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9582 if (!dp)
9583 return (DIR*)NULL;
9584 /* XXX TODO */
9585 return dp;
9586}
9587
ff276b08 9588/* duplicate a typeglob */
645c22ef 9589
1d7c1841 9590GP *
a8fc9800 9591Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9592{
9593 GP *ret;
b37c2d43 9594
1d7c1841
GS
9595 if (!gp)
9596 return (GP*)NULL;
9597 /* look for it in the table first */
9598 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9599 if (ret)
9600 return ret;
9601
9602 /* create anew and remember what it is */
a02a5408 9603 Newxz(ret, 1, GP);
1d7c1841
GS
9604 ptr_table_store(PL_ptr_table, gp, ret);
9605
9606 /* clone */
9607 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9608 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9609 ret->gp_io = io_dup_inc(gp->gp_io, param);
9610 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9611 ret->gp_av = av_dup_inc(gp->gp_av, param);
9612 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9613 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9614 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9615 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9616 ret->gp_line = gp->gp_line;
9617 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9618 return ret;
9619}
9620
645c22ef
DM
9621/* duplicate a chain of magic */
9622
1d7c1841 9623MAGIC *
a8fc9800 9624Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9625{
cb359b41
JH
9626 MAGIC *mgprev = (MAGIC*)NULL;
9627 MAGIC *mgret;
1d7c1841
GS
9628 if (!mg)
9629 return (MAGIC*)NULL;
9630 /* look for it in the table first */
9631 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9632 if (mgret)
9633 return mgret;
9634
9635 for (; mg; mg = mg->mg_moremagic) {
9636 MAGIC *nmg;
a02a5408 9637 Newxz(nmg, 1, MAGIC);
cb359b41 9638 if (mgprev)
1d7c1841 9639 mgprev->mg_moremagic = nmg;
cb359b41
JH
9640 else
9641 mgret = nmg;
1d7c1841
GS
9642 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9643 nmg->mg_private = mg->mg_private;
9644 nmg->mg_type = mg->mg_type;
9645 nmg->mg_flags = mg->mg_flags;
14befaf4 9646 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9647 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9648 }
05bd4103 9649 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9650 /* The backref AV has its reference count deliberately bumped by
9651 1. */
9652 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9653 }
8d2f4536
NC
9654 else if (mg->mg_type == PERL_MAGIC_symtab) {
9655 nmg->mg_obj = mg->mg_obj;
9656 }
1d7c1841
GS
9657 else {
9658 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9659 ? sv_dup_inc(mg->mg_obj, param)
9660 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9661 }
9662 nmg->mg_len = mg->mg_len;
9663 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9664 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9665 if (mg->mg_len > 0) {
1d7c1841 9666 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9667 if (mg->mg_type == PERL_MAGIC_overload_table &&
9668 AMT_AMAGIC((AMT*)mg->mg_ptr))
9669 {
c445ea15 9670 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9671 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9672 I32 i;
9673 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9674 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9675 }
9676 }
9677 }
9678 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9679 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9680 }
68795e93
NIS
9681 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9682 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9683 }
1d7c1841
GS
9684 mgprev = nmg;
9685 }
9686 return mgret;
9687}
9688
645c22ef
DM
9689/* create a new pointer-mapping table */
9690
1d7c1841
GS
9691PTR_TBL_t *
9692Perl_ptr_table_new(pTHX)
9693{
9694 PTR_TBL_t *tbl;
96a5add6
AL
9695 PERL_UNUSED_CONTEXT;
9696
a02a5408 9697 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9698 tbl->tbl_max = 511;
9699 tbl->tbl_items = 0;
a02a5408 9700 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9701 return tbl;
9702}
9703
7119fd33
NC
9704#define PTR_TABLE_HASH(ptr) \
9705 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9706
93e68bfb
JC
9707/*
9708 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9709 following define) and at call to new_body_inline made below in
9710 Perl_ptr_table_store()
9711 */
9712
9713#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9714
645c22ef
DM
9715/* map an existing pointer using a table */
9716
7bf61b54 9717STATIC PTR_TBL_ENT_t *
b0e6ae5b 9718S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9719 PTR_TBL_ENT_t *tblent;
4373e329 9720 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9721 assert(tbl);
9722 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9723 for (; tblent; tblent = tblent->next) {
9724 if (tblent->oldval == sv)
7bf61b54 9725 return tblent;
1d7c1841 9726 }
d4c19fe8 9727 return NULL;
7bf61b54
NC
9728}
9729
9730void *
9731Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9732{
b0e6ae5b 9733 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9734 PERL_UNUSED_CONTEXT;
d4c19fe8 9735 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9736}
9737
645c22ef
DM
9738/* add a new entry to a pointer-mapping table */
9739
1d7c1841 9740void
44f8325f 9741Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9742{
0c9fdfe0 9743 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9744 PERL_UNUSED_CONTEXT;
1d7c1841 9745
7bf61b54
NC
9746 if (tblent) {
9747 tblent->newval = newsv;
9748 } else {
9749 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9750
d2a0f284
JC
9751 new_body_inline(tblent, PTE_SVSLOT);
9752
7bf61b54
NC
9753 tblent->oldval = oldsv;
9754 tblent->newval = newsv;
9755 tblent->next = tbl->tbl_ary[entry];
9756 tbl->tbl_ary[entry] = tblent;
9757 tbl->tbl_items++;
9758 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9759 ptr_table_split(tbl);
1d7c1841 9760 }
1d7c1841
GS
9761}
9762
645c22ef
DM
9763/* double the hash bucket size of an existing ptr table */
9764
1d7c1841
GS
9765void
9766Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9767{
9768 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9769 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9770 UV newsize = oldsize * 2;
9771 UV i;
96a5add6 9772 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9773
9774 Renew(ary, newsize, PTR_TBL_ENT_t*);
9775 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9776 tbl->tbl_max = --newsize;
9777 tbl->tbl_ary = ary;
9778 for (i=0; i < oldsize; i++, ary++) {
9779 PTR_TBL_ENT_t **curentp, **entp, *ent;
9780 if (!*ary)
9781 continue;
9782 curentp = ary + oldsize;
9783 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9784 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9785 *entp = ent->next;
9786 ent->next = *curentp;
9787 *curentp = ent;
9788 continue;
9789 }
9790 else
9791 entp = &ent->next;
9792 }
9793 }
9794}
9795
645c22ef
DM
9796/* remove all the entries from a ptr table */
9797
a0739874
DM
9798void
9799Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9800{
d5cefff9 9801 if (tbl && tbl->tbl_items) {
c445ea15 9802 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9803 UV riter = tbl->tbl_max;
a0739874 9804
d5cefff9
NC
9805 do {
9806 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9807
d5cefff9 9808 while (entry) {
00b6aa41 9809 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9810 entry = entry->next;
9811 del_pte(oentry);
9812 }
9813 } while (riter--);
a0739874 9814
d5cefff9
NC
9815 tbl->tbl_items = 0;
9816 }
a0739874
DM
9817}
9818
645c22ef
DM
9819/* clear and free a ptr table */
9820
a0739874
DM
9821void
9822Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9823{
9824 if (!tbl) {
9825 return;
9826 }
9827 ptr_table_clear(tbl);
9828 Safefree(tbl->tbl_ary);
9829 Safefree(tbl);
9830}
9831
5bd07a3d 9832
83841fad 9833void
eb86f8b3 9834Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9835{
9836 if (SvROK(sstr)) {
b162af07
SP
9837 SvRV_set(dstr, SvWEAKREF(sstr)
9838 ? sv_dup(SvRV(sstr), param)
9839 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9840
83841fad 9841 }
3f7c398e 9842 else if (SvPVX_const(sstr)) {
83841fad
NIS
9843 /* Has something there */
9844 if (SvLEN(sstr)) {
68795e93 9845 /* Normal PV - clone whole allocated space */
3f7c398e 9846 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9847 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9848 /* Not that normal - actually sstr is copy on write.
9849 But we are a true, independant SV, so: */
9850 SvREADONLY_off(dstr);
9851 SvFAKE_off(dstr);
9852 }
68795e93 9853 }
83841fad
NIS
9854 else {
9855 /* Special case - not normally malloced for some reason */
f7877b28
NC
9856 if (isGV_with_GP(sstr)) {
9857 /* Don't need to do anything here. */
9858 }
9859 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9860 /* A "shared" PV - clone it as "shared" PV */
9861 SvPV_set(dstr,
9862 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9863 param)));
83841fad
NIS
9864 }
9865 else {
9866 /* Some other special case - random pointer */
f880fe2f 9867 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9868 }
83841fad
NIS
9869 }
9870 }
9871 else {
4608196e 9872 /* Copy the NULL */
f880fe2f 9873 if (SvTYPE(dstr) == SVt_RV)
b162af07 9874 SvRV_set(dstr, NULL);
f880fe2f 9875 else
6136c704 9876 SvPV_set(dstr, NULL);
83841fad
NIS
9877 }
9878}
9879
662fb8b2
NC
9880/* duplicate an SV of any type (including AV, HV etc) */
9881
1d7c1841 9882SV *
eb86f8b3 9883Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9884{
27da23d5 9885 dVAR;
1d7c1841
GS
9886 SV *dstr;
9887
9888 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9889 return NULL;
1d7c1841
GS
9890 /* look for it in the table first */
9891 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9892 if (dstr)
9893 return dstr;
9894
0405e91e
AB
9895 if(param->flags & CLONEf_JOIN_IN) {
9896 /** We are joining here so we don't want do clone
9897 something that is bad **/
eb86f8b3
AL
9898 if (SvTYPE(sstr) == SVt_PVHV) {
9899 const char * const hvname = HvNAME_get(sstr);
9900 if (hvname)
9901 /** don't clone stashes if they already exist **/
9902 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9903 }
9904 }
9905
1d7c1841
GS
9906 /* create anew and remember what it is */
9907 new_SV(dstr);
fd0854ff
DM
9908
9909#ifdef DEBUG_LEAKING_SCALARS
9910 dstr->sv_debug_optype = sstr->sv_debug_optype;
9911 dstr->sv_debug_line = sstr->sv_debug_line;
9912 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9913 dstr->sv_debug_cloned = 1;
fd0854ff 9914 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9915#endif
9916
1d7c1841
GS
9917 ptr_table_store(PL_ptr_table, sstr, dstr);
9918
9919 /* clone */
9920 SvFLAGS(dstr) = SvFLAGS(sstr);
9921 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9922 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9923
9924#ifdef DEBUGGING
3f7c398e 9925 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9926 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9927 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9928#endif
9929
9660f481
DM
9930 /* don't clone objects whose class has asked us not to */
9931 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9932 SvFLAGS(dstr) &= ~SVTYPEMASK;
9933 SvOBJECT_off(dstr);
9934 return dstr;
9935 }
9936
1d7c1841
GS
9937 switch (SvTYPE(sstr)) {
9938 case SVt_NULL:
9939 SvANY(dstr) = NULL;
9940 break;
9941 case SVt_IV:
339049b0 9942 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9943 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9944 break;
9945 case SVt_NV:
9946 SvANY(dstr) = new_XNV();
9d6ce603 9947 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9948 break;
9949 case SVt_RV:
339049b0 9950 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9951 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9952 break;
662fb8b2
NC
9953 default:
9954 {
9955 /* These are all the types that need complex bodies allocating. */
662fb8b2 9956 void *new_body;
2bcc16b3
NC
9957 const svtype sv_type = SvTYPE(sstr);
9958 const struct body_details *const sv_type_details
9959 = bodies_by_type + sv_type;
662fb8b2 9960
93e68bfb 9961 switch (sv_type) {
662fb8b2 9962 default:
bb263b4e 9963 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9964 break;
9965
662fb8b2
NC
9966 case SVt_PVGV:
9967 if (GvUNIQUE((GV*)sstr)) {
bb263b4e 9968 /*EMPTY*/; /* Do sharing here, and fall through */
662fb8b2 9969 }
c22188b4
NC
9970 case SVt_PVIO:
9971 case SVt_PVFM:
9972 case SVt_PVHV:
9973 case SVt_PVAV:
93e68bfb 9974 case SVt_PVBM:
662fb8b2 9975 case SVt_PVCV:
662fb8b2 9976 case SVt_PVLV:
662fb8b2 9977 case SVt_PVMG:
662fb8b2 9978 case SVt_PVNV:
662fb8b2 9979 case SVt_PVIV:
662fb8b2 9980 case SVt_PV:
d2a0f284 9981 assert(sv_type_details->body_size);
c22188b4 9982 if (sv_type_details->arena) {
d2a0f284 9983 new_body_inline(new_body, sv_type);
c22188b4 9984 new_body
b9502f15 9985 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9986 } else {
9987 new_body = new_NOARENA(sv_type_details);
9988 }
1d7c1841 9989 }
662fb8b2
NC
9990 assert(new_body);
9991 SvANY(dstr) = new_body;
9992
2bcc16b3 9993#ifndef PURIFY
b9502f15
NC
9994 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
9995 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 9996 sv_type_details->copy, char);
2bcc16b3
NC
9997#else
9998 Copy(((char*)SvANY(sstr)),
9999 ((char*)SvANY(dstr)),
d2a0f284 10000 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10001#endif
662fb8b2 10002
f7877b28
NC
10003 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10004 && !isGV_with_GP(dstr))
662fb8b2
NC
10005 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10006
10007 /* The Copy above means that all the source (unduplicated) pointers
10008 are now in the destination. We can check the flags and the
10009 pointers in either, but it's possible that there's less cache
10010 missing by always going for the destination.
10011 FIXME - instrument and check that assumption */
f32993d6 10012 if (sv_type >= SVt_PVMG) {
e736a858 10013 HV *ourstash;
2892acdb 10014 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
10015 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
10016 } else if (SvMAGIC(dstr))
662fb8b2
NC
10017 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10018 if (SvSTASH(dstr))
10019 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10020 }
662fb8b2 10021
f32993d6
NC
10022 /* The cast silences a GCC warning about unhandled types. */
10023 switch ((int)sv_type) {
662fb8b2
NC
10024 case SVt_PV:
10025 break;
10026 case SVt_PVIV:
10027 break;
10028 case SVt_PVNV:
10029 break;
10030 case SVt_PVMG:
10031 break;
10032 case SVt_PVBM:
10033 break;
10034 case SVt_PVLV:
10035 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10036 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10037 LvTARG(dstr) = dstr;
10038 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10039 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10040 else
10041 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10042 break;
10043 case SVt_PVGV:
acda4c6a
NC
10044 if (GvNAME_HEK(dstr))
10045 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 10046
e15faf7d
NC
10047 /* Don't call sv_add_backref here as it's going to be created
10048 as part of the magic cloning of the symbol table. */
f7877b28
NC
10049 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10050 if(isGV_with_GP(sstr)) {
10051 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10052 at the point of this comment. */
10053 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10054 (void)GpREFCNT_inc(GvGP(dstr));
10055 } else
10056 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10057 break;
10058 case SVt_PVIO:
10059 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10060 if (IoOFP(dstr) == IoIFP(sstr))
10061 IoOFP(dstr) = IoIFP(dstr);
10062 else
10063 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10064 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10065 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10066 /* I have no idea why fake dirp (rsfps)
10067 should be treated differently but otherwise
10068 we end up with leaks -- sky*/
10069 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10070 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10071 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10072 } else {
10073 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10074 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10075 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10076 if (IoDIRP(dstr)) {
10077 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10078 } else {
bb263b4e 10079 /*EMPTY*/;
100ce7e1
NC
10080 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10081 }
662fb8b2
NC
10082 }
10083 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10084 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10085 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10086 break;
10087 case SVt_PVAV:
10088 if (AvARRAY((AV*)sstr)) {
10089 SV **dst_ary, **src_ary;
10090 SSize_t items = AvFILLp((AV*)sstr) + 1;
10091
10092 src_ary = AvARRAY((AV*)sstr);
a02a5408 10093 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10094 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10095 SvPV_set(dstr, (char*)dst_ary);
10096 AvALLOC((AV*)dstr) = dst_ary;
10097 if (AvREAL((AV*)sstr)) {
10098 while (items-- > 0)
10099 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10100 }
10101 else {
10102 while (items-- > 0)
10103 *dst_ary++ = sv_dup(*src_ary++, param);
10104 }
10105 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10106 while (items-- > 0) {
10107 *dst_ary++ = &PL_sv_undef;
10108 }
bfcb3514 10109 }
662fb8b2 10110 else {
bd61b366 10111 SvPV_set(dstr, NULL);
662fb8b2 10112 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10113 }
662fb8b2
NC
10114 break;
10115 case SVt_PVHV:
10116 {
cbbf8932 10117 HEK *hvname = NULL;
662fb8b2
NC
10118
10119 if (HvARRAY((HV*)sstr)) {
10120 STRLEN i = 0;
10121 const bool sharekeys = !!HvSHAREKEYS(sstr);
10122 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10123 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10124 char *darray;
a02a5408 10125 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10126 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10127 char);
10128 HvARRAY(dstr) = (HE**)darray;
10129 while (i <= sxhv->xhv_max) {
5c4138a0 10130 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10131 HvARRAY(dstr)[i] = source
10132 ? he_dup(source, sharekeys, param) : 0;
10133 ++i;
10134 }
10135 if (SvOOK(sstr)) {
00b6aa41
AL
10136 struct xpvhv_aux * const saux = HvAUX(sstr);
10137 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10138 /* This flag isn't copied. */
10139 /* SvOOK_on(hv) attacks the IV flags. */
10140 SvFLAGS(dstr) |= SVf_OOK;
10141
10142 hvname = saux->xhv_name;
dd690478
NC
10143 daux->xhv_name
10144 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10145
10146 daux->xhv_riter = saux->xhv_riter;
10147 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10148 ? he_dup(saux->xhv_eiter,
10149 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10150 daux->xhv_backreferences = saux->xhv_backreferences
10151 ? (AV*) SvREFCNT_inc(
10152 sv_dup((SV*)saux->
10153 xhv_backreferences,
10154 param))
10155 : 0;
662fb8b2
NC
10156 }
10157 }
10158 else {
bd61b366 10159 SvPV_set(dstr, NULL);
662fb8b2
NC
10160 }
10161 /* Record stashes for possible cloning in Perl_clone(). */
10162 if(hvname)
10163 av_push(param->stashes, dstr);
10164 }
10165 break;
662fb8b2 10166 case SVt_PVCV:
bb172083
NC
10167 if (!(param->flags & CLONEf_COPY_STACKS)) {
10168 CvDEPTH(dstr) = 0;
10169 }
10170 case SVt_PVFM:
662fb8b2
NC
10171 /* NOTE: not refcounted */
10172 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10173 OP_REFCNT_LOCK;
d04ba589
NC
10174 if (!CvISXSUB(dstr))
10175 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10176 OP_REFCNT_UNLOCK;
cfae286e 10177 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10178 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10179 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10180 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10181 }
10182 /* don't dup if copying back - CvGV isn't refcounted, so the
10183 * duped GV may never be freed. A bit of a hack! DAPM */
10184 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10185 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10186 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10187 CvOUTSIDE(dstr) =
10188 CvWEAKOUTSIDE(sstr)
10189 ? cv_dup( CvOUTSIDE(dstr), param)
10190 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10191 if (!CvISXSUB(dstr))
662fb8b2
NC
10192 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10193 break;
bfcb3514 10194 }
1d7c1841 10195 }
1d7c1841
GS
10196 }
10197
10198 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10199 ++PL_sv_objcount;
10200
10201 return dstr;
d2d73c3e 10202 }
1d7c1841 10203
645c22ef
DM
10204/* duplicate a context */
10205
1d7c1841 10206PERL_CONTEXT *
a8fc9800 10207Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10208{
10209 PERL_CONTEXT *ncxs;
10210
10211 if (!cxs)
10212 return (PERL_CONTEXT*)NULL;
10213
10214 /* look for it in the table first */
10215 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10216 if (ncxs)
10217 return ncxs;
10218
10219 /* create anew and remember what it is */
a02a5408 10220 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10221 ptr_table_store(PL_ptr_table, cxs, ncxs);
10222
10223 while (ix >= 0) {
c445ea15
AL
10224 PERL_CONTEXT * const cx = &cxs[ix];
10225 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10226 ncx->cx_type = cx->cx_type;
10227 if (CxTYPE(cx) == CXt_SUBST) {
10228 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10229 }
10230 else {
10231 ncx->blk_oldsp = cx->blk_oldsp;
10232 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10233 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10234 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10235 ncx->blk_oldpm = cx->blk_oldpm;
10236 ncx->blk_gimme = cx->blk_gimme;
10237 switch (CxTYPE(cx)) {
10238 case CXt_SUB:
10239 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10240 ? cv_dup_inc(cx->blk_sub.cv, param)
10241 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10242 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10243 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10244 : NULL);
d2d73c3e 10245 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10246 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10247 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10248 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10249 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10250 break;
10251 case CXt_EVAL:
10252 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10253 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10254 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10255 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10256 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10257 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10258 break;
10259 case CXt_LOOP:
10260 ncx->blk_loop.label = cx->blk_loop.label;
10261 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10262 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10263 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10264 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10265 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10266 ? cx->blk_loop.iterdata
d2d73c3e 10267 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10268 ncx->blk_loop.oldcomppad
10269 = (PAD*)ptr_table_fetch(PL_ptr_table,
10270 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10271 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10272 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10273 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10274 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10275 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10276 break;
10277 case CXt_FORMAT:
d2d73c3e
AB
10278 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10279 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10280 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10281 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10282 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10283 break;
10284 case CXt_BLOCK:
10285 case CXt_NULL:
10286 break;
10287 }
10288 }
10289 --ix;
10290 }
10291 return ncxs;
10292}
10293
645c22ef
DM
10294/* duplicate a stack info structure */
10295
1d7c1841 10296PERL_SI *
a8fc9800 10297Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10298{
10299 PERL_SI *nsi;
10300
10301 if (!si)
10302 return (PERL_SI*)NULL;
10303
10304 /* look for it in the table first */
10305 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10306 if (nsi)
10307 return nsi;
10308
10309 /* create anew and remember what it is */
a02a5408 10310 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10311 ptr_table_store(PL_ptr_table, si, nsi);
10312
d2d73c3e 10313 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10314 nsi->si_cxix = si->si_cxix;
10315 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10316 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10317 nsi->si_type = si->si_type;
d2d73c3e
AB
10318 nsi->si_prev = si_dup(si->si_prev, param);
10319 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10320 nsi->si_markoff = si->si_markoff;
10321
10322 return nsi;
10323}
10324
10325#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10326#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10327#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10328#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10329#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10330#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10331#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10332#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10333#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10334#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10335#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10336#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10337#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10338#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10339
10340/* XXXXX todo */
10341#define pv_dup_inc(p) SAVEPV(p)
10342#define pv_dup(p) SAVEPV(p)
10343#define svp_dup_inc(p,pp) any_dup(p,pp)
10344
645c22ef
DM
10345/* map any object to the new equivent - either something in the
10346 * ptr table, or something in the interpreter structure
10347 */
10348
1d7c1841 10349void *
53c1dcc0 10350Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10351{
10352 void *ret;
10353
10354 if (!v)
10355 return (void*)NULL;
10356
10357 /* look for it in the table first */
10358 ret = ptr_table_fetch(PL_ptr_table, v);
10359 if (ret)
10360 return ret;
10361
10362 /* see if it is part of the interpreter structure */
10363 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10364 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10365 else {
1d7c1841 10366 ret = v;
05ec9bb3 10367 }
1d7c1841
GS
10368
10369 return ret;
10370}
10371
645c22ef
DM
10372/* duplicate the save stack */
10373
1d7c1841 10374ANY *
a8fc9800 10375Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10376{
53c1dcc0
AL
10377 ANY * const ss = proto_perl->Tsavestack;
10378 const I32 max = proto_perl->Tsavestack_max;
10379 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10380 ANY *nss;
10381 SV *sv;
10382 GV *gv;
10383 AV *av;
10384 HV *hv;
10385 void* ptr;
10386 int intval;
10387 long longval;
10388 GP *gp;
10389 IV iv;
c4e33207 10390 char *c = NULL;
1d7c1841 10391 void (*dptr) (void*);
acfe0abc 10392 void (*dxptr) (pTHX_ void*);
1d7c1841 10393
a02a5408 10394 Newxz(nss, max, ANY);
1d7c1841
GS
10395
10396 while (ix > 0) {
b464bac0 10397 I32 i = POPINT(ss,ix);
1d7c1841
GS
10398 TOPINT(nss,ix) = i;
10399 switch (i) {
10400 case SAVEt_ITEM: /* normal string */
a41cc44e 10401 case SAVEt_SV: /* scalar reference */
1d7c1841 10402 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10403 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10404 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10405 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10406 break;
05ec9bb3
NIS
10407 case SAVEt_SHARED_PVREF: /* char* in shared space */
10408 c = (char*)POPPTR(ss,ix);
10409 TOPPTR(nss,ix) = savesharedpv(c);
10410 ptr = POPPTR(ss,ix);
10411 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10412 break;
1d7c1841
GS
10413 case SAVEt_GENERIC_SVREF: /* generic sv */
10414 case SAVEt_SVREF: /* scalar reference */
10415 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10416 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10417 ptr = POPPTR(ss,ix);
10418 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10419 break;
a41cc44e 10420 case SAVEt_HV: /* hash reference */
1d7c1841 10421 case SAVEt_AV: /* array reference */
337d28f5
NC
10422 sv = POPPTR(ss,ix);
10423 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10424 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10425 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10426 break;
10427 case SAVEt_INT: /* int reference */
10428 ptr = POPPTR(ss,ix);
10429 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10430 intval = (int)POPINT(ss,ix);
10431 TOPINT(nss,ix) = intval;
10432 break;
10433 case SAVEt_LONG: /* long reference */
10434 ptr = POPPTR(ss,ix);
10435 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10436 longval = (long)POPLONG(ss,ix);
10437 TOPLONG(nss,ix) = longval;
10438 break;
10439 case SAVEt_I32: /* I32 reference */
10440 case SAVEt_I16: /* I16 reference */
10441 case SAVEt_I8: /* I8 reference */
88effcc9 10442 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10443 ptr = POPPTR(ss,ix);
10444 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10445 i = POPINT(ss,ix);
10446 TOPINT(nss,ix) = i;
10447 break;
10448 case SAVEt_IV: /* IV reference */
10449 ptr = POPPTR(ss,ix);
10450 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10451 iv = POPIV(ss,ix);
10452 TOPIV(nss,ix) = iv;
10453 break;
a41cc44e
NC
10454 case SAVEt_HPTR: /* HV* reference */
10455 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10456 case SAVEt_SPTR: /* SV* reference */
10457 ptr = POPPTR(ss,ix);
10458 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10459 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10460 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10461 break;
10462 case SAVEt_VPTR: /* random* reference */
10463 ptr = POPPTR(ss,ix);
10464 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10465 ptr = POPPTR(ss,ix);
10466 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10467 break;
b03d03b0 10468 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10469 case SAVEt_PPTR: /* char* reference */
10470 ptr = POPPTR(ss,ix);
10471 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10472 c = (char*)POPPTR(ss,ix);
10473 TOPPTR(nss,ix) = pv_dup(c);
10474 break;
1d7c1841
GS
10475 case SAVEt_NSTAB:
10476 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10477 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10478 break;
10479 case SAVEt_GP: /* scalar reference */
10480 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10481 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10482 (void)GpREFCNT_inc(gp);
10483 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10484 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10485 c = (char*)POPPTR(ss,ix);
10486 TOPPTR(nss,ix) = pv_dup(c);
10487 iv = POPIV(ss,ix);
10488 TOPIV(nss,ix) = iv;
10489 iv = POPIV(ss,ix);
10490 TOPIV(nss,ix) = iv;
10491 break;
10492 case SAVEt_FREESV:
26d9b02f 10493 case SAVEt_MORTALIZESV:
1d7c1841 10494 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10495 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10496 break;
10497 case SAVEt_FREEOP:
10498 ptr = POPPTR(ss,ix);
10499 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10500 /* these are assumed to be refcounted properly */
53c1dcc0 10501 OP *o;
1d7c1841
GS
10502 switch (((OP*)ptr)->op_type) {
10503 case OP_LEAVESUB:
10504 case OP_LEAVESUBLV:
10505 case OP_LEAVEEVAL:
10506 case OP_LEAVE:
10507 case OP_SCOPE:
10508 case OP_LEAVEWRITE:
e977893f
GS
10509 TOPPTR(nss,ix) = ptr;
10510 o = (OP*)ptr;
10511 OpREFCNT_inc(o);
1d7c1841
GS
10512 break;
10513 default:
5f66b61c 10514 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10515 break;
10516 }
10517 }
10518 else
5f66b61c 10519 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10520 break;
10521 case SAVEt_FREEPV:
10522 c = (char*)POPPTR(ss,ix);
10523 TOPPTR(nss,ix) = pv_dup_inc(c);
10524 break;
10525 case SAVEt_CLEARSV:
10526 longval = POPLONG(ss,ix);
10527 TOPLONG(nss,ix) = longval;
10528 break;
10529 case SAVEt_DELETE:
10530 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10531 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10532 c = (char*)POPPTR(ss,ix);
10533 TOPPTR(nss,ix) = pv_dup_inc(c);
10534 i = POPINT(ss,ix);
10535 TOPINT(nss,ix) = i;
10536 break;
10537 case SAVEt_DESTRUCTOR:
10538 ptr = POPPTR(ss,ix);
10539 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10540 dptr = POPDPTR(ss,ix);
8141890a
JH
10541 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10542 any_dup(FPTR2DPTR(void *, dptr),
10543 proto_perl));
1d7c1841
GS
10544 break;
10545 case SAVEt_DESTRUCTOR_X:
10546 ptr = POPPTR(ss,ix);
10547 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10548 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10549 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10550 any_dup(FPTR2DPTR(void *, dxptr),
10551 proto_perl));
1d7c1841
GS
10552 break;
10553 case SAVEt_REGCONTEXT:
10554 case SAVEt_ALLOC:
10555 i = POPINT(ss,ix);
10556 TOPINT(nss,ix) = i;
10557 ix -= i;
10558 break;
10559 case SAVEt_STACK_POS: /* Position on Perl stack */
10560 i = POPINT(ss,ix);
10561 TOPINT(nss,ix) = i;
10562 break;
10563 case SAVEt_AELEM: /* array element */
10564 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10565 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10566 i = POPINT(ss,ix);
10567 TOPINT(nss,ix) = i;
10568 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10569 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10570 break;
10571 case SAVEt_HELEM: /* hash element */
10572 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10573 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10574 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10575 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10576 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10577 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10578 break;
10579 case SAVEt_OP:
10580 ptr = POPPTR(ss,ix);
10581 TOPPTR(nss,ix) = ptr;
10582 break;
10583 case SAVEt_HINTS:
10584 i = POPINT(ss,ix);
10585 TOPINT(nss,ix) = i;
b3ca2e83 10586 ptr = POPPTR(ss,ix);
080ac856 10587 if (ptr) {
7b6dd8c3 10588 HINTS_REFCNT_LOCK;
080ac856 10589 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10590 HINTS_REFCNT_UNLOCK;
10591 }
cbb1fbea 10592 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10593 if (i & HINT_LOCALIZE_HH) {
10594 hv = (HV*)POPPTR(ss,ix);
10595 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10596 }
1d7c1841 10597 break;
c4410b1b
GS
10598 case SAVEt_COMPPAD:
10599 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10600 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10601 break;
c3564e5c
GS
10602 case SAVEt_PADSV:
10603 longval = (long)POPLONG(ss,ix);
10604 TOPLONG(nss,ix) = longval;
10605 ptr = POPPTR(ss,ix);
10606 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10607 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10608 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10609 break;
a1bb4754 10610 case SAVEt_BOOL:
38d8b13e 10611 ptr = POPPTR(ss,ix);
b9609c01 10612 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10613 longval = (long)POPBOOL(ss,ix);
b9609c01 10614 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10615 break;
8bd2680e
MHM
10616 case SAVEt_SET_SVFLAGS:
10617 i = POPINT(ss,ix);
10618 TOPINT(nss,ix) = i;
10619 i = POPINT(ss,ix);
10620 TOPINT(nss,ix) = i;
10621 sv = (SV*)POPPTR(ss,ix);
10622 TOPPTR(nss,ix) = sv_dup(sv, param);
10623 break;
5bfb7d0e
NC
10624 case SAVEt_RE_STATE:
10625 {
10626 const struct re_save_state *const old_state
10627 = (struct re_save_state *)
10628 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10629 struct re_save_state *const new_state
10630 = (struct re_save_state *)
10631 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10632
10633 Copy(old_state, new_state, 1, struct re_save_state);
10634 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10635
10636 new_state->re_state_bostr
10637 = pv_dup(old_state->re_state_bostr);
10638 new_state->re_state_reginput
10639 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10640 new_state->re_state_regeol
10641 = pv_dup(old_state->re_state_regeol);
10642 new_state->re_state_regstartp
10643 = any_dup(old_state->re_state_regstartp, proto_perl);
10644 new_state->re_state_regendp
10645 = any_dup(old_state->re_state_regendp, proto_perl);
10646 new_state->re_state_reglastparen
10647 = any_dup(old_state->re_state_reglastparen, proto_perl);
10648 new_state->re_state_reglastcloseparen
10649 = any_dup(old_state->re_state_reglastcloseparen,
10650 proto_perl);
5bfb7d0e
NC
10651 /* XXX This just has to be broken. The old save_re_context
10652 code did SAVEGENERICPV(PL_reg_start_tmp);
10653 PL_reg_start_tmp is char **.
10654 Look above to what the dup code does for
10655 SAVEt_GENERIC_PVREF
10656 It can never have worked.
10657 So this is merely a faithful copy of the exiting bug: */
10658 new_state->re_state_reg_start_tmp
10659 = (char **) pv_dup((char *)
10660 old_state->re_state_reg_start_tmp);
10661 /* I assume that it only ever "worked" because no-one called
10662 (pseudo)fork while the regexp engine had re-entered itself.
10663 */
5bfb7d0e
NC
10664#ifdef PERL_OLD_COPY_ON_WRITE
10665 new_state->re_state_nrs
10666 = sv_dup(old_state->re_state_nrs, param);
10667#endif
10668 new_state->re_state_reg_magic
10669 = any_dup(old_state->re_state_reg_magic, proto_perl);
10670 new_state->re_state_reg_oldcurpm
10671 = any_dup(old_state->re_state_reg_oldcurpm, proto_perl);
10672 new_state->re_state_reg_curpm
10673 = any_dup(old_state->re_state_reg_curpm, proto_perl);
10674 new_state->re_state_reg_oldsaved
10675 = pv_dup(old_state->re_state_reg_oldsaved);
10676 new_state->re_state_reg_poscache
10677 = pv_dup(old_state->re_state_reg_poscache);
10678#ifdef DEBUGGING
10679 new_state->re_state_reg_starttry
10680 = pv_dup(old_state->re_state_reg_starttry);
10681#endif
10682 break;
10683 }
68da3b2f
NC
10684 case SAVEt_COMPILE_WARNINGS:
10685 ptr = POPPTR(ss,ix);
10686 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10687 break;
1d7c1841 10688 default:
ca05af4a 10689 Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i);
1d7c1841
GS
10690 }
10691 }
10692
bd81e77b
NC
10693 return nss;
10694}
10695
10696
10697/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10698 * flag to the result. This is done for each stash before cloning starts,
10699 * so we know which stashes want their objects cloned */
10700
10701static void
10702do_mark_cloneable_stash(pTHX_ SV *sv)
10703{
10704 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10705 if (hvname) {
10706 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10707 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10708 if (cloner && GvCV(cloner)) {
10709 dSP;
10710 UV status;
10711
10712 ENTER;
10713 SAVETMPS;
10714 PUSHMARK(SP);
10715 XPUSHs(sv_2mortal(newSVhek(hvname)));
10716 PUTBACK;
10717 call_sv((SV*)GvCV(cloner), G_SCALAR);
10718 SPAGAIN;
10719 status = POPu;
10720 PUTBACK;
10721 FREETMPS;
10722 LEAVE;
10723 if (status)
10724 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10725 }
10726 }
10727}
10728
10729
10730
10731/*
10732=for apidoc perl_clone
10733
10734Create and return a new interpreter by cloning the current one.
10735
10736perl_clone takes these flags as parameters:
10737
10738CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10739without it we only clone the data and zero the stacks,
10740with it we copy the stacks and the new perl interpreter is
10741ready to run at the exact same point as the previous one.
10742The pseudo-fork code uses COPY_STACKS while the
10743threads->new doesn't.
10744
10745CLONEf_KEEP_PTR_TABLE
10746perl_clone keeps a ptr_table with the pointer of the old
10747variable as a key and the new variable as a value,
10748this allows it to check if something has been cloned and not
10749clone it again but rather just use the value and increase the
10750refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10751the ptr_table using the function
10752C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10753reason to keep it around is if you want to dup some of your own
10754variable who are outside the graph perl scans, example of this
10755code is in threads.xs create
10756
10757CLONEf_CLONE_HOST
10758This is a win32 thing, it is ignored on unix, it tells perls
10759win32host code (which is c++) to clone itself, this is needed on
10760win32 if you want to run two threads at the same time,
10761if you just want to do some stuff in a separate perl interpreter
10762and then throw it away and return to the original one,
10763you don't need to do anything.
10764
10765=cut
10766*/
10767
10768/* XXX the above needs expanding by someone who actually understands it ! */
10769EXTERN_C PerlInterpreter *
10770perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10771
10772PerlInterpreter *
10773perl_clone(PerlInterpreter *proto_perl, UV flags)
10774{
10775 dVAR;
10776#ifdef PERL_IMPLICIT_SYS
10777
10778 /* perlhost.h so we need to call into it
10779 to clone the host, CPerlHost should have a c interface, sky */
10780
10781 if (flags & CLONEf_CLONE_HOST) {
10782 return perl_clone_host(proto_perl,flags);
10783 }
10784 return perl_clone_using(proto_perl, flags,
10785 proto_perl->IMem,
10786 proto_perl->IMemShared,
10787 proto_perl->IMemParse,
10788 proto_perl->IEnv,
10789 proto_perl->IStdIO,
10790 proto_perl->ILIO,
10791 proto_perl->IDir,
10792 proto_perl->ISock,
10793 proto_perl->IProc);
10794}
10795
10796PerlInterpreter *
10797perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10798 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10799 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10800 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10801 struct IPerlDir* ipD, struct IPerlSock* ipS,
10802 struct IPerlProc* ipP)
10803{
10804 /* XXX many of the string copies here can be optimized if they're
10805 * constants; they need to be allocated as common memory and just
10806 * their pointers copied. */
10807
10808 IV i;
10809 CLONE_PARAMS clone_params;
5f66b61c 10810 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10811
5f66b61c 10812 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10813 /* for each stash, determine whether its objects should be cloned */
10814 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10815 PERL_SET_THX(my_perl);
10816
10817# ifdef DEBUGGING
7e337ee0 10818 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10819 PL_op = NULL;
10820 PL_curcop = NULL;
bd81e77b
NC
10821 PL_markstack = 0;
10822 PL_scopestack = 0;
10823 PL_savestack = 0;
10824 PL_savestack_ix = 0;
10825 PL_savestack_max = -1;
10826 PL_sig_pending = 0;
10827 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10828# else /* !DEBUGGING */
10829 Zero(my_perl, 1, PerlInterpreter);
10830# endif /* DEBUGGING */
10831
10832 /* host pointers */
10833 PL_Mem = ipM;
10834 PL_MemShared = ipMS;
10835 PL_MemParse = ipMP;
10836 PL_Env = ipE;
10837 PL_StdIO = ipStd;
10838 PL_LIO = ipLIO;
10839 PL_Dir = ipD;
10840 PL_Sock = ipS;
10841 PL_Proc = ipP;
10842#else /* !PERL_IMPLICIT_SYS */
10843 IV i;
10844 CLONE_PARAMS clone_params;
10845 CLONE_PARAMS* param = &clone_params;
5f66b61c 10846 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10847 /* for each stash, determine whether its objects should be cloned */
10848 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10849 PERL_SET_THX(my_perl);
10850
10851# ifdef DEBUGGING
7e337ee0 10852 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10853 PL_op = NULL;
10854 PL_curcop = NULL;
bd81e77b
NC
10855 PL_markstack = 0;
10856 PL_scopestack = 0;
10857 PL_savestack = 0;
10858 PL_savestack_ix = 0;
10859 PL_savestack_max = -1;
10860 PL_sig_pending = 0;
10861 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10862# else /* !DEBUGGING */
10863 Zero(my_perl, 1, PerlInterpreter);
10864# endif /* DEBUGGING */
10865#endif /* PERL_IMPLICIT_SYS */
10866 param->flags = flags;
10867 param->proto_perl = proto_perl;
10868
7cb608b5
NC
10869 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10870
fdda85ca 10871 PL_body_arenas = NULL;
bd81e77b
NC
10872 Zero(&PL_body_roots, 1, PL_body_roots);
10873
10874 PL_nice_chunk = NULL;
10875 PL_nice_chunk_size = 0;
10876 PL_sv_count = 0;
10877 PL_sv_objcount = 0;
a0714e2c
SS
10878 PL_sv_root = NULL;
10879 PL_sv_arenaroot = NULL;
bd81e77b
NC
10880
10881 PL_debug = proto_perl->Idebug;
10882
10883 PL_hash_seed = proto_perl->Ihash_seed;
10884 PL_rehash_seed = proto_perl->Irehash_seed;
10885
10886#ifdef USE_REENTRANT_API
10887 /* XXX: things like -Dm will segfault here in perlio, but doing
10888 * PERL_SET_CONTEXT(proto_perl);
10889 * breaks too many other things
10890 */
10891 Perl_reentrant_init(aTHX);
10892#endif
10893
10894 /* create SV map for pointer relocation */
10895 PL_ptr_table = ptr_table_new();
10896
10897 /* initialize these special pointers as early as possible */
10898 SvANY(&PL_sv_undef) = NULL;
10899 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10900 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10901 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10902
10903 SvANY(&PL_sv_no) = new_XPVNV();
10904 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10905 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10906 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10907 SvPV_set(&PL_sv_no, SAVEPVN(PL_No, 0));
10908 SvCUR_set(&PL_sv_no, 0);
10909 SvLEN_set(&PL_sv_no, 1);
10910 SvIV_set(&PL_sv_no, 0);
10911 SvNV_set(&PL_sv_no, 0);
10912 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10913
10914 SvANY(&PL_sv_yes) = new_XPVNV();
10915 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10916 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10917 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10918 SvPV_set(&PL_sv_yes, SAVEPVN(PL_Yes, 1));
10919 SvCUR_set(&PL_sv_yes, 1);
10920 SvLEN_set(&PL_sv_yes, 2);
10921 SvIV_set(&PL_sv_yes, 1);
10922 SvNV_set(&PL_sv_yes, 1);
10923 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10924
10925 /* create (a non-shared!) shared string table */
10926 PL_strtab = newHV();
10927 HvSHAREKEYS_off(PL_strtab);
10928 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10929 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10930
10931 PL_compiling = proto_perl->Icompiling;
10932
10933 /* These two PVs will be free'd special way so must set them same way op.c does */
10934 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10935 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10936
10937 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10938 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10939
10940 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10941 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
bd81e77b
NC
10942 if (!specialCopIO(PL_compiling.cop_io))
10943 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
cbb1fbea
NC
10944 if (PL_compiling.cop_hints) {
10945 HINTS_REFCNT_LOCK;
10946 PL_compiling.cop_hints->refcounted_he_refcnt++;
10947 HINTS_REFCNT_UNLOCK;
10948 }
bd81e77b
NC
10949 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10950
10951 /* pseudo environmental stuff */
10952 PL_origargc = proto_perl->Iorigargc;
10953 PL_origargv = proto_perl->Iorigargv;
10954
10955 param->stashes = newAV(); /* Setup array of objects to call clone on */
10956
10957 /* Set tainting stuff before PerlIO_debug can possibly get called */
10958 PL_tainting = proto_perl->Itainting;
10959 PL_taint_warn = proto_perl->Itaint_warn;
10960
10961#ifdef PERLIO_LAYERS
10962 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10963 PerlIO_clone(aTHX_ proto_perl, param);
10964#endif
10965
10966 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10967 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10968 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10969 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10970 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10971 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10972
10973 /* switches */
10974 PL_minus_c = proto_perl->Iminus_c;
10975 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10976 PL_localpatches = proto_perl->Ilocalpatches;
10977 PL_splitstr = proto_perl->Isplitstr;
10978 PL_preprocess = proto_perl->Ipreprocess;
10979 PL_minus_n = proto_perl->Iminus_n;
10980 PL_minus_p = proto_perl->Iminus_p;
10981 PL_minus_l = proto_perl->Iminus_l;
10982 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10983 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10984 PL_minus_F = proto_perl->Iminus_F;
10985 PL_doswitches = proto_perl->Idoswitches;
10986 PL_dowarn = proto_perl->Idowarn;
10987 PL_doextract = proto_perl->Idoextract;
10988 PL_sawampersand = proto_perl->Isawampersand;
10989 PL_unsafe = proto_perl->Iunsafe;
10990 PL_inplace = SAVEPV(proto_perl->Iinplace);
10991 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
10992 PL_perldb = proto_perl->Iperldb;
10993 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
10994 PL_exit_flags = proto_perl->Iexit_flags;
10995
10996 /* magical thingies */
10997 /* XXX time(&PL_basetime) when asked for? */
10998 PL_basetime = proto_perl->Ibasetime;
10999 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11000
11001 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
11002 PL_statusvalue = proto_perl->Istatusvalue;
11003#ifdef VMS
11004 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11005#else
11006 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11007#endif
11008 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11009
11010 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
11011 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
11012 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
11013
11014 /* Clone the regex array */
11015 PL_regex_padav = newAV();
11016 {
11017 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 11018 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 11019 IV i;
7f466ec7 11020 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 11021 for(i = 1; i <= len; i++) {
7a5b473e
AL
11022 const SV * const regex = regexen[i];
11023 SV * const sv =
11024 SvREPADTMP(regex)
11025 ? sv_dup_inc(regex, param)
11026 : SvREFCNT_inc(
11027 newSViv(PTR2IV(re_dup(
11028 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11029 ;
11030 av_push(PL_regex_padav, sv);
bd81e77b
NC
11031 }
11032 }
11033 PL_regex_pad = AvARRAY(PL_regex_padav);
11034
11035 /* shortcuts to various I/O objects */
11036 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11037 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11038 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11039 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11040 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11041 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11042
bd81e77b
NC
11043 /* shortcuts to regexp stuff */
11044 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11045
bd81e77b
NC
11046 /* shortcuts to misc objects */
11047 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11048
bd81e77b
NC
11049 /* shortcuts to debugging objects */
11050 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11051 PL_DBline = gv_dup(proto_perl->IDBline, param);
11052 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11053 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11054 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11055 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11056 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11057 PL_lineary = av_dup(proto_perl->Ilineary, param);
11058 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11059
bd81e77b
NC
11060 /* symbol tables */
11061 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11062 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11063 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11064 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11065 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11066
11067 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11068 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11069 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11070 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11071 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11072 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11073
11074 PL_sub_generation = proto_perl->Isub_generation;
11075
11076 /* funky return mechanisms */
11077 PL_forkprocess = proto_perl->Iforkprocess;
11078
11079 /* subprocess state */
11080 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11081
11082 /* internal state */
11083 PL_maxo = proto_perl->Imaxo;
11084 if (proto_perl->Iop_mask)
11085 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11086 else
bd61b366 11087 PL_op_mask = NULL;
bd81e77b
NC
11088 /* PL_asserting = proto_perl->Iasserting; */
11089
11090 /* current interpreter roots */
11091 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
11092 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
11093 PL_main_start = proto_perl->Imain_start;
11094 PL_eval_root = proto_perl->Ieval_root;
11095 PL_eval_start = proto_perl->Ieval_start;
11096
11097 /* runtime control stuff */
11098 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11099 PL_copline = proto_perl->Icopline;
11100
11101 PL_filemode = proto_perl->Ifilemode;
11102 PL_lastfd = proto_perl->Ilastfd;
11103 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11104 PL_Argv = NULL;
bd61b366 11105 PL_Cmd = NULL;
bd81e77b
NC
11106 PL_gensym = proto_perl->Igensym;
11107 PL_preambled = proto_perl->Ipreambled;
11108 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11109 PL_laststatval = proto_perl->Ilaststatval;
11110 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11111 PL_mess_sv = NULL;
bd81e77b
NC
11112
11113 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11114
11115 /* interpreter atexit processing */
11116 PL_exitlistlen = proto_perl->Iexitlistlen;
11117 if (PL_exitlistlen) {
11118 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11119 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11120 }
bd81e77b
NC
11121 else
11122 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11123
11124 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11125 if (PL_my_cxt_size) {
f16dd614
DM
11126 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11127 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11128 }
11129 else
11130 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11131 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11132 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11133 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11134
11135 PL_profiledata = NULL;
11136 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11137 /* PL_rsfp_filters entries have fake IoDIRP() */
11138 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11139
bd81e77b 11140 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11141
bd81e77b 11142 PAD_CLONE_VARS(proto_perl, param);
9660f481 11143
bd81e77b
NC
11144#ifdef HAVE_INTERP_INTERN
11145 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11146#endif
645c22ef 11147
bd81e77b
NC
11148 /* more statics moved here */
11149 PL_generation = proto_perl->Igeneration;
11150 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11151
bd81e77b
NC
11152 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11153 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11154
bd81e77b
NC
11155 PL_uid = proto_perl->Iuid;
11156 PL_euid = proto_perl->Ieuid;
11157 PL_gid = proto_perl->Igid;
11158 PL_egid = proto_perl->Iegid;
11159 PL_nomemok = proto_perl->Inomemok;
11160 PL_an = proto_perl->Ian;
11161 PL_evalseq = proto_perl->Ievalseq;
11162 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11163 PL_origalen = proto_perl->Iorigalen;
11164#ifdef PERL_USES_PL_PIDSTATUS
11165 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11166#endif
11167 PL_osname = SAVEPV(proto_perl->Iosname);
11168 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11169
bd81e77b 11170 PL_runops = proto_perl->Irunops;
6a78b4db 11171
bd81e77b 11172 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11173
bd81e77b
NC
11174#ifdef CSH
11175 PL_cshlen = proto_perl->Icshlen;
11176 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11177#endif
645c22ef 11178
bd81e77b
NC
11179 PL_lex_state = proto_perl->Ilex_state;
11180 PL_lex_defer = proto_perl->Ilex_defer;
11181 PL_lex_expect = proto_perl->Ilex_expect;
11182 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11183 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11184 PL_lex_starts = proto_perl->Ilex_starts;
11185 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11186 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11187 PL_lex_op = proto_perl->Ilex_op;
11188 PL_lex_inpat = proto_perl->Ilex_inpat;
11189 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11190 PL_lex_brackets = proto_perl->Ilex_brackets;
11191 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11192 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11193 PL_lex_casemods = proto_perl->Ilex_casemods;
11194 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11195 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11196
5db06880
NC
11197#ifdef PERL_MAD
11198 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11199 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11200 PL_realtokenstart = proto_perl->Irealtokenstart;
11201 PL_faketokens = proto_perl->Ifaketokens;
11202 PL_thismad = proto_perl->Ithismad;
11203 PL_thistoken = proto_perl->Ithistoken;
11204 PL_thisopen = proto_perl->Ithisopen;
11205 PL_thisstuff = proto_perl->Ithisstuff;
11206 PL_thisclose = proto_perl->Ithisclose;
11207 PL_thiswhite = proto_perl->Ithiswhite;
11208 PL_nextwhite = proto_perl->Inextwhite;
11209 PL_skipwhite = proto_perl->Iskipwhite;
11210 PL_endwhite = proto_perl->Iendwhite;
11211 PL_curforce = proto_perl->Icurforce;
5db06880 11212#else
bd81e77b
NC
11213 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11214 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11215 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11216#endif
c43294b8 11217
bd81e77b
NC
11218 /* XXX This is probably masking the deeper issue of why
11219 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11220 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11221 * (A little debugging with a watchpoint on it may help.)
11222 */
11223 if (SvANY(proto_perl->Ilinestr)) {
11224 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11225 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11226 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11227 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11228 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11229 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11230 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11231 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11232 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11233 }
11234 else {
561b68a9 11235 PL_linestr = newSV(79);
bd81e77b
NC
11236 sv_upgrade(PL_linestr,SVt_PVIV);
11237 sv_setpvn(PL_linestr,"",0);
11238 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11239 }
11240 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11241 PL_pending_ident = proto_perl->Ipending_ident;
11242 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11243
11244 PL_expect = proto_perl->Iexpect;
11245
11246 PL_multi_start = proto_perl->Imulti_start;
11247 PL_multi_end = proto_perl->Imulti_end;
11248 PL_multi_open = proto_perl->Imulti_open;
11249 PL_multi_close = proto_perl->Imulti_close;
11250
11251 PL_error_count = proto_perl->Ierror_count;
11252 PL_subline = proto_perl->Isubline;
11253 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11254
bd81e77b
NC
11255 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11256 if (SvANY(proto_perl->Ilinestr)) {
11257 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11258 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11259 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11260 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11261 PL_last_lop_op = proto_perl->Ilast_lop_op;
11262 }
11263 else {
11264 PL_last_uni = SvPVX(PL_linestr);
11265 PL_last_lop = SvPVX(PL_linestr);
11266 PL_last_lop_op = 0;
11267 }
11268 PL_in_my = proto_perl->Iin_my;
11269 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11270#ifdef FCRYPT
11271 PL_cryptseen = proto_perl->Icryptseen;
11272#endif
1d7c1841 11273
bd81e77b 11274 PL_hints = proto_perl->Ihints;
1d7c1841 11275
bd81e77b 11276 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11277
bd81e77b
NC
11278#ifdef USE_LOCALE_COLLATE
11279 PL_collation_ix = proto_perl->Icollation_ix;
11280 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11281 PL_collation_standard = proto_perl->Icollation_standard;
11282 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11283 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11284#endif /* USE_LOCALE_COLLATE */
1d7c1841 11285
bd81e77b
NC
11286#ifdef USE_LOCALE_NUMERIC
11287 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11288 PL_numeric_standard = proto_perl->Inumeric_standard;
11289 PL_numeric_local = proto_perl->Inumeric_local;
11290 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11291#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11292
bd81e77b
NC
11293 /* utf8 character classes */
11294 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11295 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11296 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11297 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11298 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11299 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11300 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11301 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11302 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11303 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11304 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11305 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11306 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11307 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11308 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11309 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11310 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11311 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11312 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11313 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11314
bd81e77b
NC
11315 /* Did the locale setup indicate UTF-8? */
11316 PL_utf8locale = proto_perl->Iutf8locale;
11317 /* Unicode features (see perlrun/-C) */
11318 PL_unicode = proto_perl->Iunicode;
1d7c1841 11319
bd81e77b
NC
11320 /* Pre-5.8 signals control */
11321 PL_signals = proto_perl->Isignals;
1d7c1841 11322
bd81e77b
NC
11323 /* times() ticks per second */
11324 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11325
bd81e77b
NC
11326 /* Recursion stopper for PerlIO_find_layer */
11327 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11328
bd81e77b
NC
11329 /* sort() routine */
11330 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11331
bd81e77b
NC
11332 /* Not really needed/useful since the reenrant_retint is "volatile",
11333 * but do it for consistency's sake. */
11334 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11335
bd81e77b
NC
11336 /* Hooks to shared SVs and locks. */
11337 PL_sharehook = proto_perl->Isharehook;
11338 PL_lockhook = proto_perl->Ilockhook;
11339 PL_unlockhook = proto_perl->Iunlockhook;
11340 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11341
bd81e77b
NC
11342 PL_runops_std = proto_perl->Irunops_std;
11343 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11344
bd81e77b
NC
11345#ifdef THREADS_HAVE_PIDS
11346 PL_ppid = proto_perl->Ippid;
11347#endif
1d7c1841 11348
bd81e77b 11349 /* swatch cache */
5c284bb0 11350 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11351 PL_last_swash_klen = 0;
11352 PL_last_swash_key[0]= '\0';
11353 PL_last_swash_tmps = (U8*)NULL;
11354 PL_last_swash_slen = 0;
1d7c1841 11355
bd81e77b
NC
11356 PL_glob_index = proto_perl->Iglob_index;
11357 PL_srand_called = proto_perl->Isrand_called;
11358 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11359 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11360
bd81e77b
NC
11361 if (proto_perl->Ipsig_pend) {
11362 Newxz(PL_psig_pend, SIG_SIZE, int);
11363 }
11364 else {
11365 PL_psig_pend = (int*)NULL;
11366 }
05ec9bb3 11367
bd81e77b
NC
11368 if (proto_perl->Ipsig_ptr) {
11369 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11370 Newxz(PL_psig_name, SIG_SIZE, SV*);
11371 for (i = 1; i < SIG_SIZE; i++) {
11372 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11373 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11374 }
11375 }
11376 else {
11377 PL_psig_ptr = (SV**)NULL;
11378 PL_psig_name = (SV**)NULL;
11379 }
05ec9bb3 11380
bd81e77b 11381 /* thrdvar.h stuff */
1d7c1841 11382
bd81e77b
NC
11383 if (flags & CLONEf_COPY_STACKS) {
11384 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11385 PL_tmps_ix = proto_perl->Ttmps_ix;
11386 PL_tmps_max = proto_perl->Ttmps_max;
11387 PL_tmps_floor = proto_perl->Ttmps_floor;
11388 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11389 i = 0;
11390 while (i <= PL_tmps_ix) {
11391 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11392 ++i;
11393 }
d2d73c3e 11394
bd81e77b
NC
11395 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11396 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11397 Newxz(PL_markstack, i, I32);
11398 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11399 - proto_perl->Tmarkstack);
11400 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11401 - proto_perl->Tmarkstack);
11402 Copy(proto_perl->Tmarkstack, PL_markstack,
11403 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11404
bd81e77b
NC
11405 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11406 * NOTE: unlike the others! */
11407 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11408 PL_scopestack_max = proto_perl->Tscopestack_max;
11409 Newxz(PL_scopestack, PL_scopestack_max, I32);
11410 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11411
bd81e77b
NC
11412 /* NOTE: si_dup() looks at PL_markstack */
11413 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11414
bd81e77b
NC
11415 /* PL_curstack = PL_curstackinfo->si_stack; */
11416 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11417 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11418
bd81e77b
NC
11419 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11420 PL_stack_base = AvARRAY(PL_curstack);
11421 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11422 - proto_perl->Tstack_base);
11423 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11424
bd81e77b
NC
11425 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11426 * NOTE: unlike the others! */
11427 PL_savestack_ix = proto_perl->Tsavestack_ix;
11428 PL_savestack_max = proto_perl->Tsavestack_max;
11429 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11430 PL_savestack = ss_dup(proto_perl, param);
11431 }
11432 else {
11433 init_stacks();
11434 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11435
11436 /* although we're not duplicating the tmps stack, we should still
11437 * add entries for any SVs on the tmps stack that got cloned by a
11438 * non-refcount means (eg a temp in @_); otherwise they will be
11439 * orphaned
11440 */
11441 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11442 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11443 proto_perl->Ttmps_stack[i]);
11444 if (nsv && !SvREFCNT(nsv)) {
11445 EXTEND_MORTAL(1);
b37c2d43 11446 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11447 }
11448 }
bd81e77b 11449 }
1d7c1841 11450
bd81e77b
NC
11451 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11452 PL_top_env = &PL_start_env;
1d7c1841 11453
bd81e77b 11454 PL_op = proto_perl->Top;
4a4c6fe3 11455
a0714e2c 11456 PL_Sv = NULL;
bd81e77b
NC
11457 PL_Xpv = (XPV*)NULL;
11458 PL_na = proto_perl->Tna;
1fcf4c12 11459
bd81e77b
NC
11460 PL_statbuf = proto_perl->Tstatbuf;
11461 PL_statcache = proto_perl->Tstatcache;
11462 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11463 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11464#ifdef HAS_TIMES
11465 PL_timesbuf = proto_perl->Ttimesbuf;
11466#endif
1d7c1841 11467
bd81e77b
NC
11468 PL_tainted = proto_perl->Ttainted;
11469 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11470 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11471 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11472 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11473 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11474 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11475 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11476 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11477 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11478
bd81e77b
NC
11479 PL_restartop = proto_perl->Trestartop;
11480 PL_in_eval = proto_perl->Tin_eval;
11481 PL_delaymagic = proto_perl->Tdelaymagic;
11482 PL_dirty = proto_perl->Tdirty;
11483 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11484
bd81e77b 11485 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11486 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11487 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11488 PL_lastgotoprobe = NULL;
bd81e77b 11489 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11490
bd81e77b
NC
11491 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11492 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11493 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11494 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11495 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11496 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11497
bd81e77b 11498 /* regex stuff */
1d7c1841 11499
bd81e77b
NC
11500 PL_screamfirst = NULL;
11501 PL_screamnext = NULL;
11502 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11503 PL_lastscream = NULL;
1d7c1841 11504
bd81e77b 11505 PL_watchaddr = NULL;
bd61b366 11506 PL_watchok = NULL;
1d7c1841 11507
bd81e77b 11508 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11509 PL_colorset = 0; /* reinits PL_colors[] */
11510 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11511
bd81e77b
NC
11512 /* RE engine - function pointers */
11513 PL_regcompp = proto_perl->Tregcompp;
11514 PL_regexecp = proto_perl->Tregexecp;
11515 PL_regint_start = proto_perl->Tregint_start;
11516 PL_regint_string = proto_perl->Tregint_string;
11517 PL_regfree = proto_perl->Tregfree;
46ab3289 11518 Zero(&PL_reg_state, 1, struct re_save_state);
bd81e77b 11519 PL_reginterp_cnt = 0;
5d9a96ca 11520 PL_regmatch_slab = NULL;
1d7c1841 11521
bd81e77b
NC
11522 /* Pluggable optimizer */
11523 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11524
bd81e77b 11525 PL_stashcache = newHV();
1d7c1841 11526
bd81e77b
NC
11527 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11528 ptr_table_free(PL_ptr_table);
11529 PL_ptr_table = NULL;
11530 }
1d7c1841 11531
bd81e77b
NC
11532 /* Call the ->CLONE method, if it exists, for each of the stashes
11533 identified by sv_dup() above.
11534 */
11535 while(av_len(param->stashes) != -1) {
11536 HV* const stash = (HV*) av_shift(param->stashes);
11537 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11538 if (cloner && GvCV(cloner)) {
11539 dSP;
11540 ENTER;
11541 SAVETMPS;
11542 PUSHMARK(SP);
11543 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11544 PUTBACK;
11545 call_sv((SV*)GvCV(cloner), G_DISCARD);
11546 FREETMPS;
11547 LEAVE;
11548 }
1d7c1841 11549 }
1d7c1841 11550
bd81e77b 11551 SvREFCNT_dec(param->stashes);
1d7c1841 11552
bd81e77b
NC
11553 /* orphaned? eg threads->new inside BEGIN or use */
11554 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11555 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11556 SAVEFREESV(PL_compcv);
11557 }
dd2155a4 11558
bd81e77b
NC
11559 return my_perl;
11560}
1d7c1841 11561
bd81e77b 11562#endif /* USE_ITHREADS */
1d7c1841 11563
bd81e77b
NC
11564/*
11565=head1 Unicode Support
1d7c1841 11566
bd81e77b 11567=for apidoc sv_recode_to_utf8
1d7c1841 11568
bd81e77b
NC
11569The encoding is assumed to be an Encode object, on entry the PV
11570of the sv is assumed to be octets in that encoding, and the sv
11571will be converted into Unicode (and UTF-8).
1d7c1841 11572
bd81e77b
NC
11573If the sv already is UTF-8 (or if it is not POK), or if the encoding
11574is not a reference, nothing is done to the sv. If the encoding is not
11575an C<Encode::XS> Encoding object, bad things will happen.
11576(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11577
bd81e77b 11578The PV of the sv is returned.
1d7c1841 11579
bd81e77b 11580=cut */
1d7c1841 11581
bd81e77b
NC
11582char *
11583Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11584{
11585 dVAR;
11586 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11587 SV *uni;
11588 STRLEN len;
11589 const char *s;
11590 dSP;
11591 ENTER;
11592 SAVETMPS;
11593 save_re_context();
11594 PUSHMARK(sp);
11595 EXTEND(SP, 3);
11596 XPUSHs(encoding);
11597 XPUSHs(sv);
11598/*
11599 NI-S 2002/07/09
11600 Passing sv_yes is wrong - it needs to be or'ed set of constants
11601 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11602 remove converted chars from source.
1d7c1841 11603
bd81e77b 11604 Both will default the value - let them.
1d7c1841 11605
bd81e77b
NC
11606 XPUSHs(&PL_sv_yes);
11607*/
11608 PUTBACK;
11609 call_method("decode", G_SCALAR);
11610 SPAGAIN;
11611 uni = POPs;
11612 PUTBACK;
11613 s = SvPV_const(uni, len);
11614 if (s != SvPVX_const(sv)) {
11615 SvGROW(sv, len + 1);
11616 Move(s, SvPVX(sv), len + 1, char);
11617 SvCUR_set(sv, len);
11618 }
11619 FREETMPS;
11620 LEAVE;
11621 SvUTF8_on(sv);
11622 return SvPVX(sv);
389edf32 11623 }
bd81e77b
NC
11624 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11625}
1d7c1841 11626
bd81e77b
NC
11627/*
11628=for apidoc sv_cat_decode
1d7c1841 11629
bd81e77b
NC
11630The encoding is assumed to be an Encode object, the PV of the ssv is
11631assumed to be octets in that encoding and decoding the input starts
11632from the position which (PV + *offset) pointed to. The dsv will be
11633concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11634when the string tstr appears in decoding output or the input ends on
11635the PV of the ssv. The value which the offset points will be modified
11636to the last input position on the ssv.
1d7c1841 11637
bd81e77b 11638Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11639
bd81e77b
NC
11640=cut */
11641
11642bool
11643Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11644 SV *ssv, int *offset, char *tstr, int tlen)
11645{
11646 dVAR;
11647 bool ret = FALSE;
11648 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11649 SV *offsv;
11650 dSP;
11651 ENTER;
11652 SAVETMPS;
11653 save_re_context();
11654 PUSHMARK(sp);
11655 EXTEND(SP, 6);
11656 XPUSHs(encoding);
11657 XPUSHs(dsv);
11658 XPUSHs(ssv);
11659 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11660 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11661 PUTBACK;
11662 call_method("cat_decode", G_SCALAR);
11663 SPAGAIN;
11664 ret = SvTRUE(TOPs);
11665 *offset = SvIV(offsv);
11666 PUTBACK;
11667 FREETMPS;
11668 LEAVE;
389edf32 11669 }
bd81e77b
NC
11670 else
11671 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11672 return ret;
1d7c1841 11673
bd81e77b 11674}
1d7c1841 11675
bd81e77b
NC
11676/* ---------------------------------------------------------------------
11677 *
11678 * support functions for report_uninit()
11679 */
1d7c1841 11680
bd81e77b
NC
11681/* the maxiumum size of array or hash where we will scan looking
11682 * for the undefined element that triggered the warning */
1d7c1841 11683
bd81e77b 11684#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11685
bd81e77b
NC
11686/* Look for an entry in the hash whose value has the same SV as val;
11687 * If so, return a mortal copy of the key. */
1d7c1841 11688
bd81e77b
NC
11689STATIC SV*
11690S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11691{
11692 dVAR;
11693 register HE **array;
11694 I32 i;
6c3182a5 11695
bd81e77b
NC
11696 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11697 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11698 return NULL;
6c3182a5 11699
bd81e77b 11700 array = HvARRAY(hv);
6c3182a5 11701
bd81e77b
NC
11702 for (i=HvMAX(hv); i>0; i--) {
11703 register HE *entry;
11704 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11705 if (HeVAL(entry) != val)
11706 continue;
11707 if ( HeVAL(entry) == &PL_sv_undef ||
11708 HeVAL(entry) == &PL_sv_placeholder)
11709 continue;
11710 if (!HeKEY(entry))
a0714e2c 11711 return NULL;
bd81e77b
NC
11712 if (HeKLEN(entry) == HEf_SVKEY)
11713 return sv_mortalcopy(HeKEY_sv(entry));
11714 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11715 }
11716 }
a0714e2c 11717 return NULL;
bd81e77b 11718}
6c3182a5 11719
bd81e77b
NC
11720/* Look for an entry in the array whose value has the same SV as val;
11721 * If so, return the index, otherwise return -1. */
6c3182a5 11722
bd81e77b
NC
11723STATIC I32
11724S_find_array_subscript(pTHX_ AV *av, SV* val)
11725{
97aff369 11726 dVAR;
bd81e77b
NC
11727 SV** svp;
11728 I32 i;
11729 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11730 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11731 return -1;
57c6e6d2 11732
bd81e77b
NC
11733 svp = AvARRAY(av);
11734 for (i=AvFILLp(av); i>=0; i--) {
11735 if (svp[i] == val && svp[i] != &PL_sv_undef)
11736 return i;
11737 }
11738 return -1;
11739}
15a5279a 11740
bd81e77b
NC
11741/* S_varname(): return the name of a variable, optionally with a subscript.
11742 * If gv is non-zero, use the name of that global, along with gvtype (one
11743 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11744 * targ. Depending on the value of the subscript_type flag, return:
11745 */
bce260cd 11746
bd81e77b
NC
11747#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11748#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11749#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11750#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11751
bd81e77b
NC
11752STATIC SV*
11753S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11754 SV* keyname, I32 aindex, int subscript_type)
11755{
1d7c1841 11756
bd81e77b
NC
11757 SV * const name = sv_newmortal();
11758 if (gv) {
11759 char buffer[2];
11760 buffer[0] = gvtype;
11761 buffer[1] = 0;
1d7c1841 11762
bd81e77b 11763 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11764
bd81e77b 11765 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11766
bd81e77b
NC
11767 if ((unsigned int)SvPVX(name)[1] <= 26) {
11768 buffer[0] = '^';
11769 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11770
bd81e77b
NC
11771 /* Swap the 1 unprintable control character for the 2 byte pretty
11772 version - ie substr($name, 1, 1) = $buffer; */
11773 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11774 }
bd81e77b
NC
11775 }
11776 else {
11777 U32 unused;
11778 CV * const cv = find_runcv(&unused);
11779 SV *sv;
11780 AV *av;
1d7c1841 11781
bd81e77b 11782 if (!cv || !CvPADLIST(cv))
a0714e2c 11783 return NULL;
bd81e77b
NC
11784 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11785 sv = *av_fetch(av, targ, FALSE);
11786 /* SvLEN in a pad name is not to be trusted */
11787 sv_setpv(name, SvPV_nolen_const(sv));
11788 }
1d7c1841 11789
bd81e77b 11790 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11791 SV * const sv = newSV(0);
bd81e77b
NC
11792 *SvPVX(name) = '$';
11793 Perl_sv_catpvf(aTHX_ name, "{%s}",
11794 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11795 SvREFCNT_dec(sv);
11796 }
11797 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11798 *SvPVX(name) = '$';
11799 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11800 }
11801 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11802 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11803
bd81e77b
NC
11804 return name;
11805}
1d7c1841 11806
1d7c1841 11807
bd81e77b
NC
11808/*
11809=for apidoc find_uninit_var
1d7c1841 11810
bd81e77b
NC
11811Find the name of the undefined variable (if any) that caused the operator o
11812to issue a "Use of uninitialized value" warning.
11813If match is true, only return a name if it's value matches uninit_sv.
11814So roughly speaking, if a unary operator (such as OP_COS) generates a
11815warning, then following the direct child of the op may yield an
11816OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11817other hand, with OP_ADD there are two branches to follow, so we only print
11818the variable name if we get an exact match.
1d7c1841 11819
bd81e77b 11820The name is returned as a mortal SV.
1d7c1841 11821
bd81e77b
NC
11822Assumes that PL_op is the op that originally triggered the error, and that
11823PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11824
bd81e77b
NC
11825=cut
11826*/
1d7c1841 11827
bd81e77b
NC
11828STATIC SV *
11829S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11830{
11831 dVAR;
11832 SV *sv;
11833 AV *av;
11834 GV *gv;
11835 OP *o, *o2, *kid;
1d7c1841 11836
bd81e77b
NC
11837 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11838 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11839 return NULL;
1d7c1841 11840
bd81e77b 11841 switch (obase->op_type) {
1d7c1841 11842
bd81e77b
NC
11843 case OP_RV2AV:
11844 case OP_RV2HV:
11845 case OP_PADAV:
11846 case OP_PADHV:
11847 {
11848 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11849 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11850 I32 index = 0;
a0714e2c 11851 SV *keysv = NULL;
bd81e77b 11852 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11853
bd81e77b
NC
11854 if (pad) { /* @lex, %lex */
11855 sv = PAD_SVl(obase->op_targ);
a0714e2c 11856 gv = NULL;
bd81e77b
NC
11857 }
11858 else {
11859 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11860 /* @global, %global */
11861 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11862 if (!gv)
11863 break;
11864 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11865 }
11866 else /* @{expr}, %{expr} */
11867 return find_uninit_var(cUNOPx(obase)->op_first,
11868 uninit_sv, match);
11869 }
1d7c1841 11870
bd81e77b
NC
11871 /* attempt to find a match within the aggregate */
11872 if (hash) {
d4c19fe8 11873 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11874 if (keysv)
11875 subscript_type = FUV_SUBSCRIPT_HASH;
11876 }
11877 else {
11878 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv);
11879 if (index >= 0)
11880 subscript_type = FUV_SUBSCRIPT_ARRAY;
11881 }
1d7c1841 11882
bd81e77b
NC
11883 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11884 break;
1d7c1841 11885
bd81e77b
NC
11886 return varname(gv, hash ? '%' : '@', obase->op_targ,
11887 keysv, index, subscript_type);
11888 }
1d7c1841 11889
bd81e77b
NC
11890 case OP_PADSV:
11891 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11892 break;
a0714e2c
SS
11893 return varname(NULL, '$', obase->op_targ,
11894 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11895
bd81e77b
NC
11896 case OP_GVSV:
11897 gv = cGVOPx_gv(obase);
11898 if (!gv || (match && GvSV(gv) != uninit_sv))
11899 break;
a0714e2c 11900 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11901
bd81e77b
NC
11902 case OP_AELEMFAST:
11903 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11904 if (match) {
11905 SV **svp;
11906 av = (AV*)PAD_SV(obase->op_targ);
11907 if (!av || SvRMAGICAL(av))
11908 break;
11909 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11910 if (!svp || *svp != uninit_sv)
11911 break;
11912 }
a0714e2c
SS
11913 return varname(NULL, '$', obase->op_targ,
11914 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11915 }
11916 else {
11917 gv = cGVOPx_gv(obase);
11918 if (!gv)
11919 break;
11920 if (match) {
11921 SV **svp;
11922 av = GvAV(gv);
11923 if (!av || SvRMAGICAL(av))
11924 break;
11925 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11926 if (!svp || *svp != uninit_sv)
11927 break;
11928 }
11929 return varname(gv, '$', 0,
a0714e2c 11930 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11931 }
11932 break;
1d7c1841 11933
bd81e77b
NC
11934 case OP_EXISTS:
11935 o = cUNOPx(obase)->op_first;
11936 if (!o || o->op_type != OP_NULL ||
11937 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11938 break;
11939 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11940
bd81e77b
NC
11941 case OP_AELEM:
11942 case OP_HELEM:
11943 if (PL_op == obase)
11944 /* $a[uninit_expr] or $h{uninit_expr} */
11945 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11946
a0714e2c 11947 gv = NULL;
bd81e77b
NC
11948 o = cBINOPx(obase)->op_first;
11949 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11950
bd81e77b 11951 /* get the av or hv, and optionally the gv */
a0714e2c 11952 sv = NULL;
bd81e77b
NC
11953 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11954 sv = PAD_SV(o->op_targ);
11955 }
11956 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11957 && cUNOPo->op_first->op_type == OP_GV)
11958 {
11959 gv = cGVOPx_gv(cUNOPo->op_first);
11960 if (!gv)
11961 break;
11962 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11963 }
11964 if (!sv)
11965 break;
11966
11967 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11968 /* index is constant */
11969 if (match) {
11970 if (SvMAGICAL(sv))
11971 break;
11972 if (obase->op_type == OP_HELEM) {
11973 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11974 if (!he || HeVAL(he) != uninit_sv)
11975 break;
11976 }
11977 else {
00b6aa41 11978 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11979 if (!svp || *svp != uninit_sv)
11980 break;
11981 }
11982 }
11983 if (obase->op_type == OP_HELEM)
11984 return varname(gv, '%', o->op_targ,
11985 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11986 else
a0714e2c 11987 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 11988 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11989 }
11990 else {
11991 /* index is an expression;
11992 * attempt to find a match within the aggregate */
11993 if (obase->op_type == OP_HELEM) {
d4c19fe8 11994 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11995 if (keysv)
11996 return varname(gv, '%', o->op_targ,
11997 keysv, 0, FUV_SUBSCRIPT_HASH);
11998 }
11999 else {
d4c19fe8 12000 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
12001 if (index >= 0)
12002 return varname(gv, '@', o->op_targ,
a0714e2c 12003 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12004 }
12005 if (match)
12006 break;
12007 return varname(gv,
12008 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12009 ? '@' : '%',
a0714e2c 12010 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 12011 }
bd81e77b 12012 break;
dc507217 12013
bd81e77b
NC
12014 case OP_AASSIGN:
12015 /* only examine RHS */
12016 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 12017
bd81e77b
NC
12018 case OP_OPEN:
12019 o = cUNOPx(obase)->op_first;
12020 if (o->op_type == OP_PUSHMARK)
12021 o = o->op_sibling;
1d7c1841 12022
bd81e77b
NC
12023 if (!o->op_sibling) {
12024 /* one-arg version of open is highly magical */
a0ae6670 12025
bd81e77b
NC
12026 if (o->op_type == OP_GV) { /* open FOO; */
12027 gv = cGVOPx_gv(o);
12028 if (match && GvSV(gv) != uninit_sv)
12029 break;
12030 return varname(gv, '$', 0,
a0714e2c 12031 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12032 }
12033 /* other possibilities not handled are:
12034 * open $x; or open my $x; should return '${*$x}'
12035 * open expr; should return '$'.expr ideally
12036 */
12037 break;
12038 }
12039 goto do_op;
ccfc67b7 12040
bd81e77b
NC
12041 /* ops where $_ may be an implicit arg */
12042 case OP_TRANS:
12043 case OP_SUBST:
12044 case OP_MATCH:
12045 if ( !(obase->op_flags & OPf_STACKED)) {
12046 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12047 ? PAD_SVl(obase->op_targ)
12048 : DEFSV))
12049 {
12050 sv = sv_newmortal();
12051 sv_setpvn(sv, "$_", 2);
12052 return sv;
12053 }
12054 }
12055 goto do_op;
9f4817db 12056
bd81e77b
NC
12057 case OP_PRTF:
12058 case OP_PRINT:
12059 /* skip filehandle as it can't produce 'undef' warning */
12060 o = cUNOPx(obase)->op_first;
12061 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12062 o = o->op_sibling->op_sibling;
12063 goto do_op2;
9f4817db 12064
9f4817db 12065
bd81e77b
NC
12066 case OP_RV2SV:
12067 case OP_CUSTOM:
12068 case OP_ENTERSUB:
12069 match = 1; /* XS or custom code could trigger random warnings */
12070 goto do_op;
9f4817db 12071
bd81e77b
NC
12072 case OP_SCHOMP:
12073 case OP_CHOMP:
12074 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12075 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12076 /*FALLTHROUGH*/
5d170f3a 12077
bd81e77b
NC
12078 default:
12079 do_op:
12080 if (!(obase->op_flags & OPf_KIDS))
12081 break;
12082 o = cUNOPx(obase)->op_first;
12083
12084 do_op2:
12085 if (!o)
12086 break;
f9893866 12087
bd81e77b
NC
12088 /* if all except one arg are constant, or have no side-effects,
12089 * or are optimized away, then it's unambiguous */
5f66b61c 12090 o2 = NULL;
bd81e77b 12091 for (kid=o; kid; kid = kid->op_sibling) {
d4f41c01 12092 SV *sv;
bd81e77b 12093 if (kid &&
d4f41c01
NC
12094 ( (kid->op_type == OP_CONST && (sv = cSVOPx_sv(kid))
12095 && SvOK(sv))
bd81e77b
NC
12096 || (kid->op_type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12097 || (kid->op_type == OP_PUSHMARK)
12098 )
12099 )
12100 continue;
12101 if (o2) { /* more than one found */
5f66b61c 12102 o2 = NULL;
bd81e77b
NC
12103 break;
12104 }
12105 o2 = kid;
12106 }
12107 if (o2)
12108 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12109
bd81e77b
NC
12110 /* scan all args */
12111 while (o) {
12112 sv = find_uninit_var(o, uninit_sv, 1);
12113 if (sv)
12114 return sv;
12115 o = o->op_sibling;
d0063567 12116 }
bd81e77b 12117 break;
f9893866 12118 }
a0714e2c 12119 return NULL;
9f4817db
JH
12120}
12121
220e2d4e 12122
bd81e77b
NC
12123/*
12124=for apidoc report_uninit
68795e93 12125
bd81e77b 12126Print appropriate "Use of uninitialized variable" warning
220e2d4e 12127
bd81e77b
NC
12128=cut
12129*/
220e2d4e 12130
bd81e77b
NC
12131void
12132Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12133{
97aff369 12134 dVAR;
bd81e77b 12135 if (PL_op) {
a0714e2c 12136 SV* varname = NULL;
bd81e77b
NC
12137 if (uninit_sv) {
12138 varname = find_uninit_var(PL_op, uninit_sv,0);
12139 if (varname)
12140 sv_insert(varname, 0, 0, " ", 1);
12141 }
12142 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12143 varname ? SvPV_nolen_const(varname) : "",
12144 " in ", OP_DESC(PL_op));
220e2d4e 12145 }
a73e8557 12146 else
bd81e77b
NC
12147 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12148 "", "", "");
220e2d4e 12149}
f9893866 12150
241d1a3b
NC
12151/*
12152 * Local variables:
12153 * c-indentation-style: bsd
12154 * c-basic-offset: 4
12155 * indent-tabs-mode: t
12156 * End:
12157 *
37442d52
RGS
12158 * ex: set ts=8 sts=4 sw=4 noet:
12159 */