This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
win32/Makefile changes for Windows 64 bit
[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 3887/*
47518d95 3888=for apidoc sv_usepvn_flags
954c1994 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
c1c21316
NC
3893by C<malloc>. The string length, C<len>, must be supplied. By default
3894this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
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"
c1c21316
NC
3897that pointer (e.g. ptr + 1) be used.
3898
3899If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
3900SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
3901I<may> be skipped. (i.e. the buffer is actually at least 1 byte longer than
3902C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
3903
3904=cut
3905*/
3906
ef50df4b 3907void
47518d95 3908Perl_sv_usepvn_flags(pTHX_ SV *sv, char *ptr, STRLEN len, U32 flags)
463ee0b2 3909{
97aff369 3910 dVAR;
1936d2a7 3911 STRLEN allocate;
765f542d 3912 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3913 SvUPGRADE(sv, SVt_PV);
463ee0b2 3914 if (!ptr) {
a0d0e21e 3915 (void)SvOK_off(sv);
47518d95
NC
3916 if (flags & SV_SMAGIC)
3917 SvSETMAGIC(sv);
463ee0b2
LW
3918 return;
3919 }
3f7c398e 3920 if (SvPVX_const(sv))
8bd4d4c5 3921 SvPV_free(sv);
1936d2a7 3922
2e90b4cd
NC
3923 if (flags & SV_HAS_TRAILING_NUL)
3924 assert(ptr[len] == '\0');
3925
c1c21316 3926 allocate = (flags & SV_HAS_TRAILING_NUL)
8f01dc65 3927 ? len + 1: PERL_STRLEN_ROUNDUP(len + 1);
69d25b4f
NC
3928#ifdef DEBUGGING
3929 {
3930 /* Force a move to shake out bugs in callers. */
3931 char *new_ptr = safemalloc(allocate);
3932 Copy(ptr, new_ptr, len, char);
3933 PoisonFree(ptr,len,char);
3934 Safefree(ptr);
3935 ptr = new_ptr;
3936 }
3937#else
c1c21316
NC
3938 if (!(flags & SV_HAS_TRAILING_NUL)) {
3939 ptr = saferealloc (ptr, allocate);
3940 }
69d25b4f 3941#endif
f880fe2f 3942 SvPV_set(sv, ptr);
463ee0b2 3943 SvCUR_set(sv, len);
1936d2a7 3944 SvLEN_set(sv, allocate);
c1c21316
NC
3945 if (!(flags & SV_HAS_TRAILING_NUL)) {
3946 *SvEND(sv) = '\0';
3947 }
1aa99e6b 3948 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3949 SvTAINT(sv);
47518d95
NC
3950 if (flags & SV_SMAGIC)
3951 SvSETMAGIC(sv);
ef50df4b
GS
3952}
3953
f8c7b90f 3954#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3955/* Need to do this *after* making the SV normal, as we need the buffer
3956 pointer to remain valid until after we've copied it. If we let go too early,
3957 another thread could invalidate it by unsharing last of the same hash key
3958 (which it can do by means other than releasing copy-on-write Svs)
3959 or by changing the other copy-on-write SVs in the loop. */
3960STATIC void
bdd68bc3 3961S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3962{
3963 if (len) { /* this SV was SvIsCOW_normal(sv) */
3964 /* we need to find the SV pointing to us. */
cf5629ad 3965 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 3966
765f542d
NC
3967 if (current == sv) {
3968 /* The SV we point to points back to us (there were only two of us
3969 in the loop.)
3970 Hence other SV is no longer copy on write either. */
3971 SvFAKE_off(after);
3972 SvREADONLY_off(after);
3973 } else {
3974 /* We need to follow the pointers around the loop. */
3975 SV *next;
3976 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3977 assert (next);
3978 current = next;
3979 /* don't loop forever if the structure is bust, and we have
3980 a pointer into a closed loop. */
3981 assert (current != after);
3f7c398e 3982 assert (SvPVX_const(current) == pvx);
765f542d
NC
3983 }
3984 /* Make the SV before us point to the SV after us. */
a29f6d03 3985 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3986 }
3987 } else {
bdd68bc3 3988 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3989 }
3990}
3991
3992int
3993Perl_sv_release_IVX(pTHX_ register SV *sv)
3994{
3995 if (SvIsCOW(sv))
3996 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
3997 SvOOK_off(sv);
3998 return 0;
765f542d
NC
3999}
4000#endif
645c22ef
DM
4001/*
4002=for apidoc sv_force_normal_flags
4003
4004Undo various types of fakery on an SV: if the PV is a shared string, make
4005a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4006an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4007we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4008then a copy-on-write scalar drops its PV buffer (if any) and becomes
4009SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4010set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4011C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4012with flags set to 0.
645c22ef
DM
4013
4014=cut
4015*/
4016
6fc92669 4017void
840a7b70 4018Perl_sv_force_normal_flags(pTHX_ register SV *sv, U32 flags)
0f15f207 4019{
97aff369 4020 dVAR;
f8c7b90f 4021#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4022 if (SvREADONLY(sv)) {
4023 /* At this point I believe I should acquire a global SV mutex. */
4024 if (SvFAKE(sv)) {
b64e5050 4025 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4026 const STRLEN len = SvLEN(sv);
4027 const STRLEN cur = SvCUR(sv);
a28509cc 4028 SV * const next = SV_COW_NEXT_SV(sv); /* next COW sv in the loop. */
46187eeb
NC
4029 if (DEBUG_C_TEST) {
4030 PerlIO_printf(Perl_debug_log,
4031 "Copy on write: Force normal %ld\n",
4032 (long) flags);
e419cbc5 4033 sv_dump(sv);
46187eeb 4034 }
765f542d
NC
4035 SvFAKE_off(sv);
4036 SvREADONLY_off(sv);
9f653bb5 4037 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4038 SvPV_set(sv, NULL);
87a1ef3d 4039 SvLEN_set(sv, 0);
765f542d
NC
4040 if (flags & SV_COW_DROP_PV) {
4041 /* OK, so we don't need to copy our buffer. */
4042 SvPOK_off(sv);
4043 } else {
4044 SvGROW(sv, cur + 1);
4045 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4046 SvCUR_set(sv, cur);
765f542d
NC
4047 *SvEND(sv) = '\0';
4048 }
bdd68bc3 4049 sv_release_COW(sv, pvx, len, next);
46187eeb 4050 if (DEBUG_C_TEST) {
e419cbc5 4051 sv_dump(sv);
46187eeb 4052 }
765f542d 4053 }
923e4eb5 4054 else if (IN_PERL_RUNTIME)
765f542d
NC
4055 Perl_croak(aTHX_ PL_no_modify);
4056 /* At this point I believe that I can drop the global SV mutex. */
4057 }
4058#else
2213622d 4059 if (SvREADONLY(sv)) {
1c846c1f 4060 if (SvFAKE(sv)) {
b64e5050 4061 const char * const pvx = SvPVX_const(sv);
66a1b24b 4062 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4063 SvFAKE_off(sv);
4064 SvREADONLY_off(sv);
bd61b366 4065 SvPV_set(sv, NULL);
66a1b24b 4066 SvLEN_set(sv, 0);
1c846c1f 4067 SvGROW(sv, len + 1);
706aa1c9 4068 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4069 *SvEND(sv) = '\0';
bdd68bc3 4070 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4071 }
923e4eb5 4072 else if (IN_PERL_RUNTIME)
cea2e8a9 4073 Perl_croak(aTHX_ PL_no_modify);
0f15f207 4074 }
765f542d 4075#endif
2213622d 4076 if (SvROK(sv))
840a7b70 4077 sv_unref_flags(sv, flags);
6fc92669
GS
4078 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4079 sv_unglob(sv);
0f15f207 4080}
1c846c1f 4081
645c22ef 4082/*
954c1994
GS
4083=for apidoc sv_chop
4084
1c846c1f 4085Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4086SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4087the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4088string. Uses the "OOK hack".
3f7c398e 4089Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4090refer to the same chunk of data.
954c1994
GS
4091
4092=cut
4093*/
4094
79072805 4095void
f54cb97a 4096Perl_sv_chop(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
4097{
4098 register STRLEN delta;
a0d0e21e 4099 if (!ptr || !SvPOKp(sv))
79072805 4100 return;
3f7c398e 4101 delta = ptr - SvPVX_const(sv);
2213622d 4102 SV_CHECK_THINKFIRST(sv);
79072805
LW
4103 if (SvTYPE(sv) < SVt_PVIV)
4104 sv_upgrade(sv,SVt_PVIV);
4105
4106 if (!SvOOK(sv)) {
50483b2c 4107 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4108 const char *pvx = SvPVX_const(sv);
a28509cc 4109 const STRLEN len = SvCUR(sv);
50483b2c 4110 SvGROW(sv, len + 1);
706aa1c9 4111 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4112 *SvEND(sv) = '\0';
4113 }
45977657 4114 SvIV_set(sv, 0);
a4bfb290
AB
4115 /* Same SvOOK_on but SvOOK_on does a SvIOK_off
4116 and we do that anyway inside the SvNIOK_off
4117 */
7a5fa8a2 4118 SvFLAGS(sv) |= SVf_OOK;
79072805 4119 }
a4bfb290 4120 SvNIOK_off(sv);
b162af07
SP
4121 SvLEN_set(sv, SvLEN(sv) - delta);
4122 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4123 SvPV_set(sv, SvPVX(sv) + delta);
45977657 4124 SvIV_set(sv, SvIVX(sv) + delta);
79072805
LW
4125}
4126
954c1994
GS
4127/*
4128=for apidoc sv_catpvn
4129
4130Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4131C<len> indicates number of bytes to copy. If the SV has the UTF-8
4132status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4133Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4134
8d6d96c1
HS
4135=for apidoc sv_catpvn_flags
4136
4137Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4138C<len> indicates number of bytes to copy. If the SV has the UTF-8
4139status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4140If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4141appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4142in terms of this function.
4143
4144=cut
4145*/
4146
4147void
4148Perl_sv_catpvn_flags(pTHX_ register SV *dsv, register const char *sstr, register STRLEN slen, I32 flags)
4149{
97aff369 4150 dVAR;
8d6d96c1 4151 STRLEN dlen;
fabdb6c0 4152 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4153
8d6d96c1
HS
4154 SvGROW(dsv, dlen + slen + 1);
4155 if (sstr == dstr)
3f7c398e 4156 sstr = SvPVX_const(dsv);
8d6d96c1 4157 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4158 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4159 *SvEND(dsv) = '\0';
4160 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4161 SvTAINT(dsv);
bddd5118
NC
4162 if (flags & SV_SMAGIC)
4163 SvSETMAGIC(dsv);
79072805
LW
4164}
4165
954c1994 4166/*
954c1994
GS
4167=for apidoc sv_catsv
4168
13e8c8e3
JH
4169Concatenates the string from SV C<ssv> onto the end of the string in
4170SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4171not 'set' magic. See C<sv_catsv_mg>.
954c1994 4172
8d6d96c1
HS
4173=for apidoc sv_catsv_flags
4174
4175Concatenates the string from SV C<ssv> onto the end of the string in
4176SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4177bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4178and C<sv_catsv_nomg> are implemented in terms of this function.
4179
4180=cut */
4181
ef50df4b 4182void
8d6d96c1 4183Perl_sv_catsv_flags(pTHX_ SV *dsv, register SV *ssv, I32 flags)
79072805 4184{
97aff369 4185 dVAR;
bddd5118 4186 if (ssv) {
00b6aa41
AL
4187 STRLEN slen;
4188 const char *spv = SvPV_const(ssv, slen);
4189 if (spv) {
bddd5118
NC
4190 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4191 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4192 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4193 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4194 dsv->sv_flags doesn't have that bit set.
4fd84b44 4195 Andy Dougherty 12 Oct 2001
bddd5118
NC
4196 */
4197 const I32 sutf8 = DO_UTF8(ssv);
4198 I32 dutf8;
13e8c8e3 4199
bddd5118
NC
4200 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4201 mg_get(dsv);
4202 dutf8 = DO_UTF8(dsv);
8d6d96c1 4203
bddd5118
NC
4204 if (dutf8 != sutf8) {
4205 if (dutf8) {
4206 /* Not modifying source SV, so taking a temporary copy. */
00b6aa41 4207 SV* const csv = sv_2mortal(newSVpvn(spv, slen));
13e8c8e3 4208
bddd5118
NC
4209 sv_utf8_upgrade(csv);
4210 spv = SvPV_const(csv, slen);
4211 }
4212 else
4213 sv_utf8_upgrade_nomg(dsv);
13e8c8e3 4214 }
bddd5118 4215 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4216 }
560a288e 4217 }
bddd5118
NC
4218 if (flags & SV_SMAGIC)
4219 SvSETMAGIC(dsv);
79072805
LW
4220}
4221
954c1994 4222/*
954c1994
GS
4223=for apidoc sv_catpv
4224
4225Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4226If the SV has the UTF-8 status set, then the bytes appended should be
4227valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4228
d5ce4a7c 4229=cut */
954c1994 4230
ef50df4b 4231void
0c981600 4232Perl_sv_catpv(pTHX_ register SV *sv, register const char *ptr)
79072805 4233{
97aff369 4234 dVAR;
79072805 4235 register STRLEN len;
463ee0b2 4236 STRLEN tlen;
748a9306 4237 char *junk;
79072805 4238
0c981600 4239 if (!ptr)
79072805 4240 return;
748a9306 4241 junk = SvPV_force(sv, tlen);
0c981600 4242 len = strlen(ptr);
463ee0b2 4243 SvGROW(sv, tlen + len + 1);
0c981600 4244 if (ptr == junk)
3f7c398e 4245 ptr = SvPVX_const(sv);
0c981600 4246 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4247 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4248 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4249 SvTAINT(sv);
79072805
LW
4250}
4251
954c1994
GS
4252/*
4253=for apidoc sv_catpv_mg
4254
4255Like C<sv_catpv>, but also handles 'set' magic.
4256
4257=cut
4258*/
4259
ef50df4b 4260void
0c981600 4261Perl_sv_catpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b 4262{
0c981600 4263 sv_catpv(sv,ptr);
ef50df4b
GS
4264 SvSETMAGIC(sv);
4265}
4266
645c22ef
DM
4267/*
4268=for apidoc newSV
4269
561b68a9
SH
4270Creates a new SV. A non-zero C<len> parameter indicates the number of
4271bytes of preallocated string space the SV should have. An extra byte for a
4272trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4273space is allocated.) The reference count for the new SV is set to 1.
4274
4275In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4276parameter, I<x>, a debug aid which allowed callers to identify themselves.
4277This aid has been superseded by a new build option, PERL_MEM_LOG (see
4278L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4279modules supporting older perls.
645c22ef
DM
4280
4281=cut
4282*/
4283
79072805 4284SV *
864dbfa3 4285Perl_newSV(pTHX_ STRLEN len)
79072805 4286{
97aff369 4287 dVAR;
79072805 4288 register SV *sv;
1c846c1f 4289
4561caa4 4290 new_SV(sv);
79072805
LW
4291 if (len) {
4292 sv_upgrade(sv, SVt_PV);
4293 SvGROW(sv, len + 1);
4294 }
4295 return sv;
4296}
954c1994 4297/*
92110913 4298=for apidoc sv_magicext
954c1994 4299
68795e93 4300Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4301supplied vtable and returns a pointer to the magic added.
92110913 4302
2d8d5d5a
SH
4303Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4304In particular, you can add magic to SvREADONLY SVs, and add more than
4305one instance of the same 'how'.
645c22ef 4306
2d8d5d5a
SH
4307If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4308stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4309special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4310to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4311
2d8d5d5a 4312(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4313
4314=cut
4315*/
92110913 4316MAGIC *
92e67595 4317Perl_sv_magicext(pTHX_ SV* sv, SV* obj, int how, MGVTBL *vtable,
92110913 4318 const char* name, I32 namlen)
79072805 4319{
97aff369 4320 dVAR;
79072805 4321 MAGIC* mg;
68795e93 4322
92110913 4323 if (SvTYPE(sv) < SVt_PVMG) {
862a34c6 4324 SvUPGRADE(sv, SVt_PVMG);
463ee0b2 4325 }
a02a5408 4326 Newxz(mg, 1, MAGIC);
79072805 4327 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4328 SvMAGIC_set(sv, mg);
75f9d97a 4329
05f95b08
SB
4330 /* Sometimes a magic contains a reference loop, where the sv and
4331 object refer to each other. To prevent a reference loop that
4332 would prevent such objects being freed, we look for such loops
4333 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4334
4335 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4336 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4337
4338 */
14befaf4
DM
4339 if (!obj || obj == sv ||
4340 how == PERL_MAGIC_arylen ||
4341 how == PERL_MAGIC_qr ||
8d2f4536 4342 how == PERL_MAGIC_symtab ||
75f9d97a
JH
4343 (SvTYPE(obj) == SVt_PVGV &&
4344 (GvSV(obj) == sv || GvHV(obj) == (HV*)sv || GvAV(obj) == (AV*)sv ||
4345 GvCV(obj) == (CV*)sv || GvIOp(obj) == (IO*)sv ||
2628be26 4346 GvFORM(obj) == (CV*)sv)))
75f9d97a 4347 {
8990e307 4348 mg->mg_obj = obj;
75f9d97a 4349 }
85e6fe83 4350 else {
b37c2d43 4351 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4352 mg->mg_flags |= MGf_REFCOUNTED;
4353 }
b5ccf5f2
YST
4354
4355 /* Normal self-ties simply pass a null object, and instead of
4356 using mg_obj directly, use the SvTIED_obj macro to produce a
4357 new RV as needed. For glob "self-ties", we are tieing the PVIO
4358 with an RV obj pointing to the glob containing the PVIO. In
4359 this case, to avoid a reference loop, we need to weaken the
4360 reference.
4361 */
4362
4363 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4364 obj && SvROK(obj) && GvIO(SvRV(obj)) == (IO*)sv)
4365 {
4366 sv_rvweaken(obj);
4367 }
4368
79072805 4369 mg->mg_type = how;
565764a8 4370 mg->mg_len = namlen;
9cbac4c7 4371 if (name) {
92110913 4372 if (namlen > 0)
1edc1566 4373 mg->mg_ptr = savepvn(name, namlen);
c6ee37c5 4374 else if (namlen == HEf_SVKEY)
b37c2d43 4375 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV*)name);
68795e93 4376 else
92110913 4377 mg->mg_ptr = (char *) name;
9cbac4c7 4378 }
92110913 4379 mg->mg_virtual = vtable;
68795e93 4380
92110913
NIS
4381 mg_magical(sv);
4382 if (SvGMAGICAL(sv))
4383 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4384 return mg;
4385}
4386
4387/*
4388=for apidoc sv_magic
1c846c1f 4389
92110913
NIS
4390Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4391then adds a new magic item of type C<how> to the head of the magic list.
4392
2d8d5d5a
SH
4393See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4394handling of the C<name> and C<namlen> arguments.
4395
4509d3fb
SB
4396You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4397to add more than one instance of the same 'how'.
4398
92110913
NIS
4399=cut
4400*/
4401
4402void
4403Perl_sv_magic(pTHX_ register SV *sv, SV *obj, int how, const char *name, I32 namlen)
68795e93 4404{
97aff369 4405 dVAR;
92e67595 4406 MGVTBL *vtable;
92110913 4407 MAGIC* mg;
92110913 4408
f8c7b90f 4409#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4410 if (SvIsCOW(sv))
4411 sv_force_normal_flags(sv, 0);
4412#endif
92110913 4413 if (SvREADONLY(sv)) {
d8084ca5
DM
4414 if (
4415 /* its okay to attach magic to shared strings; the subsequent
4416 * upgrade to PVMG will unshare the string */
4417 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4418
4419 && IN_PERL_RUNTIME
92110913
NIS
4420 && how != PERL_MAGIC_regex_global
4421 && how != PERL_MAGIC_bm
4422 && how != PERL_MAGIC_fm
4423 && how != PERL_MAGIC_sv
e6469971 4424 && how != PERL_MAGIC_backref
92110913
NIS
4425 )
4426 {
4427 Perl_croak(aTHX_ PL_no_modify);
4428 }
4429 }
4430 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
4431 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
4432 /* sv_magic() refuses to add a magic of the same 'how' as an
4433 existing one
92110913 4434 */
2a509ed3 4435 if (how == PERL_MAGIC_taint) {
92110913 4436 mg->mg_len |= 1;
2a509ed3
NC
4437 /* Any scalar which already had taint magic on which someone
4438 (erroneously?) did SvIOK_on() or similar will now be
4439 incorrectly sporting public "OK" flags. */
4440 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4441 }
92110913
NIS
4442 return;
4443 }
4444 }
68795e93 4445
79072805 4446 switch (how) {
14befaf4 4447 case PERL_MAGIC_sv:
92110913 4448 vtable = &PL_vtbl_sv;
79072805 4449 break;
14befaf4 4450 case PERL_MAGIC_overload:
92110913 4451 vtable = &PL_vtbl_amagic;
a0d0e21e 4452 break;
14befaf4 4453 case PERL_MAGIC_overload_elem:
92110913 4454 vtable = &PL_vtbl_amagicelem;
a0d0e21e 4455 break;
14befaf4 4456 case PERL_MAGIC_overload_table:
92110913 4457 vtable = &PL_vtbl_ovrld;
a0d0e21e 4458 break;
14befaf4 4459 case PERL_MAGIC_bm:
92110913 4460 vtable = &PL_vtbl_bm;
79072805 4461 break;
14befaf4 4462 case PERL_MAGIC_regdata:
92110913 4463 vtable = &PL_vtbl_regdata;
6cef1e77 4464 break;
14befaf4 4465 case PERL_MAGIC_regdatum:
92110913 4466 vtable = &PL_vtbl_regdatum;
6cef1e77 4467 break;
14befaf4 4468 case PERL_MAGIC_env:
92110913 4469 vtable = &PL_vtbl_env;
79072805 4470 break;
14befaf4 4471 case PERL_MAGIC_fm:
92110913 4472 vtable = &PL_vtbl_fm;
55497cff 4473 break;
14befaf4 4474 case PERL_MAGIC_envelem:
92110913 4475 vtable = &PL_vtbl_envelem;
79072805 4476 break;
14befaf4 4477 case PERL_MAGIC_regex_global:
92110913 4478 vtable = &PL_vtbl_mglob;
93a17b20 4479 break;
14befaf4 4480 case PERL_MAGIC_isa:
92110913 4481 vtable = &PL_vtbl_isa;
463ee0b2 4482 break;
14befaf4 4483 case PERL_MAGIC_isaelem:
92110913 4484 vtable = &PL_vtbl_isaelem;
463ee0b2 4485 break;
14befaf4 4486 case PERL_MAGIC_nkeys:
92110913 4487 vtable = &PL_vtbl_nkeys;
16660edb 4488 break;
14befaf4 4489 case PERL_MAGIC_dbfile:
aec46f14 4490 vtable = NULL;
93a17b20 4491 break;
14befaf4 4492 case PERL_MAGIC_dbline:
92110913 4493 vtable = &PL_vtbl_dbline;
79072805 4494 break;
36477c24 4495#ifdef USE_LOCALE_COLLATE
14befaf4 4496 case PERL_MAGIC_collxfrm:
92110913 4497 vtable = &PL_vtbl_collxfrm;
bbce6d69 4498 break;
36477c24 4499#endif /* USE_LOCALE_COLLATE */
14befaf4 4500 case PERL_MAGIC_tied:
92110913 4501 vtable = &PL_vtbl_pack;
463ee0b2 4502 break;
14befaf4
DM
4503 case PERL_MAGIC_tiedelem:
4504 case PERL_MAGIC_tiedscalar:
92110913 4505 vtable = &PL_vtbl_packelem;
463ee0b2 4506 break;
14befaf4 4507 case PERL_MAGIC_qr:
92110913 4508 vtable = &PL_vtbl_regexp;
c277df42 4509 break;
b3ca2e83
NC
4510 case PERL_MAGIC_hints:
4511 /* As this vtable is all NULL, we can reuse it. */
14befaf4 4512 case PERL_MAGIC_sig:
92110913 4513 vtable = &PL_vtbl_sig;
79072805 4514 break;
14befaf4 4515 case PERL_MAGIC_sigelem:
92110913 4516 vtable = &PL_vtbl_sigelem;
79072805 4517 break;
14befaf4 4518 case PERL_MAGIC_taint:
92110913 4519 vtable = &PL_vtbl_taint;
463ee0b2 4520 break;
14befaf4 4521 case PERL_MAGIC_uvar:
92110913 4522 vtable = &PL_vtbl_uvar;
79072805 4523 break;
14befaf4 4524 case PERL_MAGIC_vec:
92110913 4525 vtable = &PL_vtbl_vec;
79072805 4526 break;
a3874608 4527 case PERL_MAGIC_arylen_p:
bfcb3514 4528 case PERL_MAGIC_rhash:
8d2f4536 4529 case PERL_MAGIC_symtab:
ece467f9 4530 case PERL_MAGIC_vstring:
aec46f14 4531 vtable = NULL;
ece467f9 4532 break;
7e8c5dac
HS
4533 case PERL_MAGIC_utf8:
4534 vtable = &PL_vtbl_utf8;
4535 break;
14befaf4 4536 case PERL_MAGIC_substr:
92110913 4537 vtable = &PL_vtbl_substr;
79072805 4538 break;
14befaf4 4539 case PERL_MAGIC_defelem:
92110913 4540 vtable = &PL_vtbl_defelem;
5f05dabc 4541 break;
14befaf4 4542 case PERL_MAGIC_arylen:
92110913 4543 vtable = &PL_vtbl_arylen;
79072805 4544 break;
14befaf4 4545 case PERL_MAGIC_pos:
92110913 4546 vtable = &PL_vtbl_pos;
a0d0e21e 4547 break;
14befaf4 4548 case PERL_MAGIC_backref:
92110913 4549 vtable = &PL_vtbl_backref;
810b8aa5 4550 break;
b3ca2e83
NC
4551 case PERL_MAGIC_hintselem:
4552 vtable = &PL_vtbl_hintselem;
4553 break;
14befaf4
DM
4554 case PERL_MAGIC_ext:
4555 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
4556 /* Useful for attaching extension internal data to perl vars. */
4557 /* Note that multiple extensions may clash if magical scalars */
4558 /* etc holding private data from one are passed to another. */
aec46f14 4559 vtable = NULL;
a0d0e21e 4560 break;
79072805 4561 default:
14befaf4 4562 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 4563 }
68795e93 4564
92110913 4565 /* Rest of work is done else where */
aec46f14 4566 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 4567
92110913
NIS
4568 switch (how) {
4569 case PERL_MAGIC_taint:
4570 mg->mg_len = 1;
4571 break;
4572 case PERL_MAGIC_ext:
4573 case PERL_MAGIC_dbfile:
4574 SvRMAGICAL_on(sv);
4575 break;
4576 }
463ee0b2
LW
4577}
4578
c461cf8f
JH
4579/*
4580=for apidoc sv_unmagic
4581
645c22ef 4582Removes all magic of type C<type> from an SV.
c461cf8f
JH
4583
4584=cut
4585*/
4586
463ee0b2 4587int
864dbfa3 4588Perl_sv_unmagic(pTHX_ SV *sv, int type)
463ee0b2
LW
4589{
4590 MAGIC* mg;
4591 MAGIC** mgp;
91bba347 4592 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 4593 return 0;
064cf529 4594 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
4595 for (mg = *mgp; mg; mg = *mgp) {
4596 if (mg->mg_type == type) {
e1ec3a88 4597 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 4598 *mgp = mg->mg_moremagic;
1d7c1841 4599 if (vtbl && vtbl->svt_free)
fc0dc3b3 4600 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 4601 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 4602 if (mg->mg_len > 0)
1edc1566 4603 Safefree(mg->mg_ptr);
565764a8 4604 else if (mg->mg_len == HEf_SVKEY)
1edc1566 4605 SvREFCNT_dec((SV*)mg->mg_ptr);
d2923cdd 4606 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 4607 Safefree(mg->mg_ptr);
9cbac4c7 4608 }
a0d0e21e
LW
4609 if (mg->mg_flags & MGf_REFCOUNTED)
4610 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
4611 Safefree(mg);
4612 }
4613 else
4614 mgp = &mg->mg_moremagic;
79072805 4615 }
91bba347 4616 if (!SvMAGIC(sv)) {
463ee0b2 4617 SvMAGICAL_off(sv);
c268c2a6 4618 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
86f55936 4619 SvMAGIC_set(sv, NULL);
463ee0b2
LW
4620 }
4621
4622 return 0;
79072805
LW
4623}
4624
c461cf8f
JH
4625/*
4626=for apidoc sv_rvweaken
4627
645c22ef
DM
4628Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
4629referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
4630push a back-reference to this RV onto the array of backreferences
4631associated with that magic.
c461cf8f
JH
4632
4633=cut
4634*/
4635
810b8aa5 4636SV *
864dbfa3 4637Perl_sv_rvweaken(pTHX_ SV *sv)
810b8aa5
GS
4638{
4639 SV *tsv;
4640 if (!SvOK(sv)) /* let undefs pass */
4641 return sv;
4642 if (!SvROK(sv))
cea2e8a9 4643 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 4644 else if (SvWEAKREF(sv)) {
810b8aa5 4645 if (ckWARN(WARN_MISC))
9014280d 4646 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
4647 return sv;
4648 }
4649 tsv = SvRV(sv);
e15faf7d 4650 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 4651 SvWEAKREF_on(sv);
1c846c1f 4652 SvREFCNT_dec(tsv);
810b8aa5
GS
4653 return sv;
4654}
4655
645c22ef
DM
4656/* Give tsv backref magic if it hasn't already got it, then push a
4657 * back-reference to sv onto the array associated with the backref magic.
4658 */
4659
e15faf7d
NC
4660void
4661Perl_sv_add_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4662{
97aff369 4663 dVAR;
810b8aa5 4664 AV *av;
86f55936
NC
4665
4666 if (SvTYPE(tsv) == SVt_PVHV) {
4667 AV **const avp = Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
4668
4669 av = *avp;
4670 if (!av) {
4671 /* There is no AV in the offical place - try a fixup. */
4672 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
4673
4674 if (mg) {
4675 /* Aha. They've got it stowed in magic. Bring it back. */
4676 av = (AV*)mg->mg_obj;
4677 /* Stop mg_free decreasing the refernce count. */
4678 mg->mg_obj = NULL;
4679 /* Stop mg_free even calling the destructor, given that
4680 there's no AV to free up. */
4681 mg->mg_virtual = 0;
4682 sv_unmagic(tsv, PERL_MAGIC_backref);
4683 } else {
4684 av = newAV();
4685 AvREAL_off(av);
b37c2d43 4686 SvREFCNT_inc_simple_void(av);
86f55936
NC
4687 }
4688 *avp = av;
4689 }
4690 } else {
4691 const MAGIC *const mg
4692 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4693 if (mg)
4694 av = (AV*)mg->mg_obj;
4695 else {
4696 av = newAV();
4697 AvREAL_off(av);
4698 sv_magic(tsv, (SV*)av, PERL_MAGIC_backref, NULL, 0);
4699 /* av now has a refcnt of 2, which avoids it getting freed
4700 * before us during global cleanup. The extra ref is removed
4701 * by magic_killbackrefs() when tsv is being freed */
4702 }
810b8aa5 4703 }
d91d49e8 4704 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
4705 av_extend(av, AvFILLp(av)+1);
4706 }
4707 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
4708}
4709
645c22ef
DM
4710/* delete a back-reference to ourselves from the backref magic associated
4711 * with the SV we point to.
4712 */
4713
1c846c1f 4714STATIC void
e15faf7d 4715S_sv_del_backref(pTHX_ SV *tsv, SV *sv)
810b8aa5 4716{
97aff369 4717 dVAR;
86f55936 4718 AV *av = NULL;
810b8aa5
GS
4719 SV **svp;
4720 I32 i;
86f55936
NC
4721
4722 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
4723 av = *Perl_hv_backreferences_p(aTHX_ (HV*)tsv);
5b285ea4
NC
4724 /* We mustn't attempt to "fix up" the hash here by moving the
4725 backreference array back to the hv_aux structure, as that is stored
4726 in the main HvARRAY(), and hfreentries assumes that no-one
4727 reallocates HvARRAY() while it is running. */
86f55936
NC
4728 }
4729 if (!av) {
4730 const MAGIC *const mg
4731 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
4732 if (mg)
4733 av = (AV *)mg->mg_obj;
4734 }
4735 if (!av) {
e15faf7d
NC
4736 if (PL_in_clean_all)
4737 return;
cea2e8a9 4738 Perl_croak(aTHX_ "panic: del_backref");
86f55936
NC
4739 }
4740
4741 if (SvIS_FREED(av))
4742 return;
4743
810b8aa5 4744 svp = AvARRAY(av);
6a76db8b
NC
4745 /* We shouldn't be in here more than once, but for paranoia reasons lets
4746 not assume this. */
4747 for (i = AvFILLp(av); i >= 0; i--) {
4748 if (svp[i] == sv) {
4749 const SSize_t fill = AvFILLp(av);
4750 if (i != fill) {
4751 /* We weren't the last entry.
4752 An unordered list has this property that you can take the
4753 last element off the end to fill the hole, and it's still
4754 an unordered list :-)
4755 */
4756 svp[i] = svp[fill];
4757 }
a0714e2c 4758 svp[fill] = NULL;
6a76db8b
NC
4759 AvFILLp(av) = fill - 1;
4760 }
4761 }
810b8aa5
GS
4762}
4763
86f55936
NC
4764int
4765Perl_sv_kill_backrefs(pTHX_ SV *sv, AV *av)
4766{
4767 SV **svp = AvARRAY(av);
4768
4769 PERL_UNUSED_ARG(sv);
4770
4771 /* Not sure why the av can get freed ahead of its sv, but somehow it does
4772 in ext/B/t/bytecode.t test 15 (involving print <DATA>) */
4773 if (svp && !SvIS_FREED(av)) {
4774 SV *const *const last = svp + AvFILLp(av);
4775
4776 while (svp <= last) {
4777 if (*svp) {
4778 SV *const referrer = *svp;
4779 if (SvWEAKREF(referrer)) {
4780 /* XXX Should we check that it hasn't changed? */
4781 SvRV_set(referrer, 0);
4782 SvOK_off(referrer);
4783 SvWEAKREF_off(referrer);
4784 } else if (SvTYPE(referrer) == SVt_PVGV ||
4785 SvTYPE(referrer) == SVt_PVLV) {
4786 /* You lookin' at me? */
4787 assert(GvSTASH(referrer));
4788 assert(GvSTASH(referrer) == (HV*)sv);
4789 GvSTASH(referrer) = 0;
4790 } else {
4791 Perl_croak(aTHX_
4792 "panic: magic_killbackrefs (flags=%"UVxf")",
4793 (UV)SvFLAGS(referrer));
4794 }
4795
a0714e2c 4796 *svp = NULL;
86f55936
NC
4797 }
4798 svp++;
4799 }
4800 }
4801 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
4802 return 0;
4803}
4804
954c1994
GS
4805/*
4806=for apidoc sv_insert
4807
4808Inserts a string at the specified offset/length within the SV. Similar to
4809the Perl substr() function.
4810
4811=cut
4812*/
4813
79072805 4814void
e1ec3a88 4815Perl_sv_insert(pTHX_ SV *bigstr, STRLEN offset, STRLEN len, const char *little, STRLEN littlelen)
79072805 4816{
97aff369 4817 dVAR;
79072805
LW
4818 register char *big;
4819 register char *mid;
4820 register char *midend;
4821 register char *bigend;
4822 register I32 i;
6ff81951 4823 STRLEN curlen;
1c846c1f 4824
79072805 4825
8990e307 4826 if (!bigstr)
cea2e8a9 4827 Perl_croak(aTHX_ "Can't modify non-existent substring");
6ff81951 4828 SvPV_force(bigstr, curlen);
60fa28ff 4829 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
4830 if (offset + len > curlen) {
4831 SvGROW(bigstr, offset+len+1);
93524f2b 4832 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
4833 SvCUR_set(bigstr, offset+len);
4834 }
79072805 4835
69b47968 4836 SvTAINT(bigstr);
79072805
LW
4837 i = littlelen - len;
4838 if (i > 0) { /* string might grow */
a0d0e21e 4839 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
4840 mid = big + offset + len;
4841 midend = bigend = big + SvCUR(bigstr);
4842 bigend += i;
4843 *bigend = '\0';
4844 while (midend > mid) /* shove everything down */
4845 *--bigend = *--midend;
4846 Move(little,big+offset,littlelen,char);
b162af07 4847 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
4848 SvSETMAGIC(bigstr);
4849 return;
4850 }
4851 else if (i == 0) {
463ee0b2 4852 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
4853 SvSETMAGIC(bigstr);
4854 return;
4855 }
4856
463ee0b2 4857 big = SvPVX(bigstr);
79072805
LW
4858 mid = big + offset;
4859 midend = mid + len;
4860 bigend = big + SvCUR(bigstr);
4861
4862 if (midend > bigend)
cea2e8a9 4863 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
4864
4865 if (mid - big > bigend - midend) { /* faster to shorten from end */
4866 if (littlelen) {
4867 Move(little, mid, littlelen,char);
4868 mid += littlelen;
4869 }
4870 i = bigend - midend;
4871 if (i > 0) {
4872 Move(midend, mid, i,char);
4873 mid += i;
4874 }
4875 *mid = '\0';
4876 SvCUR_set(bigstr, mid - big);
4877 }
155aba94 4878 else if ((i = mid - big)) { /* faster from front */
79072805
LW
4879 midend -= littlelen;
4880 mid = midend;
4881 sv_chop(bigstr,midend-i);
4882 big += i;
4883 while (i--)
4884 *--midend = *--big;
4885 if (littlelen)
4886 Move(little, mid, littlelen,char);
4887 }
4888 else if (littlelen) {
4889 midend -= littlelen;
4890 sv_chop(bigstr,midend);
4891 Move(little,midend,littlelen,char);
4892 }
4893 else {
4894 sv_chop(bigstr,midend);
4895 }
4896 SvSETMAGIC(bigstr);
4897}
4898
c461cf8f
JH
4899/*
4900=for apidoc sv_replace
4901
4902Make the first argument a copy of the second, then delete the original.
645c22ef
DM
4903The target SV physically takes over ownership of the body of the source SV
4904and inherits its flags; however, the target keeps any magic it owns,
4905and any magic in the source is discarded.
ff276b08 4906Note that this is a rather specialist SV copying operation; most of the
645c22ef 4907time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
4908
4909=cut
4910*/
79072805
LW
4911
4912void
864dbfa3 4913Perl_sv_replace(pTHX_ register SV *sv, register SV *nsv)
79072805 4914{
97aff369 4915 dVAR;
a3b680e6 4916 const U32 refcnt = SvREFCNT(sv);
765f542d 4917 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 4918 if (SvREFCNT(nsv) != 1) {
7437becc 4919 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
30e5c352
NC
4920 UVuf " != 1)", (UV) SvREFCNT(nsv));
4921 }
93a17b20 4922 if (SvMAGICAL(sv)) {
a0d0e21e
LW
4923 if (SvMAGICAL(nsv))
4924 mg_free(nsv);
4925 else
4926 sv_upgrade(nsv, SVt_PVMG);
b162af07 4927 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 4928 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 4929 SvMAGICAL_off(sv);
b162af07 4930 SvMAGIC_set(sv, NULL);
93a17b20 4931 }
79072805
LW
4932 SvREFCNT(sv) = 0;
4933 sv_clear(sv);
477f5d66 4934 assert(!SvREFCNT(sv));
fd0854ff
DM
4935#ifdef DEBUG_LEAKING_SCALARS
4936 sv->sv_flags = nsv->sv_flags;
4937 sv->sv_any = nsv->sv_any;
4938 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 4939 sv->sv_u = nsv->sv_u;
fd0854ff 4940#else
79072805 4941 StructCopy(nsv,sv,SV);
fd0854ff 4942#endif
7b2c381c
NC
4943 /* Currently could join these into one piece of pointer arithmetic, but
4944 it would be unclear. */
4945 if(SvTYPE(sv) == SVt_IV)
4946 SvANY(sv)
339049b0 4947 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c 4948 else if (SvTYPE(sv) == SVt_RV) {
339049b0 4949 SvANY(sv) = &sv->sv_u.svu_rv;
7b2c381c
NC
4950 }
4951
fd0854ff 4952
f8c7b90f 4953#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
4954 if (SvIsCOW_normal(nsv)) {
4955 /* We need to follow the pointers around the loop to make the
4956 previous SV point to sv, rather than nsv. */
4957 SV *next;
4958 SV *current = nsv;
4959 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
4960 assert(next);
4961 current = next;
3f7c398e 4962 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
4963 }
4964 /* Make the SV before us point to the SV after us. */
4965 if (DEBUG_C_TEST) {
4966 PerlIO_printf(Perl_debug_log, "previous is\n");
4967 sv_dump(current);
a29f6d03
NC
4968 PerlIO_printf(Perl_debug_log,
4969 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
4970 (UV) SV_COW_NEXT_SV(current), (UV) sv);
4971 }
a29f6d03 4972 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
4973 }
4974#endif
79072805 4975 SvREFCNT(sv) = refcnt;
1edc1566 4976 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 4977 SvREFCNT(nsv) = 0;
463ee0b2 4978 del_SV(nsv);
79072805
LW
4979}
4980
c461cf8f
JH
4981/*
4982=for apidoc sv_clear
4983
645c22ef
DM
4984Clear an SV: call any destructors, free up any memory used by the body,
4985and free the body itself. The SV's head is I<not> freed, although
4986its type is set to all 1's so that it won't inadvertently be assumed
4987to be live during global destruction etc.
4988This function should only be called when REFCNT is zero. Most of the time
4989you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
4990instead.
c461cf8f
JH
4991
4992=cut
4993*/
4994
79072805 4995void
864dbfa3 4996Perl_sv_clear(pTHX_ register SV *sv)
79072805 4997{
27da23d5 4998 dVAR;
82bb6deb 4999 const U32 type = SvTYPE(sv);
8edfc514
NC
5000 const struct body_details *const sv_type_details
5001 = bodies_by_type + type;
82bb6deb 5002
79072805
LW
5003 assert(sv);
5004 assert(SvREFCNT(sv) == 0);
5005
d2a0f284
JC
5006 if (type <= SVt_IV) {
5007 /* See the comment in sv.h about the collusion between this early
5008 return and the overloading of the NULL and IV slots in the size
5009 table. */
82bb6deb 5010 return;
d2a0f284 5011 }
82bb6deb 5012
ed6116ce 5013 if (SvOBJECT(sv)) {
3280af22 5014 if (PL_defstash) { /* Still have a symbol table? */
39644a26 5015 dSP;
893645bd 5016 HV* stash;
d460ef45 5017 do {
b464bac0 5018 CV* destructor;
4e8e7886 5019 stash = SvSTASH(sv);
32251b26 5020 destructor = StashHANDLER(stash,DESTROY);
4e8e7886 5021 if (destructor) {
1b6737cc 5022 SV* const tmpref = newRV(sv);
5cc433a6 5023 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5024 ENTER;
e788e7d3 5025 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5026 EXTEND(SP, 2);
5027 PUSHMARK(SP);
5cc433a6 5028 PUSHs(tmpref);
4e8e7886 5029 PUTBACK;
44389ee9 5030 call_sv((SV*)destructor, G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5031
5032
d3acc0f7 5033 POPSTACK;
3095d977 5034 SPAGAIN;
4e8e7886 5035 LEAVE;
5cc433a6
AB
5036 if(SvREFCNT(tmpref) < 2) {
5037 /* tmpref is not kept alive! */
5038 SvREFCNT(sv)--;
b162af07 5039 SvRV_set(tmpref, NULL);
5cc433a6
AB
5040 SvROK_off(tmpref);
5041 }
5042 SvREFCNT_dec(tmpref);
4e8e7886
GS
5043 }
5044 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5045
6f44e0a4
JP
5046
5047 if (SvREFCNT(sv)) {
5048 if (PL_in_clean_objs)
cea2e8a9 5049 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5050 HvNAME_get(stash));
6f44e0a4
JP
5051 /* DESTROY gave object new lease on life */
5052 return;
5053 }
a0d0e21e 5054 }
4e8e7886 5055
a0d0e21e 5056 if (SvOBJECT(sv)) {
4e8e7886 5057 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5058 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5059 if (type != SVt_PVIO)
3280af22 5060 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5061 }
463ee0b2 5062 }
82bb6deb 5063 if (type >= SVt_PVMG) {
e736a858
NC
5064 HV *ourstash;
5065 if ((type == SVt_PVMG || type == SVt_PVGV) &&
5066 (ourstash = OURSTASH(sv))) {
5067 SvREFCNT_dec(ourstash);
5068 } else if (SvMAGIC(sv))
524189f1 5069 mg_free(sv);
00b1698f 5070 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5071 SvREFCNT_dec(SvSTASH(sv));
5072 }
82bb6deb 5073 switch (type) {
8990e307 5074 case SVt_PVIO:
df0bd2f4
GS
5075 if (IoIFP(sv) &&
5076 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5077 IoIFP(sv) != PerlIO_stdout() &&
5078 IoIFP(sv) != PerlIO_stderr())
93578b34 5079 {
f2b5be74 5080 io_close((IO*)sv, FALSE);
93578b34 5081 }
1d7c1841 5082 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5083 PerlDir_close(IoDIRP(sv));
1d7c1841 5084 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5085 Safefree(IoTOP_NAME(sv));
5086 Safefree(IoFMT_NAME(sv));
5087 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5088 goto freescalar;
79072805 5089 case SVt_PVBM:
a0d0e21e 5090 goto freescalar;
79072805 5091 case SVt_PVCV:
748a9306 5092 case SVt_PVFM:
85e6fe83 5093 cv_undef((CV*)sv);
a0d0e21e 5094 goto freescalar;
79072805 5095 case SVt_PVHV:
86f55936 5096 Perl_hv_kill_backrefs(aTHX_ (HV*)sv);
85e6fe83 5097 hv_undef((HV*)sv);
a0d0e21e 5098 break;
79072805 5099 case SVt_PVAV:
85e6fe83 5100 av_undef((AV*)sv);
a0d0e21e 5101 break;
02270b4e 5102 case SVt_PVLV:
dd28f7bb
DM
5103 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5104 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5105 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5106 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5107 }
5108 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5109 SvREFCNT_dec(LvTARG(sv));
02270b4e 5110 goto freescalar;
a0d0e21e 5111 case SVt_PVGV:
1edc1566 5112 gp_free((GV*)sv);
acda4c6a
NC
5113 if (GvNAME_HEK(sv)) {
5114 unshare_hek(GvNAME_HEK(sv));
5115 }
893645bd
NC
5116 /* If we're in a stash, we don't own a reference to it. However it does
5117 have a back reference to us, which needs to be cleared. */
5118 if (GvSTASH(sv))
5119 sv_del_backref((SV*)GvSTASH(sv), sv);
79072805 5120 case SVt_PVMG:
79072805
LW
5121 case SVt_PVNV:
5122 case SVt_PVIV:
a0d0e21e 5123 freescalar:
5228ca4e
NC
5124 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5125 if (SvOOK(sv)) {
93524f2b 5126 SvPV_set(sv, SvPVX_mutable(sv) - SvIVX(sv));
5228ca4e
NC
5127 /* Don't even bother with turning off the OOK flag. */
5128 }
79072805 5129 case SVt_PV:
a0d0e21e 5130 case SVt_RV:
810b8aa5 5131 if (SvROK(sv)) {
b37c2d43 5132 SV * const target = SvRV(sv);
810b8aa5 5133 if (SvWEAKREF(sv))
e15faf7d 5134 sv_del_backref(target, sv);
810b8aa5 5135 else
e15faf7d 5136 SvREFCNT_dec(target);
810b8aa5 5137 }
f8c7b90f 5138#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5139 else if (SvPVX_const(sv)) {
765f542d
NC
5140 if (SvIsCOW(sv)) {
5141 /* I believe I need to grab the global SV mutex here and
5142 then recheck the COW status. */
46187eeb
NC
5143 if (DEBUG_C_TEST) {
5144 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5145 sv_dump(sv);
46187eeb 5146 }
bdd68bc3
NC
5147 sv_release_COW(sv, SvPVX_const(sv), SvLEN(sv),
5148 SV_COW_NEXT_SV(sv));
765f542d
NC
5149 /* And drop it here. */
5150 SvFAKE_off(sv);
5151 } else if (SvLEN(sv)) {
3f7c398e 5152 Safefree(SvPVX_const(sv));
765f542d
NC
5153 }
5154 }
5155#else
3f7c398e 5156 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5157 Safefree(SvPVX_mutable(sv));
3f7c398e 5158 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5159 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5160 SvFAKE_off(sv);
5161 }
765f542d 5162#endif
79072805
LW
5163 break;
5164 case SVt_NV:
79072805
LW
5165 break;
5166 }
5167
893645bd
NC
5168 SvFLAGS(sv) &= SVf_BREAK;
5169 SvFLAGS(sv) |= SVTYPEMASK;
5170
8edfc514 5171 if (sv_type_details->arena) {
b9502f15 5172 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5173 &PL_body_roots[type]);
5174 }
d2a0f284 5175 else if (sv_type_details->body_size) {
8edfc514
NC
5176 my_safefree(SvANY(sv));
5177 }
79072805
LW
5178}
5179
645c22ef
DM
5180/*
5181=for apidoc sv_newref
5182
5183Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5184instead.
5185
5186=cut
5187*/
5188
79072805 5189SV *
864dbfa3 5190Perl_sv_newref(pTHX_ SV *sv)
79072805 5191{
96a5add6 5192 PERL_UNUSED_CONTEXT;
463ee0b2 5193 if (sv)
4db098f4 5194 (SvREFCNT(sv))++;
79072805
LW
5195 return sv;
5196}
5197
c461cf8f
JH
5198/*
5199=for apidoc sv_free
5200
645c22ef
DM
5201Decrement an SV's reference count, and if it drops to zero, call
5202C<sv_clear> to invoke destructors and free up any memory used by
5203the body; finally, deallocate the SV's head itself.
5204Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5205
5206=cut
5207*/
5208
79072805 5209void
864dbfa3 5210Perl_sv_free(pTHX_ SV *sv)
79072805 5211{
27da23d5 5212 dVAR;
79072805
LW
5213 if (!sv)
5214 return;
a0d0e21e
LW
5215 if (SvREFCNT(sv) == 0) {
5216 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5217 /* this SV's refcnt has been artificially decremented to
5218 * trigger cleanup */
a0d0e21e 5219 return;
3280af22 5220 if (PL_in_clean_all) /* All is fair */
1edc1566 5221 return;
d689ffdd
JP
5222 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5223 /* make sure SvREFCNT(sv)==0 happens very seldom */
5224 SvREFCNT(sv) = (~(U32)0)/2;
5225 return;
5226 }
41e4abd8 5227 if (ckWARN_d(WARN_INTERNAL)) {
d5dede04 5228 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
5229 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5230 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5231#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5232 Perl_dump_sv_child(aTHX_ sv);
5233#endif
5234 }
79072805
LW
5235 return;
5236 }
4db098f4 5237 if (--(SvREFCNT(sv)) > 0)
8990e307 5238 return;
8c4d3c90
NC
5239 Perl_sv_free2(aTHX_ sv);
5240}
5241
5242void
5243Perl_sv_free2(pTHX_ SV *sv)
5244{
27da23d5 5245 dVAR;
463ee0b2
LW
5246#ifdef DEBUGGING
5247 if (SvTEMP(sv)) {
0453d815 5248 if (ckWARN_d(WARN_DEBUGGING))
9014280d 5249 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
472d47bc
SB
5250 "Attempt to free temp prematurely: SV 0x%"UVxf
5251 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5252 return;
79072805 5253 }
463ee0b2 5254#endif
d689ffdd
JP
5255 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5256 /* make sure SvREFCNT(sv)==0 happens very seldom */
5257 SvREFCNT(sv) = (~(U32)0)/2;
5258 return;
5259 }
79072805 5260 sv_clear(sv);
477f5d66
CS
5261 if (! SvREFCNT(sv))
5262 del_SV(sv);
79072805
LW
5263}
5264
954c1994
GS
5265/*
5266=for apidoc sv_len
5267
645c22ef
DM
5268Returns the length of the string in the SV. Handles magic and type
5269coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5270
5271=cut
5272*/
5273
79072805 5274STRLEN
864dbfa3 5275Perl_sv_len(pTHX_ register SV *sv)
79072805 5276{
463ee0b2 5277 STRLEN len;
79072805
LW
5278
5279 if (!sv)
5280 return 0;
5281
8990e307 5282 if (SvGMAGICAL(sv))
565764a8 5283 len = mg_length(sv);
8990e307 5284 else
4d84ee25 5285 (void)SvPV_const(sv, len);
463ee0b2 5286 return len;
79072805
LW
5287}
5288
c461cf8f
JH
5289/*
5290=for apidoc sv_len_utf8
5291
5292Returns the number of characters in the string in an SV, counting wide
1e54db1a 5293UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5294
5295=cut
5296*/
5297
7e8c5dac
HS
5298/*
5299 * The length is cached in PERL_UTF8_magic, in the mg_len field. Also the
9564a3bd
NC
5300 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5301 * (Note that the mg_len is not the length of the mg_ptr field.
5302 * This allows the cache to store the character length of the string without
5303 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5304 *
7e8c5dac
HS
5305 */
5306
a0ed51b3 5307STRLEN
864dbfa3 5308Perl_sv_len_utf8(pTHX_ register SV *sv)
a0ed51b3 5309{
a0ed51b3
LW
5310 if (!sv)
5311 return 0;
5312
a0ed51b3 5313 if (SvGMAGICAL(sv))
b76347f2 5314 return mg_length(sv);
a0ed51b3 5315 else
b76347f2 5316 {
26346457 5317 STRLEN len;
e62f0680 5318 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5319
26346457
NC
5320 if (PL_utf8cache) {
5321 STRLEN ulen;
5322 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : 0;
5323
5324 if (mg && mg->mg_len != -1) {
5325 ulen = mg->mg_len;
5326 if (PL_utf8cache < 0) {
5327 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5328 if (real != ulen) {
5329 /* Need to turn the assertions off otherwise we may
5330 recurse infinitely while printing error messages.
5331 */
5332 SAVEI8(PL_utf8cache);
5333 PL_utf8cache = 0;
5334 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVf
ec07b5e0 5335 " real %"UVf" for %"SVf,
26346457
NC
5336 (UV) ulen, (UV) real, sv);
5337 }
5338 }
5339 }
5340 else {
5341 ulen = Perl_utf8_length(aTHX_ s, s + len);
5342 if (!SvREADONLY(sv)) {
5343 if (!mg) {
5344 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5345 &PL_vtbl_utf8, 0, 0);
5346 }
cb9e20bb 5347 assert(mg);
26346457 5348 mg->mg_len = ulen;
cb9e20bb 5349 }
cb9e20bb 5350 }
26346457 5351 return ulen;
7e8c5dac 5352 }
26346457 5353 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
5354 }
5355}
5356
9564a3bd
NC
5357/* Walk forwards to find the byte corresponding to the passed in UTF-8
5358 offset. */
bdf30dd6
NC
5359static STRLEN
5360S_sv_pos_u2b_forwards(pTHX_ const U8 *const start, const U8 *const send,
5361 STRLEN uoffset)
5362{
5363 const U8 *s = start;
5364
d4c19fe8
AL
5365 PERL_UNUSED_CONTEXT;
5366
bdf30dd6
NC
5367 while (s < send && uoffset--)
5368 s += UTF8SKIP(s);
5369 if (s > send) {
5370 /* This is the existing behaviour. Possibly it should be a croak, as
5371 it's actually a bounds error */
5372 s = send;
5373 }
5374 return s - start;
5375}
5376
9564a3bd
NC
5377/* Given the length of the string in both bytes and UTF-8 characters, decide
5378 whether to walk forwards or backwards to find the byte corresponding to
5379 the passed in UTF-8 offset. */
c336ad0b
NC
5380static STRLEN
5381S_sv_pos_u2b_midway(pTHX_ const U8 *const start, const U8 *send,
5382 STRLEN uoffset, STRLEN uend)
5383{
5384 STRLEN backw = uend - uoffset;
5385 if (uoffset < 2 * backw) {
25a8a4ef 5386 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
5387 forward (that's where the 2 * backw comes from).
5388 (The real figure of course depends on the UTF-8 data.) */
5389 return S_sv_pos_u2b_forwards(aTHX_ start, send, uoffset);
5390 }
5391
5392 while (backw--) {
5393 send--;
5394 while (UTF8_IS_CONTINUATION(*send))
5395 send--;
5396 }
5397 return send - start;
5398}
5399
9564a3bd
NC
5400/* For the string representation of the given scalar, find the byte
5401 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
5402 give another position in the string, *before* the sought offset, which
5403 (which is always true, as 0, 0 is a valid pair of positions), which should
5404 help reduce the amount of linear searching.
5405 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
5406 will be used to reduce the amount of linear searching. The cache will be
5407 created if necessary, and the found value offered to it for update. */
28ccbf94
NC
5408static STRLEN
5409S_sv_pos_u2b_cached(pTHX_ SV *sv, MAGIC **mgp, const U8 *const start,
5410 const U8 *const send, STRLEN uoffset,
5411 STRLEN uoffset0, STRLEN boffset0) {
7087a21c 5412 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
5413 bool found = FALSE;
5414
75c33c12
NC
5415 assert (uoffset >= uoffset0);
5416
c336ad0b 5417 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
0905937d 5418 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
d8b2e1f9
NC
5419 if ((*mgp)->mg_ptr) {
5420 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
5421 if (cache[0] == uoffset) {
5422 /* An exact match. */
5423 return cache[1];
5424 }
ab455f60
NC
5425 if (cache[2] == uoffset) {
5426 /* An exact match. */
5427 return cache[3];
5428 }
668af93f
NC
5429
5430 if (cache[0] < uoffset) {
d8b2e1f9
NC
5431 /* The cache already knows part of the way. */
5432 if (cache[0] > uoffset0) {
5433 /* The cache knows more than the passed in pair */
5434 uoffset0 = cache[0];
5435 boffset0 = cache[1];
5436 }
5437 if ((*mgp)->mg_len != -1) {
5438 /* And we know the end too. */
5439 boffset = boffset0
5440 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5441 uoffset - uoffset0,
5442 (*mgp)->mg_len - uoffset0);
5443 } else {
5444 boffset = boffset0
5445 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5446 send, uoffset - uoffset0);
5447 }
dd7c5fd3
NC
5448 }
5449 else if (cache[2] < uoffset) {
5450 /* We're between the two cache entries. */
5451 if (cache[2] > uoffset0) {
5452 /* and the cache knows more than the passed in pair */
5453 uoffset0 = cache[2];
5454 boffset0 = cache[3];
5455 }
5456
668af93f
NC
5457 boffset = boffset0
5458 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5459 start + cache[1],
5460 uoffset - uoffset0,
5461 cache[0] - uoffset0);
dd7c5fd3
NC
5462 } else {
5463 boffset = boffset0
5464 + S_sv_pos_u2b_midway(aTHX_ start + boffset0,
5465 start + cache[3],
5466 uoffset - uoffset0,
5467 cache[2] - uoffset0);
d8b2e1f9 5468 }
668af93f 5469 found = TRUE;
d8b2e1f9
NC
5470 }
5471 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
5472 /* If we can take advantage of a passed in offset, do so. */
5473 /* In fact, offset0 is either 0, or less than offset, so don't
5474 need to worry about the other possibility. */
5475 boffset = boffset0
5476 + S_sv_pos_u2b_midway(aTHX_ start + boffset0, send,
5477 uoffset - uoffset0,
5478 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
5479 found = TRUE;
5480 }
28ccbf94 5481 }
c336ad0b
NC
5482
5483 if (!found || PL_utf8cache < 0) {
75c33c12
NC
5484 const STRLEN real_boffset
5485 = boffset0 + S_sv_pos_u2b_forwards(aTHX_ start + boffset0,
5486 send, uoffset - uoffset0);
5487
c336ad0b
NC
5488 if (found && PL_utf8cache < 0) {
5489 if (real_boffset != boffset) {
5490 /* Need to turn the assertions off otherwise we may recurse
5491 infinitely while printing error messages. */
5492 SAVEI8(PL_utf8cache);
5493 PL_utf8cache = 0;
5494 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVf
5495 " real %"UVf" for %"SVf,
5496 (UV) boffset, (UV) real_boffset, sv);
5497 }
5498 }
5499 boffset = real_boffset;
28ccbf94 5500 }
0905937d 5501
ab455f60 5502 S_utf8_mg_pos_cache_update(aTHX_ sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
5503 return boffset;
5504}
5505
9564a3bd
NC
5506
5507/*
5508=for apidoc sv_pos_u2b
5509
5510Converts the value pointed to by offsetp from a count of UTF-8 chars from
5511the start of the string, to a count of the equivalent number of bytes; if
5512lenp is non-zero, it does the same to lenp, but this time starting from
5513the offset, rather than from the start of the string. Handles magic and
5514type coercion.
5515
5516=cut
5517*/
5518
5519/*
5520 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
5521 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5522 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
5523 *
5524 */
5525
a0ed51b3 5526void
864dbfa3 5527Perl_sv_pos_u2b(pTHX_ register SV *sv, I32* offsetp, I32* lenp)
a0ed51b3 5528{
245d4a47 5529 const U8 *start;
a0ed51b3
LW
5530 STRLEN len;
5531
5532 if (!sv)
5533 return;
5534
245d4a47 5535 start = (U8*)SvPV_const(sv, len);
7e8c5dac 5536 if (len) {
bdf30dd6
NC
5537 STRLEN uoffset = (STRLEN) *offsetp;
5538 const U8 * const send = start + len;
0905937d 5539 MAGIC *mg = NULL;
28ccbf94
NC
5540 STRLEN boffset = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send,
5541 uoffset, 0, 0);
bdf30dd6
NC
5542
5543 *offsetp = (I32) boffset;
5544
5545 if (lenp) {
28ccbf94
NC
5546 /* Convert the relative offset to absolute. */
5547 STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
5548 STRLEN boffset2
5549 = S_sv_pos_u2b_cached(aTHX_ sv, &mg, start, send, uoffset2,
5550 uoffset, boffset) - boffset;
bdf30dd6 5551
28ccbf94 5552 *lenp = boffset2;
bdf30dd6 5553 }
7e8c5dac
HS
5554 }
5555 else {
5556 *offsetp = 0;
5557 if (lenp)
5558 *lenp = 0;
a0ed51b3 5559 }
e23c8137 5560
a0ed51b3
LW
5561 return;
5562}
5563
9564a3bd
NC
5564/* Create and update the UTF8 magic offset cache, with the proffered utf8/
5565 byte length pairing. The (byte) length of the total SV is passed in too,
5566 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
5567 may not have updated SvCUR, so we can't rely on reading it directly.
5568
5569 The proffered utf8/byte length pairing isn't used if the cache already has
5570 two pairs, and swapping either for the proffered pair would increase the
5571 RMS of the intervals between known byte offsets.
5572
5573 The cache itself consists of 4 STRLEN values
5574 0: larger UTF-8 offset
5575 1: corresponding byte offset
5576 2: smaller UTF-8 offset
5577 3: corresponding byte offset
5578
5579 Unused cache pairs have the value 0, 0.
5580 Keeping the cache "backwards" means that the invariant of
5581 cache[0] >= cache[2] is maintained even with empty slots, which means that
5582 the code that uses it doesn't need to worry if only 1 entry has actually
5583 been set to non-zero. It also makes the "position beyond the end of the
5584 cache" logic much simpler, as the first slot is always the one to start
5585 from.
645c22ef 5586*/
ec07b5e0 5587static void
ab455f60
NC
5588S_utf8_mg_pos_cache_update(pTHX_ SV *sv, MAGIC **mgp, STRLEN byte, STRLEN utf8,
5589 STRLEN blen)
ec07b5e0
NC
5590{
5591 STRLEN *cache;
5592 if (SvREADONLY(sv))
5593 return;
5594
5595 if (!*mgp) {
5596 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
5597 0);
5598 (*mgp)->mg_len = -1;
5599 }
5600 assert(*mgp);
5601
5602 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
5603 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
5604 (*mgp)->mg_ptr = (char *) cache;
5605 }
5606 assert(cache);
5607
5608 if (PL_utf8cache < 0) {
ef816a78 5609 const U8 *start = (const U8 *) SvPVX_const(sv);
0905937d
NC
5610 const U8 *const end = start + byte;
5611 STRLEN realutf8 = 0;
5612
5613 while (start < end) {
5614 start += UTF8SKIP(start);
5615 realutf8++;
5616 }
5617
5618 /* Can't use S_sv_pos_b2u_forwards as it will scream warnings on
5619 surrogates. FIXME - is it inconsistent that b2u warns, but u2b
5620 doesn't? I don't know whether this difference was introduced with
5621 the caching code in 5.8.1. */
ec07b5e0
NC
5622
5623 if (realutf8 != utf8) {
5624 /* Need to turn the assertions off otherwise we may recurse
5625 infinitely while printing error messages. */
5626 SAVEI8(PL_utf8cache);
5627 PL_utf8cache = 0;
5628 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVf
5629 " real %"UVf" for %"SVf, (UV) utf8, (UV) realutf8, sv);
5630 }
5631 }
ab455f60
NC
5632
5633 /* Cache is held with the later position first, to simplify the code
5634 that deals with unbounded ends. */
5635
5636 ASSERT_UTF8_CACHE(cache);
5637 if (cache[1] == 0) {
5638 /* Cache is totally empty */
5639 cache[0] = utf8;
5640 cache[1] = byte;
5641 } else if (cache[3] == 0) {
5642 if (byte > cache[1]) {
5643 /* New one is larger, so goes first. */
5644 cache[2] = cache[0];
5645 cache[3] = cache[1];
5646 cache[0] = utf8;
5647 cache[1] = byte;
5648 } else {
5649 cache[2] = utf8;
5650 cache[3] = byte;
5651 }
5652 } else {
5653#define THREEWAY_SQUARE(a,b,c,d) \
5654 ((float)((d) - (c))) * ((float)((d) - (c))) \
5655 + ((float)((c) - (b))) * ((float)((c) - (b))) \
5656 + ((float)((b) - (a))) * ((float)((b) - (a)))
5657
5658 /* Cache has 2 slots in use, and we know three potential pairs.
5659 Keep the two that give the lowest RMS distance. Do the
5660 calcualation in bytes simply because we always know the byte
5661 length. squareroot has the same ordering as the positive value,
5662 so don't bother with the actual square root. */
5663 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
5664 if (byte > cache[1]) {
5665 /* New position is after the existing pair of pairs. */
5666 const float keep_earlier
5667 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5668 const float keep_later
5669 = THREEWAY_SQUARE(0, cache[1], byte, blen);
5670
5671 if (keep_later < keep_earlier) {
5672 if (keep_later < existing) {
5673 cache[2] = cache[0];
5674 cache[3] = cache[1];
5675 cache[0] = utf8;
5676 cache[1] = byte;
5677 }
5678 }
5679 else {
5680 if (keep_earlier < existing) {
5681 cache[0] = utf8;
5682 cache[1] = byte;
5683 }
5684 }
5685 }
57d7fbf1
NC
5686 else if (byte > cache[3]) {
5687 /* New position is between the existing pair of pairs. */
5688 const float keep_earlier
5689 = THREEWAY_SQUARE(0, cache[3], byte, blen);
5690 const float keep_later
5691 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5692
5693 if (keep_later < keep_earlier) {
5694 if (keep_later < existing) {
5695 cache[2] = utf8;
5696 cache[3] = byte;
5697 }
5698 }
5699 else {
5700 if (keep_earlier < existing) {
5701 cache[0] = utf8;
5702 cache[1] = byte;
5703 }
5704 }
5705 }
5706 else {
5707 /* New position is before the existing pair of pairs. */
5708 const float keep_earlier
5709 = THREEWAY_SQUARE(0, byte, cache[3], blen);
5710 const float keep_later
5711 = THREEWAY_SQUARE(0, byte, cache[1], blen);
5712
5713 if (keep_later < keep_earlier) {
5714 if (keep_later < existing) {
5715 cache[2] = utf8;
5716 cache[3] = byte;
5717 }
5718 }
5719 else {
5720 if (keep_earlier < existing) {
5721 cache[0] = cache[2];
5722 cache[1] = cache[3];
5723 cache[2] = utf8;
5724 cache[3] = byte;
5725 }
5726 }
5727 }
ab455f60 5728 }
0905937d 5729 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
5730}
5731
5732/* If we don't know the character offset of the end of a region, our only
5733 option is to walk forwards to the target byte offset. */
5734static STRLEN
5735S_sv_pos_b2u_forwards(pTHX_ const U8 *s, const U8 *const target)
5736{
5737 STRLEN len = 0;
5738 while (s < target) {
5739 STRLEN n = 1;
5740
5741 /* Call utf8n_to_uvchr() to validate the sequence
5742 * (unless a simple non-UTF character) */
5743 if (!UTF8_IS_INVARIANT(*s))
5744 utf8n_to_uvchr(s, UTF8SKIP(s), &n, 0);
5745 if (n > 0) {
5746 s += n;
5747 len++;
5748 }
5749 else
5750 break;
5751 }
5752 return len;
5753}
5754
5755/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
5756 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
5757 backward is half the speed of walking forward. */
ec07b5e0
NC
5758static STRLEN
5759S_sv_pos_b2u_midway(pTHX_ const U8 *s, const U8 *const target, const U8 *end,
5760 STRLEN endu)
5761{
5762 const STRLEN forw = target - s;
5763 STRLEN backw = end - target;
5764
5765 if (forw < 2 * backw) {
5766 return S_sv_pos_b2u_forwards(aTHX_ s, target);
5767 }
5768
5769 while (end > target) {
5770 end--;
5771 while (UTF8_IS_CONTINUATION(*end)) {
5772 end--;
5773 }
5774 endu--;
5775 }
5776 return endu;
5777}
5778
9564a3bd
NC
5779/*
5780=for apidoc sv_pos_b2u
5781
5782Converts the value pointed to by offsetp from a count of bytes from the
5783start of the string, to a count of the equivalent number of UTF-8 chars.
5784Handles magic and type coercion.
5785
5786=cut
5787*/
5788
5789/*
5790 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
5791 * PERL_UTF8_magic of the sv to store the mapping between UTF-8 and
5792 * byte offsets.
5793 *
5794 */
a0ed51b3 5795void
7e8c5dac 5796Perl_sv_pos_b2u(pTHX_ register SV* sv, I32* offsetp)
a0ed51b3 5797{
83003860 5798 const U8* s;
ec07b5e0 5799 const STRLEN byte = *offsetp;
7087a21c 5800 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 5801 STRLEN blen;
ec07b5e0
NC
5802 MAGIC* mg = NULL;
5803 const U8* send;
a922f900 5804 bool found = FALSE;
a0ed51b3
LW
5805
5806 if (!sv)
5807 return;
5808
ab455f60 5809 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 5810
ab455f60 5811 if (blen < byte)
ec07b5e0 5812 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 5813
ec07b5e0 5814 send = s + byte;
a67d7df9 5815
ffca234a
NC
5816 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
5817 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
5818 if (mg->mg_ptr) {
d4c19fe8 5819 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 5820 if (cache[1] == byte) {
ec07b5e0
NC
5821 /* An exact match. */
5822 *offsetp = cache[0];
ec07b5e0 5823 return;
7e8c5dac 5824 }
ab455f60
NC
5825 if (cache[3] == byte) {
5826 /* An exact match. */
5827 *offsetp = cache[2];
5828 return;
5829 }
668af93f
NC
5830
5831 if (cache[1] < byte) {
ec07b5e0 5832 /* We already know part of the way. */
b9f984a5
NC
5833 if (mg->mg_len != -1) {
5834 /* Actually, we know the end too. */
5835 len = cache[0]
5836 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 5837 s + blen, mg->mg_len - cache[0]);
b9f984a5
NC
5838 } else {
5839 len = cache[0]
5840 + S_sv_pos_b2u_forwards(aTHX_ s + cache[1], send);
5841 }
7e8c5dac 5842 }
9f985e4c
NC
5843 else if (cache[3] < byte) {
5844 /* We're between the two cached pairs, so we do the calculation
5845 offset by the byte/utf-8 positions for the earlier pair,
5846 then add the utf-8 characters from the string start to
5847 there. */
5848 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
5849 s + cache[1], cache[0] - cache[2])
5850 + cache[2];
5851
5852 }
5853 else { /* cache[3] > byte */
5854 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
5855 cache[2]);
7e8c5dac 5856
7e8c5dac 5857 }
ec07b5e0 5858 ASSERT_UTF8_CACHE(cache);
a922f900 5859 found = TRUE;
ffca234a 5860 } else if (mg->mg_len != -1) {
ab455f60 5861 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 5862 found = TRUE;
7e8c5dac 5863 }
a0ed51b3 5864 }
a922f900
NC
5865 if (!found || PL_utf8cache < 0) {
5866 const STRLEN real_len = S_sv_pos_b2u_forwards(aTHX_ s, send);
5867
5868 if (found && PL_utf8cache < 0) {
5869 if (len != real_len) {
5870 /* Need to turn the assertions off otherwise we may recurse
5871 infinitely while printing error messages. */
5872 SAVEI8(PL_utf8cache);
5873 PL_utf8cache = 0;
5874 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVf
5875 " real %"UVf" for %"SVf,
5876 (UV) len, (UV) real_len, sv);
5877 }
5878 }
5879 len = real_len;
ec07b5e0
NC
5880 }
5881 *offsetp = len;
5882
ab455f60 5883 S_utf8_mg_pos_cache_update(aTHX_ sv, &mg, byte, len, blen);
a0ed51b3
LW
5884}
5885
954c1994
GS
5886/*
5887=for apidoc sv_eq
5888
5889Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
5890identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5891coerce its args to strings if necessary.
954c1994
GS
5892
5893=cut
5894*/
5895
79072805 5896I32
e01b9e88 5897Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 5898{
97aff369 5899 dVAR;
e1ec3a88 5900 const char *pv1;
463ee0b2 5901 STRLEN cur1;
e1ec3a88 5902 const char *pv2;
463ee0b2 5903 STRLEN cur2;
e01b9e88 5904 I32 eq = 0;
bd61b366 5905 char *tpv = NULL;
a0714e2c 5906 SV* svrecode = NULL;
79072805 5907
e01b9e88 5908 if (!sv1) {
79072805
LW
5909 pv1 = "";
5910 cur1 = 0;
5911 }
463ee0b2 5912 else
4d84ee25 5913 pv1 = SvPV_const(sv1, cur1);
79072805 5914
e01b9e88
SC
5915 if (!sv2){
5916 pv2 = "";
5917 cur2 = 0;
92d29cee 5918 }
e01b9e88 5919 else
4d84ee25 5920 pv2 = SvPV_const(sv2, cur2);
79072805 5921
cf48d248 5922 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
5923 /* Differing utf8ness.
5924 * Do not UTF8size the comparands as a side-effect. */
5925 if (PL_encoding) {
5926 if (SvUTF8(sv1)) {
553e1bcc
AT
5927 svrecode = newSVpvn(pv2, cur2);
5928 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5929 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
5930 }
5931 else {
553e1bcc
AT
5932 svrecode = newSVpvn(pv1, cur1);
5933 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 5934 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
5935 }
5936 /* Now both are in UTF-8. */
0a1bd7ac
DM
5937 if (cur1 != cur2) {
5938 SvREFCNT_dec(svrecode);
799ef3cb 5939 return FALSE;
0a1bd7ac 5940 }
799ef3cb
JH
5941 }
5942 else {
5943 bool is_utf8 = TRUE;
5944
5945 if (SvUTF8(sv1)) {
5946 /* sv1 is the UTF-8 one,
5947 * if is equal it must be downgrade-able */
9d4ba2ae 5948 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
5949 &cur1, &is_utf8);
5950 if (pv != pv1)
553e1bcc 5951 pv1 = tpv = pv;
799ef3cb
JH
5952 }
5953 else {
5954 /* sv2 is the UTF-8 one,
5955 * if is equal it must be downgrade-able */
9d4ba2ae 5956 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
5957 &cur2, &is_utf8);
5958 if (pv != pv2)
553e1bcc 5959 pv2 = tpv = pv;
799ef3cb
JH
5960 }
5961 if (is_utf8) {
5962 /* Downgrade not possible - cannot be eq */
bf694877 5963 assert (tpv == 0);
799ef3cb
JH
5964 return FALSE;
5965 }
5966 }
cf48d248
JH
5967 }
5968
5969 if (cur1 == cur2)
765f542d 5970 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 5971
b37c2d43 5972 SvREFCNT_dec(svrecode);
553e1bcc
AT
5973 if (tpv)
5974 Safefree(tpv);
cf48d248 5975
e01b9e88 5976 return eq;
79072805
LW
5977}
5978
954c1994
GS
5979/*
5980=for apidoc sv_cmp
5981
5982Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
5983string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
5984C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
5985coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
5986
5987=cut
5988*/
5989
79072805 5990I32
e01b9e88 5991Perl_sv_cmp(pTHX_ register SV *sv1, register SV *sv2)
79072805 5992{
97aff369 5993 dVAR;
560a288e 5994 STRLEN cur1, cur2;
e1ec3a88 5995 const char *pv1, *pv2;
bd61b366 5996 char *tpv = NULL;
cf48d248 5997 I32 cmp;
a0714e2c 5998 SV *svrecode = NULL;
560a288e 5999
e01b9e88
SC
6000 if (!sv1) {
6001 pv1 = "";
560a288e
GS
6002 cur1 = 0;
6003 }
e01b9e88 6004 else
4d84ee25 6005 pv1 = SvPV_const(sv1, cur1);
560a288e 6006
553e1bcc 6007 if (!sv2) {
e01b9e88 6008 pv2 = "";
560a288e
GS
6009 cur2 = 0;
6010 }
e01b9e88 6011 else
4d84ee25 6012 pv2 = SvPV_const(sv2, cur2);
79072805 6013
cf48d248 6014 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6015 /* Differing utf8ness.
6016 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6017 if (SvUTF8(sv1)) {
799ef3cb 6018 if (PL_encoding) {
553e1bcc
AT
6019 svrecode = newSVpvn(pv2, cur2);
6020 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6021 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6022 }
6023 else {
e1ec3a88 6024 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6025 }
cf48d248
JH
6026 }
6027 else {
799ef3cb 6028 if (PL_encoding) {
553e1bcc
AT
6029 svrecode = newSVpvn(pv1, cur1);
6030 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6031 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6032 }
6033 else {
e1ec3a88 6034 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6035 }
cf48d248
JH
6036 }
6037 }
6038
e01b9e88 6039 if (!cur1) {
cf48d248 6040 cmp = cur2 ? -1 : 0;
e01b9e88 6041 } else if (!cur2) {
cf48d248
JH
6042 cmp = 1;
6043 } else {
e1ec3a88 6044 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6045
6046 if (retval) {
cf48d248 6047 cmp = retval < 0 ? -1 : 1;
e01b9e88 6048 } else if (cur1 == cur2) {
cf48d248
JH
6049 cmp = 0;
6050 } else {
6051 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6052 }
cf48d248 6053 }
16660edb 6054
b37c2d43 6055 SvREFCNT_dec(svrecode);
553e1bcc
AT
6056 if (tpv)
6057 Safefree(tpv);
cf48d248
JH
6058
6059 return cmp;
bbce6d69 6060}
16660edb 6061
c461cf8f
JH
6062/*
6063=for apidoc sv_cmp_locale
6064
645c22ef
DM
6065Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6066'use bytes' aware, handles get magic, and will coerce its args to strings
6067if necessary. See also C<sv_cmp_locale>. See also C<sv_cmp>.
c461cf8f
JH
6068
6069=cut
6070*/
6071
bbce6d69 6072I32
864dbfa3 6073Perl_sv_cmp_locale(pTHX_ register SV *sv1, register SV *sv2)
bbce6d69 6074{
97aff369 6075 dVAR;
36477c24 6076#ifdef USE_LOCALE_COLLATE
16660edb 6077
bbce6d69 6078 char *pv1, *pv2;
6079 STRLEN len1, len2;
6080 I32 retval;
16660edb 6081
3280af22 6082 if (PL_collation_standard)
bbce6d69 6083 goto raw_compare;
16660edb 6084
bbce6d69 6085 len1 = 0;
8ac85365 6086 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6087 len2 = 0;
8ac85365 6088 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6089
bbce6d69 6090 if (!pv1 || !len1) {
6091 if (pv2 && len2)
6092 return -1;
6093 else
6094 goto raw_compare;
6095 }
6096 else {
6097 if (!pv2 || !len2)
6098 return 1;
6099 }
16660edb 6100
bbce6d69 6101 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6102
bbce6d69 6103 if (retval)
16660edb 6104 return retval < 0 ? -1 : 1;
6105
bbce6d69 6106 /*
6107 * When the result of collation is equality, that doesn't mean
6108 * that there are no differences -- some locales exclude some
6109 * characters from consideration. So to avoid false equalities,
6110 * we use the raw string as a tiebreaker.
6111 */
16660edb 6112
bbce6d69 6113 raw_compare:
5f66b61c 6114 /*FALLTHROUGH*/
16660edb 6115
36477c24 6116#endif /* USE_LOCALE_COLLATE */
16660edb 6117
bbce6d69 6118 return sv_cmp(sv1, sv2);
6119}
79072805 6120
645c22ef 6121
36477c24 6122#ifdef USE_LOCALE_COLLATE
645c22ef 6123
7a4c00b4 6124/*
645c22ef
DM
6125=for apidoc sv_collxfrm
6126
6127Add Collate Transform magic to an SV if it doesn't already have it.
6128
6129Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6130scalar data of the variable, but transformed to such a format that a normal
6131memory comparison can be used to compare the data according to the locale
6132settings.
6133
6134=cut
6135*/
6136
bbce6d69 6137char *
864dbfa3 6138Perl_sv_collxfrm(pTHX_ SV *sv, STRLEN *nxp)
bbce6d69 6139{
97aff369 6140 dVAR;
7a4c00b4 6141 MAGIC *mg;
16660edb 6142
14befaf4 6143 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6144 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6145 const char *s;
6146 char *xf;
bbce6d69 6147 STRLEN len, xlen;
6148
7a4c00b4 6149 if (mg)
6150 Safefree(mg->mg_ptr);
93524f2b 6151 s = SvPV_const(sv, len);
bbce6d69 6152 if ((xf = mem_collxfrm(s, len, &xlen))) {
ff0cee69 6153 if (SvREADONLY(sv)) {
6154 SAVEFREEPV(xf);
6155 *nxp = xlen;
3280af22 6156 return xf + sizeof(PL_collation_ix);
ff0cee69 6157 }
7a4c00b4 6158 if (! mg) {
d83f0a82
NC
6159#ifdef PERL_OLD_COPY_ON_WRITE
6160 if (SvIsCOW(sv))
6161 sv_force_normal_flags(sv, 0);
6162#endif
6163 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6164 0, 0);
7a4c00b4 6165 assert(mg);
bbce6d69 6166 }
7a4c00b4 6167 mg->mg_ptr = xf;
565764a8 6168 mg->mg_len = xlen;
7a4c00b4 6169 }
6170 else {
ff0cee69 6171 if (mg) {
6172 mg->mg_ptr = NULL;
565764a8 6173 mg->mg_len = -1;
ff0cee69 6174 }
bbce6d69 6175 }
6176 }
7a4c00b4 6177 if (mg && mg->mg_ptr) {
565764a8 6178 *nxp = mg->mg_len;
3280af22 6179 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6180 }
6181 else {
6182 *nxp = 0;
6183 return NULL;
16660edb 6184 }
79072805
LW
6185}
6186
36477c24 6187#endif /* USE_LOCALE_COLLATE */
bbce6d69 6188
c461cf8f
JH
6189/*
6190=for apidoc sv_gets
6191
6192Get a line from the filehandle and store it into the SV, optionally
6193appending to the currently-stored string.
6194
6195=cut
6196*/
6197
79072805 6198char *
864dbfa3 6199Perl_sv_gets(pTHX_ register SV *sv, register PerlIO *fp, I32 append)
79072805 6200{
97aff369 6201 dVAR;
e1ec3a88 6202 const char *rsptr;
c07a80fd 6203 STRLEN rslen;
6204 register STDCHAR rslast;
6205 register STDCHAR *bp;
6206 register I32 cnt;
9c5ffd7c 6207 I32 i = 0;
8bfdd7d9 6208 I32 rspara = 0;
c07a80fd 6209
bc44a8a2
NC
6210 if (SvTHINKFIRST(sv))
6211 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6212 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6213 from <>.
6214 However, perlbench says it's slower, because the existing swipe code
6215 is faster than copy on write.
6216 Swings and roundabouts. */
862a34c6 6217 SvUPGRADE(sv, SVt_PV);
99491443 6218
ff68c719 6219 SvSCREAM_off(sv);
efd8b2ba
AE
6220
6221 if (append) {
6222 if (PerlIO_isutf8(fp)) {
6223 if (!SvUTF8(sv)) {
6224 sv_utf8_upgrade_nomg(sv);
6225 sv_pos_u2b(sv,&append,0);
6226 }
6227 } else if (SvUTF8(sv)) {
561b68a9 6228 SV * const tsv = newSV(0);
efd8b2ba
AE
6229 sv_gets(tsv, fp, 0);
6230 sv_utf8_upgrade_nomg(tsv);
6231 SvCUR_set(sv,append);
6232 sv_catsv(sv,tsv);
6233 sv_free(tsv);
6234 goto return_string_or_null;
6235 }
6236 }
6237
6238 SvPOK_only(sv);
6239 if (PerlIO_isutf8(fp))
6240 SvUTF8_on(sv);
c07a80fd 6241
923e4eb5 6242 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6243 /* we always read code in line mode */
6244 rsptr = "\n";
6245 rslen = 1;
6246 }
6247 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6248 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6249 of amount we are going to read -- may result in mallocing
6250 more memory than we really need if the layers below reduce
6251 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6252 */
e311fd51 6253 Stat_t st;
e468d35b 6254 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6255 const Off_t offset = PerlIO_tell(fp);
58f1856e 6256 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6257 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6258 }
6259 }
c07a80fd 6260 rsptr = NULL;
6261 rslen = 0;
6262 }
3280af22 6263 else if (RsRECORD(PL_rs)) {
e311fd51 6264 I32 bytesread;
5b2b9c68 6265 char *buffer;
acbd132f 6266 U32 recsize;
5b2b9c68
HM
6267
6268 /* Grab the size of the record we're getting */
acbd132f 6269 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6270 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6271 /* Go yank in */
6272#ifdef VMS
6273 /* VMS wants read instead of fread, because fread doesn't respect */
6274 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6275 /* doing, but we've got no other real choice - except avoid stdio
6276 as implementation - perhaps write a :vms layer ?
6277 */
5b2b9c68
HM
6278 bytesread = PerlLIO_read(PerlIO_fileno(fp), buffer, recsize);
6279#else
6280 bytesread = PerlIO_read(fp, buffer, recsize);
6281#endif
27e6ca2d
AE
6282 if (bytesread < 0)
6283 bytesread = 0;
e311fd51 6284 SvCUR_set(sv, bytesread += append);
e670df4e 6285 buffer[bytesread] = '\0';
efd8b2ba 6286 goto return_string_or_null;
5b2b9c68 6287 }
3280af22 6288 else if (RsPARA(PL_rs)) {
c07a80fd 6289 rsptr = "\n\n";
6290 rslen = 2;
8bfdd7d9 6291 rspara = 1;
c07a80fd 6292 }
7d59b7e4
NIS
6293 else {
6294 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6295 if (PerlIO_isutf8(fp)) {
6296 rsptr = SvPVutf8(PL_rs, rslen);
6297 }
6298 else {
6299 if (SvUTF8(PL_rs)) {
6300 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6301 Perl_croak(aTHX_ "Wide character in $/");
6302 }
6303 }
93524f2b 6304 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6305 }
6306 }
6307
c07a80fd 6308 rslast = rslen ? rsptr[rslen - 1] : '\0';
6309
8bfdd7d9 6310 if (rspara) { /* have to do this both before and after */
79072805 6311 do { /* to make sure file boundaries work right */
760ac839 6312 if (PerlIO_eof(fp))
a0d0e21e 6313 return 0;
760ac839 6314 i = PerlIO_getc(fp);
79072805 6315 if (i != '\n') {
a0d0e21e
LW
6316 if (i == -1)
6317 return 0;
760ac839 6318 PerlIO_ungetc(fp,i);
79072805
LW
6319 break;
6320 }
6321 } while (i != EOF);
6322 }
c07a80fd 6323
760ac839
LW
6324 /* See if we know enough about I/O mechanism to cheat it ! */
6325
6326 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 6327 of abstracting out stdio interface. One call should be cheap
760ac839
LW
6328 enough here - and may even be a macro allowing compile
6329 time optimization.
6330 */
6331
6332 if (PerlIO_fast_gets(fp)) {
6333
6334 /*
6335 * We're going to steal some values from the stdio struct
6336 * and put EVERYTHING in the innermost loop into registers.
6337 */
6338 register STDCHAR *ptr;
6339 STRLEN bpx;
6340 I32 shortbuffered;
6341
16660edb 6342#if defined(VMS) && defined(PERLIO_IS_STDIO)
6343 /* An ungetc()d char is handled separately from the regular
6344 * buffer, so we getc() it back out and stuff it in the buffer.
6345 */
6346 i = PerlIO_getc(fp);
6347 if (i == EOF) return 0;
6348 *(--((*fp)->_ptr)) = (unsigned char) i;
6349 (*fp)->_cnt++;
6350#endif
c07a80fd 6351
c2960299 6352 /* Here is some breathtakingly efficient cheating */
c07a80fd 6353
a20bf0c3 6354 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 6355 /* make sure we have the room */
7a5fa8a2 6356 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 6357 /* Not room for all of it
7a5fa8a2 6358 if we are looking for a separator and room for some
e468d35b
NIS
6359 */
6360 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 6361 /* just process what we have room for */
79072805
LW
6362 shortbuffered = cnt - SvLEN(sv) + append + 1;
6363 cnt -= shortbuffered;
6364 }
6365 else {
6366 shortbuffered = 0;
bbce6d69 6367 /* remember that cnt can be negative */
eb160463 6368 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
6369 }
6370 }
7a5fa8a2 6371 else
79072805 6372 shortbuffered = 0;
3f7c398e 6373 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 6374 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 6375 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6376 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 6377 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 6378 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6379 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6380 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
6381 for (;;) {
6382 screamer:
93a17b20 6383 if (cnt > 0) {
c07a80fd 6384 if (rslen) {
760ac839
LW
6385 while (cnt > 0) { /* this | eat */
6386 cnt--;
c07a80fd 6387 if ((*bp++ = *ptr++) == rslast) /* really | dust */
6388 goto thats_all_folks; /* screams | sed :-) */
6389 }
6390 }
6391 else {
1c846c1f
NIS
6392 Copy(ptr, bp, cnt, char); /* this | eat */
6393 bp += cnt; /* screams | dust */
c07a80fd 6394 ptr += cnt; /* louder | sed :-) */
a5f75d66 6395 cnt = 0;
93a17b20 6396 }
79072805
LW
6397 }
6398
748a9306 6399 if (shortbuffered) { /* oh well, must extend */
79072805
LW
6400 cnt = shortbuffered;
6401 shortbuffered = 0;
3f7c398e 6402 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6403 SvCUR_set(sv, bpx);
6404 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 6405 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
6406 continue;
6407 }
6408
16660edb 6409 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
6410 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
6411 PTR2UV(ptr),(long)cnt));
cc00df79 6412 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 6413#if 0
16660edb 6414 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6415 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6416 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6417 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6418#endif
1c846c1f 6419 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 6420 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
6421 another abstraction. */
760ac839 6422 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 6423#if 0
16660edb 6424 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6425 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6426 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6427 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 6428#endif
a20bf0c3
JH
6429 cnt = PerlIO_get_cnt(fp);
6430 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 6431 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6432 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 6433
748a9306
LW
6434 if (i == EOF) /* all done for ever? */
6435 goto thats_really_all_folks;
6436
3f7c398e 6437 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
6438 SvCUR_set(sv, bpx);
6439 SvGROW(sv, bpx + cnt + 2);
3f7c398e 6440 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 6441
eb160463 6442 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 6443
c07a80fd 6444 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 6445 goto thats_all_folks;
79072805
LW
6446 }
6447
6448thats_all_folks:
3f7c398e 6449 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 6450 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 6451 goto screamer; /* go back to the fray */
79072805
LW
6452thats_really_all_folks:
6453 if (shortbuffered)
6454 cnt += shortbuffered;
16660edb 6455 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6456 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 6457 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 6458 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 6459 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 6460 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 6461 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 6462 *bp = '\0';
3f7c398e 6463 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 6464 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 6465 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 6466 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
6467 }
6468 else
79072805 6469 {
6edd2cd5 6470 /*The big, slow, and stupid way. */
27da23d5 6471#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 6472 STDCHAR *buf = NULL;
a02a5408 6473 Newx(buf, 8192, STDCHAR);
6edd2cd5 6474 assert(buf);
4d2c4e07 6475#else
6edd2cd5 6476 STDCHAR buf[8192];
4d2c4e07 6477#endif
79072805 6478
760ac839 6479screamer2:
c07a80fd 6480 if (rslen) {
00b6aa41 6481 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 6482 bp = buf;
eb160463 6483 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
6484 ; /* keep reading */
6485 cnt = bp - buf;
c07a80fd 6486 }
6487 else {
760ac839 6488 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 6489 /* Accomodate broken VAXC compiler, which applies U8 cast to
6490 * both args of ?: operator, causing EOF to change into 255
6491 */
37be0adf 6492 if (cnt > 0)
cbe9e203
JH
6493 i = (U8)buf[cnt - 1];
6494 else
37be0adf 6495 i = EOF;
c07a80fd 6496 }
79072805 6497
cbe9e203
JH
6498 if (cnt < 0)
6499 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
6500 if (append)
6501 sv_catpvn(sv, (char *) buf, cnt);
6502 else
6503 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 6504
6505 if (i != EOF && /* joy */
6506 (!rslen ||
6507 SvCUR(sv) < rslen ||
3f7c398e 6508 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
6509 {
6510 append = -1;
63e4d877
CS
6511 /*
6512 * If we're reading from a TTY and we get a short read,
6513 * indicating that the user hit his EOF character, we need
6514 * to notice it now, because if we try to read from the TTY
6515 * again, the EOF condition will disappear.
6516 *
6517 * The comparison of cnt to sizeof(buf) is an optimization
6518 * that prevents unnecessary calls to feof().
6519 *
6520 * - jik 9/25/96
6521 */
6522 if (!(cnt < sizeof(buf) && PerlIO_eof(fp)))
6523 goto screamer2;
79072805 6524 }
6edd2cd5 6525
27da23d5 6526#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
6527 Safefree(buf);
6528#endif
79072805
LW
6529 }
6530
8bfdd7d9 6531 if (rspara) { /* have to do this both before and after */
c07a80fd 6532 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 6533 i = PerlIO_getc(fp);
79072805 6534 if (i != '\n') {
760ac839 6535 PerlIO_ungetc(fp,i);
79072805
LW
6536 break;
6537 }
6538 }
6539 }
c07a80fd 6540
efd8b2ba 6541return_string_or_null:
bd61b366 6542 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
6543}
6544
954c1994
GS
6545/*
6546=for apidoc sv_inc
6547
645c22ef
DM
6548Auto-increment of the value in the SV, doing string to numeric conversion
6549if necessary. Handles 'get' magic.
954c1994
GS
6550
6551=cut
6552*/
6553
79072805 6554void
864dbfa3 6555Perl_sv_inc(pTHX_ register SV *sv)
79072805 6556{
97aff369 6557 dVAR;
79072805 6558 register char *d;
463ee0b2 6559 int flags;
79072805
LW
6560
6561 if (!sv)
6562 return;
5b295bef 6563 SvGETMAGIC(sv);
ed6116ce 6564 if (SvTHINKFIRST(sv)) {
765f542d
NC
6565 if (SvIsCOW(sv))
6566 sv_force_normal_flags(sv, 0);
0f15f207 6567 if (SvREADONLY(sv)) {
923e4eb5 6568 if (IN_PERL_RUNTIME)
cea2e8a9 6569 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6570 }
a0d0e21e 6571 if (SvROK(sv)) {
b5be31e9 6572 IV i;
9e7bc3e8
JD
6573 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
6574 return;
56431972 6575 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6576 sv_unref(sv);
6577 sv_setiv(sv, i);
a0d0e21e 6578 }
ed6116ce 6579 }
8990e307 6580 flags = SvFLAGS(sv);
28e5dec8
JH
6581 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
6582 /* It's (privately or publicly) a float, but not tested as an
6583 integer, so test it to see. */
d460ef45 6584 (void) SvIV(sv);
28e5dec8
JH
6585 flags = SvFLAGS(sv);
6586 }
6587 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6588 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6589#ifdef PERL_PRESERVE_IVUV
28e5dec8 6590 oops_its_int:
59d8ce62 6591#endif
25da4f38
IZ
6592 if (SvIsUV(sv)) {
6593 if (SvUVX(sv) == UV_MAX)
a1e868e7 6594 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
6595 else
6596 (void)SvIOK_only_UV(sv);
607fa7f2 6597 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
6598 } else {
6599 if (SvIVX(sv) == IV_MAX)
28e5dec8 6600 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
6601 else {
6602 (void)SvIOK_only(sv);
45977657 6603 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 6604 }
55497cff 6605 }
79072805
LW
6606 return;
6607 }
28e5dec8
JH
6608 if (flags & SVp_NOK) {
6609 (void)SvNOK_only(sv);
9d6ce603 6610 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6611 return;
6612 }
6613
3f7c398e 6614 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 6615 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 6616 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 6617 (void)SvIOK_only(sv);
45977657 6618 SvIV_set(sv, 1);
79072805
LW
6619 return;
6620 }
463ee0b2 6621 d = SvPVX(sv);
79072805
LW
6622 while (isALPHA(*d)) d++;
6623 while (isDIGIT(*d)) d++;
6624 if (*d) {
28e5dec8 6625#ifdef PERL_PRESERVE_IVUV
d1be9408 6626 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
6627 warnings. Probably ought to make the sv_iv_please() that does
6628 the conversion if possible, and silently. */
504618e9 6629 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6630 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6631 /* Need to try really hard to see if it's an integer.
6632 9.22337203685478e+18 is an integer.
6633 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6634 so $a="9.22337203685478e+18"; $a+0; $a++
6635 needs to be the same as $a="9.22337203685478e+18"; $a++
6636 or we go insane. */
d460ef45 6637
28e5dec8
JH
6638 (void) sv_2iv(sv);
6639 if (SvIOK(sv))
6640 goto oops_its_int;
6641
6642 /* sv_2iv *should* have made this an NV */
6643 if (flags & SVp_NOK) {
6644 (void)SvNOK_only(sv);
9d6ce603 6645 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
6646 return;
6647 }
6648 /* I don't think we can get here. Maybe I should assert this
6649 And if we do get here I suspect that sv_setnv will croak. NWC
6650 Fall through. */
6651#if defined(USE_LONG_DOUBLE)
6652 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 6653 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6654#else
1779d84d 6655 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 6656 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6657#endif
6658 }
6659#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6660 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
6661 return;
6662 }
6663 d--;
3f7c398e 6664 while (d >= SvPVX_const(sv)) {
79072805
LW
6665 if (isDIGIT(*d)) {
6666 if (++*d <= '9')
6667 return;
6668 *(d--) = '0';
6669 }
6670 else {
9d116dd7
JH
6671#ifdef EBCDIC
6672 /* MKS: The original code here died if letters weren't consecutive.
6673 * at least it didn't have to worry about non-C locales. The
6674 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 6675 * arranged in order (although not consecutively) and that only
9d116dd7
JH
6676 * [A-Za-z] are accepted by isALPHA in the C locale.
6677 */
6678 if (*d != 'z' && *d != 'Z') {
6679 do { ++*d; } while (!isALPHA(*d));
6680 return;
6681 }
6682 *(d--) -= 'z' - 'a';
6683#else
79072805
LW
6684 ++*d;
6685 if (isALPHA(*d))
6686 return;
6687 *(d--) -= 'z' - 'a' + 1;
9d116dd7 6688#endif
79072805
LW
6689 }
6690 }
6691 /* oh,oh, the number grew */
6692 SvGROW(sv, SvCUR(sv) + 2);
b162af07 6693 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 6694 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
6695 *d = d[-1];
6696 if (isDIGIT(d[1]))
6697 *d = '1';
6698 else
6699 *d = d[1];
6700}
6701
954c1994
GS
6702/*
6703=for apidoc sv_dec
6704
645c22ef
DM
6705Auto-decrement of the value in the SV, doing string to numeric conversion
6706if necessary. Handles 'get' magic.
954c1994
GS
6707
6708=cut
6709*/
6710
79072805 6711void
864dbfa3 6712Perl_sv_dec(pTHX_ register SV *sv)
79072805 6713{
97aff369 6714 dVAR;
463ee0b2
LW
6715 int flags;
6716
79072805
LW
6717 if (!sv)
6718 return;
5b295bef 6719 SvGETMAGIC(sv);
ed6116ce 6720 if (SvTHINKFIRST(sv)) {
765f542d
NC
6721 if (SvIsCOW(sv))
6722 sv_force_normal_flags(sv, 0);
0f15f207 6723 if (SvREADONLY(sv)) {
923e4eb5 6724 if (IN_PERL_RUNTIME)
cea2e8a9 6725 Perl_croak(aTHX_ PL_no_modify);
0f15f207 6726 }
a0d0e21e 6727 if (SvROK(sv)) {
b5be31e9 6728 IV i;
9e7bc3e8
JD
6729 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
6730 return;
56431972 6731 i = PTR2IV(SvRV(sv));
b5be31e9
SM
6732 sv_unref(sv);
6733 sv_setiv(sv, i);
a0d0e21e 6734 }
ed6116ce 6735 }
28e5dec8
JH
6736 /* Unlike sv_inc we don't have to worry about string-never-numbers
6737 and keeping them magic. But we mustn't warn on punting */
8990e307 6738 flags = SvFLAGS(sv);
28e5dec8
JH
6739 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
6740 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 6741#ifdef PERL_PRESERVE_IVUV
28e5dec8 6742 oops_its_int:
59d8ce62 6743#endif
25da4f38
IZ
6744 if (SvIsUV(sv)) {
6745 if (SvUVX(sv) == 0) {
6746 (void)SvIOK_only(sv);
45977657 6747 SvIV_set(sv, -1);
25da4f38
IZ
6748 }
6749 else {
6750 (void)SvIOK_only_UV(sv);
f4eee32f 6751 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 6752 }
25da4f38
IZ
6753 } else {
6754 if (SvIVX(sv) == IV_MIN)
65202027 6755 sv_setnv(sv, (NV)IV_MIN - 1.0);
25da4f38
IZ
6756 else {
6757 (void)SvIOK_only(sv);
45977657 6758 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 6759 }
55497cff 6760 }
6761 return;
6762 }
28e5dec8 6763 if (flags & SVp_NOK) {
9d6ce603 6764 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6765 (void)SvNOK_only(sv);
6766 return;
6767 }
8990e307 6768 if (!(flags & SVp_POK)) {
ef088171
NC
6769 if ((flags & SVTYPEMASK) < SVt_PVIV)
6770 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
6771 SvIV_set(sv, -1);
6772 (void)SvIOK_only(sv);
79072805
LW
6773 return;
6774 }
28e5dec8
JH
6775#ifdef PERL_PRESERVE_IVUV
6776 {
504618e9 6777 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
6778 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
6779 /* Need to try really hard to see if it's an integer.
6780 9.22337203685478e+18 is an integer.
6781 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
6782 so $a="9.22337203685478e+18"; $a+0; $a--
6783 needs to be the same as $a="9.22337203685478e+18"; $a--
6784 or we go insane. */
d460ef45 6785
28e5dec8
JH
6786 (void) sv_2iv(sv);
6787 if (SvIOK(sv))
6788 goto oops_its_int;
6789
6790 /* sv_2iv *should* have made this an NV */
6791 if (flags & SVp_NOK) {
6792 (void)SvNOK_only(sv);
9d6ce603 6793 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
6794 return;
6795 }
6796 /* I don't think we can get here. Maybe I should assert this
6797 And if we do get here I suspect that sv_setnv will croak. NWC
6798 Fall through. */
6799#if defined(USE_LONG_DOUBLE)
6800 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 6801 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 6802#else
1779d84d 6803 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 6804 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
6805#endif
6806 }
6807 }
6808#endif /* PERL_PRESERVE_IVUV */
3f7c398e 6809 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
6810}
6811
954c1994
GS
6812/*
6813=for apidoc sv_mortalcopy
6814
645c22ef 6815Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
6816The new SV is marked as mortal. It will be destroyed "soon", either by an
6817explicit call to FREETMPS, or by an implicit call at places such as
6818statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
6819
6820=cut
6821*/
6822
79072805
LW
6823/* Make a string that will exist for the duration of the expression
6824 * evaluation. Actually, it may have to last longer than that, but
6825 * hopefully we won't free it until it has been assigned to a
6826 * permanent location. */
6827
6828SV *
864dbfa3 6829Perl_sv_mortalcopy(pTHX_ SV *oldstr)
79072805 6830{
97aff369 6831 dVAR;
463ee0b2 6832 register SV *sv;
b881518d 6833
4561caa4 6834 new_SV(sv);
79072805 6835 sv_setsv(sv,oldstr);
677b06e3
GS
6836 EXTEND_MORTAL(1);
6837 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307
LW
6838 SvTEMP_on(sv);
6839 return sv;
6840}
6841
954c1994
GS
6842/*
6843=for apidoc sv_newmortal
6844
645c22ef 6845Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
6846set to 1. It will be destroyed "soon", either by an explicit call to
6847FREETMPS, or by an implicit call at places such as statement boundaries.
6848See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
6849
6850=cut
6851*/
6852
8990e307 6853SV *
864dbfa3 6854Perl_sv_newmortal(pTHX)
8990e307 6855{
97aff369 6856 dVAR;
8990e307
LW
6857 register SV *sv;
6858
4561caa4 6859 new_SV(sv);
8990e307 6860 SvFLAGS(sv) = SVs_TEMP;
677b06e3
GS
6861 EXTEND_MORTAL(1);
6862 PL_tmps_stack[++PL_tmps_ix] = sv;
79072805
LW
6863 return sv;
6864}
6865
954c1994
GS
6866/*
6867=for apidoc sv_2mortal
6868
d4236ebc
DM
6869Marks an existing SV as mortal. The SV will be destroyed "soon", either
6870by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
6871statement boundaries. SvTEMP() is turned on which means that the SV's
6872string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
6873and C<sv_mortalcopy>.
954c1994
GS
6874
6875=cut
6876*/
6877
79072805 6878SV *
864dbfa3 6879Perl_sv_2mortal(pTHX_ register SV *sv)
79072805 6880{
27da23d5 6881 dVAR;
79072805 6882 if (!sv)
7a5b473e 6883 return NULL;
d689ffdd 6884 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 6885 return sv;
677b06e3
GS
6886 EXTEND_MORTAL(1);
6887 PL_tmps_stack[++PL_tmps_ix] = sv;
8990e307 6888 SvTEMP_on(sv);
79072805
LW
6889 return sv;
6890}
6891
954c1994
GS
6892/*
6893=for apidoc newSVpv
6894
6895Creates a new SV and copies a string into it. The reference count for the
6896SV is set to 1. If C<len> is zero, Perl will compute the length using
6897strlen(). For efficiency, consider using C<newSVpvn> instead.
6898
6899=cut
6900*/
6901
79072805 6902SV *
864dbfa3 6903Perl_newSVpv(pTHX_ const char *s, STRLEN len)
79072805 6904{
97aff369 6905 dVAR;
463ee0b2 6906 register SV *sv;
79072805 6907
4561caa4 6908 new_SV(sv);
616d8c9c 6909 sv_setpvn(sv,s,len ? len : strlen(s));
79072805
LW
6910 return sv;
6911}
6912
954c1994
GS
6913/*
6914=for apidoc newSVpvn
6915
6916Creates a new SV and copies a string into it. The reference count for the
1c846c1f 6917SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 6918string. You are responsible for ensuring that the source string is at least
9e09f5f2 6919C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
6920
6921=cut
6922*/
6923
9da1e3b5 6924SV *
864dbfa3 6925Perl_newSVpvn(pTHX_ const char *s, STRLEN len)
9da1e3b5 6926{
97aff369 6927 dVAR;
9da1e3b5
MUN
6928 register SV *sv;
6929
6930 new_SV(sv);
9da1e3b5
MUN
6931 sv_setpvn(sv,s,len);
6932 return sv;
6933}
6934
bd08039b
NC
6935
6936/*
926f8064 6937=for apidoc newSVhek
bd08039b
NC
6938
6939Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
6940point to the shared string table where possible. Returns a new (undefined)
6941SV if the hek is NULL.
bd08039b
NC
6942
6943=cut
6944*/
6945
6946SV *
c1b02ed8 6947Perl_newSVhek(pTHX_ const HEK *hek)
bd08039b 6948{
97aff369 6949 dVAR;
5aaec2b4
NC
6950 if (!hek) {
6951 SV *sv;
6952
6953 new_SV(sv);
6954 return sv;
6955 }
6956
bd08039b
NC
6957 if (HEK_LEN(hek) == HEf_SVKEY) {
6958 return newSVsv(*(SV**)HEK_KEY(hek));
6959 } else {
6960 const int flags = HEK_FLAGS(hek);
6961 if (flags & HVhek_WASUTF8) {
6962 /* Trouble :-)
6963 Andreas would like keys he put in as utf8 to come back as utf8
6964 */
6965 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
6966 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
6967 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
6968
6969 SvUTF8_on (sv);
6970 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
6971 return sv;
6972 } else if (flags & HVhek_REHASH) {
6973 /* We don't have a pointer to the hv, so we have to replicate the
6974 flag into every HEK. This hv is using custom a hasing
6975 algorithm. Hence we can't return a shared string scalar, as
6976 that would contain the (wrong) hash value, and might get passed
6977 into an hv routine with a regular hash */
6978
b64e5050 6979 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
6980 if (HEK_UTF8(hek))
6981 SvUTF8_on (sv);
6982 return sv;
6983 }
6984 /* This will be overwhelminly the most common case. */
409dfe77
NC
6985 {
6986 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
6987 more efficient than sharepvn(). */
6988 SV *sv;
6989
6990 new_SV(sv);
6991 sv_upgrade(sv, SVt_PV);
6992 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
6993 SvCUR_set(sv, HEK_LEN(hek));
6994 SvLEN_set(sv, 0);
6995 SvREADONLY_on(sv);
6996 SvFAKE_on(sv);
6997 SvPOK_on(sv);
6998 if (HEK_UTF8(hek))
6999 SvUTF8_on(sv);
7000 return sv;
7001 }
bd08039b
NC
7002 }
7003}
7004
1c846c1f
NIS
7005/*
7006=for apidoc newSVpvn_share
7007
3f7c398e 7008Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef
DM
7009table. If the string does not already exist in the table, it is created
7010first. Turns on READONLY and FAKE. The string's hash is stored in the UV
7011slot of the SV; if the C<hash> parameter is non-zero, that value is used;
7012otherwise the hash is computed. The idea here is that as the string table
3f7c398e 7013is used for shared hash keys these strings will have SvPVX_const == HeKEY and
645c22ef 7014hash lookup will avoid string compare.
1c846c1f
NIS
7015
7016=cut
7017*/
7018
7019SV *
c3654f1a 7020Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7021{
97aff369 7022 dVAR;
1c846c1f 7023 register SV *sv;
c3654f1a 7024 bool is_utf8 = FALSE;
a51caccf
NC
7025 const char *const orig_src = src;
7026
c3654f1a 7027 if (len < 0) {
77caf834 7028 STRLEN tmplen = -len;
c3654f1a 7029 is_utf8 = TRUE;
75a54232 7030 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7031 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7032 len = tmplen;
7033 }
1c846c1f 7034 if (!hash)
5afd6d42 7035 PERL_HASH(hash, src, len);
1c846c1f 7036 new_SV(sv);
bdd68bc3 7037 sv_upgrade(sv, SVt_PV);
f880fe2f 7038 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7039 SvCUR_set(sv, len);
b162af07 7040 SvLEN_set(sv, 0);
1c846c1f
NIS
7041 SvREADONLY_on(sv);
7042 SvFAKE_on(sv);
7043 SvPOK_on(sv);
c3654f1a
IH
7044 if (is_utf8)
7045 SvUTF8_on(sv);
a51caccf
NC
7046 if (src != orig_src)
7047 Safefree(src);
1c846c1f
NIS
7048 return sv;
7049}
7050
645c22ef 7051
cea2e8a9 7052#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7053
7054/* pTHX_ magic can't cope with varargs, so this is a no-context
7055 * version of the main function, (which may itself be aliased to us).
7056 * Don't access this version directly.
7057 */
7058
46fc3d4c 7059SV *
cea2e8a9 7060Perl_newSVpvf_nocontext(const char* pat, ...)
46fc3d4c 7061{
cea2e8a9 7062 dTHX;
46fc3d4c 7063 register SV *sv;
7064 va_list args;
46fc3d4c 7065 va_start(args, pat);
c5be433b 7066 sv = vnewSVpvf(pat, &args);
46fc3d4c 7067 va_end(args);
7068 return sv;
7069}
cea2e8a9 7070#endif
46fc3d4c 7071
954c1994
GS
7072/*
7073=for apidoc newSVpvf
7074
645c22ef 7075Creates a new SV and initializes it with the string formatted like
954c1994
GS
7076C<sprintf>.
7077
7078=cut
7079*/
7080
cea2e8a9
GS
7081SV *
7082Perl_newSVpvf(pTHX_ const char* pat, ...)
7083{
7084 register SV *sv;
7085 va_list args;
cea2e8a9 7086 va_start(args, pat);
c5be433b 7087 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7088 va_end(args);
7089 return sv;
7090}
46fc3d4c 7091
645c22ef
DM
7092/* backend for newSVpvf() and newSVpvf_nocontext() */
7093
79072805 7094SV *
c5be433b
GS
7095Perl_vnewSVpvf(pTHX_ const char* pat, va_list* args)
7096{
97aff369 7097 dVAR;
c5be433b
GS
7098 register SV *sv;
7099 new_SV(sv);
4608196e 7100 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7101 return sv;
7102}
7103
954c1994
GS
7104/*
7105=for apidoc newSVnv
7106
7107Creates a new SV and copies a floating point value into it.
7108The reference count for the SV is set to 1.
7109
7110=cut
7111*/
7112
c5be433b 7113SV *
65202027 7114Perl_newSVnv(pTHX_ NV n)
79072805 7115{
97aff369 7116 dVAR;
463ee0b2 7117 register SV *sv;
79072805 7118
4561caa4 7119 new_SV(sv);
79072805
LW
7120 sv_setnv(sv,n);
7121 return sv;
7122}
7123
954c1994
GS
7124/*
7125=for apidoc newSViv
7126
7127Creates a new SV and copies an integer into it. The reference count for the
7128SV is set to 1.
7129
7130=cut
7131*/
7132
79072805 7133SV *
864dbfa3 7134Perl_newSViv(pTHX_ IV i)
79072805 7135{
97aff369 7136 dVAR;
463ee0b2 7137 register SV *sv;
79072805 7138
4561caa4 7139 new_SV(sv);
79072805
LW
7140 sv_setiv(sv,i);
7141 return sv;
7142}
7143
954c1994 7144/*
1a3327fb
JH
7145=for apidoc newSVuv
7146
7147Creates a new SV and copies an unsigned integer into it.
7148The reference count for the SV is set to 1.
7149
7150=cut
7151*/
7152
7153SV *
7154Perl_newSVuv(pTHX_ UV u)
7155{
97aff369 7156 dVAR;
1a3327fb
JH
7157 register SV *sv;
7158
7159 new_SV(sv);
7160 sv_setuv(sv,u);
7161 return sv;
7162}
7163
7164/*
954c1994
GS
7165=for apidoc newRV_noinc
7166
7167Creates an RV wrapper for an SV. The reference count for the original
7168SV is B<not> incremented.
7169
7170=cut
7171*/
7172
2304df62 7173SV *
864dbfa3 7174Perl_newRV_noinc(pTHX_ SV *tmpRef)
2304df62 7175{
97aff369 7176 dVAR;
2304df62
AD
7177 register SV *sv;
7178
4561caa4 7179 new_SV(sv);
2304df62 7180 sv_upgrade(sv, SVt_RV);
76e3520e 7181 SvTEMP_off(tmpRef);
b162af07 7182 SvRV_set(sv, tmpRef);
2304df62 7183 SvROK_on(sv);
2304df62
AD
7184 return sv;
7185}
7186
ff276b08 7187/* newRV_inc is the official function name to use now.
645c22ef
DM
7188 * newRV_inc is in fact #defined to newRV in sv.h
7189 */
7190
5f05dabc 7191SV *
7f466ec7 7192Perl_newRV(pTHX_ SV *sv)
5f05dabc 7193{
97aff369 7194 dVAR;
7f466ec7 7195 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7196}
5f05dabc 7197
954c1994
GS
7198/*
7199=for apidoc newSVsv
7200
7201Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7202(Uses C<sv_setsv>).
954c1994
GS
7203
7204=cut
7205*/
7206
79072805 7207SV *
864dbfa3 7208Perl_newSVsv(pTHX_ register SV *old)
79072805 7209{
97aff369 7210 dVAR;
463ee0b2 7211 register SV *sv;
79072805
LW
7212
7213 if (!old)
7a5b473e 7214 return NULL;
8990e307 7215 if (SvTYPE(old) == SVTYPEMASK) {
0453d815 7216 if (ckWARN_d(WARN_INTERNAL))
9014280d 7217 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 7218 return NULL;
79072805 7219 }
4561caa4 7220 new_SV(sv);
e90aabeb
NC
7221 /* SV_GMAGIC is the default for sv_setv()
7222 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7223 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7224 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 7225 return sv;
79072805
LW
7226}
7227
645c22ef
DM
7228/*
7229=for apidoc sv_reset
7230
7231Underlying implementation for the C<reset> Perl function.
7232Note that the perl-level function is vaguely deprecated.
7233
7234=cut
7235*/
7236
79072805 7237void
e1ec3a88 7238Perl_sv_reset(pTHX_ register const char *s, HV *stash)
79072805 7239{
27da23d5 7240 dVAR;
4802d5d7 7241 char todo[PERL_UCHAR_MAX+1];
79072805 7242
49d8d3a1
MB
7243 if (!stash)
7244 return;
7245
79072805 7246 if (!*s) { /* reset ?? searches */
aec46f14 7247 MAGIC * const mg = mg_find((SV *)stash, PERL_MAGIC_symtab);
8d2f4536
NC
7248 if (mg) {
7249 PMOP *pm = (PMOP *) mg->mg_obj;
7250 while (pm) {
7251 pm->op_pmdynflags &= ~PMdf_USED;
7252 pm = pm->op_pmnext;
7253 }
79072805
LW
7254 }
7255 return;
7256 }
7257
7258 /* reset variables */
7259
7260 if (!HvARRAY(stash))
7261 return;
463ee0b2
LW
7262
7263 Zero(todo, 256, char);
79072805 7264 while (*s) {
b464bac0
AL
7265 I32 max;
7266 I32 i = (unsigned char)*s;
79072805
LW
7267 if (s[1] == '-') {
7268 s += 2;
7269 }
4802d5d7 7270 max = (unsigned char)*s++;
79072805 7271 for ( ; i <= max; i++) {
463ee0b2
LW
7272 todo[i] = 1;
7273 }
a0d0e21e 7274 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 7275 HE *entry;
79072805 7276 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
7277 entry;
7278 entry = HeNEXT(entry))
7279 {
b464bac0
AL
7280 register GV *gv;
7281 register SV *sv;
7282
1edc1566 7283 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 7284 continue;
1edc1566 7285 gv = (GV*)HeVAL(entry);
79072805 7286 sv = GvSV(gv);
e203899d
NC
7287 if (sv) {
7288 if (SvTHINKFIRST(sv)) {
7289 if (!SvREADONLY(sv) && SvROK(sv))
7290 sv_unref(sv);
7291 /* XXX Is this continue a bug? Why should THINKFIRST
7292 exempt us from resetting arrays and hashes? */
7293 continue;
7294 }
7295 SvOK_off(sv);
7296 if (SvTYPE(sv) >= SVt_PV) {
7297 SvCUR_set(sv, 0);
bd61b366 7298 if (SvPVX_const(sv) != NULL)
e203899d
NC
7299 *SvPVX(sv) = '\0';
7300 SvTAINT(sv);
7301 }
79072805
LW
7302 }
7303 if (GvAV(gv)) {
7304 av_clear(GvAV(gv));
7305 }
bfcb3514 7306 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
7307#if defined(VMS)
7308 Perl_die(aTHX_ "Can't reset %%ENV on this system");
7309#else /* ! VMS */
463ee0b2 7310 hv_clear(GvHV(gv));
b0269e46
AB
7311# if defined(USE_ENVIRON_ARRAY)
7312 if (gv == PL_envgv)
7313 my_clearenv();
7314# endif /* USE_ENVIRON_ARRAY */
7315#endif /* VMS */
79072805
LW
7316 }
7317 }
7318 }
7319 }
7320}
7321
645c22ef
DM
7322/*
7323=for apidoc sv_2io
7324
7325Using various gambits, try to get an IO from an SV: the IO slot if its a
7326GV; or the recursive result if we're an RV; or the IO slot of the symbol
7327named after the PV if we're a string.
7328
7329=cut
7330*/
7331
46fc3d4c 7332IO*
864dbfa3 7333Perl_sv_2io(pTHX_ SV *sv)
46fc3d4c 7334{
7335 IO* io;
7336 GV* gv;
7337
7338 switch (SvTYPE(sv)) {
7339 case SVt_PVIO:
7340 io = (IO*)sv;
7341 break;
7342 case SVt_PVGV:
7343 gv = (GV*)sv;
7344 io = GvIO(gv);
7345 if (!io)
cea2e8a9 7346 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
46fc3d4c 7347 break;
7348 default:
7349 if (!SvOK(sv))
cea2e8a9 7350 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 7351 if (SvROK(sv))
7352 return sv_2io(SvRV(sv));
f776e3cd 7353 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 7354 if (gv)
7355 io = GvIO(gv);
7356 else
7357 io = 0;
7358 if (!io)
35c1215d 7359 Perl_croak(aTHX_ "Bad filehandle: %"SVf, sv);
46fc3d4c 7360 break;
7361 }
7362 return io;
7363}
7364
645c22ef
DM
7365/*
7366=for apidoc sv_2cv
7367
7368Using various gambits, try to get a CV from an SV; in addition, try if
7369possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
f2c0649b 7370The flags in C<lref> are passed to sv_fetchsv.
645c22ef
DM
7371
7372=cut
7373*/
7374
79072805 7375CV *
864dbfa3 7376Perl_sv_2cv(pTHX_ SV *sv, HV **st, GV **gvp, I32 lref)
79072805 7377{
27da23d5 7378 dVAR;
a0714e2c 7379 GV *gv = NULL;
601f1833 7380 CV *cv = NULL;
79072805 7381
85dec29a
NC
7382 if (!sv) {
7383 *st = NULL;
7384 *gvp = NULL;
7385 return NULL;
7386 }
79072805 7387 switch (SvTYPE(sv)) {
79072805
LW
7388 case SVt_PVCV:
7389 *st = CvSTASH(sv);
a0714e2c 7390 *gvp = NULL;
79072805
LW
7391 return (CV*)sv;
7392 case SVt_PVHV:
7393 case SVt_PVAV:
ef58ba18 7394 *st = NULL;
a0714e2c 7395 *gvp = NULL;
601f1833 7396 return NULL;
8990e307
LW
7397 case SVt_PVGV:
7398 gv = (GV*)sv;
a0d0e21e 7399 *gvp = gv;
8990e307
LW
7400 *st = GvESTASH(gv);
7401 goto fix_gv;
7402
79072805 7403 default:
5b295bef 7404 SvGETMAGIC(sv);
a0d0e21e 7405 if (SvROK(sv)) {
823a54a3 7406 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
f5284f61
IZ
7407 tryAMAGICunDEREF(to_cv);
7408
62f274bf
GS
7409 sv = SvRV(sv);
7410 if (SvTYPE(sv) == SVt_PVCV) {
7411 cv = (CV*)sv;
a0714e2c 7412 *gvp = NULL;
62f274bf
GS
7413 *st = CvSTASH(cv);
7414 return cv;
7415 }
7416 else if(isGV(sv))
7417 gv = (GV*)sv;
7418 else
cea2e8a9 7419 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 7420 }
62f274bf 7421 else if (isGV(sv))
79072805
LW
7422 gv = (GV*)sv;
7423 else
7a5fd60d 7424 gv = gv_fetchsv(sv, lref, SVt_PVCV);
79072805 7425 *gvp = gv;
ef58ba18
NC
7426 if (!gv) {
7427 *st = NULL;
601f1833 7428 return NULL;
ef58ba18 7429 }
e26df76a
NC
7430 /* Some flags to gv_fetchsv mean don't really create the GV */
7431 if (SvTYPE(gv) != SVt_PVGV) {
7432 *st = NULL;
7433 return NULL;
7434 }
79072805 7435 *st = GvESTASH(gv);
8990e307 7436 fix_gv:
8ebc5c01 7437 if (lref && !GvCVu(gv)) {
4633a7c4 7438 SV *tmpsv;
748a9306 7439 ENTER;
561b68a9 7440 tmpsv = newSV(0);
bd61b366 7441 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
7442 /* XXX this is probably not what they think they're getting.
7443 * It has the same effect as "sub name;", i.e. just a forward
7444 * declaration! */
774d564b 7445 newSUB(start_subparse(FALSE, 0),
4633a7c4 7446 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 7447 NULL, NULL);
748a9306 7448 LEAVE;
8ebc5c01 7449 if (!GvCVu(gv))
35c1215d
NC
7450 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
7451 sv);
8990e307 7452 }
8ebc5c01 7453 return GvCVu(gv);
79072805
LW
7454 }
7455}
7456
c461cf8f
JH
7457/*
7458=for apidoc sv_true
7459
7460Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
7461Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
7462instead use an in-line version.
c461cf8f
JH
7463
7464=cut
7465*/
7466
79072805 7467I32
864dbfa3 7468Perl_sv_true(pTHX_ register SV *sv)
79072805 7469{
8990e307
LW
7470 if (!sv)
7471 return 0;
79072805 7472 if (SvPOK(sv)) {
823a54a3
AL
7473 register const XPV* const tXpv = (XPV*)SvANY(sv);
7474 if (tXpv &&
c2f1de04 7475 (tXpv->xpv_cur > 1 ||
339049b0 7476 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
7477 return 1;
7478 else
7479 return 0;
7480 }
7481 else {
7482 if (SvIOK(sv))
463ee0b2 7483 return SvIVX(sv) != 0;
79072805
LW
7484 else {
7485 if (SvNOK(sv))
463ee0b2 7486 return SvNVX(sv) != 0.0;
79072805 7487 else
463ee0b2 7488 return sv_2bool(sv);
79072805
LW
7489 }
7490 }
7491}
79072805 7492
645c22ef 7493/*
c461cf8f
JH
7494=for apidoc sv_pvn_force
7495
7496Get a sensible string out of the SV somehow.
645c22ef
DM
7497A private implementation of the C<SvPV_force> macro for compilers which
7498can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 7499
8d6d96c1
HS
7500=for apidoc sv_pvn_force_flags
7501
7502Get a sensible string out of the SV somehow.
7503If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
7504appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
7505implemented in terms of this function.
645c22ef
DM
7506You normally want to use the various wrapper macros instead: see
7507C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
7508
7509=cut
7510*/
7511
7512char *
7513Perl_sv_pvn_force_flags(pTHX_ SV *sv, STRLEN *lp, I32 flags)
7514{
97aff369 7515 dVAR;
6fc92669 7516 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 7517 sv_force_normal_flags(sv, 0);
1c846c1f 7518
a0d0e21e 7519 if (SvPOK(sv)) {
13c5b33c
NC
7520 if (lp)
7521 *lp = SvCUR(sv);
a0d0e21e
LW
7522 }
7523 else {
a3b680e6 7524 char *s;
13c5b33c
NC
7525 STRLEN len;
7526
4d84ee25 7527 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 7528 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
7529 if (PL_op)
7530 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 7531 ref, OP_NAME(PL_op));
4d84ee25 7532 else
b64e5050 7533 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 7534 }
b64e5050 7535 if (SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
cea2e8a9 7536 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 7537 OP_NAME(PL_op));
b64e5050 7538 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
7539 if (lp)
7540 *lp = len;
7541
3f7c398e 7542 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
7543 if (SvROK(sv))
7544 sv_unref(sv);
862a34c6 7545 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 7546 SvGROW(sv, len + 1);
706aa1c9 7547 Move(s,SvPVX(sv),len,char);
a0d0e21e
LW
7548 SvCUR_set(sv, len);
7549 *SvEND(sv) = '\0';
7550 }
7551 if (!SvPOK(sv)) {
7552 SvPOK_on(sv); /* validate pointer */
7553 SvTAINT(sv);
1d7c1841 7554 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 7555 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
7556 }
7557 }
4d84ee25 7558 return SvPVX_mutable(sv);
a0d0e21e
LW
7559}
7560
645c22ef 7561/*
645c22ef
DM
7562=for apidoc sv_pvbyten_force
7563
0feed65a 7564The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
7565
7566=cut
7567*/
7568
7340a771
GS
7569char *
7570Perl_sv_pvbyten_force(pTHX_ SV *sv, STRLEN *lp)
7571{
46ec2f14 7572 sv_pvn_force(sv,lp);
ffebcc3e 7573 sv_utf8_downgrade(sv,0);
46ec2f14
TS
7574 *lp = SvCUR(sv);
7575 return SvPVX(sv);
7340a771
GS
7576}
7577
645c22ef 7578/*
c461cf8f
JH
7579=for apidoc sv_pvutf8n_force
7580
0feed65a 7581The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
7582
7583=cut
7584*/
7585
7340a771
GS
7586char *
7587Perl_sv_pvutf8n_force(pTHX_ SV *sv, STRLEN *lp)
7588{
46ec2f14 7589 sv_pvn_force(sv,lp);
560a288e 7590 sv_utf8_upgrade(sv);
46ec2f14
TS
7591 *lp = SvCUR(sv);
7592 return SvPVX(sv);
7340a771
GS
7593}
7594
c461cf8f
JH
7595/*
7596=for apidoc sv_reftype
7597
7598Returns a string describing what the SV is a reference to.
7599
7600=cut
7601*/
7602
1cb0ed9b 7603char *
bfed75c6 7604Perl_sv_reftype(pTHX_ const SV *sv, int ob)
a0d0e21e 7605{
07409e01
NC
7606 /* The fact that I don't need to downcast to char * everywhere, only in ?:
7607 inside return suggests a const propagation bug in g++. */
c86bf373 7608 if (ob && SvOBJECT(sv)) {
1b6737cc 7609 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 7610 return name ? name : (char *) "__ANON__";
c86bf373 7611 }
a0d0e21e
LW
7612 else {
7613 switch (SvTYPE(sv)) {
7614 case SVt_NULL:
7615 case SVt_IV:
7616 case SVt_NV:
7617 case SVt_RV:
7618 case SVt_PV:
7619 case SVt_PVIV:
7620 case SVt_PVNV:
7621 case SVt_PVMG:
7622 case SVt_PVBM:
1cb0ed9b 7623 if (SvVOK(sv))
439cb1c4 7624 return "VSTRING";
a0d0e21e
LW
7625 if (SvROK(sv))
7626 return "REF";
7627 else
7628 return "SCALAR";
1cb0ed9b 7629
07409e01 7630 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
7631 /* tied lvalues should appear to be
7632 * scalars for backwards compatitbility */
7633 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 7634 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
7635 case SVt_PVAV: return "ARRAY";
7636 case SVt_PVHV: return "HASH";
7637 case SVt_PVCV: return "CODE";
7638 case SVt_PVGV: return "GLOB";
1d2dff63 7639 case SVt_PVFM: return "FORMAT";
27f9d8f3 7640 case SVt_PVIO: return "IO";
a0d0e21e
LW
7641 default: return "UNKNOWN";
7642 }
7643 }
7644}
7645
954c1994
GS
7646/*
7647=for apidoc sv_isobject
7648
7649Returns a boolean indicating whether the SV is an RV pointing to a blessed
7650object. If the SV is not an RV, or if the object is not blessed, then this
7651will return false.
7652
7653=cut
7654*/
7655
463ee0b2 7656int
864dbfa3 7657Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 7658{
68dc0745 7659 if (!sv)
7660 return 0;
5b295bef 7661 SvGETMAGIC(sv);
85e6fe83
LW
7662 if (!SvROK(sv))
7663 return 0;
7664 sv = (SV*)SvRV(sv);
7665 if (!SvOBJECT(sv))
7666 return 0;
7667 return 1;
7668}
7669
954c1994
GS
7670/*
7671=for apidoc sv_isa
7672
7673Returns a boolean indicating whether the SV is blessed into the specified
7674class. This does not check for subtypes; use C<sv_derived_from> to verify
7675an inheritance relationship.
7676
7677=cut
7678*/
7679
85e6fe83 7680int
864dbfa3 7681Perl_sv_isa(pTHX_ SV *sv, const char *name)
463ee0b2 7682{
bfcb3514 7683 const char *hvname;
68dc0745 7684 if (!sv)
7685 return 0;
5b295bef 7686 SvGETMAGIC(sv);
ed6116ce 7687 if (!SvROK(sv))
463ee0b2 7688 return 0;
ed6116ce
LW
7689 sv = (SV*)SvRV(sv);
7690 if (!SvOBJECT(sv))
463ee0b2 7691 return 0;
bfcb3514
NC
7692 hvname = HvNAME_get(SvSTASH(sv));
7693 if (!hvname)
e27ad1f2 7694 return 0;
463ee0b2 7695
bfcb3514 7696 return strEQ(hvname, name);
463ee0b2
LW
7697}
7698
954c1994
GS
7699/*
7700=for apidoc newSVrv
7701
7702Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
7703it will be upgraded to one. If C<classname> is non-null then the new SV will
7704be blessed in the specified package. The new SV is returned and its
7705reference count is 1.
7706
7707=cut
7708*/
7709
463ee0b2 7710SV*
864dbfa3 7711Perl_newSVrv(pTHX_ SV *rv, const char *classname)
463ee0b2 7712{
97aff369 7713 dVAR;
463ee0b2
LW
7714 SV *sv;
7715
4561caa4 7716 new_SV(sv);
51cf62d8 7717
765f542d 7718 SV_CHECK_THINKFIRST_COW_DROP(rv);
51cf62d8 7719 SvAMAGIC_off(rv);
51cf62d8 7720
0199fce9 7721 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 7722 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
7723 SvREFCNT(rv) = 0;
7724 sv_clear(rv);
7725 SvFLAGS(rv) = 0;
7726 SvREFCNT(rv) = refcnt;
0199fce9 7727
dc5494d2
NC
7728 sv_upgrade(rv, SVt_RV);
7729 } else if (SvROK(rv)) {
7730 SvREFCNT_dec(SvRV(rv));
7731 } else if (SvTYPE(rv) < SVt_RV)
0199fce9
JD
7732 sv_upgrade(rv, SVt_RV);
7733 else if (SvTYPE(rv) > SVt_RV) {
8bd4d4c5 7734 SvPV_free(rv);
0199fce9
JD
7735 SvCUR_set(rv, 0);
7736 SvLEN_set(rv, 0);
7737 }
51cf62d8 7738
0c34ef67 7739 SvOK_off(rv);
b162af07 7740 SvRV_set(rv, sv);
ed6116ce 7741 SvROK_on(rv);
463ee0b2 7742
a0d0e21e 7743 if (classname) {
1b6737cc 7744 HV* const stash = gv_stashpv(classname, TRUE);
a0d0e21e
LW
7745 (void)sv_bless(rv, stash);
7746 }
7747 return sv;
7748}
7749
954c1994
GS
7750/*
7751=for apidoc sv_setref_pv
7752
7753Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
7754argument will be upgraded to an RV. That RV will be modified to point to
7755the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
7756into the SV. The C<classname> argument indicates the package for the
bd61b366 7757blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7758will have a reference count of 1, and the RV will be returned.
954c1994
GS
7759
7760Do not use with other Perl types such as HV, AV, SV, CV, because those
7761objects will become corrupted by the pointer copy process.
7762
7763Note that C<sv_setref_pvn> copies the string while this copies the pointer.
7764
7765=cut
7766*/
7767
a0d0e21e 7768SV*
864dbfa3 7769Perl_sv_setref_pv(pTHX_ SV *rv, const char *classname, void *pv)
a0d0e21e 7770{
97aff369 7771 dVAR;
189b2af5 7772 if (!pv) {
3280af22 7773 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
7774 SvSETMAGIC(rv);
7775 }
a0d0e21e 7776 else
56431972 7777 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
7778 return rv;
7779}
7780
954c1994
GS
7781/*
7782=for apidoc sv_setref_iv
7783
7784Copies an integer into a new SV, optionally blessing the SV. The C<rv>
7785argument will be upgraded to an RV. That RV will be modified to point to
7786the new SV. The C<classname> argument indicates the package for the
bd61b366 7787blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7788will have a reference count of 1, and the RV will be returned.
954c1994
GS
7789
7790=cut
7791*/
7792
a0d0e21e 7793SV*
864dbfa3 7794Perl_sv_setref_iv(pTHX_ SV *rv, const char *classname, IV iv)
a0d0e21e
LW
7795{
7796 sv_setiv(newSVrv(rv,classname), iv);
7797 return rv;
7798}
7799
954c1994 7800/*
e1c57cef
JH
7801=for apidoc sv_setref_uv
7802
7803Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
7804argument will be upgraded to an RV. That RV will be modified to point to
7805the new SV. The C<classname> argument indicates the package for the
bd61b366 7806blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7807will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
7808
7809=cut
7810*/
7811
7812SV*
7813Perl_sv_setref_uv(pTHX_ SV *rv, const char *classname, UV uv)
7814{
7815 sv_setuv(newSVrv(rv,classname), uv);
7816 return rv;
7817}
7818
7819/*
954c1994
GS
7820=for apidoc sv_setref_nv
7821
7822Copies a double into a new SV, optionally blessing the SV. The C<rv>
7823argument will be upgraded to an RV. That RV will be modified to point to
7824the new SV. The C<classname> argument indicates the package for the
bd61b366 7825blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 7826will have a reference count of 1, and the RV will be returned.
954c1994
GS
7827
7828=cut
7829*/
7830
a0d0e21e 7831SV*
65202027 7832Perl_sv_setref_nv(pTHX_ SV *rv, const char *classname, NV nv)
a0d0e21e
LW
7833{
7834 sv_setnv(newSVrv(rv,classname), nv);
7835 return rv;
7836}
463ee0b2 7837
954c1994
GS
7838/*
7839=for apidoc sv_setref_pvn
7840
7841Copies a string into a new SV, optionally blessing the SV. The length of the
7842string must be specified with C<n>. The C<rv> argument will be upgraded to
7843an RV. That RV will be modified to point to the new SV. The C<classname>
7844argument indicates the package for the blessing. Set C<classname> to
bd61b366 7845C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 7846of 1, and the RV will be returned.
954c1994
GS
7847
7848Note that C<sv_setref_pv> copies the pointer while this copies the string.
7849
7850=cut
7851*/
7852
a0d0e21e 7853SV*
1b6737cc 7854Perl_sv_setref_pvn(pTHX_ SV *rv, const char *classname, const char *pv, STRLEN n)
a0d0e21e
LW
7855{
7856 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
7857 return rv;
7858}
7859
954c1994
GS
7860/*
7861=for apidoc sv_bless
7862
7863Blesses an SV into a specified package. The SV must be an RV. The package
7864must be designated by its stash (see C<gv_stashpv()>). The reference count
7865of the SV is unaffected.
7866
7867=cut
7868*/
7869
a0d0e21e 7870SV*
864dbfa3 7871Perl_sv_bless(pTHX_ SV *sv, HV *stash)
a0d0e21e 7872{
97aff369 7873 dVAR;
76e3520e 7874 SV *tmpRef;
a0d0e21e 7875 if (!SvROK(sv))
cea2e8a9 7876 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
7877 tmpRef = SvRV(sv);
7878 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
7879 if (SvREADONLY(tmpRef))
cea2e8a9 7880 Perl_croak(aTHX_ PL_no_modify);
76e3520e
GS
7881 if (SvOBJECT(tmpRef)) {
7882 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7883 --PL_sv_objcount;
76e3520e 7884 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 7885 }
a0d0e21e 7886 }
76e3520e
GS
7887 SvOBJECT_on(tmpRef);
7888 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 7889 ++PL_sv_objcount;
862a34c6 7890 SvUPGRADE(tmpRef, SVt_PVMG);
b37c2d43 7891 SvSTASH_set(tmpRef, (HV*)SvREFCNT_inc_simple(stash));
a0d0e21e 7892
2e3febc6
CS
7893 if (Gv_AMG(stash))
7894 SvAMAGIC_on(sv);
7895 else
7896 SvAMAGIC_off(sv);
a0d0e21e 7897
1edbfb88
AB
7898 if(SvSMAGICAL(tmpRef))
7899 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
7900 mg_set(tmpRef);
7901
7902
ecdeb87c 7903
a0d0e21e
LW
7904 return sv;
7905}
7906
645c22ef 7907/* Downgrades a PVGV to a PVMG.
645c22ef
DM
7908 */
7909
76e3520e 7910STATIC void
cea2e8a9 7911S_sv_unglob(pTHX_ SV *sv)
a0d0e21e 7912{
97aff369 7913 dVAR;
850fabdf 7914 void *xpvmg;
b37c2d43 7915 SV * const temp = sv_newmortal();
850fabdf 7916
a0d0e21e
LW
7917 assert(SvTYPE(sv) == SVt_PVGV);
7918 SvFAKE_off(sv);
180488f8
NC
7919 gv_efullname3(temp, (GV *) sv, "*");
7920
f7877b28 7921 if (GvGP(sv)) {
1edc1566 7922 gp_free((GV*)sv);
f7877b28 7923 }
e826b3c7 7924 if (GvSTASH(sv)) {
e15faf7d 7925 sv_del_backref((SV*)GvSTASH(sv), sv);
5c284bb0 7926 GvSTASH(sv) = NULL;
e826b3c7 7927 }
a5f75d66 7928 GvMULTI_off(sv);
acda4c6a
NC
7929 if (GvNAME_HEK(sv)) {
7930 unshare_hek(GvNAME_HEK(sv));
7931 }
dedf8e73 7932 SvSCREAM_off(sv);
850fabdf
GS
7933
7934 /* need to keep SvANY(sv) in the right arena */
7935 xpvmg = new_XPVMG();
7936 StructCopy(SvANY(sv), xpvmg, XPVMG);
7937 del_XPVGV(SvANY(sv));
7938 SvANY(sv) = xpvmg;
7939
a0d0e21e
LW
7940 SvFLAGS(sv) &= ~SVTYPEMASK;
7941 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
7942
7943 /* Intentionally not calling any local SET magic, as this isn't so much a
7944 set operation as merely an internal storage change. */
7945 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
7946}
7947
954c1994 7948/*
840a7b70 7949=for apidoc sv_unref_flags
954c1994
GS
7950
7951Unsets the RV status of the SV, and decrements the reference count of
7952whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
7953as a reversal of C<newSVrv>. The C<cflags> argument can contain
7954C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
7955(otherwise the decrementing is conditional on the reference count being
7956different from one or the reference being a readonly SV).
7889fe52 7957See C<SvROK_off>.
954c1994
GS
7958
7959=cut
7960*/
7961
ed6116ce 7962void
e15faf7d 7963Perl_sv_unref_flags(pTHX_ SV *ref, U32 flags)
ed6116ce 7964{
b64e5050 7965 SV* const target = SvRV(ref);
810b8aa5 7966
e15faf7d
NC
7967 if (SvWEAKREF(ref)) {
7968 sv_del_backref(target, ref);
7969 SvWEAKREF_off(ref);
7970 SvRV_set(ref, NULL);
810b8aa5
GS
7971 return;
7972 }
e15faf7d
NC
7973 SvRV_set(ref, NULL);
7974 SvROK_off(ref);
7975 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 7976 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
7977 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
7978 SvREFCNT_dec(target);
840a7b70 7979 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 7980 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 7981}
8990e307 7982
840a7b70 7983/*
645c22ef
DM
7984=for apidoc sv_untaint
7985
7986Untaint an SV. Use C<SvTAINTED_off> instead.
7987=cut
7988*/
7989
bbce6d69 7990void
864dbfa3 7991Perl_sv_untaint(pTHX_ SV *sv)
bbce6d69 7992{
13f57bf8 7993 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 7994 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 7995 if (mg)
565764a8 7996 mg->mg_len &= ~1;
36477c24 7997 }
bbce6d69 7998}
7999
645c22ef
DM
8000/*
8001=for apidoc sv_tainted
8002
8003Test an SV for taintedness. Use C<SvTAINTED> instead.
8004=cut
8005*/
8006
bbce6d69 8007bool
864dbfa3 8008Perl_sv_tainted(pTHX_ SV *sv)
bbce6d69 8009{
13f57bf8 8010 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8011 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8012 if (mg && (mg->mg_len & 1) )
36477c24 8013 return TRUE;
8014 }
8015 return FALSE;
bbce6d69 8016}
8017
09540bc3
JH
8018/*
8019=for apidoc sv_setpviv
8020
8021Copies an integer into the given SV, also updating its string value.
8022Does not handle 'set' magic. See C<sv_setpviv_mg>.
8023
8024=cut
8025*/
8026
8027void
8028Perl_sv_setpviv(pTHX_ SV *sv, IV iv)
8029{
8030 char buf[TYPE_CHARS(UV)];
8031 char *ebuf;
b64e5050 8032 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3
JH
8033
8034 sv_setpvn(sv, ptr, ebuf - ptr);
8035}
8036
8037/*
8038=for apidoc sv_setpviv_mg
8039
8040Like C<sv_setpviv>, but also handles 'set' magic.
8041
8042=cut
8043*/
8044
8045void
8046Perl_sv_setpviv_mg(pTHX_ SV *sv, IV iv)
8047{
df7eb254 8048 sv_setpviv(sv, iv);
09540bc3
JH
8049 SvSETMAGIC(sv);
8050}
8051
cea2e8a9 8052#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8053
8054/* pTHX_ magic can't cope with varargs, so this is a no-context
8055 * version of the main function, (which may itself be aliased to us).
8056 * Don't access this version directly.
8057 */
8058
cea2e8a9
GS
8059void
8060Perl_sv_setpvf_nocontext(SV *sv, const char* pat, ...)
8061{
8062 dTHX;
8063 va_list args;
8064 va_start(args, pat);
c5be433b 8065 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8066 va_end(args);
8067}
8068
645c22ef
DM
8069/* pTHX_ magic can't cope with varargs, so this is a no-context
8070 * version of the main function, (which may itself be aliased to us).
8071 * Don't access this version directly.
8072 */
cea2e8a9
GS
8073
8074void
8075Perl_sv_setpvf_mg_nocontext(SV *sv, const char* pat, ...)
8076{
8077 dTHX;
8078 va_list args;
8079 va_start(args, pat);
c5be433b 8080 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8081 va_end(args);
cea2e8a9
GS
8082}
8083#endif
8084
954c1994
GS
8085/*
8086=for apidoc sv_setpvf
8087
bffc3d17
SH
8088Works like C<sv_catpvf> but copies the text into the SV instead of
8089appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8090
8091=cut
8092*/
8093
46fc3d4c 8094void
864dbfa3 8095Perl_sv_setpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8096{
8097 va_list args;
46fc3d4c 8098 va_start(args, pat);
c5be433b 8099 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8100 va_end(args);
8101}
8102
bffc3d17
SH
8103/*
8104=for apidoc sv_vsetpvf
8105
8106Works like C<sv_vcatpvf> but copies the text into the SV instead of
8107appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8108
8109Usually used via its frontend C<sv_setpvf>.
8110
8111=cut
8112*/
645c22ef 8113
c5be433b
GS
8114void
8115Perl_sv_vsetpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8116{
4608196e 8117 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8118}
ef50df4b 8119
954c1994
GS
8120/*
8121=for apidoc sv_setpvf_mg
8122
8123Like C<sv_setpvf>, but also handles 'set' magic.
8124
8125=cut
8126*/
8127
ef50df4b 8128void
864dbfa3 8129Perl_sv_setpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8130{
8131 va_list args;
ef50df4b 8132 va_start(args, pat);
c5be433b 8133 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 8134 va_end(args);
c5be433b
GS
8135}
8136
bffc3d17
SH
8137/*
8138=for apidoc sv_vsetpvf_mg
8139
8140Like C<sv_vsetpvf>, but also handles 'set' magic.
8141
8142Usually used via its frontend C<sv_setpvf_mg>.
8143
8144=cut
8145*/
645c22ef 8146
c5be433b
GS
8147void
8148Perl_sv_vsetpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8149{
4608196e 8150 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8151 SvSETMAGIC(sv);
8152}
8153
cea2e8a9 8154#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8155
8156/* pTHX_ magic can't cope with varargs, so this is a no-context
8157 * version of the main function, (which may itself be aliased to us).
8158 * Don't access this version directly.
8159 */
8160
cea2e8a9
GS
8161void
8162Perl_sv_catpvf_nocontext(SV *sv, const char* pat, ...)
8163{
8164 dTHX;
8165 va_list args;
8166 va_start(args, pat);
c5be433b 8167 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
8168 va_end(args);
8169}
8170
645c22ef
DM
8171/* pTHX_ magic can't cope with varargs, so this is a no-context
8172 * version of the main function, (which may itself be aliased to us).
8173 * Don't access this version directly.
8174 */
8175
cea2e8a9
GS
8176void
8177Perl_sv_catpvf_mg_nocontext(SV *sv, const char* pat, ...)
8178{
8179 dTHX;
8180 va_list args;
8181 va_start(args, pat);
c5be433b 8182 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 8183 va_end(args);
cea2e8a9
GS
8184}
8185#endif
8186
954c1994
GS
8187/*
8188=for apidoc sv_catpvf
8189
d5ce4a7c
GA
8190Processes its arguments like C<sprintf> and appends the formatted
8191output to an SV. If the appended data contains "wide" characters
8192(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
8193and characters >255 formatted with %c), the original SV might get
bffc3d17 8194upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
8195C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
8196valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 8197
d5ce4a7c 8198=cut */
954c1994 8199
46fc3d4c 8200void
864dbfa3 8201Perl_sv_catpvf(pTHX_ SV *sv, const char* pat, ...)
46fc3d4c 8202{
8203 va_list args;
46fc3d4c 8204 va_start(args, pat);
c5be433b 8205 sv_vcatpvf(sv, pat, &args);
46fc3d4c 8206 va_end(args);
8207}
8208
bffc3d17
SH
8209/*
8210=for apidoc sv_vcatpvf
8211
8212Processes its arguments like C<vsprintf> and appends the formatted output
8213to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
8214
8215Usually used via its frontend C<sv_catpvf>.
8216
8217=cut
8218*/
645c22ef 8219
ef50df4b 8220void
c5be433b
GS
8221Perl_sv_vcatpvf(pTHX_ SV *sv, const char* pat, va_list* args)
8222{
4608196e 8223 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
8224}
8225
954c1994
GS
8226/*
8227=for apidoc sv_catpvf_mg
8228
8229Like C<sv_catpvf>, but also handles 'set' magic.
8230
8231=cut
8232*/
8233
c5be433b 8234void
864dbfa3 8235Perl_sv_catpvf_mg(pTHX_ SV *sv, const char* pat, ...)
ef50df4b
GS
8236{
8237 va_list args;
ef50df4b 8238 va_start(args, pat);
c5be433b 8239 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 8240 va_end(args);
c5be433b
GS
8241}
8242
bffc3d17
SH
8243/*
8244=for apidoc sv_vcatpvf_mg
8245
8246Like C<sv_vcatpvf>, but also handles 'set' magic.
8247
8248Usually used via its frontend C<sv_catpvf_mg>.
8249
8250=cut
8251*/
645c22ef 8252
c5be433b
GS
8253void
8254Perl_sv_vcatpvf_mg(pTHX_ SV *sv, const char* pat, va_list* args)
8255{
4608196e 8256 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
8257 SvSETMAGIC(sv);
8258}
8259
954c1994
GS
8260/*
8261=for apidoc sv_vsetpvfn
8262
bffc3d17 8263Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
8264appending it.
8265
bffc3d17 8266Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 8267
954c1994
GS
8268=cut
8269*/
8270
46fc3d4c 8271void
7d5ea4e7 8272Perl_sv_vsetpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8273{
8274 sv_setpvn(sv, "", 0);
7d5ea4e7 8275 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 8276}
8277
2d00ba3b 8278STATIC I32
9dd79c3f 8279S_expect_number(pTHX_ char** pattern)
211dfcf1 8280{
97aff369 8281 dVAR;
211dfcf1
HS
8282 I32 var = 0;
8283 switch (**pattern) {
8284 case '1': case '2': case '3':
8285 case '4': case '5': case '6':
8286 case '7': case '8': case '9':
2fba7546
GA
8287 var = *(*pattern)++ - '0';
8288 while (isDIGIT(**pattern)) {
5f66b61c 8289 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
8290 if (tmp < var)
8291 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
8292 var = tmp;
8293 }
211dfcf1
HS
8294 }
8295 return var;
8296}
211dfcf1 8297
c445ea15
AL
8298STATIC char *
8299S_F0convert(NV nv, char *endbuf, STRLEN *len)
4151a5fe 8300{
a3b680e6 8301 const int neg = nv < 0;
4151a5fe 8302 UV uv;
4151a5fe
IZ
8303
8304 if (neg)
8305 nv = -nv;
8306 if (nv < UV_MAX) {
b464bac0 8307 char *p = endbuf;
4151a5fe 8308 nv += 0.5;
028f8eaa 8309 uv = (UV)nv;
4151a5fe
IZ
8310 if (uv & 1 && uv == nv)
8311 uv--; /* Round to even */
8312 do {
a3b680e6 8313 const unsigned dig = uv % 10;
4151a5fe
IZ
8314 *--p = '0' + dig;
8315 } while (uv /= 10);
8316 if (neg)
8317 *--p = '-';
8318 *len = endbuf - p;
8319 return p;
8320 }
bd61b366 8321 return NULL;
4151a5fe
IZ
8322}
8323
8324
954c1994
GS
8325/*
8326=for apidoc sv_vcatpvfn
8327
8328Processes its arguments like C<vsprintf> and appends the formatted output
8329to an SV. Uses an array of SVs if the C style variable argument list is
8330missing (NULL). When running with taint checks enabled, indicates via
8331C<maybe_tainted> if results are untrustworthy (often due to the use of
8332locales).
8333
bffc3d17 8334Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 8335
954c1994
GS
8336=cut
8337*/
8338
8896765a
RB
8339
8340#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
8341 vecstr = (U8*)SvPV_const(vecsv,veclen);\
8342 vec_utf8 = DO_UTF8(vecsv);
8343
1ef29b0e
RGS
8344/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
8345
46fc3d4c 8346void
7d5ea4e7 8347Perl_sv_vcatpvfn(pTHX_ SV *sv, const char *pat, STRLEN patlen, va_list *args, SV **svargs, I32 svmax, bool *maybe_tainted)
46fc3d4c 8348{
97aff369 8349 dVAR;
46fc3d4c 8350 char *p;
8351 char *q;
a3b680e6 8352 const char *patend;
fc36a67e 8353 STRLEN origlen;
46fc3d4c 8354 I32 svix = 0;
27da23d5 8355 static const char nullstr[] = "(null)";
a0714e2c 8356 SV *argsv = NULL;
b464bac0
AL
8357 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
8358 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 8359 SV *nsv = NULL;
4151a5fe
IZ
8360 /* Times 4: a decimal digit takes more than 3 binary digits.
8361 * NV_DIG: mantissa takes than many decimal digits.
8362 * Plus 32: Playing safe. */
8363 char ebuf[IV_DIG * 4 + NV_DIG + 32];
8364 /* large enough for "%#.#f" --chip */
8365 /* what about long double NVs? --jhi */
db79b45b 8366
53c1dcc0
AL
8367 PERL_UNUSED_ARG(maybe_tainted);
8368
46fc3d4c 8369 /* no matter what, this is a string now */
fc36a67e 8370 (void)SvPV_force(sv, origlen);
46fc3d4c 8371
8896765a 8372 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 8373 if (patlen == 0)
8374 return;
0dbb1585 8375 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
8376 if (args) {
8377 const char * const s = va_arg(*args, char*);
8378 sv_catpv(sv, s ? s : nullstr);
8379 }
8380 else if (svix < svmax) {
8381 sv_catsv(sv, *svargs);
2d03de9c
AL
8382 }
8383 return;
0dbb1585 8384 }
8896765a
RB
8385 if (args && patlen == 3 && pat[0] == '%' &&
8386 pat[1] == '-' && pat[2] == 'p') {
8387 argsv = va_arg(*args, SV*);
8388 sv_catsv(sv, argsv);
8896765a 8389 return;
46fc3d4c 8390 }
8391
1d917b39 8392#ifndef USE_LONG_DOUBLE
4151a5fe 8393 /* special-case "%.<number>[gf]" */
7af36d83 8394 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
8395 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
8396 unsigned digits = 0;
8397 const char *pp;
8398
8399 pp = pat + 2;
8400 while (*pp >= '0' && *pp <= '9')
8401 digits = 10 * digits + (*pp++ - '0');
028f8eaa 8402 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
8403 NV nv;
8404
7af36d83 8405 if (svix < svmax)
4151a5fe
IZ
8406 nv = SvNV(*svargs);
8407 else
8408 return;
8409 if (*pp == 'g') {
2873255c
NC
8410 /* Add check for digits != 0 because it seems that some
8411 gconverts are buggy in this case, and we don't yet have
8412 a Configure test for this. */
8413 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
8414 /* 0, point, slack */
2e59c212 8415 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
8416 sv_catpv(sv, ebuf);
8417 if (*ebuf) /* May return an empty string for digits==0 */
8418 return;
8419 }
8420 } else if (!digits) {
8421 STRLEN l;
8422
8423 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
8424 sv_catpvn(sv, p, l);
8425 return;
8426 }
8427 }
8428 }
8429 }
1d917b39 8430#endif /* !USE_LONG_DOUBLE */
4151a5fe 8431
2cf2cfc6 8432 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 8433 has_utf8 = TRUE;
2cf2cfc6 8434
46fc3d4c 8435 patend = (char*)pat + patlen;
8436 for (p = (char*)pat; p < patend; p = q) {
8437 bool alt = FALSE;
8438 bool left = FALSE;
b22c7a20 8439 bool vectorize = FALSE;
211dfcf1 8440 bool vectorarg = FALSE;
2cf2cfc6 8441 bool vec_utf8 = FALSE;
46fc3d4c 8442 char fill = ' ';
8443 char plus = 0;
8444 char intsize = 0;
8445 STRLEN width = 0;
fc36a67e 8446 STRLEN zeros = 0;
46fc3d4c 8447 bool has_precis = FALSE;
8448 STRLEN precis = 0;
c445ea15 8449 const I32 osvix = svix;
2cf2cfc6 8450 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
8451#ifdef HAS_LDBL_SPRINTF_BUG
8452 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 8453 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
8454 bool fix_ldbl_sprintf_bug = FALSE;
8455#endif
205f51d8 8456
46fc3d4c 8457 char esignbuf[4];
89ebb4a3 8458 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 8459 STRLEN esignlen = 0;
8460
bd61b366 8461 const char *eptr = NULL;
fc36a67e 8462 STRLEN elen = 0;
a0714e2c 8463 SV *vecsv = NULL;
4608196e 8464 const U8 *vecstr = NULL;
b22c7a20 8465 STRLEN veclen = 0;
934abaf1 8466 char c = 0;
46fc3d4c 8467 int i;
9c5ffd7c 8468 unsigned base = 0;
8c8eb53c
RB
8469 IV iv = 0;
8470 UV uv = 0;
9e5b023a
JH
8471 /* we need a long double target in case HAS_LONG_DOUBLE but
8472 not USE_LONG_DOUBLE
8473 */
35fff930 8474#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
8475 long double nv;
8476#else
65202027 8477 NV nv;
9e5b023a 8478#endif
46fc3d4c 8479 STRLEN have;
8480 STRLEN need;
8481 STRLEN gap;
7af36d83 8482 const char *dotstr = ".";
b22c7a20 8483 STRLEN dotstrlen = 1;
211dfcf1 8484 I32 efix = 0; /* explicit format parameter index */
eb3fce90 8485 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
8486 I32 epix = 0; /* explicit precision index */
8487 I32 evix = 0; /* explicit vector index */
eb3fce90 8488 bool asterisk = FALSE;
46fc3d4c 8489
211dfcf1 8490 /* echo everything up to the next format specification */
46fc3d4c 8491 for (q = p; q < patend && *q != '%'; ++q) ;
8492 if (q > p) {
db79b45b
JH
8493 if (has_utf8 && !pat_utf8)
8494 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
8495 else
8496 sv_catpvn(sv, p, q - p);
46fc3d4c 8497 p = q;
8498 }
8499 if (q++ >= patend)
8500 break;
8501
211dfcf1
HS
8502/*
8503 We allow format specification elements in this order:
8504 \d+\$ explicit format parameter index
8505 [-+ 0#]+ flags
a472f209 8506 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 8507 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
8508 \d+|\*(\d+\$)? width using optional (optionally specified) arg
8509 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
8510 [hlqLV] size
8896765a
RB
8511 [%bcdefginopsuxDFOUX] format (mandatory)
8512*/
8513
8514 if (args) {
8515/*
8516 As of perl5.9.3, printf format checking is on by default.
8517 Internally, perl uses %p formats to provide an escape to
8518 some extended formatting. This block deals with those
8519 extensions: if it does not match, (char*)q is reset and
8520 the normal format processing code is used.
8521
8522 Currently defined extensions are:
8523 %p include pointer address (standard)
8524 %-p (SVf) include an SV (previously %_)
8525 %-<num>p include an SV with precision <num>
8526 %1p (VDf) include a v-string (as %vd)
8527 %<num>p reserved for future extensions
8528
8529 Robin Barker 2005-07-14
211dfcf1 8530*/
8896765a
RB
8531 char* r = q;
8532 bool sv = FALSE;
8533 STRLEN n = 0;
8534 if (*q == '-')
8535 sv = *q++;
c445ea15 8536 n = expect_number(&q);
8896765a
RB
8537 if (*q++ == 'p') {
8538 if (sv) { /* SVf */
8539 if (n) {
8540 precis = n;
8541 has_precis = TRUE;
8542 }
8543 argsv = va_arg(*args, SV*);
8544 eptr = SvPVx_const(argsv, elen);
8545 if (DO_UTF8(argsv))
8546 is_utf8 = TRUE;
8547 goto string;
8548 }
8549#if vdNUMBER
8550 else if (n == vdNUMBER) { /* VDf */
8551 vectorize = TRUE;
8552 VECTORIZE_ARGS
8553 goto format_vd;
8554 }
8555#endif
8556 else if (n) {
8557 if (ckWARN_d(WARN_INTERNAL))
8558 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8559 "internal %%<num>p might conflict with future printf extensions");
8560 }
8561 }
8562 q = r;
8563 }
8564
c445ea15 8565 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
8566 if (*q == '$') {
8567 ++q;
8568 efix = width;
8569 } else {
8570 goto gotwidth;
8571 }
8572 }
8573
fc36a67e 8574 /* FLAGS */
8575
46fc3d4c 8576 while (*q) {
8577 switch (*q) {
8578 case ' ':
8579 case '+':
8580 plus = *q++;
8581 continue;
8582
8583 case '-':
8584 left = TRUE;
8585 q++;
8586 continue;
8587
8588 case '0':
8589 fill = *q++;
8590 continue;
8591
8592 case '#':
8593 alt = TRUE;
8594 q++;
8595 continue;
8596
fc36a67e 8597 default:
8598 break;
8599 }
8600 break;
8601 }
46fc3d4c 8602
211dfcf1 8603 tryasterisk:
eb3fce90 8604 if (*q == '*') {
211dfcf1 8605 q++;
c445ea15 8606 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
8607 if (*q++ != '$')
8608 goto unknown;
eb3fce90 8609 asterisk = TRUE;
211dfcf1
HS
8610 }
8611 if (*q == 'v') {
eb3fce90 8612 q++;
211dfcf1
HS
8613 if (vectorize)
8614 goto unknown;
9cbac4c7 8615 if ((vectorarg = asterisk)) {
211dfcf1
HS
8616 evix = ewix;
8617 ewix = 0;
8618 asterisk = FALSE;
8619 }
8620 vectorize = TRUE;
8621 goto tryasterisk;
eb3fce90
JH
8622 }
8623
211dfcf1 8624 if (!asterisk)
858a90f9 8625 {
7a5fa8a2 8626 if( *q == '0' )
f3583277 8627 fill = *q++;
c445ea15 8628 width = expect_number(&q);
858a90f9 8629 }
211dfcf1
HS
8630
8631 if (vectorize) {
8632 if (vectorarg) {
8633 if (args)
8634 vecsv = va_arg(*args, SV*);
7ad96abb
NC
8635 else if (evix) {
8636 vecsv = (evix > 0 && evix <= svmax)
8637 ? svargs[evix-1] : &PL_sv_undef;
8638 } else {
8639 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
8640 }
245d4a47 8641 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
8642 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
8643 bad with tied or overloaded values that return UTF8. */
211dfcf1 8644 if (DO_UTF8(vecsv))
2cf2cfc6 8645 is_utf8 = TRUE;
640283f5
NC
8646 else if (has_utf8) {
8647 vecsv = sv_mortalcopy(vecsv);
8648 sv_utf8_upgrade(vecsv);
8649 dotstr = SvPV_const(vecsv, dotstrlen);
8650 is_utf8 = TRUE;
8651 }
211dfcf1
HS
8652 }
8653 if (args) {
8896765a 8654 VECTORIZE_ARGS
eb3fce90 8655 }
7ad96abb 8656 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 8657 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 8658 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 8659 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
8660
8661 /* if this is a version object, we need to convert
8662 * back into v-string notation and then let the
8663 * vectorize happen normally
d7aa5382 8664 */
96b8f7ce
JP
8665 if (sv_derived_from(vecsv, "version")) {
8666 char *version = savesvpv(vecsv);
34ba6322
SP
8667 if ( hv_exists((HV*)SvRV(vecsv), "alpha", 5 ) ) {
8668 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
8669 "vector argument not supported with alpha versions");
8670 goto unknown;
8671 }
96b8f7ce
JP
8672 vecsv = sv_newmortal();
8673 /* scan_vstring is expected to be called during
8674 * tokenization, so we need to fake up the end
8675 * of the buffer for it
8676 */
8677 PL_bufend = version + veclen;
8678 scan_vstring(version, vecsv);
8679 vecstr = (U8*)SvPV_const(vecsv, veclen);
8680 vec_utf8 = DO_UTF8(vecsv);
8681 Safefree(version);
d7aa5382 8682 }
211dfcf1
HS
8683 }
8684 else {
8685 vecstr = (U8*)"";
8686 veclen = 0;
8687 }
eb3fce90 8688 }
fc36a67e 8689
eb3fce90 8690 if (asterisk) {
fc36a67e 8691 if (args)
8692 i = va_arg(*args, int);
8693 else
eb3fce90
JH
8694 i = (ewix ? ewix <= svmax : svix < svmax) ?
8695 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8696 left |= (i < 0);
8697 width = (i < 0) ? -i : i;
fc36a67e 8698 }
211dfcf1 8699 gotwidth:
fc36a67e 8700
8701 /* PRECISION */
46fc3d4c 8702
fc36a67e 8703 if (*q == '.') {
8704 q++;
8705 if (*q == '*') {
211dfcf1 8706 q++;
c445ea15 8707 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
8708 goto unknown;
8709 /* XXX: todo, support specified precision parameter */
8710 if (epix)
211dfcf1 8711 goto unknown;
46fc3d4c 8712 if (args)
8713 i = va_arg(*args, int);
8714 else
eb3fce90
JH
8715 i = (ewix ? ewix <= svmax : svix < svmax)
8716 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 8717 precis = (i < 0) ? 0 : i;
fc36a67e 8718 }
8719 else {
8720 precis = 0;
8721 while (isDIGIT(*q))
8722 precis = precis * 10 + (*q++ - '0');
8723 }
8724 has_precis = TRUE;
8725 }
46fc3d4c 8726
fc36a67e 8727 /* SIZE */
46fc3d4c 8728
fc36a67e 8729 switch (*q) {
c623ac67
GS
8730#ifdef WIN32
8731 case 'I': /* Ix, I32x, and I64x */
8732# ifdef WIN64
8733 if (q[1] == '6' && q[2] == '4') {
8734 q += 3;
8735 intsize = 'q';
8736 break;
8737 }
8738# endif
8739 if (q[1] == '3' && q[2] == '2') {
8740 q += 3;
8741 break;
8742 }
8743# ifdef WIN64
8744 intsize = 'q';
8745# endif
8746 q++;
8747 break;
8748#endif
9e5b023a 8749#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 8750 case 'L': /* Ld */
5f66b61c 8751 /*FALLTHROUGH*/
e5c81feb 8752#ifdef HAS_QUAD
6f9bb7fd 8753 case 'q': /* qd */
9e5b023a 8754#endif
6f9bb7fd
GS
8755 intsize = 'q';
8756 q++;
8757 break;
8758#endif
fc36a67e 8759 case 'l':
9e5b023a 8760#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 8761 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 8762 intsize = 'q';
8763 q += 2;
46fc3d4c 8764 break;
cf2093f6 8765 }
fc36a67e 8766#endif
5f66b61c 8767 /*FALLTHROUGH*/
fc36a67e 8768 case 'h':
5f66b61c 8769 /*FALLTHROUGH*/
fc36a67e 8770 case 'V':
8771 intsize = *q++;
46fc3d4c 8772 break;
8773 }
8774
fc36a67e 8775 /* CONVERSION */
8776
211dfcf1
HS
8777 if (*q == '%') {
8778 eptr = q++;
8779 elen = 1;
26372e71
GA
8780 if (vectorize) {
8781 c = '%';
8782 goto unknown;
8783 }
211dfcf1
HS
8784 goto string;
8785 }
8786
26372e71 8787 if (!vectorize && !args) {
86c51f8b
NC
8788 if (efix) {
8789 const I32 i = efix-1;
8790 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
8791 } else {
8792 argsv = (svix >= 0 && svix < svmax)
8793 ? svargs[svix++] : &PL_sv_undef;
8794 }
863811b2 8795 }
211dfcf1 8796
46fc3d4c 8797 switch (c = *q++) {
8798
8799 /* STRINGS */
8800
46fc3d4c 8801 case 'c':
26372e71
GA
8802 if (vectorize)
8803 goto unknown;
8804 uv = (args) ? va_arg(*args, int) : SvIVx(argsv);
1bd104fb
JH
8805 if ((uv > 255 ||
8806 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 8807 && !IN_BYTES) {
dfe13c55 8808 eptr = (char*)utf8buf;
9041c2e3 8809 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 8810 is_utf8 = TRUE;
7e2040f0
GS
8811 }
8812 else {
8813 c = (char)uv;
8814 eptr = &c;
8815 elen = 1;
a0ed51b3 8816 }
46fc3d4c 8817 goto string;
8818
46fc3d4c 8819 case 's':
26372e71
GA
8820 if (vectorize)
8821 goto unknown;
8822 if (args) {
fc36a67e 8823 eptr = va_arg(*args, char*);
c635e13b 8824 if (eptr)
1d7c1841
GS
8825#ifdef MACOS_TRADITIONAL
8826 /* On MacOS, %#s format is used for Pascal strings */
8827 if (alt)
8828 elen = *eptr++;
8829 else
8830#endif
c635e13b 8831 elen = strlen(eptr);
8832 else {
27da23d5 8833 eptr = (char *)nullstr;
c635e13b 8834 elen = sizeof nullstr - 1;
8835 }
46fc3d4c 8836 }
211dfcf1 8837 else {
4d84ee25 8838 eptr = SvPVx_const(argsv, elen);
7e2040f0 8839 if (DO_UTF8(argsv)) {
a0ed51b3
LW
8840 if (has_precis && precis < elen) {
8841 I32 p = precis;
7e2040f0 8842 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
8843 precis = p;
8844 }
8845 if (width) { /* fudge width (can't fudge elen) */
7e2040f0 8846 width += elen - sv_len_utf8(argsv);
a0ed51b3 8847 }
2cf2cfc6 8848 is_utf8 = TRUE;
a0ed51b3
LW
8849 }
8850 }
fc36a67e 8851
46fc3d4c 8852 string:
8853 if (has_precis && elen > precis)
8854 elen = precis;
8855 break;
8856
8857 /* INTEGERS */
8858
fc36a67e 8859 case 'p':
be75b157 8860 if (alt || vectorize)
c2e66d9e 8861 goto unknown;
211dfcf1 8862 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 8863 base = 16;
8864 goto integer;
8865
46fc3d4c 8866 case 'D':
29fe7a80 8867#ifdef IV_IS_QUAD
22f3ae8c 8868 intsize = 'q';
29fe7a80 8869#else
46fc3d4c 8870 intsize = 'l';
29fe7a80 8871#endif
5f66b61c 8872 /*FALLTHROUGH*/
46fc3d4c 8873 case 'd':
8874 case 'i':
8896765a
RB
8875#if vdNUMBER
8876 format_vd:
8877#endif
b22c7a20 8878 if (vectorize) {
ba210ebe 8879 STRLEN ulen;
211dfcf1
HS
8880 if (!veclen)
8881 continue;
2cf2cfc6
A
8882 if (vec_utf8)
8883 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8884 UTF8_ALLOW_ANYUV);
b22c7a20 8885 else {
e83d50c9 8886 uv = *vecstr;
b22c7a20
GS
8887 ulen = 1;
8888 }
8889 vecstr += ulen;
8890 veclen -= ulen;
e83d50c9
JP
8891 if (plus)
8892 esignbuf[esignlen++] = plus;
b22c7a20
GS
8893 }
8894 else if (args) {
46fc3d4c 8895 switch (intsize) {
8896 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 8897 case 'l': iv = va_arg(*args, long); break;
fc36a67e 8898 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 8899 default: iv = va_arg(*args, int); break;
cf2093f6
JH
8900#ifdef HAS_QUAD
8901 case 'q': iv = va_arg(*args, Quad_t); break;
8902#endif
46fc3d4c 8903 }
8904 }
8905 else {
b10c0dba 8906 IV tiv = SvIVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8907 switch (intsize) {
b10c0dba
MHM
8908 case 'h': iv = (short)tiv; break;
8909 case 'l': iv = (long)tiv; break;
8910 case 'V':
8911 default: iv = tiv; break;
cf2093f6 8912#ifdef HAS_QUAD
b10c0dba 8913 case 'q': iv = (Quad_t)tiv; break;
cf2093f6 8914#endif
46fc3d4c 8915 }
8916 }
e83d50c9
JP
8917 if ( !vectorize ) /* we already set uv above */
8918 {
8919 if (iv >= 0) {
8920 uv = iv;
8921 if (plus)
8922 esignbuf[esignlen++] = plus;
8923 }
8924 else {
8925 uv = -iv;
8926 esignbuf[esignlen++] = '-';
8927 }
46fc3d4c 8928 }
8929 base = 10;
8930 goto integer;
8931
fc36a67e 8932 case 'U':
29fe7a80 8933#ifdef IV_IS_QUAD
22f3ae8c 8934 intsize = 'q';
29fe7a80 8935#else
fc36a67e 8936 intsize = 'l';
29fe7a80 8937#endif
5f66b61c 8938 /*FALLTHROUGH*/
fc36a67e 8939 case 'u':
8940 base = 10;
8941 goto uns_integer;
8942
4f19785b
WSI
8943 case 'b':
8944 base = 2;
8945 goto uns_integer;
8946
46fc3d4c 8947 case 'O':
29fe7a80 8948#ifdef IV_IS_QUAD
22f3ae8c 8949 intsize = 'q';
29fe7a80 8950#else
46fc3d4c 8951 intsize = 'l';
29fe7a80 8952#endif
5f66b61c 8953 /*FALLTHROUGH*/
46fc3d4c 8954 case 'o':
8955 base = 8;
8956 goto uns_integer;
8957
8958 case 'X':
46fc3d4c 8959 case 'x':
8960 base = 16;
46fc3d4c 8961
8962 uns_integer:
b22c7a20 8963 if (vectorize) {
ba210ebe 8964 STRLEN ulen;
b22c7a20 8965 vector:
211dfcf1
HS
8966 if (!veclen)
8967 continue;
2cf2cfc6
A
8968 if (vec_utf8)
8969 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
8970 UTF8_ALLOW_ANYUV);
b22c7a20 8971 else {
a05b299f 8972 uv = *vecstr;
b22c7a20
GS
8973 ulen = 1;
8974 }
8975 vecstr += ulen;
8976 veclen -= ulen;
8977 }
8978 else if (args) {
46fc3d4c 8979 switch (intsize) {
8980 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 8981 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 8982 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 8983 default: uv = va_arg(*args, unsigned); break;
cf2093f6 8984#ifdef HAS_QUAD
9e3321a5 8985 case 'q': uv = va_arg(*args, Uquad_t); break;
cf2093f6 8986#endif
46fc3d4c 8987 }
8988 }
8989 else {
b10c0dba 8990 UV tuv = SvUVx(argsv); /* work around GCC bug #13488 */
46fc3d4c 8991 switch (intsize) {
b10c0dba
MHM
8992 case 'h': uv = (unsigned short)tuv; break;
8993 case 'l': uv = (unsigned long)tuv; break;
8994 case 'V':
8995 default: uv = tuv; break;
cf2093f6 8996#ifdef HAS_QUAD
b10c0dba 8997 case 'q': uv = (Uquad_t)tuv; break;
cf2093f6 8998#endif
46fc3d4c 8999 }
9000 }
9001
9002 integer:
4d84ee25
NC
9003 {
9004 char *ptr = ebuf + sizeof ebuf;
9005 switch (base) {
9006 unsigned dig;
9007 case 16:
9008 if (!uv)
9009 alt = FALSE;
9010 p = (char*)((c == 'X')
9011 ? "0123456789ABCDEF" : "0123456789abcdef");
9012 do {
9013 dig = uv & 15;
9014 *--ptr = p[dig];
9015 } while (uv >>= 4);
9016 if (alt) {
9017 esignbuf[esignlen++] = '0';
9018 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9019 }
9020 break;
9021 case 8:
9022 do {
9023 dig = uv & 7;
9024 *--ptr = '0' + dig;
9025 } while (uv >>= 3);
9026 if (alt && *ptr != '0')
9027 *--ptr = '0';
9028 break;
9029 case 2:
ed2b91d2
GA
9030 if (!uv)
9031 alt = FALSE;
4d84ee25
NC
9032 do {
9033 dig = uv & 1;
9034 *--ptr = '0' + dig;
9035 } while (uv >>= 1);
9036 if (alt) {
9037 esignbuf[esignlen++] = '0';
9038 esignbuf[esignlen++] = 'b';
9039 }
9040 break;
9041 default: /* it had better be ten or less */
9042 do {
9043 dig = uv % base;
9044 *--ptr = '0' + dig;
9045 } while (uv /= base);
9046 break;
46fc3d4c 9047 }
4d84ee25
NC
9048 elen = (ebuf + sizeof ebuf) - ptr;
9049 eptr = ptr;
9050 if (has_precis) {
9051 if (precis > elen)
9052 zeros = precis - elen;
9053 else if (precis == 0 && elen == 1 && *eptr == '0')
9054 elen = 0;
eda88b6d 9055 }
c10ed8b9 9056 }
46fc3d4c 9057 break;
9058
9059 /* FLOATING POINT */
9060
fc36a67e 9061 case 'F':
9062 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9063 /*FALLTHROUGH*/
46fc3d4c 9064 case 'e': case 'E':
fc36a67e 9065 case 'f':
46fc3d4c 9066 case 'g': case 'G':
26372e71
GA
9067 if (vectorize)
9068 goto unknown;
46fc3d4c 9069
9070 /* This is evil, but floating point is even more evil */
9071
9e5b023a
JH
9072 /* for SV-style calling, we can only get NV
9073 for C-style calling, we assume %f is double;
9074 for simplicity we allow any of %Lf, %llf, %qf for long double
9075 */
9076 switch (intsize) {
9077 case 'V':
9078#if defined(USE_LONG_DOUBLE)
9079 intsize = 'q';
9080#endif
9081 break;
8a2e3f14 9082/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 9083 case 'l':
5f66b61c 9084 /*FALLTHROUGH*/
9e5b023a
JH
9085 default:
9086#if defined(USE_LONG_DOUBLE)
9087 intsize = args ? 0 : 'q';
9088#endif
9089 break;
9090 case 'q':
9091#if defined(HAS_LONG_DOUBLE)
9092 break;
9093#else
5f66b61c 9094 /*FALLTHROUGH*/
9e5b023a
JH
9095#endif
9096 case 'h':
9e5b023a
JH
9097 goto unknown;
9098 }
9099
9100 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 9101 nv = (args) ?
35fff930
JH
9102#if LONG_DOUBLESIZE > DOUBLESIZE
9103 intsize == 'q' ?
205f51d8
AS
9104 va_arg(*args, long double) :
9105 va_arg(*args, double)
35fff930 9106#else
205f51d8 9107 va_arg(*args, double)
35fff930 9108#endif
9e5b023a 9109 : SvNVx(argsv);
fc36a67e 9110
9111 need = 0;
9112 if (c != 'e' && c != 'E') {
9113 i = PERL_INT_MIN;
9e5b023a
JH
9114 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9115 will cast our (long double) to (double) */
73b309ea 9116 (void)Perl_frexp(nv, &i);
fc36a67e 9117 if (i == PERL_INT_MIN)
cea2e8a9 9118 Perl_die(aTHX_ "panic: frexp");
c635e13b 9119 if (i > 0)
fc36a67e 9120 need = BIT_DIGITS(i);
9121 }
9122 need += has_precis ? precis : 6; /* known default */
20f6aaab 9123
fc36a67e 9124 if (need < width)
9125 need = width;
9126
20f6aaab
AS
9127#ifdef HAS_LDBL_SPRINTF_BUG
9128 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
9129 with sfio - Allen <allens@cpan.org> */
9130
9131# ifdef DBL_MAX
9132# define MY_DBL_MAX DBL_MAX
9133# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9134# if DOUBLESIZE >= 8
9135# define MY_DBL_MAX 1.7976931348623157E+308L
9136# else
9137# define MY_DBL_MAX 3.40282347E+38L
9138# endif
9139# endif
9140
9141# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
9142# define MY_DBL_MAX_BUG 1L
20f6aaab 9143# else
205f51d8 9144# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 9145# endif
20f6aaab 9146
205f51d8
AS
9147# ifdef DBL_MIN
9148# define MY_DBL_MIN DBL_MIN
9149# else /* XXX guessing! -Allen */
9150# if DOUBLESIZE >= 8
9151# define MY_DBL_MIN 2.2250738585072014E-308L
9152# else
9153# define MY_DBL_MIN 1.17549435E-38L
9154# endif
9155# endif
20f6aaab 9156
205f51d8
AS
9157 if ((intsize == 'q') && (c == 'f') &&
9158 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
9159 (need < DBL_DIG)) {
9160 /* it's going to be short enough that
9161 * long double precision is not needed */
9162
9163 if ((nv <= 0L) && (nv >= -0L))
9164 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
9165 else {
9166 /* would use Perl_fp_class as a double-check but not
9167 * functional on IRIX - see perl.h comments */
9168
9169 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
9170 /* It's within the range that a double can represent */
9171#if defined(DBL_MAX) && !defined(DBL_MIN)
9172 if ((nv >= ((long double)1/DBL_MAX)) ||
9173 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 9174#endif
205f51d8 9175 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 9176 }
205f51d8
AS
9177 }
9178 if (fix_ldbl_sprintf_bug == TRUE) {
9179 double temp;
9180
9181 intsize = 0;
9182 temp = (double)nv;
9183 nv = (NV)temp;
9184 }
20f6aaab 9185 }
205f51d8
AS
9186
9187# undef MY_DBL_MAX
9188# undef MY_DBL_MAX_BUG
9189# undef MY_DBL_MIN
9190
20f6aaab
AS
9191#endif /* HAS_LDBL_SPRINTF_BUG */
9192
46fc3d4c 9193 need += 20; /* fudge factor */
80252599
GS
9194 if (PL_efloatsize < need) {
9195 Safefree(PL_efloatbuf);
9196 PL_efloatsize = need + 20; /* more fudge */
a02a5408 9197 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 9198 PL_efloatbuf[0] = '\0';
46fc3d4c 9199 }
9200
4151a5fe
IZ
9201 if ( !(width || left || plus || alt) && fill != '0'
9202 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
9203 /* See earlier comment about buggy Gconvert when digits,
9204 aka precis is 0 */
9205 if ( c == 'g' && precis) {
2e59c212 9206 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
9207 /* May return an empty string for digits==0 */
9208 if (*PL_efloatbuf) {
9209 elen = strlen(PL_efloatbuf);
4151a5fe 9210 goto float_converted;
4150c189 9211 }
4151a5fe
IZ
9212 } else if ( c == 'f' && !precis) {
9213 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
9214 break;
9215 }
9216 }
4d84ee25
NC
9217 {
9218 char *ptr = ebuf + sizeof ebuf;
9219 *--ptr = '\0';
9220 *--ptr = c;
9221 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 9222#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
9223 if (intsize == 'q') {
9224 /* Copy the one or more characters in a long double
9225 * format before the 'base' ([efgEFG]) character to
9226 * the format string. */
9227 static char const prifldbl[] = PERL_PRIfldbl;
9228 char const *p = prifldbl + sizeof(prifldbl) - 3;
9229 while (p >= prifldbl) { *--ptr = *p--; }
9230 }
65202027 9231#endif
4d84ee25
NC
9232 if (has_precis) {
9233 base = precis;
9234 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9235 *--ptr = '.';
9236 }
9237 if (width) {
9238 base = width;
9239 do { *--ptr = '0' + (base % 10); } while (base /= 10);
9240 }
9241 if (fill == '0')
9242 *--ptr = fill;
9243 if (left)
9244 *--ptr = '-';
9245 if (plus)
9246 *--ptr = plus;
9247 if (alt)
9248 *--ptr = '#';
9249 *--ptr = '%';
9250
9251 /* No taint. Otherwise we are in the strange situation
9252 * where printf() taints but print($float) doesn't.
9253 * --jhi */
9e5b023a 9254#if defined(HAS_LONG_DOUBLE)
4150c189
NC
9255 elen = ((intsize == 'q')
9256 ? my_sprintf(PL_efloatbuf, ptr, nv)
9257 : my_sprintf(PL_efloatbuf, ptr, (double)nv));
9e5b023a 9258#else
4150c189 9259 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 9260#endif
4d84ee25 9261 }
4151a5fe 9262 float_converted:
80252599 9263 eptr = PL_efloatbuf;
46fc3d4c 9264 break;
9265
fc36a67e 9266 /* SPECIAL */
9267
9268 case 'n':
26372e71
GA
9269 if (vectorize)
9270 goto unknown;
fc36a67e 9271 i = SvCUR(sv) - origlen;
26372e71 9272 if (args) {
c635e13b 9273 switch (intsize) {
9274 case 'h': *(va_arg(*args, short*)) = i; break;
9275 default: *(va_arg(*args, int*)) = i; break;
9276 case 'l': *(va_arg(*args, long*)) = i; break;
9277 case 'V': *(va_arg(*args, IV*)) = i; break;
cf2093f6
JH
9278#ifdef HAS_QUAD
9279 case 'q': *(va_arg(*args, Quad_t*)) = i; break;
9280#endif
c635e13b 9281 }
fc36a67e 9282 }
9dd79c3f 9283 else
211dfcf1 9284 sv_setuv_mg(argsv, (UV)i);
fc36a67e 9285 continue; /* not "break" */
9286
9287 /* UNKNOWN */
9288
46fc3d4c 9289 default:
fc36a67e 9290 unknown:
041457d9
DM
9291 if (!args
9292 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
9293 && ckWARN(WARN_PRINTF))
9294 {
c4420975 9295 SV * const msg = sv_newmortal();
35c1215d
NC
9296 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
9297 (PL_op->op_type == OP_PRTF) ? "" : "s");
0f4b6630 9298 if (c) {
0f4b6630 9299 if (isPRINT(c))
1c846c1f 9300 Perl_sv_catpvf(aTHX_ msg,
0f4b6630
JH
9301 "\"%%%c\"", c & 0xFF);
9302 else
9303 Perl_sv_catpvf(aTHX_ msg,
57def98f 9304 "\"%%\\%03"UVof"\"",
0f4b6630 9305 (UV)c & 0xFF);
0f4b6630 9306 } else
396482e1 9307 sv_catpvs(msg, "end of string");
9014280d 9308 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, msg); /* yes, this is reentrant */
c635e13b 9309 }
fb73857a 9310
9311 /* output mangled stuff ... */
9312 if (c == '\0')
9313 --q;
46fc3d4c 9314 eptr = p;
9315 elen = q - p;
fb73857a 9316
9317 /* ... right here, because formatting flags should not apply */
9318 SvGROW(sv, SvCUR(sv) + elen + 1);
9319 p = SvEND(sv);
4459522c 9320 Copy(eptr, p, elen, char);
fb73857a 9321 p += elen;
9322 *p = '\0';
3f7c398e 9323 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 9324 svix = osvix;
fb73857a 9325 continue; /* not "break" */
46fc3d4c 9326 }
9327
6c94ec8b
HS
9328 /* calculate width before utf8_upgrade changes it */
9329 have = esignlen + zeros + elen;
ed2b91d2
GA
9330 if (have < zeros)
9331 Perl_croak_nocontext(PL_memory_wrap);
6c94ec8b 9332
d2876be5
JH
9333 if (is_utf8 != has_utf8) {
9334 if (is_utf8) {
9335 if (SvCUR(sv))
9336 sv_utf8_upgrade(sv);
9337 }
9338 else {
53c1dcc0 9339 SV * const nsv = sv_2mortal(newSVpvn(eptr, elen));
d2876be5 9340 sv_utf8_upgrade(nsv);
93524f2b 9341 eptr = SvPVX_const(nsv);
d2876be5
JH
9342 elen = SvCUR(nsv);
9343 }
9344 SvGROW(sv, SvCUR(sv) + elen + 1);
9345 p = SvEND(sv);
9346 *p = '\0';
9347 }
6af65485 9348
46fc3d4c 9349 need = (have > width ? have : width);
9350 gap = need - have;
9351
d2641cbd
PC
9352 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
9353 Perl_croak_nocontext(PL_memory_wrap);
b22c7a20 9354 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 9355 p = SvEND(sv);
9356 if (esignlen && fill == '0') {
53c1dcc0 9357 int i;
eb160463 9358 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9359 *p++ = esignbuf[i];
9360 }
9361 if (gap && !left) {
9362 memset(p, fill, gap);
9363 p += gap;
9364 }
9365 if (esignlen && fill != '0') {
53c1dcc0 9366 int i;
eb160463 9367 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 9368 *p++ = esignbuf[i];
9369 }
fc36a67e 9370 if (zeros) {
53c1dcc0 9371 int i;
fc36a67e 9372 for (i = zeros; i; i--)
9373 *p++ = '0';
9374 }
46fc3d4c 9375 if (elen) {
4459522c 9376 Copy(eptr, p, elen, char);
46fc3d4c 9377 p += elen;
9378 }
9379 if (gap && left) {
9380 memset(p, ' ', gap);
9381 p += gap;
9382 }
b22c7a20
GS
9383 if (vectorize) {
9384 if (veclen) {
4459522c 9385 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
9386 p += dotstrlen;
9387 }
9388 else
9389 vectorize = FALSE; /* done iterating over vecstr */
9390 }
2cf2cfc6
A
9391 if (is_utf8)
9392 has_utf8 = TRUE;
9393 if (has_utf8)
7e2040f0 9394 SvUTF8_on(sv);
46fc3d4c 9395 *p = '\0';
3f7c398e 9396 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
9397 if (vectorize) {
9398 esignlen = 0;
9399 goto vector;
9400 }
46fc3d4c 9401 }
9402}
51371543 9403
645c22ef
DM
9404/* =========================================================================
9405
9406=head1 Cloning an interpreter
9407
9408All the macros and functions in this section are for the private use of
9409the main function, perl_clone().
9410
9411The foo_dup() functions make an exact copy of an existing foo thinngy.
9412During the course of a cloning, a hash table is used to map old addresses
9413to new addresses. The table is created and manipulated with the
9414ptr_table_* functions.
9415
9416=cut
9417
9418============================================================================*/
9419
9420
1d7c1841
GS
9421#if defined(USE_ITHREADS)
9422
d4c19fe8 9423/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
9424#ifndef GpREFCNT_inc
9425# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
9426#endif
9427
9428
a41cc44e
NC
9429/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
9430 that currently av_dup and hv_dup are the same as sv_dup. If this changes,
9431 please unmerge ss_dup. */
d2d73c3e 9432#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 9433#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
d2d73c3e
AB
9434#define av_dup(s,t) (AV*)sv_dup((SV*)s,t)
9435#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9436#define hv_dup(s,t) (HV*)sv_dup((SV*)s,t)
9437#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9438#define cv_dup(s,t) (CV*)sv_dup((SV*)s,t)
9439#define cv_dup_inc(s,t) (CV*)SvREFCNT_inc(sv_dup((SV*)s,t))
9440#define io_dup(s,t) (IO*)sv_dup((SV*)s,t)
9441#define io_dup_inc(s,t) (IO*)SvREFCNT_inc(sv_dup((SV*)s,t))
9442#define gv_dup(s,t) (GV*)sv_dup((SV*)s,t)
9443#define gv_dup_inc(s,t) (GV*)SvREFCNT_inc(sv_dup((SV*)s,t))
6136c704
AL
9444#define SAVEPV(p) ((p) ? savepv(p) : NULL)
9445#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 9446
d2d73c3e 9447
d2f185dc
AMS
9448/* Duplicate a regexp. Required reading: pregcomp() and pregfree() in
9449 regcomp.c. AMS 20010712 */
645c22ef 9450
1d7c1841 9451REGEXP *
53c1dcc0 9452Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param)
1d7c1841 9453{
27da23d5 9454 dVAR;
d2f185dc
AMS
9455 REGEXP *ret;
9456 int i, len, npar;
9457 struct reg_substr_datum *s;
9458
9459 if (!r)
9460 return (REGEXP *)NULL;
9461
9462 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
9463 return ret;
9464
9465 len = r->offsets[0];
9466 npar = r->nparens+1;
9467
a02a5408 9468 Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
d2f185dc
AMS
9469 Copy(r->program, ret->program, len+1, regnode);
9470
a02a5408 9471 Newx(ret->startp, npar, I32);
d2f185dc 9472 Copy(r->startp, ret->startp, npar, I32);
a02a5408 9473 Newx(ret->endp, npar, I32);
d2f185dc
AMS
9474 Copy(r->startp, ret->startp, npar, I32);
9475
a02a5408 9476 Newx(ret->substrs, 1, struct reg_substr_data);
d2f185dc
AMS
9477 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
9478 s->min_offset = r->substrs->data[i].min_offset;
9479 s->max_offset = r->substrs->data[i].max_offset;
9480 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
33b8afdf 9481 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
d2f185dc
AMS
9482 }
9483
70612e96 9484 ret->regstclass = NULL;
d2f185dc
AMS
9485 if (r->data) {
9486 struct reg_data *d;
e1ec3a88 9487 const int count = r->data->count;
53c1dcc0 9488 int i;
d2f185dc 9489
a02a5408 9490 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d2f185dc 9491 char, struct reg_data);
a02a5408 9492 Newx(d->what, count, U8);
d2f185dc
AMS
9493
9494 d->count = count;
9495 for (i = 0; i < count; i++) {
9496 d->what[i] = r->data->what[i];
9497 switch (d->what[i]) {
a3621e74
YO
9498 /* legal options are one of: sfpont
9499 see also regcomp.h and pregfree() */
d2f185dc
AMS
9500 case 's':
9501 d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
9502 break;
9503 case 'p':
9504 d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
9505 break;
9506 case 'f':
9507 /* This is cheating. */
a02a5408 9508 Newx(d->data[i], 1, struct regnode_charclass_class);
d2f185dc
AMS
9509 StructCopy(r->data->data[i], d->data[i],
9510 struct regnode_charclass_class);
70612e96 9511 ret->regstclass = (regnode*)d->data[i];
d2f185dc
AMS
9512 break;
9513 case 'o':
33773810
AMS
9514 /* Compiled op trees are readonly, and can thus be
9515 shared without duplication. */
b34c0dd4 9516 OP_REFCNT_LOCK;
9b978d73 9517 d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
b34c0dd4 9518 OP_REFCNT_UNLOCK;
9b978d73 9519 break;
d2f185dc
AMS
9520 case 'n':
9521 d->data[i] = r->data->data[i];
9522 break;
a3621e74
YO
9523 case 't':
9524 d->data[i] = r->data->data[i];
9525 OP_REFCNT_LOCK;
9526 ((reg_trie_data*)d->data[i])->refcount++;
9527 OP_REFCNT_UNLOCK;
9528 break;
9529 default:
9530 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
d2f185dc
AMS
9531 }
9532 }
9533
9534 ret->data = d;
9535 }
9536 else
9537 ret->data = NULL;
9538
a02a5408 9539 Newx(ret->offsets, 2*len+1, U32);
d2f185dc
AMS
9540 Copy(r->offsets, ret->offsets, 2*len+1, U32);
9541
e01c5899 9542 ret->precomp = SAVEPVN(r->precomp, r->prelen);
d2f185dc
AMS
9543 ret->refcnt = r->refcnt;
9544 ret->minlen = r->minlen;
9545 ret->prelen = r->prelen;
9546 ret->nparens = r->nparens;
9547 ret->lastparen = r->lastparen;
9548 ret->lastcloseparen = r->lastcloseparen;
9549 ret->reganch = r->reganch;
9550
70612e96
RG
9551 ret->sublen = r->sublen;
9552
9553 if (RX_MATCH_COPIED(ret))
e01c5899 9554 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
70612e96 9555 else
bd61b366 9556 ret->subbeg = NULL;
f8c7b90f 9557#ifdef PERL_OLD_COPY_ON_WRITE
a0714e2c 9558 ret->saved_copy = NULL;
9a26048b 9559#endif
70612e96 9560
d2f185dc
AMS
9561 ptr_table_store(PL_ptr_table, r, ret);
9562 return ret;
1d7c1841
GS
9563}
9564
d2d73c3e 9565/* duplicate a file handle */
645c22ef 9566
1d7c1841 9567PerlIO *
a8fc9800 9568Perl_fp_dup(pTHX_ PerlIO *fp, char type, CLONE_PARAMS *param)
1d7c1841
GS
9569{
9570 PerlIO *ret;
53c1dcc0
AL
9571
9572 PERL_UNUSED_ARG(type);
73d840c0 9573
1d7c1841
GS
9574 if (!fp)
9575 return (PerlIO*)NULL;
9576
9577 /* look for it in the table first */
9578 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
9579 if (ret)
9580 return ret;
9581
9582 /* create anew and remember what it is */
ecdeb87c 9583 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
9584 ptr_table_store(PL_ptr_table, fp, ret);
9585 return ret;
9586}
9587
645c22ef
DM
9588/* duplicate a directory handle */
9589
1d7c1841
GS
9590DIR *
9591Perl_dirp_dup(pTHX_ DIR *dp)
9592{
96a5add6 9593 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9594 if (!dp)
9595 return (DIR*)NULL;
9596 /* XXX TODO */
9597 return dp;
9598}
9599
ff276b08 9600/* duplicate a typeglob */
645c22ef 9601
1d7c1841 9602GP *
a8fc9800 9603Perl_gp_dup(pTHX_ GP *gp, CLONE_PARAMS* param)
1d7c1841
GS
9604{
9605 GP *ret;
b37c2d43 9606
1d7c1841
GS
9607 if (!gp)
9608 return (GP*)NULL;
9609 /* look for it in the table first */
9610 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
9611 if (ret)
9612 return ret;
9613
9614 /* create anew and remember what it is */
a02a5408 9615 Newxz(ret, 1, GP);
1d7c1841
GS
9616 ptr_table_store(PL_ptr_table, gp, ret);
9617
9618 /* clone */
9619 ret->gp_refcnt = 0; /* must be before any other dups! */
d2d73c3e
AB
9620 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
9621 ret->gp_io = io_dup_inc(gp->gp_io, param);
9622 ret->gp_form = cv_dup_inc(gp->gp_form, param);
9623 ret->gp_av = av_dup_inc(gp->gp_av, param);
9624 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
9625 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
9626 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 9627 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841
GS
9628 ret->gp_line = gp->gp_line;
9629 ret->gp_file = gp->gp_file; /* points to COP.cop_file */
9630 return ret;
9631}
9632
645c22ef
DM
9633/* duplicate a chain of magic */
9634
1d7c1841 9635MAGIC *
a8fc9800 9636Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS* param)
1d7c1841 9637{
cb359b41
JH
9638 MAGIC *mgprev = (MAGIC*)NULL;
9639 MAGIC *mgret;
1d7c1841
GS
9640 if (!mg)
9641 return (MAGIC*)NULL;
9642 /* look for it in the table first */
9643 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
9644 if (mgret)
9645 return mgret;
9646
9647 for (; mg; mg = mg->mg_moremagic) {
9648 MAGIC *nmg;
a02a5408 9649 Newxz(nmg, 1, MAGIC);
cb359b41 9650 if (mgprev)
1d7c1841 9651 mgprev->mg_moremagic = nmg;
cb359b41
JH
9652 else
9653 mgret = nmg;
1d7c1841
GS
9654 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
9655 nmg->mg_private = mg->mg_private;
9656 nmg->mg_type = mg->mg_type;
9657 nmg->mg_flags = mg->mg_flags;
14befaf4 9658 if (mg->mg_type == PERL_MAGIC_qr) {
d2f185dc 9659 nmg->mg_obj = (SV*)re_dup((REGEXP*)mg->mg_obj, param);
1d7c1841 9660 }
05bd4103 9661 else if(mg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
9662 /* The backref AV has its reference count deliberately bumped by
9663 1. */
9664 nmg->mg_obj = SvREFCNT_inc(av_dup_inc((AV*) mg->mg_obj, param));
05bd4103 9665 }
8d2f4536
NC
9666 else if (mg->mg_type == PERL_MAGIC_symtab) {
9667 nmg->mg_obj = mg->mg_obj;
9668 }
1d7c1841
GS
9669 else {
9670 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
d2d73c3e
AB
9671 ? sv_dup_inc(mg->mg_obj, param)
9672 : sv_dup(mg->mg_obj, param);
1d7c1841
GS
9673 }
9674 nmg->mg_len = mg->mg_len;
9675 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
14befaf4 9676 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
68795e93 9677 if (mg->mg_len > 0) {
1d7c1841 9678 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
14befaf4
DM
9679 if (mg->mg_type == PERL_MAGIC_overload_table &&
9680 AMT_AMAGIC((AMT*)mg->mg_ptr))
9681 {
c445ea15 9682 const AMT * const amtp = (AMT*)mg->mg_ptr;
0bcc34c2 9683 AMT * const namtp = (AMT*)nmg->mg_ptr;
1d7c1841
GS
9684 I32 i;
9685 for (i = 1; i < NofAMmeth; i++) {
d2d73c3e 9686 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
1d7c1841
GS
9687 }
9688 }
9689 }
9690 else if (mg->mg_len == HEf_SVKEY)
d2d73c3e 9691 nmg->mg_ptr = (char*)sv_dup_inc((SV*)mg->mg_ptr, param);
1d7c1841 9692 }
68795e93
NIS
9693 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
9694 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
9695 }
1d7c1841
GS
9696 mgprev = nmg;
9697 }
9698 return mgret;
9699}
9700
645c22ef
DM
9701/* create a new pointer-mapping table */
9702
1d7c1841
GS
9703PTR_TBL_t *
9704Perl_ptr_table_new(pTHX)
9705{
9706 PTR_TBL_t *tbl;
96a5add6
AL
9707 PERL_UNUSED_CONTEXT;
9708
a02a5408 9709 Newxz(tbl, 1, PTR_TBL_t);
1d7c1841
GS
9710 tbl->tbl_max = 511;
9711 tbl->tbl_items = 0;
a02a5408 9712 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
9713 return tbl;
9714}
9715
7119fd33
NC
9716#define PTR_TABLE_HASH(ptr) \
9717 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 9718
93e68bfb
JC
9719/*
9720 we use the PTE_SVSLOT 'reservation' made above, both here (in the
9721 following define) and at call to new_body_inline made below in
9722 Perl_ptr_table_store()
9723 */
9724
9725#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 9726
645c22ef
DM
9727/* map an existing pointer using a table */
9728
7bf61b54 9729STATIC PTR_TBL_ENT_t *
b0e6ae5b 9730S_ptr_table_find(PTR_TBL_t *tbl, const void *sv) {
1d7c1841 9731 PTR_TBL_ENT_t *tblent;
4373e329 9732 const UV hash = PTR_TABLE_HASH(sv);
1d7c1841
GS
9733 assert(tbl);
9734 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
9735 for (; tblent; tblent = tblent->next) {
9736 if (tblent->oldval == sv)
7bf61b54 9737 return tblent;
1d7c1841 9738 }
d4c19fe8 9739 return NULL;
7bf61b54
NC
9740}
9741
9742void *
9743Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *tbl, const void *sv)
9744{
b0e6ae5b 9745 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
96a5add6 9746 PERL_UNUSED_CONTEXT;
d4c19fe8 9747 return tblent ? tblent->newval : NULL;
1d7c1841
GS
9748}
9749
645c22ef
DM
9750/* add a new entry to a pointer-mapping table */
9751
1d7c1841 9752void
44f8325f 9753Perl_ptr_table_store(pTHX_ PTR_TBL_t *tbl, const void *oldsv, void *newsv)
1d7c1841 9754{
0c9fdfe0 9755 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
96a5add6 9756 PERL_UNUSED_CONTEXT;
1d7c1841 9757
7bf61b54
NC
9758 if (tblent) {
9759 tblent->newval = newsv;
9760 } else {
9761 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
9762
d2a0f284
JC
9763 new_body_inline(tblent, PTE_SVSLOT);
9764
7bf61b54
NC
9765 tblent->oldval = oldsv;
9766 tblent->newval = newsv;
9767 tblent->next = tbl->tbl_ary[entry];
9768 tbl->tbl_ary[entry] = tblent;
9769 tbl->tbl_items++;
9770 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
9771 ptr_table_split(tbl);
1d7c1841 9772 }
1d7c1841
GS
9773}
9774
645c22ef
DM
9775/* double the hash bucket size of an existing ptr table */
9776
1d7c1841
GS
9777void
9778Perl_ptr_table_split(pTHX_ PTR_TBL_t *tbl)
9779{
9780 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 9781 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
9782 UV newsize = oldsize * 2;
9783 UV i;
96a5add6 9784 PERL_UNUSED_CONTEXT;
1d7c1841
GS
9785
9786 Renew(ary, newsize, PTR_TBL_ENT_t*);
9787 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
9788 tbl->tbl_max = --newsize;
9789 tbl->tbl_ary = ary;
9790 for (i=0; i < oldsize; i++, ary++) {
9791 PTR_TBL_ENT_t **curentp, **entp, *ent;
9792 if (!*ary)
9793 continue;
9794 curentp = ary + oldsize;
9795 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 9796 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
9797 *entp = ent->next;
9798 ent->next = *curentp;
9799 *curentp = ent;
9800 continue;
9801 }
9802 else
9803 entp = &ent->next;
9804 }
9805 }
9806}
9807
645c22ef
DM
9808/* remove all the entries from a ptr table */
9809
a0739874
DM
9810void
9811Perl_ptr_table_clear(pTHX_ PTR_TBL_t *tbl)
9812{
d5cefff9 9813 if (tbl && tbl->tbl_items) {
c445ea15 9814 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 9815 UV riter = tbl->tbl_max;
a0739874 9816
d5cefff9
NC
9817 do {
9818 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 9819
d5cefff9 9820 while (entry) {
00b6aa41 9821 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
9822 entry = entry->next;
9823 del_pte(oentry);
9824 }
9825 } while (riter--);
a0739874 9826
d5cefff9
NC
9827 tbl->tbl_items = 0;
9828 }
a0739874
DM
9829}
9830
645c22ef
DM
9831/* clear and free a ptr table */
9832
a0739874
DM
9833void
9834Perl_ptr_table_free(pTHX_ PTR_TBL_t *tbl)
9835{
9836 if (!tbl) {
9837 return;
9838 }
9839 ptr_table_clear(tbl);
9840 Safefree(tbl->tbl_ary);
9841 Safefree(tbl);
9842}
9843
5bd07a3d 9844
83841fad 9845void
eb86f8b3 9846Perl_rvpv_dup(pTHX_ SV *dstr, const SV *sstr, CLONE_PARAMS* param)
83841fad
NIS
9847{
9848 if (SvROK(sstr)) {
b162af07
SP
9849 SvRV_set(dstr, SvWEAKREF(sstr)
9850 ? sv_dup(SvRV(sstr), param)
9851 : sv_dup_inc(SvRV(sstr), param));
f880fe2f 9852
83841fad 9853 }
3f7c398e 9854 else if (SvPVX_const(sstr)) {
83841fad
NIS
9855 /* Has something there */
9856 if (SvLEN(sstr)) {
68795e93 9857 /* Normal PV - clone whole allocated space */
3f7c398e 9858 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
9859 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
9860 /* Not that normal - actually sstr is copy on write.
9861 But we are a true, independant SV, so: */
9862 SvREADONLY_off(dstr);
9863 SvFAKE_off(dstr);
9864 }
68795e93 9865 }
83841fad
NIS
9866 else {
9867 /* Special case - not normally malloced for some reason */
f7877b28
NC
9868 if (isGV_with_GP(sstr)) {
9869 /* Don't need to do anything here. */
9870 }
9871 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
9872 /* A "shared" PV - clone it as "shared" PV */
9873 SvPV_set(dstr,
9874 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
9875 param)));
83841fad
NIS
9876 }
9877 else {
9878 /* Some other special case - random pointer */
f880fe2f 9879 SvPV_set(dstr, SvPVX(sstr));
d3d0e6f1 9880 }
83841fad
NIS
9881 }
9882 }
9883 else {
4608196e 9884 /* Copy the NULL */
f880fe2f 9885 if (SvTYPE(dstr) == SVt_RV)
b162af07 9886 SvRV_set(dstr, NULL);
f880fe2f 9887 else
6136c704 9888 SvPV_set(dstr, NULL);
83841fad
NIS
9889 }
9890}
9891
662fb8b2
NC
9892/* duplicate an SV of any type (including AV, HV etc) */
9893
1d7c1841 9894SV *
eb86f8b3 9895Perl_sv_dup(pTHX_ const SV *sstr, CLONE_PARAMS* param)
1d7c1841 9896{
27da23d5 9897 dVAR;
1d7c1841
GS
9898 SV *dstr;
9899
9900 if (!sstr || SvTYPE(sstr) == SVTYPEMASK)
6136c704 9901 return NULL;
1d7c1841
GS
9902 /* look for it in the table first */
9903 dstr = (SV*)ptr_table_fetch(PL_ptr_table, sstr);
9904 if (dstr)
9905 return dstr;
9906
0405e91e
AB
9907 if(param->flags & CLONEf_JOIN_IN) {
9908 /** We are joining here so we don't want do clone
9909 something that is bad **/
eb86f8b3
AL
9910 if (SvTYPE(sstr) == SVt_PVHV) {
9911 const char * const hvname = HvNAME_get(sstr);
9912 if (hvname)
9913 /** don't clone stashes if they already exist **/
9914 return (SV*)gv_stashpv(hvname,0);
0405e91e
AB
9915 }
9916 }
9917
1d7c1841
GS
9918 /* create anew and remember what it is */
9919 new_SV(dstr);
fd0854ff
DM
9920
9921#ifdef DEBUG_LEAKING_SCALARS
9922 dstr->sv_debug_optype = sstr->sv_debug_optype;
9923 dstr->sv_debug_line = sstr->sv_debug_line;
9924 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
9925 dstr->sv_debug_cloned = 1;
fd0854ff 9926 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
9927#endif
9928
1d7c1841
GS
9929 ptr_table_store(PL_ptr_table, sstr, dstr);
9930
9931 /* clone */
9932 SvFLAGS(dstr) = SvFLAGS(sstr);
9933 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
9934 SvREFCNT(dstr) = 0; /* must be before any other dups! */
9935
9936#ifdef DEBUGGING
3f7c398e 9937 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 9938 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
3f7c398e 9939 PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
9940#endif
9941
9660f481
DM
9942 /* don't clone objects whose class has asked us not to */
9943 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
9944 SvFLAGS(dstr) &= ~SVTYPEMASK;
9945 SvOBJECT_off(dstr);
9946 return dstr;
9947 }
9948
1d7c1841
GS
9949 switch (SvTYPE(sstr)) {
9950 case SVt_NULL:
9951 SvANY(dstr) = NULL;
9952 break;
9953 case SVt_IV:
339049b0 9954 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
45977657 9955 SvIV_set(dstr, SvIVX(sstr));
1d7c1841
GS
9956 break;
9957 case SVt_NV:
9958 SvANY(dstr) = new_XNV();
9d6ce603 9959 SvNV_set(dstr, SvNVX(sstr));
1d7c1841
GS
9960 break;
9961 case SVt_RV:
339049b0 9962 SvANY(dstr) = &(dstr->sv_u.svu_rv);
83841fad 9963 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
1d7c1841 9964 break;
662fb8b2
NC
9965 default:
9966 {
9967 /* These are all the types that need complex bodies allocating. */
662fb8b2 9968 void *new_body;
2bcc16b3
NC
9969 const svtype sv_type = SvTYPE(sstr);
9970 const struct body_details *const sv_type_details
9971 = bodies_by_type + sv_type;
662fb8b2 9972
93e68bfb 9973 switch (sv_type) {
662fb8b2 9974 default:
bb263b4e 9975 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
9976 break;
9977
662fb8b2
NC
9978 case SVt_PVGV:
9979 if (GvUNIQUE((GV*)sstr)) {
bb263b4e 9980 /*EMPTY*/; /* Do sharing here, and fall through */
662fb8b2 9981 }
c22188b4
NC
9982 case SVt_PVIO:
9983 case SVt_PVFM:
9984 case SVt_PVHV:
9985 case SVt_PVAV:
93e68bfb 9986 case SVt_PVBM:
662fb8b2 9987 case SVt_PVCV:
662fb8b2 9988 case SVt_PVLV:
662fb8b2 9989 case SVt_PVMG:
662fb8b2 9990 case SVt_PVNV:
662fb8b2 9991 case SVt_PVIV:
662fb8b2 9992 case SVt_PV:
d2a0f284 9993 assert(sv_type_details->body_size);
c22188b4 9994 if (sv_type_details->arena) {
d2a0f284 9995 new_body_inline(new_body, sv_type);
c22188b4 9996 new_body
b9502f15 9997 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
9998 } else {
9999 new_body = new_NOARENA(sv_type_details);
10000 }
1d7c1841 10001 }
662fb8b2
NC
10002 assert(new_body);
10003 SvANY(dstr) = new_body;
10004
2bcc16b3 10005#ifndef PURIFY
b9502f15
NC
10006 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10007 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 10008 sv_type_details->copy, char);
2bcc16b3
NC
10009#else
10010 Copy(((char*)SvANY(sstr)),
10011 ((char*)SvANY(dstr)),
d2a0f284 10012 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10013#endif
662fb8b2 10014
f7877b28
NC
10015 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10016 && !isGV_with_GP(dstr))
662fb8b2
NC
10017 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10018
10019 /* The Copy above means that all the source (unduplicated) pointers
10020 are now in the destination. We can check the flags and the
10021 pointers in either, but it's possible that there's less cache
10022 missing by always going for the destination.
10023 FIXME - instrument and check that assumption */
f32993d6 10024 if (sv_type >= SVt_PVMG) {
e736a858 10025 HV *ourstash;
2892acdb 10026 if ((sv_type == SVt_PVMG) && (ourstash = OURSTASH(dstr))) {
e736a858
NC
10027 OURSTASH_set(dstr, hv_dup_inc(ourstash, param));
10028 } else if (SvMAGIC(dstr))
662fb8b2
NC
10029 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10030 if (SvSTASH(dstr))
10031 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 10032 }
662fb8b2 10033
f32993d6
NC
10034 /* The cast silences a GCC warning about unhandled types. */
10035 switch ((int)sv_type) {
662fb8b2
NC
10036 case SVt_PV:
10037 break;
10038 case SVt_PVIV:
10039 break;
10040 case SVt_PVNV:
10041 break;
10042 case SVt_PVMG:
10043 break;
10044 case SVt_PVBM:
10045 break;
10046 case SVt_PVLV:
10047 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10048 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10049 LvTARG(dstr) = dstr;
10050 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10051 LvTARG(dstr) = (SV*)he_dup((HE*)LvTARG(dstr), 0, param);
10052 else
10053 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10054 break;
10055 case SVt_PVGV:
acda4c6a
NC
10056 if (GvNAME_HEK(dstr))
10057 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
f5c1e807 10058
e15faf7d
NC
10059 /* Don't call sv_add_backref here as it's going to be created
10060 as part of the magic cloning of the symbol table. */
f7877b28
NC
10061 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10062 if(isGV_with_GP(sstr)) {
10063 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10064 at the point of this comment. */
10065 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10066 (void)GpREFCNT_inc(GvGP(dstr));
10067 } else
10068 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
10069 break;
10070 case SVt_PVIO:
10071 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10072 if (IoOFP(dstr) == IoIFP(sstr))
10073 IoOFP(dstr) = IoIFP(dstr);
10074 else
10075 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10076 /* PL_rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
10077 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10078 /* I have no idea why fake dirp (rsfps)
10079 should be treated differently but otherwise
10080 we end up with leaks -- sky*/
10081 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10082 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10083 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10084 } else {
10085 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10086 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10087 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
10088 if (IoDIRP(dstr)) {
10089 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10090 } else {
bb263b4e 10091 /*EMPTY*/;
100ce7e1
NC
10092 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10093 }
662fb8b2
NC
10094 }
10095 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10096 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10097 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
10098 break;
10099 case SVt_PVAV:
10100 if (AvARRAY((AV*)sstr)) {
10101 SV **dst_ary, **src_ary;
10102 SSize_t items = AvFILLp((AV*)sstr) + 1;
10103
10104 src_ary = AvARRAY((AV*)sstr);
a02a5408 10105 Newxz(dst_ary, AvMAX((AV*)sstr)+1, SV*);
662fb8b2
NC
10106 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
10107 SvPV_set(dstr, (char*)dst_ary);
10108 AvALLOC((AV*)dstr) = dst_ary;
10109 if (AvREAL((AV*)sstr)) {
10110 while (items-- > 0)
10111 *dst_ary++ = sv_dup_inc(*src_ary++, param);
10112 }
10113 else {
10114 while (items-- > 0)
10115 *dst_ary++ = sv_dup(*src_ary++, param);
10116 }
10117 items = AvMAX((AV*)sstr) - AvFILLp((AV*)sstr);
10118 while (items-- > 0) {
10119 *dst_ary++ = &PL_sv_undef;
10120 }
bfcb3514 10121 }
662fb8b2 10122 else {
bd61b366 10123 SvPV_set(dstr, NULL);
662fb8b2 10124 AvALLOC((AV*)dstr) = (SV**)NULL;
b79f7545 10125 }
662fb8b2
NC
10126 break;
10127 case SVt_PVHV:
10128 {
cbbf8932 10129 HEK *hvname = NULL;
662fb8b2
NC
10130
10131 if (HvARRAY((HV*)sstr)) {
10132 STRLEN i = 0;
10133 const bool sharekeys = !!HvSHAREKEYS(sstr);
10134 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
10135 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
10136 char *darray;
a02a5408 10137 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
662fb8b2
NC
10138 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
10139 char);
10140 HvARRAY(dstr) = (HE**)darray;
10141 while (i <= sxhv->xhv_max) {
5c4138a0 10142 const HE *source = HvARRAY(sstr)[i];
662fb8b2
NC
10143 HvARRAY(dstr)[i] = source
10144 ? he_dup(source, sharekeys, param) : 0;
10145 ++i;
10146 }
10147 if (SvOOK(sstr)) {
00b6aa41
AL
10148 struct xpvhv_aux * const saux = HvAUX(sstr);
10149 struct xpvhv_aux * const daux = HvAUX(dstr);
662fb8b2
NC
10150 /* This flag isn't copied. */
10151 /* SvOOK_on(hv) attacks the IV flags. */
10152 SvFLAGS(dstr) |= SVf_OOK;
10153
10154 hvname = saux->xhv_name;
dd690478
NC
10155 daux->xhv_name
10156 = hvname ? hek_dup(hvname, param) : hvname;
662fb8b2
NC
10157
10158 daux->xhv_riter = saux->xhv_riter;
10159 daux->xhv_eiter = saux->xhv_eiter
dd690478
NC
10160 ? he_dup(saux->xhv_eiter,
10161 (bool)!!HvSHAREKEYS(sstr), param) : 0;
86f55936
NC
10162 daux->xhv_backreferences = saux->xhv_backreferences
10163 ? (AV*) SvREFCNT_inc(
10164 sv_dup((SV*)saux->
10165 xhv_backreferences,
10166 param))
10167 : 0;
662fb8b2
NC
10168 }
10169 }
10170 else {
bd61b366 10171 SvPV_set(dstr, NULL);
662fb8b2
NC
10172 }
10173 /* Record stashes for possible cloning in Perl_clone(). */
10174 if(hvname)
10175 av_push(param->stashes, dstr);
10176 }
10177 break;
662fb8b2 10178 case SVt_PVCV:
bb172083
NC
10179 if (!(param->flags & CLONEf_COPY_STACKS)) {
10180 CvDEPTH(dstr) = 0;
10181 }
10182 case SVt_PVFM:
662fb8b2
NC
10183 /* NOTE: not refcounted */
10184 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
10185 OP_REFCNT_LOCK;
d04ba589
NC
10186 if (!CvISXSUB(dstr))
10187 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 10188 OP_REFCNT_UNLOCK;
cfae286e 10189 if (CvCONST(dstr) && CvISXSUB(dstr)) {
662fb8b2
NC
10190 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
10191 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
10192 sv_dup_inc((SV *)CvXSUBANY(dstr).any_ptr, param);
10193 }
10194 /* don't dup if copying back - CvGV isn't refcounted, so the
10195 * duped GV may never be freed. A bit of a hack! DAPM */
10196 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 10197 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
10198 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
10199 CvOUTSIDE(dstr) =
10200 CvWEAKOUTSIDE(sstr)
10201 ? cv_dup( CvOUTSIDE(dstr), param)
10202 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 10203 if (!CvISXSUB(dstr))
662fb8b2
NC
10204 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
10205 break;
bfcb3514 10206 }
1d7c1841 10207 }
1d7c1841
GS
10208 }
10209
10210 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
10211 ++PL_sv_objcount;
10212
10213 return dstr;
d2d73c3e 10214 }
1d7c1841 10215
645c22ef
DM
10216/* duplicate a context */
10217
1d7c1841 10218PERL_CONTEXT *
a8fc9800 10219Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
10220{
10221 PERL_CONTEXT *ncxs;
10222
10223 if (!cxs)
10224 return (PERL_CONTEXT*)NULL;
10225
10226 /* look for it in the table first */
10227 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
10228 if (ncxs)
10229 return ncxs;
10230
10231 /* create anew and remember what it is */
a02a5408 10232 Newxz(ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
10233 ptr_table_store(PL_ptr_table, cxs, ncxs);
10234
10235 while (ix >= 0) {
c445ea15
AL
10236 PERL_CONTEXT * const cx = &cxs[ix];
10237 PERL_CONTEXT * const ncx = &ncxs[ix];
1d7c1841
GS
10238 ncx->cx_type = cx->cx_type;
10239 if (CxTYPE(cx) == CXt_SUBST) {
10240 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
10241 }
10242 else {
10243 ncx->blk_oldsp = cx->blk_oldsp;
10244 ncx->blk_oldcop = cx->blk_oldcop;
1d7c1841
GS
10245 ncx->blk_oldmarksp = cx->blk_oldmarksp;
10246 ncx->blk_oldscopesp = cx->blk_oldscopesp;
10247 ncx->blk_oldpm = cx->blk_oldpm;
10248 ncx->blk_gimme = cx->blk_gimme;
10249 switch (CxTYPE(cx)) {
10250 case CXt_SUB:
10251 ncx->blk_sub.cv = (cx->blk_sub.olddepth == 0
d2d73c3e
AB
10252 ? cv_dup_inc(cx->blk_sub.cv, param)
10253 : cv_dup(cx->blk_sub.cv,param));
1d7c1841 10254 ncx->blk_sub.argarray = (cx->blk_sub.hasargs
d2d73c3e 10255 ? av_dup_inc(cx->blk_sub.argarray, param)
7d49f689 10256 : NULL);
d2d73c3e 10257 ncx->blk_sub.savearray = av_dup_inc(cx->blk_sub.savearray, param);
1d7c1841
GS
10258 ncx->blk_sub.olddepth = cx->blk_sub.olddepth;
10259 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
10260 ncx->blk_sub.lval = cx->blk_sub.lval;
f39bc417 10261 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10262 break;
10263 case CXt_EVAL:
10264 ncx->blk_eval.old_in_eval = cx->blk_eval.old_in_eval;
10265 ncx->blk_eval.old_op_type = cx->blk_eval.old_op_type;
b47cad08 10266 ncx->blk_eval.old_namesv = sv_dup_inc(cx->blk_eval.old_namesv, param);
1d7c1841 10267 ncx->blk_eval.old_eval_root = cx->blk_eval.old_eval_root;
d2d73c3e 10268 ncx->blk_eval.cur_text = sv_dup(cx->blk_eval.cur_text, param);
f39bc417 10269 ncx->blk_eval.retop = cx->blk_eval.retop;
1d7c1841
GS
10270 break;
10271 case CXt_LOOP:
10272 ncx->blk_loop.label = cx->blk_loop.label;
10273 ncx->blk_loop.resetsp = cx->blk_loop.resetsp;
10274 ncx->blk_loop.redo_op = cx->blk_loop.redo_op;
10275 ncx->blk_loop.next_op = cx->blk_loop.next_op;
10276 ncx->blk_loop.last_op = cx->blk_loop.last_op;
10277 ncx->blk_loop.iterdata = (CxPADLOOP(cx)
10278 ? cx->blk_loop.iterdata
d2d73c3e 10279 : gv_dup((GV*)cx->blk_loop.iterdata, param));
f3548bdc
DM
10280 ncx->blk_loop.oldcomppad
10281 = (PAD*)ptr_table_fetch(PL_ptr_table,
10282 cx->blk_loop.oldcomppad);
d2d73c3e
AB
10283 ncx->blk_loop.itersave = sv_dup_inc(cx->blk_loop.itersave, param);
10284 ncx->blk_loop.iterlval = sv_dup_inc(cx->blk_loop.iterlval, param);
10285 ncx->blk_loop.iterary = av_dup_inc(cx->blk_loop.iterary, param);
1d7c1841
GS
10286 ncx->blk_loop.iterix = cx->blk_loop.iterix;
10287 ncx->blk_loop.itermax = cx->blk_loop.itermax;
10288 break;
10289 case CXt_FORMAT:
d2d73c3e
AB
10290 ncx->blk_sub.cv = cv_dup(cx->blk_sub.cv, param);
10291 ncx->blk_sub.gv = gv_dup(cx->blk_sub.gv, param);
10292 ncx->blk_sub.dfoutgv = gv_dup_inc(cx->blk_sub.dfoutgv, param);
1d7c1841 10293 ncx->blk_sub.hasargs = cx->blk_sub.hasargs;
f39bc417 10294 ncx->blk_sub.retop = cx->blk_sub.retop;
1d7c1841
GS
10295 break;
10296 case CXt_BLOCK:
10297 case CXt_NULL:
10298 break;
10299 }
10300 }
10301 --ix;
10302 }
10303 return ncxs;
10304}
10305
645c22ef
DM
10306/* duplicate a stack info structure */
10307
1d7c1841 10308PERL_SI *
a8fc9800 10309Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
10310{
10311 PERL_SI *nsi;
10312
10313 if (!si)
10314 return (PERL_SI*)NULL;
10315
10316 /* look for it in the table first */
10317 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
10318 if (nsi)
10319 return nsi;
10320
10321 /* create anew and remember what it is */
a02a5408 10322 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
10323 ptr_table_store(PL_ptr_table, si, nsi);
10324
d2d73c3e 10325 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
10326 nsi->si_cxix = si->si_cxix;
10327 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 10328 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 10329 nsi->si_type = si->si_type;
d2d73c3e
AB
10330 nsi->si_prev = si_dup(si->si_prev, param);
10331 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
10332 nsi->si_markoff = si->si_markoff;
10333
10334 return nsi;
10335}
10336
10337#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
10338#define TOPINT(ss,ix) ((ss)[ix].any_i32)
10339#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
10340#define TOPLONG(ss,ix) ((ss)[ix].any_long)
10341#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
10342#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
10343#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
10344#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
10345#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
10346#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
10347#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
10348#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
10349#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
10350#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
10351
10352/* XXXXX todo */
10353#define pv_dup_inc(p) SAVEPV(p)
10354#define pv_dup(p) SAVEPV(p)
10355#define svp_dup_inc(p,pp) any_dup(p,pp)
10356
645c22ef
DM
10357/* map any object to the new equivent - either something in the
10358 * ptr table, or something in the interpreter structure
10359 */
10360
1d7c1841 10361void *
53c1dcc0 10362Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
10363{
10364 void *ret;
10365
10366 if (!v)
10367 return (void*)NULL;
10368
10369 /* look for it in the table first */
10370 ret = ptr_table_fetch(PL_ptr_table, v);
10371 if (ret)
10372 return ret;
10373
10374 /* see if it is part of the interpreter structure */
10375 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 10376 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 10377 else {
1d7c1841 10378 ret = v;
05ec9bb3 10379 }
1d7c1841
GS
10380
10381 return ret;
10382}
10383
645c22ef
DM
10384/* duplicate the save stack */
10385
1d7c1841 10386ANY *
a8fc9800 10387Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 10388{
53c1dcc0
AL
10389 ANY * const ss = proto_perl->Tsavestack;
10390 const I32 max = proto_perl->Tsavestack_max;
10391 I32 ix = proto_perl->Tsavestack_ix;
1d7c1841
GS
10392 ANY *nss;
10393 SV *sv;
10394 GV *gv;
10395 AV *av;
10396 HV *hv;
10397 void* ptr;
10398 int intval;
10399 long longval;
10400 GP *gp;
10401 IV iv;
c4e33207 10402 char *c = NULL;
1d7c1841 10403 void (*dptr) (void*);
acfe0abc 10404 void (*dxptr) (pTHX_ void*);
1d7c1841 10405
a02a5408 10406 Newxz(nss, max, ANY);
1d7c1841
GS
10407
10408 while (ix > 0) {
b464bac0 10409 I32 i = POPINT(ss,ix);
1d7c1841
GS
10410 TOPINT(nss,ix) = i;
10411 switch (i) {
10412 case SAVEt_ITEM: /* normal string */
a41cc44e 10413 case SAVEt_SV: /* scalar reference */
1d7c1841 10414 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10415 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10416 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10417 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10418 break;
05ec9bb3
NIS
10419 case SAVEt_SHARED_PVREF: /* char* in shared space */
10420 c = (char*)POPPTR(ss,ix);
10421 TOPPTR(nss,ix) = savesharedpv(c);
10422 ptr = POPPTR(ss,ix);
10423 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10424 break;
1d7c1841
GS
10425 case SAVEt_GENERIC_SVREF: /* generic sv */
10426 case SAVEt_SVREF: /* scalar reference */
10427 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10428 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10429 ptr = POPPTR(ss,ix);
10430 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
10431 break;
a41cc44e 10432 case SAVEt_HV: /* hash reference */
1d7c1841 10433 case SAVEt_AV: /* array reference */
337d28f5
NC
10434 sv = POPPTR(ss,ix);
10435 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10436 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10437 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10438 break;
10439 case SAVEt_INT: /* int reference */
10440 ptr = POPPTR(ss,ix);
10441 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10442 intval = (int)POPINT(ss,ix);
10443 TOPINT(nss,ix) = intval;
10444 break;
10445 case SAVEt_LONG: /* long reference */
10446 ptr = POPPTR(ss,ix);
10447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10448 longval = (long)POPLONG(ss,ix);
10449 TOPLONG(nss,ix) = longval;
10450 break;
10451 case SAVEt_I32: /* I32 reference */
10452 case SAVEt_I16: /* I16 reference */
10453 case SAVEt_I8: /* I8 reference */
88effcc9 10454 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
10455 ptr = POPPTR(ss,ix);
10456 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10457 i = POPINT(ss,ix);
10458 TOPINT(nss,ix) = i;
10459 break;
10460 case SAVEt_IV: /* IV reference */
10461 ptr = POPPTR(ss,ix);
10462 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10463 iv = POPIV(ss,ix);
10464 TOPIV(nss,ix) = iv;
10465 break;
a41cc44e
NC
10466 case SAVEt_HPTR: /* HV* reference */
10467 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
10468 case SAVEt_SPTR: /* SV* reference */
10469 ptr = POPPTR(ss,ix);
10470 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10471 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10472 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
10473 break;
10474 case SAVEt_VPTR: /* random* reference */
10475 ptr = POPPTR(ss,ix);
10476 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10477 ptr = POPPTR(ss,ix);
10478 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10479 break;
b03d03b0 10480 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
10481 case SAVEt_PPTR: /* char* reference */
10482 ptr = POPPTR(ss,ix);
10483 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10484 c = (char*)POPPTR(ss,ix);
10485 TOPPTR(nss,ix) = pv_dup(c);
10486 break;
1d7c1841
GS
10487 case SAVEt_NSTAB:
10488 gv = (GV*)POPPTR(ss,ix);
d2d73c3e 10489 TOPPTR(nss,ix) = gv_dup(gv, param);
1d7c1841
GS
10490 break;
10491 case SAVEt_GP: /* scalar reference */
10492 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 10493 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841
GS
10494 (void)GpREFCNT_inc(gp);
10495 gv = (GV*)POPPTR(ss,ix);
2ed3c8fc 10496 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841
GS
10497 c = (char*)POPPTR(ss,ix);
10498 TOPPTR(nss,ix) = pv_dup(c);
10499 iv = POPIV(ss,ix);
10500 TOPIV(nss,ix) = iv;
10501 iv = POPIV(ss,ix);
10502 TOPIV(nss,ix) = iv;
10503 break;
10504 case SAVEt_FREESV:
26d9b02f 10505 case SAVEt_MORTALIZESV:
1d7c1841 10506 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10507 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10508 break;
10509 case SAVEt_FREEOP:
10510 ptr = POPPTR(ss,ix);
10511 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
10512 /* these are assumed to be refcounted properly */
53c1dcc0 10513 OP *o;
1d7c1841
GS
10514 switch (((OP*)ptr)->op_type) {
10515 case OP_LEAVESUB:
10516 case OP_LEAVESUBLV:
10517 case OP_LEAVEEVAL:
10518 case OP_LEAVE:
10519 case OP_SCOPE:
10520 case OP_LEAVEWRITE:
e977893f
GS
10521 TOPPTR(nss,ix) = ptr;
10522 o = (OP*)ptr;
10523 OpREFCNT_inc(o);
1d7c1841
GS
10524 break;
10525 default:
5f66b61c 10526 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10527 break;
10528 }
10529 }
10530 else
5f66b61c 10531 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
10532 break;
10533 case SAVEt_FREEPV:
10534 c = (char*)POPPTR(ss,ix);
10535 TOPPTR(nss,ix) = pv_dup_inc(c);
10536 break;
10537 case SAVEt_CLEARSV:
10538 longval = POPLONG(ss,ix);
10539 TOPLONG(nss,ix) = longval;
10540 break;
10541 case SAVEt_DELETE:
10542 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10543 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10544 c = (char*)POPPTR(ss,ix);
10545 TOPPTR(nss,ix) = pv_dup_inc(c);
10546 i = POPINT(ss,ix);
10547 TOPINT(nss,ix) = i;
10548 break;
10549 case SAVEt_DESTRUCTOR:
10550 ptr = POPPTR(ss,ix);
10551 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10552 dptr = POPDPTR(ss,ix);
8141890a
JH
10553 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
10554 any_dup(FPTR2DPTR(void *, dptr),
10555 proto_perl));
1d7c1841
GS
10556 break;
10557 case SAVEt_DESTRUCTOR_X:
10558 ptr = POPPTR(ss,ix);
10559 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
10560 dxptr = POPDXPTR(ss,ix);
8141890a
JH
10561 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
10562 any_dup(FPTR2DPTR(void *, dxptr),
10563 proto_perl));
1d7c1841
GS
10564 break;
10565 case SAVEt_REGCONTEXT:
10566 case SAVEt_ALLOC:
10567 i = POPINT(ss,ix);
10568 TOPINT(nss,ix) = i;
10569 ix -= i;
10570 break;
10571 case SAVEt_STACK_POS: /* Position on Perl stack */
10572 i = POPINT(ss,ix);
10573 TOPINT(nss,ix) = i;
10574 break;
10575 case SAVEt_AELEM: /* array element */
10576 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10577 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
10578 i = POPINT(ss,ix);
10579 TOPINT(nss,ix) = i;
10580 av = (AV*)POPPTR(ss,ix);
d2d73c3e 10581 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841
GS
10582 break;
10583 case SAVEt_HELEM: /* hash element */
10584 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10585 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10586 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10587 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 10588 hv = (HV*)POPPTR(ss,ix);
d2d73c3e 10589 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
1d7c1841
GS
10590 break;
10591 case SAVEt_OP:
10592 ptr = POPPTR(ss,ix);
10593 TOPPTR(nss,ix) = ptr;
10594 break;
10595 case SAVEt_HINTS:
10596 i = POPINT(ss,ix);
10597 TOPINT(nss,ix) = i;
b3ca2e83 10598 ptr = POPPTR(ss,ix);
080ac856 10599 if (ptr) {
7b6dd8c3 10600 HINTS_REFCNT_LOCK;
080ac856 10601 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
10602 HINTS_REFCNT_UNLOCK;
10603 }
cbb1fbea 10604 TOPPTR(nss,ix) = ptr;
a8f8b6a7
NC
10605 if (i & HINT_LOCALIZE_HH) {
10606 hv = (HV*)POPPTR(ss,ix);
10607 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
10608 }
1d7c1841 10609 break;
c4410b1b
GS
10610 case SAVEt_COMPPAD:
10611 av = (AV*)POPPTR(ss,ix);
58ed4fbe 10612 TOPPTR(nss,ix) = av_dup(av, param);
c4410b1b 10613 break;
c3564e5c
GS
10614 case SAVEt_PADSV:
10615 longval = (long)POPLONG(ss,ix);
10616 TOPLONG(nss,ix) = longval;
10617 ptr = POPPTR(ss,ix);
10618 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
10619 sv = (SV*)POPPTR(ss,ix);
d2d73c3e 10620 TOPPTR(nss,ix) = sv_dup(sv, param);
c3564e5c 10621 break;
a1bb4754 10622 case SAVEt_BOOL:
38d8b13e 10623 ptr = POPPTR(ss,ix);
b9609c01 10624 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 10625 longval = (long)POPBOOL(ss,ix);
b9609c01 10626 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 10627 break;
8bd2680e
MHM
10628 case SAVEt_SET_SVFLAGS:
10629 i = POPINT(ss,ix);
10630 TOPINT(nss,ix) = i;
10631 i = POPINT(ss,ix);
10632 TOPINT(nss,ix) = i;
10633 sv = (SV*)POPPTR(ss,ix);
10634 TOPPTR(nss,ix) = sv_dup(sv, param);
10635 break;
5bfb7d0e
NC
10636 case SAVEt_RE_STATE:
10637 {
10638 const struct re_save_state *const old_state
10639 = (struct re_save_state *)
10640 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10641 struct re_save_state *const new_state
10642 = (struct re_save_state *)
10643 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
10644
10645 Copy(old_state, new_state, 1, struct re_save_state);
10646 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
10647
10648 new_state->re_state_bostr
10649 = pv_dup(old_state->re_state_bostr);
10650 new_state->re_state_reginput
10651 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
10652 new_state->re_state_regeol
10653 = pv_dup(old_state->re_state_regeol);
10654 new_state->re_state_regstartp
10655 = any_dup(old_state->re_state_regstartp, proto_perl);
10656 new_state->re_state_regendp
10657 = any_dup(old_state->re_state_regendp, proto_perl);
10658 new_state->re_state_reglastparen
10659 = any_dup(old_state->re_state_reglastparen, proto_perl);
10660 new_state->re_state_reglastcloseparen
10661 = any_dup(old_state->re_state_reglastcloseparen,
10662 proto_perl);
5bfb7d0e
NC
10663 /* XXX This just has to be broken. The old save_re_context
10664 code did SAVEGENERICPV(PL_reg_start_tmp);
10665 PL_reg_start_tmp is char **.
10666 Look above to what the dup code does for
10667 SAVEt_GENERIC_PVREF
10668 It can never have worked.
10669 So this is merely a faithful copy of the exiting bug: */
10670 new_state->re_state_reg_start_tmp
10671 = (char **) pv_dup((char *)
10672 old_state->re_state_reg_start_tmp);
10673 /* I assume that it only ever "worked" because no-one called
10674 (pseudo)fork while the regexp engine had re-entered itself.
10675 */
5bfb7d0e
NC
10676#ifdef PERL_OLD_COPY_ON_WRITE
10677 new_state->re_state_nrs
10678 = sv_dup(old_state->re_state_nrs, param);
10679#endif
10680 new_state->re_state_reg_magic
10681 = any_dup(old_state->re_state_reg_magic, proto_perl);
10682 new_state->re_state_reg_oldcurpm
10683 = any_dup(old_state->re_state_reg_oldcurpm, proto_perl);
10684 new_state->re_state_reg_curpm
10685 = any_dup(old_state->re_state_reg_curpm, proto_perl);
10686 new_state->re_state_reg_oldsaved
10687 = pv_dup(old_state->re_state_reg_oldsaved);
10688 new_state->re_state_reg_poscache
10689 = pv_dup(old_state->re_state_reg_poscache);
10690#ifdef DEBUGGING
10691 new_state->re_state_reg_starttry
10692 = pv_dup(old_state->re_state_reg_starttry);
10693#endif
10694 break;
10695 }
68da3b2f
NC
10696 case SAVEt_COMPILE_WARNINGS:
10697 ptr = POPPTR(ss,ix);
10698 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 10699 break;
1d7c1841 10700 default:
ca05af4a 10701 Perl_croak(aTHX_ "panic: ss_dup inconsistency (%"IVdf")", (IV) i);
1d7c1841
GS
10702 }
10703 }
10704
bd81e77b
NC
10705 return nss;
10706}
10707
10708
10709/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
10710 * flag to the result. This is done for each stash before cloning starts,
10711 * so we know which stashes want their objects cloned */
10712
10713static void
10714do_mark_cloneable_stash(pTHX_ SV *sv)
10715{
10716 const HEK * const hvname = HvNAME_HEK((HV*)sv);
10717 if (hvname) {
10718 GV* const cloner = gv_fetchmethod_autoload((HV*)sv, "CLONE_SKIP", 0);
10719 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
10720 if (cloner && GvCV(cloner)) {
10721 dSP;
10722 UV status;
10723
10724 ENTER;
10725 SAVETMPS;
10726 PUSHMARK(SP);
10727 XPUSHs(sv_2mortal(newSVhek(hvname)));
10728 PUTBACK;
10729 call_sv((SV*)GvCV(cloner), G_SCALAR);
10730 SPAGAIN;
10731 status = POPu;
10732 PUTBACK;
10733 FREETMPS;
10734 LEAVE;
10735 if (status)
10736 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
10737 }
10738 }
10739}
10740
10741
10742
10743/*
10744=for apidoc perl_clone
10745
10746Create and return a new interpreter by cloning the current one.
10747
10748perl_clone takes these flags as parameters:
10749
10750CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
10751without it we only clone the data and zero the stacks,
10752with it we copy the stacks and the new perl interpreter is
10753ready to run at the exact same point as the previous one.
10754The pseudo-fork code uses COPY_STACKS while the
10755threads->new doesn't.
10756
10757CLONEf_KEEP_PTR_TABLE
10758perl_clone keeps a ptr_table with the pointer of the old
10759variable as a key and the new variable as a value,
10760this allows it to check if something has been cloned and not
10761clone it again but rather just use the value and increase the
10762refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
10763the ptr_table using the function
10764C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
10765reason to keep it around is if you want to dup some of your own
10766variable who are outside the graph perl scans, example of this
10767code is in threads.xs create
10768
10769CLONEf_CLONE_HOST
10770This is a win32 thing, it is ignored on unix, it tells perls
10771win32host code (which is c++) to clone itself, this is needed on
10772win32 if you want to run two threads at the same time,
10773if you just want to do some stuff in a separate perl interpreter
10774and then throw it away and return to the original one,
10775you don't need to do anything.
10776
10777=cut
10778*/
10779
10780/* XXX the above needs expanding by someone who actually understands it ! */
10781EXTERN_C PerlInterpreter *
10782perl_clone_host(PerlInterpreter* proto_perl, UV flags);
10783
10784PerlInterpreter *
10785perl_clone(PerlInterpreter *proto_perl, UV flags)
10786{
10787 dVAR;
10788#ifdef PERL_IMPLICIT_SYS
10789
10790 /* perlhost.h so we need to call into it
10791 to clone the host, CPerlHost should have a c interface, sky */
10792
10793 if (flags & CLONEf_CLONE_HOST) {
10794 return perl_clone_host(proto_perl,flags);
10795 }
10796 return perl_clone_using(proto_perl, flags,
10797 proto_perl->IMem,
10798 proto_perl->IMemShared,
10799 proto_perl->IMemParse,
10800 proto_perl->IEnv,
10801 proto_perl->IStdIO,
10802 proto_perl->ILIO,
10803 proto_perl->IDir,
10804 proto_perl->ISock,
10805 proto_perl->IProc);
10806}
10807
10808PerlInterpreter *
10809perl_clone_using(PerlInterpreter *proto_perl, UV flags,
10810 struct IPerlMem* ipM, struct IPerlMem* ipMS,
10811 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
10812 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
10813 struct IPerlDir* ipD, struct IPerlSock* ipS,
10814 struct IPerlProc* ipP)
10815{
10816 /* XXX many of the string copies here can be optimized if they're
10817 * constants; they need to be allocated as common memory and just
10818 * their pointers copied. */
10819
10820 IV i;
10821 CLONE_PARAMS clone_params;
5f66b61c 10822 CLONE_PARAMS* const param = &clone_params;
bd81e77b 10823
5f66b61c 10824 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
bd81e77b
NC
10825 /* for each stash, determine whether its objects should be cloned */
10826 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10827 PERL_SET_THX(my_perl);
10828
10829# ifdef DEBUGGING
7e337ee0 10830 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10831 PL_op = NULL;
10832 PL_curcop = NULL;
bd81e77b
NC
10833 PL_markstack = 0;
10834 PL_scopestack = 0;
10835 PL_savestack = 0;
10836 PL_savestack_ix = 0;
10837 PL_savestack_max = -1;
10838 PL_sig_pending = 0;
10839 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10840# else /* !DEBUGGING */
10841 Zero(my_perl, 1, PerlInterpreter);
10842# endif /* DEBUGGING */
10843
10844 /* host pointers */
10845 PL_Mem = ipM;
10846 PL_MemShared = ipMS;
10847 PL_MemParse = ipMP;
10848 PL_Env = ipE;
10849 PL_StdIO = ipStd;
10850 PL_LIO = ipLIO;
10851 PL_Dir = ipD;
10852 PL_Sock = ipS;
10853 PL_Proc = ipP;
10854#else /* !PERL_IMPLICIT_SYS */
10855 IV i;
10856 CLONE_PARAMS clone_params;
10857 CLONE_PARAMS* param = &clone_params;
5f66b61c 10858 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
bd81e77b
NC
10859 /* for each stash, determine whether its objects should be cloned */
10860 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
10861 PERL_SET_THX(my_perl);
10862
10863# ifdef DEBUGGING
7e337ee0 10864 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
10865 PL_op = NULL;
10866 PL_curcop = NULL;
bd81e77b
NC
10867 PL_markstack = 0;
10868 PL_scopestack = 0;
10869 PL_savestack = 0;
10870 PL_savestack_ix = 0;
10871 PL_savestack_max = -1;
10872 PL_sig_pending = 0;
10873 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
10874# else /* !DEBUGGING */
10875 Zero(my_perl, 1, PerlInterpreter);
10876# endif /* DEBUGGING */
10877#endif /* PERL_IMPLICIT_SYS */
10878 param->flags = flags;
10879 param->proto_perl = proto_perl;
10880
7cb608b5
NC
10881 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
10882
fdda85ca 10883 PL_body_arenas = NULL;
bd81e77b
NC
10884 Zero(&PL_body_roots, 1, PL_body_roots);
10885
10886 PL_nice_chunk = NULL;
10887 PL_nice_chunk_size = 0;
10888 PL_sv_count = 0;
10889 PL_sv_objcount = 0;
a0714e2c
SS
10890 PL_sv_root = NULL;
10891 PL_sv_arenaroot = NULL;
bd81e77b
NC
10892
10893 PL_debug = proto_perl->Idebug;
10894
10895 PL_hash_seed = proto_perl->Ihash_seed;
10896 PL_rehash_seed = proto_perl->Irehash_seed;
10897
10898#ifdef USE_REENTRANT_API
10899 /* XXX: things like -Dm will segfault here in perlio, but doing
10900 * PERL_SET_CONTEXT(proto_perl);
10901 * breaks too many other things
10902 */
10903 Perl_reentrant_init(aTHX);
10904#endif
10905
10906 /* create SV map for pointer relocation */
10907 PL_ptr_table = ptr_table_new();
10908
10909 /* initialize these special pointers as early as possible */
10910 SvANY(&PL_sv_undef) = NULL;
10911 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
10912 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
10913 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
10914
10915 SvANY(&PL_sv_no) = new_XPVNV();
10916 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
10917 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10918 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10919 SvPV_set(&PL_sv_no, SAVEPVN(PL_No, 0));
10920 SvCUR_set(&PL_sv_no, 0);
10921 SvLEN_set(&PL_sv_no, 1);
10922 SvIV_set(&PL_sv_no, 0);
10923 SvNV_set(&PL_sv_no, 0);
10924 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
10925
10926 SvANY(&PL_sv_yes) = new_XPVNV();
10927 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
10928 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
10929 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
10930 SvPV_set(&PL_sv_yes, SAVEPVN(PL_Yes, 1));
10931 SvCUR_set(&PL_sv_yes, 1);
10932 SvLEN_set(&PL_sv_yes, 2);
10933 SvIV_set(&PL_sv_yes, 1);
10934 SvNV_set(&PL_sv_yes, 1);
10935 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
10936
10937 /* create (a non-shared!) shared string table */
10938 PL_strtab = newHV();
10939 HvSHAREKEYS_off(PL_strtab);
10940 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
10941 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
10942
10943 PL_compiling = proto_perl->Icompiling;
10944
10945 /* These two PVs will be free'd special way so must set them same way op.c does */
10946 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
10947 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
10948
10949 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
10950 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
10951
10952 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 10953 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
bd81e77b
NC
10954 if (!specialCopIO(PL_compiling.cop_io))
10955 PL_compiling.cop_io = sv_dup_inc(PL_compiling.cop_io, param);
cbb1fbea
NC
10956 if (PL_compiling.cop_hints) {
10957 HINTS_REFCNT_LOCK;
10958 PL_compiling.cop_hints->refcounted_he_refcnt++;
10959 HINTS_REFCNT_UNLOCK;
10960 }
bd81e77b
NC
10961 PL_curcop = (COP*)any_dup(proto_perl->Tcurcop, proto_perl);
10962
10963 /* pseudo environmental stuff */
10964 PL_origargc = proto_perl->Iorigargc;
10965 PL_origargv = proto_perl->Iorigargv;
10966
10967 param->stashes = newAV(); /* Setup array of objects to call clone on */
10968
10969 /* Set tainting stuff before PerlIO_debug can possibly get called */
10970 PL_tainting = proto_perl->Itainting;
10971 PL_taint_warn = proto_perl->Itaint_warn;
10972
10973#ifdef PERLIO_LAYERS
10974 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
10975 PerlIO_clone(aTHX_ proto_perl, param);
10976#endif
10977
10978 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
10979 PL_incgv = gv_dup(proto_perl->Iincgv, param);
10980 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
10981 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
10982 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
10983 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
10984
10985 /* switches */
10986 PL_minus_c = proto_perl->Iminus_c;
10987 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
10988 PL_localpatches = proto_perl->Ilocalpatches;
10989 PL_splitstr = proto_perl->Isplitstr;
10990 PL_preprocess = proto_perl->Ipreprocess;
10991 PL_minus_n = proto_perl->Iminus_n;
10992 PL_minus_p = proto_perl->Iminus_p;
10993 PL_minus_l = proto_perl->Iminus_l;
10994 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 10995 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
10996 PL_minus_F = proto_perl->Iminus_F;
10997 PL_doswitches = proto_perl->Idoswitches;
10998 PL_dowarn = proto_perl->Idowarn;
10999 PL_doextract = proto_perl->Idoextract;
11000 PL_sawampersand = proto_perl->Isawampersand;
11001 PL_unsafe = proto_perl->Iunsafe;
11002 PL_inplace = SAVEPV(proto_perl->Iinplace);
11003 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11004 PL_perldb = proto_perl->Iperldb;
11005 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11006 PL_exit_flags = proto_perl->Iexit_flags;
11007
11008 /* magical thingies */
11009 /* XXX time(&PL_basetime) when asked for? */
11010 PL_basetime = proto_perl->Ibasetime;
11011 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11012
11013 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
11014 PL_statusvalue = proto_perl->Istatusvalue;
11015#ifdef VMS
11016 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11017#else
11018 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11019#endif
11020 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11021
11022 sv_setpvn(PERL_DEBUG_PAD(0), "", 0); /* For regex debugging. */
11023 sv_setpvn(PERL_DEBUG_PAD(1), "", 0); /* ext/re needs these */
11024 sv_setpvn(PERL_DEBUG_PAD(2), "", 0); /* even without DEBUGGING. */
11025
11026 /* Clone the regex array */
11027 PL_regex_padav = newAV();
11028 {
11029 const I32 len = av_len((AV*)proto_perl->Iregex_padav);
7a5b473e 11030 SV* const * const regexen = AvARRAY((AV*)proto_perl->Iregex_padav);
bd81e77b 11031 IV i;
7f466ec7 11032 av_push(PL_regex_padav, sv_dup_inc_NN(regexen[0],param));
bd81e77b 11033 for(i = 1; i <= len; i++) {
7a5b473e
AL
11034 const SV * const regex = regexen[i];
11035 SV * const sv =
11036 SvREPADTMP(regex)
11037 ? sv_dup_inc(regex, param)
11038 : SvREFCNT_inc(
11039 newSViv(PTR2IV(re_dup(
11040 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11041 ;
11042 av_push(PL_regex_padav, sv);
bd81e77b
NC
11043 }
11044 }
11045 PL_regex_pad = AvARRAY(PL_regex_padav);
11046
11047 /* shortcuts to various I/O objects */
11048 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11049 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11050 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11051 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11052 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11053 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 11054
bd81e77b
NC
11055 /* shortcuts to regexp stuff */
11056 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 11057
bd81e77b
NC
11058 /* shortcuts to misc objects */
11059 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 11060
bd81e77b
NC
11061 /* shortcuts to debugging objects */
11062 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11063 PL_DBline = gv_dup(proto_perl->IDBline, param);
11064 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11065 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11066 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11067 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11068 PL_DBassertion = sv_dup(proto_perl->IDBassertion, param);
11069 PL_lineary = av_dup(proto_perl->Ilineary, param);
11070 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 11071
bd81e77b
NC
11072 /* symbol tables */
11073 PL_defstash = hv_dup_inc(proto_perl->Tdefstash, param);
11074 PL_curstash = hv_dup(proto_perl->Tcurstash, param);
11075 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11076 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11077 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11078
11079 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11080 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11081 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11082 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11083 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11084 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11085
11086 PL_sub_generation = proto_perl->Isub_generation;
11087
11088 /* funky return mechanisms */
11089 PL_forkprocess = proto_perl->Iforkprocess;
11090
11091 /* subprocess state */
11092 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11093
11094 /* internal state */
11095 PL_maxo = proto_perl->Imaxo;
11096 if (proto_perl->Iop_mask)
11097 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
11098 else
bd61b366 11099 PL_op_mask = NULL;
bd81e77b
NC
11100 /* PL_asserting = proto_perl->Iasserting; */
11101
11102 /* current interpreter roots */
11103 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
11104 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
11105 PL_main_start = proto_perl->Imain_start;
11106 PL_eval_root = proto_perl->Ieval_root;
11107 PL_eval_start = proto_perl->Ieval_start;
11108
11109 /* runtime control stuff */
11110 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
11111 PL_copline = proto_perl->Icopline;
11112
11113 PL_filemode = proto_perl->Ifilemode;
11114 PL_lastfd = proto_perl->Ilastfd;
11115 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
11116 PL_Argv = NULL;
bd61b366 11117 PL_Cmd = NULL;
bd81e77b
NC
11118 PL_gensym = proto_perl->Igensym;
11119 PL_preambled = proto_perl->Ipreambled;
11120 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
11121 PL_laststatval = proto_perl->Ilaststatval;
11122 PL_laststype = proto_perl->Ilaststype;
a0714e2c 11123 PL_mess_sv = NULL;
bd81e77b
NC
11124
11125 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
11126
11127 /* interpreter atexit processing */
11128 PL_exitlistlen = proto_perl->Iexitlistlen;
11129 if (PL_exitlistlen) {
11130 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
11131 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 11132 }
bd81e77b
NC
11133 else
11134 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
11135
11136 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 11137 if (PL_my_cxt_size) {
f16dd614
DM
11138 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
11139 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
11140 }
11141 else
11142 PL_my_cxt_list = (void**)NULL;
bd81e77b
NC
11143 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
11144 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
11145 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
11146
11147 PL_profiledata = NULL;
11148 PL_rsfp = fp_dup(proto_perl->Irsfp, '<', param);
11149 /* PL_rsfp_filters entries have fake IoDIRP() */
11150 PL_rsfp_filters = av_dup_inc(proto_perl->Irsfp_filters, param);
9660f481 11151
bd81e77b 11152 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 11153
bd81e77b 11154 PAD_CLONE_VARS(proto_perl, param);
9660f481 11155
bd81e77b
NC
11156#ifdef HAVE_INTERP_INTERN
11157 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
11158#endif
645c22ef 11159
bd81e77b
NC
11160 /* more statics moved here */
11161 PL_generation = proto_perl->Igeneration;
11162 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 11163
bd81e77b
NC
11164 PL_in_clean_objs = proto_perl->Iin_clean_objs;
11165 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 11166
bd81e77b
NC
11167 PL_uid = proto_perl->Iuid;
11168 PL_euid = proto_perl->Ieuid;
11169 PL_gid = proto_perl->Igid;
11170 PL_egid = proto_perl->Iegid;
11171 PL_nomemok = proto_perl->Inomemok;
11172 PL_an = proto_perl->Ian;
11173 PL_evalseq = proto_perl->Ievalseq;
11174 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
11175 PL_origalen = proto_perl->Iorigalen;
11176#ifdef PERL_USES_PL_PIDSTATUS
11177 PL_pidstatus = newHV(); /* XXX flag for cloning? */
11178#endif
11179 PL_osname = SAVEPV(proto_perl->Iosname);
11180 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 11181
bd81e77b 11182 PL_runops = proto_perl->Irunops;
6a78b4db 11183
bd81e77b 11184 Copy(proto_perl->Itokenbuf, PL_tokenbuf, 256, char);
6a78b4db 11185
bd81e77b
NC
11186#ifdef CSH
11187 PL_cshlen = proto_perl->Icshlen;
11188 PL_cshname = proto_perl->Icshname; /* XXX never deallocated */
11189#endif
645c22ef 11190
bd81e77b
NC
11191 PL_lex_state = proto_perl->Ilex_state;
11192 PL_lex_defer = proto_perl->Ilex_defer;
11193 PL_lex_expect = proto_perl->Ilex_expect;
11194 PL_lex_formbrack = proto_perl->Ilex_formbrack;
11195 PL_lex_dojoin = proto_perl->Ilex_dojoin;
11196 PL_lex_starts = proto_perl->Ilex_starts;
11197 PL_lex_stuff = sv_dup_inc(proto_perl->Ilex_stuff, param);
11198 PL_lex_repl = sv_dup_inc(proto_perl->Ilex_repl, param);
11199 PL_lex_op = proto_perl->Ilex_op;
11200 PL_lex_inpat = proto_perl->Ilex_inpat;
11201 PL_lex_inwhat = proto_perl->Ilex_inwhat;
11202 PL_lex_brackets = proto_perl->Ilex_brackets;
11203 i = (PL_lex_brackets < 120 ? 120 : PL_lex_brackets);
11204 PL_lex_brackstack = SAVEPVN(proto_perl->Ilex_brackstack,i);
11205 PL_lex_casemods = proto_perl->Ilex_casemods;
11206 i = (PL_lex_casemods < 12 ? 12 : PL_lex_casemods);
11207 PL_lex_casestack = SAVEPVN(proto_perl->Ilex_casestack,i);
645c22ef 11208
5db06880
NC
11209#ifdef PERL_MAD
11210 Copy(proto_perl->Inexttoke, PL_nexttoke, 5, NEXTTOKE);
11211 PL_lasttoke = proto_perl->Ilasttoke;
5336380d
NC
11212 PL_realtokenstart = proto_perl->Irealtokenstart;
11213 PL_faketokens = proto_perl->Ifaketokens;
11214 PL_thismad = proto_perl->Ithismad;
11215 PL_thistoken = proto_perl->Ithistoken;
11216 PL_thisopen = proto_perl->Ithisopen;
11217 PL_thisstuff = proto_perl->Ithisstuff;
11218 PL_thisclose = proto_perl->Ithisclose;
11219 PL_thiswhite = proto_perl->Ithiswhite;
11220 PL_nextwhite = proto_perl->Inextwhite;
11221 PL_skipwhite = proto_perl->Iskipwhite;
11222 PL_endwhite = proto_perl->Iendwhite;
11223 PL_curforce = proto_perl->Icurforce;
5db06880 11224#else
bd81e77b
NC
11225 Copy(proto_perl->Inextval, PL_nextval, 5, YYSTYPE);
11226 Copy(proto_perl->Inexttype, PL_nexttype, 5, I32);
11227 PL_nexttoke = proto_perl->Inexttoke;
5db06880 11228#endif
c43294b8 11229
bd81e77b
NC
11230 /* XXX This is probably masking the deeper issue of why
11231 * SvANY(proto_perl->Ilinestr) can be NULL at this point. For test case:
11232 * http://archive.develooper.com/perl5-porters%40perl.org/msg83298.html
11233 * (A little debugging with a watchpoint on it may help.)
11234 */
11235 if (SvANY(proto_perl->Ilinestr)) {
11236 PL_linestr = sv_dup_inc(proto_perl->Ilinestr, param);
11237 i = proto_perl->Ibufptr - SvPVX_const(proto_perl->Ilinestr);
11238 PL_bufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11239 i = proto_perl->Ioldbufptr - SvPVX_const(proto_perl->Ilinestr);
11240 PL_oldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11241 i = proto_perl->Ioldoldbufptr - SvPVX_const(proto_perl->Ilinestr);
11242 PL_oldoldbufptr = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11243 i = proto_perl->Ilinestart - SvPVX_const(proto_perl->Ilinestr);
11244 PL_linestart = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11245 }
11246 else {
561b68a9 11247 PL_linestr = newSV(79);
bd81e77b
NC
11248 sv_upgrade(PL_linestr,SVt_PVIV);
11249 sv_setpvn(PL_linestr,"",0);
11250 PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
11251 }
11252 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11253 PL_pending_ident = proto_perl->Ipending_ident;
11254 PL_sublex_info = proto_perl->Isublex_info; /* XXX not quite right */
11255
11256 PL_expect = proto_perl->Iexpect;
11257
11258 PL_multi_start = proto_perl->Imulti_start;
11259 PL_multi_end = proto_perl->Imulti_end;
11260 PL_multi_open = proto_perl->Imulti_open;
11261 PL_multi_close = proto_perl->Imulti_close;
11262
11263 PL_error_count = proto_perl->Ierror_count;
11264 PL_subline = proto_perl->Isubline;
11265 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 11266
bd81e77b
NC
11267 /* XXX See comment on SvANY(proto_perl->Ilinestr) above */
11268 if (SvANY(proto_perl->Ilinestr)) {
11269 i = proto_perl->Ilast_uni - SvPVX_const(proto_perl->Ilinestr);
11270 PL_last_uni = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11271 i = proto_perl->Ilast_lop - SvPVX_const(proto_perl->Ilinestr);
11272 PL_last_lop = SvPVX(PL_linestr) + (i < 0 ? 0 : i);
11273 PL_last_lop_op = proto_perl->Ilast_lop_op;
11274 }
11275 else {
11276 PL_last_uni = SvPVX(PL_linestr);
11277 PL_last_lop = SvPVX(PL_linestr);
11278 PL_last_lop_op = 0;
11279 }
11280 PL_in_my = proto_perl->Iin_my;
11281 PL_in_my_stash = hv_dup(proto_perl->Iin_my_stash, param);
11282#ifdef FCRYPT
11283 PL_cryptseen = proto_perl->Icryptseen;
11284#endif
1d7c1841 11285
bd81e77b 11286 PL_hints = proto_perl->Ihints;
1d7c1841 11287
bd81e77b 11288 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 11289
bd81e77b
NC
11290#ifdef USE_LOCALE_COLLATE
11291 PL_collation_ix = proto_perl->Icollation_ix;
11292 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
11293 PL_collation_standard = proto_perl->Icollation_standard;
11294 PL_collxfrm_base = proto_perl->Icollxfrm_base;
11295 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
11296#endif /* USE_LOCALE_COLLATE */
1d7c1841 11297
bd81e77b
NC
11298#ifdef USE_LOCALE_NUMERIC
11299 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
11300 PL_numeric_standard = proto_perl->Inumeric_standard;
11301 PL_numeric_local = proto_perl->Inumeric_local;
11302 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
11303#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 11304
bd81e77b
NC
11305 /* utf8 character classes */
11306 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
11307 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
11308 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
11309 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
11310 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
11311 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
11312 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
11313 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
11314 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
11315 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
11316 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
11317 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
11318 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
11319 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
11320 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
11321 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
11322 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
11323 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
11324 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
11325 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 11326
bd81e77b
NC
11327 /* Did the locale setup indicate UTF-8? */
11328 PL_utf8locale = proto_perl->Iutf8locale;
11329 /* Unicode features (see perlrun/-C) */
11330 PL_unicode = proto_perl->Iunicode;
1d7c1841 11331
bd81e77b
NC
11332 /* Pre-5.8 signals control */
11333 PL_signals = proto_perl->Isignals;
1d7c1841 11334
bd81e77b
NC
11335 /* times() ticks per second */
11336 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 11337
bd81e77b
NC
11338 /* Recursion stopper for PerlIO_find_layer */
11339 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 11340
bd81e77b
NC
11341 /* sort() routine */
11342 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 11343
bd81e77b
NC
11344 /* Not really needed/useful since the reenrant_retint is "volatile",
11345 * but do it for consistency's sake. */
11346 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 11347
bd81e77b
NC
11348 /* Hooks to shared SVs and locks. */
11349 PL_sharehook = proto_perl->Isharehook;
11350 PL_lockhook = proto_perl->Ilockhook;
11351 PL_unlockhook = proto_perl->Iunlockhook;
11352 PL_threadhook = proto_perl->Ithreadhook;
1d7c1841 11353
bd81e77b
NC
11354 PL_runops_std = proto_perl->Irunops_std;
11355 PL_runops_dbg = proto_perl->Irunops_dbg;
1d7c1841 11356
bd81e77b
NC
11357#ifdef THREADS_HAVE_PIDS
11358 PL_ppid = proto_perl->Ippid;
11359#endif
1d7c1841 11360
bd81e77b 11361 /* swatch cache */
5c284bb0 11362 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
11363 PL_last_swash_klen = 0;
11364 PL_last_swash_key[0]= '\0';
11365 PL_last_swash_tmps = (U8*)NULL;
11366 PL_last_swash_slen = 0;
1d7c1841 11367
bd81e77b
NC
11368 PL_glob_index = proto_perl->Iglob_index;
11369 PL_srand_called = proto_perl->Isrand_called;
11370 PL_uudmap['M'] = 0; /* reinits on demand */
bd61b366 11371 PL_bitcount = NULL; /* reinits on demand */
05ec9bb3 11372
bd81e77b
NC
11373 if (proto_perl->Ipsig_pend) {
11374 Newxz(PL_psig_pend, SIG_SIZE, int);
11375 }
11376 else {
11377 PL_psig_pend = (int*)NULL;
11378 }
05ec9bb3 11379
bd81e77b
NC
11380 if (proto_perl->Ipsig_ptr) {
11381 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
11382 Newxz(PL_psig_name, SIG_SIZE, SV*);
11383 for (i = 1; i < SIG_SIZE; i++) {
11384 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
11385 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
11386 }
11387 }
11388 else {
11389 PL_psig_ptr = (SV**)NULL;
11390 PL_psig_name = (SV**)NULL;
11391 }
05ec9bb3 11392
bd81e77b 11393 /* thrdvar.h stuff */
1d7c1841 11394
bd81e77b
NC
11395 if (flags & CLONEf_COPY_STACKS) {
11396 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
11397 PL_tmps_ix = proto_perl->Ttmps_ix;
11398 PL_tmps_max = proto_perl->Ttmps_max;
11399 PL_tmps_floor = proto_perl->Ttmps_floor;
11400 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
11401 i = 0;
11402 while (i <= PL_tmps_ix) {
11403 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Ttmps_stack[i], param);
11404 ++i;
11405 }
d2d73c3e 11406
bd81e77b
NC
11407 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
11408 i = proto_perl->Tmarkstack_max - proto_perl->Tmarkstack;
11409 Newxz(PL_markstack, i, I32);
11410 PL_markstack_max = PL_markstack + (proto_perl->Tmarkstack_max
11411 - proto_perl->Tmarkstack);
11412 PL_markstack_ptr = PL_markstack + (proto_perl->Tmarkstack_ptr
11413 - proto_perl->Tmarkstack);
11414 Copy(proto_perl->Tmarkstack, PL_markstack,
11415 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 11416
bd81e77b
NC
11417 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
11418 * NOTE: unlike the others! */
11419 PL_scopestack_ix = proto_perl->Tscopestack_ix;
11420 PL_scopestack_max = proto_perl->Tscopestack_max;
11421 Newxz(PL_scopestack, PL_scopestack_max, I32);
11422 Copy(proto_perl->Tscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 11423
bd81e77b
NC
11424 /* NOTE: si_dup() looks at PL_markstack */
11425 PL_curstackinfo = si_dup(proto_perl->Tcurstackinfo, param);
d2d73c3e 11426
bd81e77b
NC
11427 /* PL_curstack = PL_curstackinfo->si_stack; */
11428 PL_curstack = av_dup(proto_perl->Tcurstack, param);
11429 PL_mainstack = av_dup(proto_perl->Tmainstack, param);
1d7c1841 11430
bd81e77b
NC
11431 /* next PUSHs() etc. set *(PL_stack_sp+1) */
11432 PL_stack_base = AvARRAY(PL_curstack);
11433 PL_stack_sp = PL_stack_base + (proto_perl->Tstack_sp
11434 - proto_perl->Tstack_base);
11435 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 11436
bd81e77b
NC
11437 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
11438 * NOTE: unlike the others! */
11439 PL_savestack_ix = proto_perl->Tsavestack_ix;
11440 PL_savestack_max = proto_perl->Tsavestack_max;
11441 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
11442 PL_savestack = ss_dup(proto_perl, param);
11443 }
11444 else {
11445 init_stacks();
11446 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
11447
11448 /* although we're not duplicating the tmps stack, we should still
11449 * add entries for any SVs on the tmps stack that got cloned by a
11450 * non-refcount means (eg a temp in @_); otherwise they will be
11451 * orphaned
11452 */
11453 for (i = 0; i<= proto_perl->Ttmps_ix; i++) {
6136c704 11454 SV * const nsv = (SV*)ptr_table_fetch(PL_ptr_table,
34394ecd
DM
11455 proto_perl->Ttmps_stack[i]);
11456 if (nsv && !SvREFCNT(nsv)) {
11457 EXTEND_MORTAL(1);
b37c2d43 11458 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
34394ecd
DM
11459 }
11460 }
bd81e77b 11461 }
1d7c1841 11462
bd81e77b
NC
11463 PL_start_env = proto_perl->Tstart_env; /* XXXXXX */
11464 PL_top_env = &PL_start_env;
1d7c1841 11465
bd81e77b 11466 PL_op = proto_perl->Top;
4a4c6fe3 11467
a0714e2c 11468 PL_Sv = NULL;
bd81e77b
NC
11469 PL_Xpv = (XPV*)NULL;
11470 PL_na = proto_perl->Tna;
1fcf4c12 11471
bd81e77b
NC
11472 PL_statbuf = proto_perl->Tstatbuf;
11473 PL_statcache = proto_perl->Tstatcache;
11474 PL_statgv = gv_dup(proto_perl->Tstatgv, param);
11475 PL_statname = sv_dup_inc(proto_perl->Tstatname, param);
11476#ifdef HAS_TIMES
11477 PL_timesbuf = proto_perl->Ttimesbuf;
11478#endif
1d7c1841 11479
bd81e77b
NC
11480 PL_tainted = proto_perl->Ttainted;
11481 PL_curpm = proto_perl->Tcurpm; /* XXX No PMOP ref count */
11482 PL_rs = sv_dup_inc(proto_perl->Trs, param);
11483 PL_last_in_gv = gv_dup(proto_perl->Tlast_in_gv, param);
11484 PL_ofs_sv = sv_dup_inc(proto_perl->Tofs_sv, param);
11485 PL_defoutgv = gv_dup_inc(proto_perl->Tdefoutgv, param);
11486 PL_chopset = proto_perl->Tchopset; /* XXX never deallocated */
11487 PL_toptarget = sv_dup_inc(proto_perl->Ttoptarget, param);
11488 PL_bodytarget = sv_dup_inc(proto_perl->Tbodytarget, param);
11489 PL_formtarget = sv_dup(proto_perl->Tformtarget, param);
1d7c1841 11490
bd81e77b
NC
11491 PL_restartop = proto_perl->Trestartop;
11492 PL_in_eval = proto_perl->Tin_eval;
11493 PL_delaymagic = proto_perl->Tdelaymagic;
11494 PL_dirty = proto_perl->Tdirty;
11495 PL_localizing = proto_perl->Tlocalizing;
1d7c1841 11496
bd81e77b 11497 PL_errors = sv_dup_inc(proto_perl->Terrors, param);
4608196e 11498 PL_hv_fetch_ent_mh = NULL;
bd81e77b 11499 PL_modcount = proto_perl->Tmodcount;
5f66b61c 11500 PL_lastgotoprobe = NULL;
bd81e77b 11501 PL_dumpindent = proto_perl->Tdumpindent;
1d7c1841 11502
bd81e77b
NC
11503 PL_sortcop = (OP*)any_dup(proto_perl->Tsortcop, proto_perl);
11504 PL_sortstash = hv_dup(proto_perl->Tsortstash, param);
11505 PL_firstgv = gv_dup(proto_perl->Tfirstgv, param);
11506 PL_secondgv = gv_dup(proto_perl->Tsecondgv, param);
bd61b366 11507 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 11508 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 11509
bd81e77b 11510 /* regex stuff */
1d7c1841 11511
bd81e77b
NC
11512 PL_screamfirst = NULL;
11513 PL_screamnext = NULL;
11514 PL_maxscream = -1; /* reinits on demand */
a0714e2c 11515 PL_lastscream = NULL;
1d7c1841 11516
bd81e77b 11517 PL_watchaddr = NULL;
bd61b366 11518 PL_watchok = NULL;
1d7c1841 11519
bd81e77b 11520 PL_regdummy = proto_perl->Tregdummy;
bd81e77b
NC
11521 PL_colorset = 0; /* reinits PL_colors[] */
11522 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 11523
bd81e77b
NC
11524 /* RE engine - function pointers */
11525 PL_regcompp = proto_perl->Tregcompp;
11526 PL_regexecp = proto_perl->Tregexecp;
11527 PL_regint_start = proto_perl->Tregint_start;
11528 PL_regint_string = proto_perl->Tregint_string;
11529 PL_regfree = proto_perl->Tregfree;
46ab3289 11530 Zero(&PL_reg_state, 1, struct re_save_state);
bd81e77b 11531 PL_reginterp_cnt = 0;
5d9a96ca 11532 PL_regmatch_slab = NULL;
1d7c1841 11533
bd81e77b
NC
11534 /* Pluggable optimizer */
11535 PL_peepp = proto_perl->Tpeepp;
1d7c1841 11536
bd81e77b 11537 PL_stashcache = newHV();
1d7c1841 11538
bd81e77b
NC
11539 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
11540 ptr_table_free(PL_ptr_table);
11541 PL_ptr_table = NULL;
11542 }
1d7c1841 11543
bd81e77b
NC
11544 /* Call the ->CLONE method, if it exists, for each of the stashes
11545 identified by sv_dup() above.
11546 */
11547 while(av_len(param->stashes) != -1) {
11548 HV* const stash = (HV*) av_shift(param->stashes);
11549 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
11550 if (cloner && GvCV(cloner)) {
11551 dSP;
11552 ENTER;
11553 SAVETMPS;
11554 PUSHMARK(SP);
11555 XPUSHs(sv_2mortal(newSVhek(HvNAME_HEK(stash))));
11556 PUTBACK;
11557 call_sv((SV*)GvCV(cloner), G_DISCARD);
11558 FREETMPS;
11559 LEAVE;
11560 }
1d7c1841 11561 }
1d7c1841 11562
bd81e77b 11563 SvREFCNT_dec(param->stashes);
1d7c1841 11564
bd81e77b
NC
11565 /* orphaned? eg threads->new inside BEGIN or use */
11566 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 11567 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
11568 SAVEFREESV(PL_compcv);
11569 }
dd2155a4 11570
bd81e77b
NC
11571 return my_perl;
11572}
1d7c1841 11573
bd81e77b 11574#endif /* USE_ITHREADS */
1d7c1841 11575
bd81e77b
NC
11576/*
11577=head1 Unicode Support
1d7c1841 11578
bd81e77b 11579=for apidoc sv_recode_to_utf8
1d7c1841 11580
bd81e77b
NC
11581The encoding is assumed to be an Encode object, on entry the PV
11582of the sv is assumed to be octets in that encoding, and the sv
11583will be converted into Unicode (and UTF-8).
1d7c1841 11584
bd81e77b
NC
11585If the sv already is UTF-8 (or if it is not POK), or if the encoding
11586is not a reference, nothing is done to the sv. If the encoding is not
11587an C<Encode::XS> Encoding object, bad things will happen.
11588(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 11589
bd81e77b 11590The PV of the sv is returned.
1d7c1841 11591
bd81e77b 11592=cut */
1d7c1841 11593
bd81e77b
NC
11594char *
11595Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
11596{
11597 dVAR;
11598 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
11599 SV *uni;
11600 STRLEN len;
11601 const char *s;
11602 dSP;
11603 ENTER;
11604 SAVETMPS;
11605 save_re_context();
11606 PUSHMARK(sp);
11607 EXTEND(SP, 3);
11608 XPUSHs(encoding);
11609 XPUSHs(sv);
11610/*
11611 NI-S 2002/07/09
11612 Passing sv_yes is wrong - it needs to be or'ed set of constants
11613 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
11614 remove converted chars from source.
1d7c1841 11615
bd81e77b 11616 Both will default the value - let them.
1d7c1841 11617
bd81e77b
NC
11618 XPUSHs(&PL_sv_yes);
11619*/
11620 PUTBACK;
11621 call_method("decode", G_SCALAR);
11622 SPAGAIN;
11623 uni = POPs;
11624 PUTBACK;
11625 s = SvPV_const(uni, len);
11626 if (s != SvPVX_const(sv)) {
11627 SvGROW(sv, len + 1);
11628 Move(s, SvPVX(sv), len + 1, char);
11629 SvCUR_set(sv, len);
11630 }
11631 FREETMPS;
11632 LEAVE;
11633 SvUTF8_on(sv);
11634 return SvPVX(sv);
389edf32 11635 }
bd81e77b
NC
11636 return SvPOKp(sv) ? SvPVX(sv) : NULL;
11637}
1d7c1841 11638
bd81e77b
NC
11639/*
11640=for apidoc sv_cat_decode
1d7c1841 11641
bd81e77b
NC
11642The encoding is assumed to be an Encode object, the PV of the ssv is
11643assumed to be octets in that encoding and decoding the input starts
11644from the position which (PV + *offset) pointed to. The dsv will be
11645concatenated the decoded UTF-8 string from ssv. Decoding will terminate
11646when the string tstr appears in decoding output or the input ends on
11647the PV of the ssv. The value which the offset points will be modified
11648to the last input position on the ssv.
1d7c1841 11649
bd81e77b 11650Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 11651
bd81e77b
NC
11652=cut */
11653
11654bool
11655Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
11656 SV *ssv, int *offset, char *tstr, int tlen)
11657{
11658 dVAR;
11659 bool ret = FALSE;
11660 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
11661 SV *offsv;
11662 dSP;
11663 ENTER;
11664 SAVETMPS;
11665 save_re_context();
11666 PUSHMARK(sp);
11667 EXTEND(SP, 6);
11668 XPUSHs(encoding);
11669 XPUSHs(dsv);
11670 XPUSHs(ssv);
11671 XPUSHs(offsv = sv_2mortal(newSViv(*offset)));
11672 XPUSHs(sv_2mortal(newSVpvn(tstr, tlen)));
11673 PUTBACK;
11674 call_method("cat_decode", G_SCALAR);
11675 SPAGAIN;
11676 ret = SvTRUE(TOPs);
11677 *offset = SvIV(offsv);
11678 PUTBACK;
11679 FREETMPS;
11680 LEAVE;
389edf32 11681 }
bd81e77b
NC
11682 else
11683 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
11684 return ret;
1d7c1841 11685
bd81e77b 11686}
1d7c1841 11687
bd81e77b
NC
11688/* ---------------------------------------------------------------------
11689 *
11690 * support functions for report_uninit()
11691 */
1d7c1841 11692
bd81e77b
NC
11693/* the maxiumum size of array or hash where we will scan looking
11694 * for the undefined element that triggered the warning */
1d7c1841 11695
bd81e77b 11696#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 11697
bd81e77b
NC
11698/* Look for an entry in the hash whose value has the same SV as val;
11699 * If so, return a mortal copy of the key. */
1d7c1841 11700
bd81e77b
NC
11701STATIC SV*
11702S_find_hash_subscript(pTHX_ HV *hv, SV* val)
11703{
11704 dVAR;
11705 register HE **array;
11706 I32 i;
6c3182a5 11707
bd81e77b
NC
11708 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
11709 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 11710 return NULL;
6c3182a5 11711
bd81e77b 11712 array = HvARRAY(hv);
6c3182a5 11713
bd81e77b
NC
11714 for (i=HvMAX(hv); i>0; i--) {
11715 register HE *entry;
11716 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
11717 if (HeVAL(entry) != val)
11718 continue;
11719 if ( HeVAL(entry) == &PL_sv_undef ||
11720 HeVAL(entry) == &PL_sv_placeholder)
11721 continue;
11722 if (!HeKEY(entry))
a0714e2c 11723 return NULL;
bd81e77b
NC
11724 if (HeKLEN(entry) == HEf_SVKEY)
11725 return sv_mortalcopy(HeKEY_sv(entry));
11726 return sv_2mortal(newSVpvn(HeKEY(entry), HeKLEN(entry)));
11727 }
11728 }
a0714e2c 11729 return NULL;
bd81e77b 11730}
6c3182a5 11731
bd81e77b
NC
11732/* Look for an entry in the array whose value has the same SV as val;
11733 * If so, return the index, otherwise return -1. */
6c3182a5 11734
bd81e77b
NC
11735STATIC I32
11736S_find_array_subscript(pTHX_ AV *av, SV* val)
11737{
97aff369 11738 dVAR;
bd81e77b
NC
11739 SV** svp;
11740 I32 i;
11741 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
11742 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
11743 return -1;
57c6e6d2 11744
bd81e77b
NC
11745 svp = AvARRAY(av);
11746 for (i=AvFILLp(av); i>=0; i--) {
11747 if (svp[i] == val && svp[i] != &PL_sv_undef)
11748 return i;
11749 }
11750 return -1;
11751}
15a5279a 11752
bd81e77b
NC
11753/* S_varname(): return the name of a variable, optionally with a subscript.
11754 * If gv is non-zero, use the name of that global, along with gvtype (one
11755 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
11756 * targ. Depending on the value of the subscript_type flag, return:
11757 */
bce260cd 11758
bd81e77b
NC
11759#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
11760#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
11761#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
11762#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 11763
bd81e77b
NC
11764STATIC SV*
11765S_varname(pTHX_ GV *gv, const char gvtype, PADOFFSET targ,
11766 SV* keyname, I32 aindex, int subscript_type)
11767{
1d7c1841 11768
bd81e77b
NC
11769 SV * const name = sv_newmortal();
11770 if (gv) {
11771 char buffer[2];
11772 buffer[0] = gvtype;
11773 buffer[1] = 0;
1d7c1841 11774
bd81e77b 11775 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 11776
bd81e77b 11777 gv_fullname4(name, gv, buffer, 0);
1d7c1841 11778
bd81e77b
NC
11779 if ((unsigned int)SvPVX(name)[1] <= 26) {
11780 buffer[0] = '^';
11781 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 11782
bd81e77b
NC
11783 /* Swap the 1 unprintable control character for the 2 byte pretty
11784 version - ie substr($name, 1, 1) = $buffer; */
11785 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 11786 }
bd81e77b
NC
11787 }
11788 else {
11789 U32 unused;
11790 CV * const cv = find_runcv(&unused);
11791 SV *sv;
11792 AV *av;
1d7c1841 11793
bd81e77b 11794 if (!cv || !CvPADLIST(cv))
a0714e2c 11795 return NULL;
bd81e77b
NC
11796 av = (AV*)(*av_fetch(CvPADLIST(cv), 0, FALSE));
11797 sv = *av_fetch(av, targ, FALSE);
11798 /* SvLEN in a pad name is not to be trusted */
11799 sv_setpv(name, SvPV_nolen_const(sv));
11800 }
1d7c1841 11801
bd81e77b 11802 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 11803 SV * const sv = newSV(0);
bd81e77b
NC
11804 *SvPVX(name) = '$';
11805 Perl_sv_catpvf(aTHX_ name, "{%s}",
11806 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
11807 SvREFCNT_dec(sv);
11808 }
11809 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
11810 *SvPVX(name) = '$';
11811 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
11812 }
11813 else if (subscript_type == FUV_SUBSCRIPT_WITHIN)
89529cee 11814 Perl_sv_insert(aTHX_ name, 0, 0, STR_WITH_LEN("within "));
1d7c1841 11815
bd81e77b
NC
11816 return name;
11817}
1d7c1841 11818
1d7c1841 11819
bd81e77b
NC
11820/*
11821=for apidoc find_uninit_var
1d7c1841 11822
bd81e77b
NC
11823Find the name of the undefined variable (if any) that caused the operator o
11824to issue a "Use of uninitialized value" warning.
11825If match is true, only return a name if it's value matches uninit_sv.
11826So roughly speaking, if a unary operator (such as OP_COS) generates a
11827warning, then following the direct child of the op may yield an
11828OP_PADSV or OP_GV that gives the name of the undefined variable. On the
11829other hand, with OP_ADD there are two branches to follow, so we only print
11830the variable name if we get an exact match.
1d7c1841 11831
bd81e77b 11832The name is returned as a mortal SV.
1d7c1841 11833
bd81e77b
NC
11834Assumes that PL_op is the op that originally triggered the error, and that
11835PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 11836
bd81e77b
NC
11837=cut
11838*/
1d7c1841 11839
bd81e77b
NC
11840STATIC SV *
11841S_find_uninit_var(pTHX_ OP* obase, SV* uninit_sv, bool match)
11842{
11843 dVAR;
11844 SV *sv;
11845 AV *av;
11846 GV *gv;
11847 OP *o, *o2, *kid;
1d7c1841 11848
bd81e77b
NC
11849 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
11850 uninit_sv == &PL_sv_placeholder)))
a0714e2c 11851 return NULL;
1d7c1841 11852
bd81e77b 11853 switch (obase->op_type) {
1d7c1841 11854
bd81e77b
NC
11855 case OP_RV2AV:
11856 case OP_RV2HV:
11857 case OP_PADAV:
11858 case OP_PADHV:
11859 {
11860 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
11861 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
11862 I32 index = 0;
a0714e2c 11863 SV *keysv = NULL;
bd81e77b 11864 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 11865
bd81e77b
NC
11866 if (pad) { /* @lex, %lex */
11867 sv = PAD_SVl(obase->op_targ);
a0714e2c 11868 gv = NULL;
bd81e77b
NC
11869 }
11870 else {
11871 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
11872 /* @global, %global */
11873 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
11874 if (!gv)
11875 break;
11876 sv = hash ? (SV*)GvHV(gv): (SV*)GvAV(gv);
11877 }
11878 else /* @{expr}, %{expr} */
11879 return find_uninit_var(cUNOPx(obase)->op_first,
11880 uninit_sv, match);
11881 }
1d7c1841 11882
bd81e77b
NC
11883 /* attempt to find a match within the aggregate */
11884 if (hash) {
d4c19fe8 11885 keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
11886 if (keysv)
11887 subscript_type = FUV_SUBSCRIPT_HASH;
11888 }
11889 else {
11890 index = S_find_array_subscript(aTHX_ (AV*)sv, uninit_sv);
11891 if (index >= 0)
11892 subscript_type = FUV_SUBSCRIPT_ARRAY;
11893 }
1d7c1841 11894
bd81e77b
NC
11895 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
11896 break;
1d7c1841 11897
bd81e77b
NC
11898 return varname(gv, hash ? '%' : '@', obase->op_targ,
11899 keysv, index, subscript_type);
11900 }
1d7c1841 11901
bd81e77b
NC
11902 case OP_PADSV:
11903 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
11904 break;
a0714e2c
SS
11905 return varname(NULL, '$', obase->op_targ,
11906 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11907
bd81e77b
NC
11908 case OP_GVSV:
11909 gv = cGVOPx_gv(obase);
11910 if (!gv || (match && GvSV(gv) != uninit_sv))
11911 break;
a0714e2c 11912 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 11913
bd81e77b
NC
11914 case OP_AELEMFAST:
11915 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
11916 if (match) {
11917 SV **svp;
11918 av = (AV*)PAD_SV(obase->op_targ);
11919 if (!av || SvRMAGICAL(av))
11920 break;
11921 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11922 if (!svp || *svp != uninit_sv)
11923 break;
11924 }
a0714e2c
SS
11925 return varname(NULL, '$', obase->op_targ,
11926 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11927 }
11928 else {
11929 gv = cGVOPx_gv(obase);
11930 if (!gv)
11931 break;
11932 if (match) {
11933 SV **svp;
11934 av = GvAV(gv);
11935 if (!av || SvRMAGICAL(av))
11936 break;
11937 svp = av_fetch(av, (I32)obase->op_private, FALSE);
11938 if (!svp || *svp != uninit_sv)
11939 break;
11940 }
11941 return varname(gv, '$', 0,
a0714e2c 11942 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
11943 }
11944 break;
1d7c1841 11945
bd81e77b
NC
11946 case OP_EXISTS:
11947 o = cUNOPx(obase)->op_first;
11948 if (!o || o->op_type != OP_NULL ||
11949 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
11950 break;
11951 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 11952
bd81e77b
NC
11953 case OP_AELEM:
11954 case OP_HELEM:
11955 if (PL_op == obase)
11956 /* $a[uninit_expr] or $h{uninit_expr} */
11957 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 11958
a0714e2c 11959 gv = NULL;
bd81e77b
NC
11960 o = cBINOPx(obase)->op_first;
11961 kid = cBINOPx(obase)->op_last;
8cf8f3d1 11962
bd81e77b 11963 /* get the av or hv, and optionally the gv */
a0714e2c 11964 sv = NULL;
bd81e77b
NC
11965 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
11966 sv = PAD_SV(o->op_targ);
11967 }
11968 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
11969 && cUNOPo->op_first->op_type == OP_GV)
11970 {
11971 gv = cGVOPx_gv(cUNOPo->op_first);
11972 if (!gv)
11973 break;
11974 sv = o->op_type == OP_RV2HV ? (SV*)GvHV(gv) : (SV*)GvAV(gv);
11975 }
11976 if (!sv)
11977 break;
11978
11979 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
11980 /* index is constant */
11981 if (match) {
11982 if (SvMAGICAL(sv))
11983 break;
11984 if (obase->op_type == OP_HELEM) {
11985 HE* he = hv_fetch_ent((HV*)sv, cSVOPx_sv(kid), 0, 0);
11986 if (!he || HeVAL(he) != uninit_sv)
11987 break;
11988 }
11989 else {
00b6aa41 11990 SV * const * const svp = av_fetch((AV*)sv, SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
11991 if (!svp || *svp != uninit_sv)
11992 break;
11993 }
11994 }
11995 if (obase->op_type == OP_HELEM)
11996 return varname(gv, '%', o->op_targ,
11997 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
11998 else
a0714e2c 11999 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 12000 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12001 }
12002 else {
12003 /* index is an expression;
12004 * attempt to find a match within the aggregate */
12005 if (obase->op_type == OP_HELEM) {
d4c19fe8 12006 SV * const keysv = find_hash_subscript((HV*)sv, uninit_sv);
bd81e77b
NC
12007 if (keysv)
12008 return varname(gv, '%', o->op_targ,
12009 keysv, 0, FUV_SUBSCRIPT_HASH);
12010 }
12011 else {
d4c19fe8 12012 const I32 index = find_array_subscript((AV*)sv, uninit_sv);
bd81e77b
NC
12013 if (index >= 0)
12014 return varname(gv, '@', o->op_targ,
a0714e2c 12015 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12016 }
12017 if (match)
12018 break;
12019 return varname(gv,
12020 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12021 ? '@' : '%',
a0714e2c 12022 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 12023 }
bd81e77b 12024 break;
dc507217 12025
bd81e77b
NC
12026 case OP_AASSIGN:
12027 /* only examine RHS */
12028 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 12029
bd81e77b
NC
12030 case OP_OPEN:
12031 o = cUNOPx(obase)->op_first;
12032 if (o->op_type == OP_PUSHMARK)
12033 o = o->op_sibling;
1d7c1841 12034
bd81e77b
NC
12035 if (!o->op_sibling) {
12036 /* one-arg version of open is highly magical */
a0ae6670 12037
bd81e77b
NC
12038 if (o->op_type == OP_GV) { /* open FOO; */
12039 gv = cGVOPx_gv(o);
12040 if (match && GvSV(gv) != uninit_sv)
12041 break;
12042 return varname(gv, '$', 0,
a0714e2c 12043 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12044 }
12045 /* other possibilities not handled are:
12046 * open $x; or open my $x; should return '${*$x}'
12047 * open expr; should return '$'.expr ideally
12048 */
12049 break;
12050 }
12051 goto do_op;
ccfc67b7 12052
bd81e77b
NC
12053 /* ops where $_ may be an implicit arg */
12054 case OP_TRANS:
12055 case OP_SUBST:
12056 case OP_MATCH:
12057 if ( !(obase->op_flags & OPf_STACKED)) {
12058 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12059 ? PAD_SVl(obase->op_targ)
12060 : DEFSV))
12061 {
12062 sv = sv_newmortal();
12063 sv_setpvn(sv, "$_", 2);
12064 return sv;
12065 }
12066 }
12067 goto do_op;
9f4817db 12068
bd81e77b
NC
12069 case OP_PRTF:
12070 case OP_PRINT:
12071 /* skip filehandle as it can't produce 'undef' warning */
12072 o = cUNOPx(obase)->op_first;
12073 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12074 o = o->op_sibling->op_sibling;
12075 goto do_op2;
9f4817db 12076
9f4817db 12077
bd81e77b
NC
12078 case OP_RV2SV:
12079 case OP_CUSTOM:
12080 case OP_ENTERSUB:
12081 match = 1; /* XS or custom code could trigger random warnings */
12082 goto do_op;
9f4817db 12083
bd81e77b
NC
12084 case OP_SCHOMP:
12085 case OP_CHOMP:
12086 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
396482e1 12087 return sv_2mortal(newSVpvs("${$/}"));
5f66b61c 12088 /*FALLTHROUGH*/
5d170f3a 12089
bd81e77b
NC
12090 default:
12091 do_op:
12092 if (!(obase->op_flags & OPf_KIDS))
12093 break;
12094 o = cUNOPx(obase)->op_first;
12095
12096 do_op2:
12097 if (!o)
12098 break;
f9893866 12099
bd81e77b
NC
12100 /* if all except one arg are constant, or have no side-effects,
12101 * or are optimized away, then it's unambiguous */
5f66b61c 12102 o2 = NULL;
bd81e77b
NC
12103 for (kid=o; kid; kid = kid->op_sibling) {
12104 if (kid &&
30c8d9e4 12105 ( (kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid)))
bd81e77b
NC
12106 || (kid->op_type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
12107 || (kid->op_type == OP_PUSHMARK)
12108 )
12109 )
12110 continue;
12111 if (o2) { /* more than one found */
5f66b61c 12112 o2 = NULL;
bd81e77b
NC
12113 break;
12114 }
12115 o2 = kid;
12116 }
12117 if (o2)
12118 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 12119
bd81e77b
NC
12120 /* scan all args */
12121 while (o) {
12122 sv = find_uninit_var(o, uninit_sv, 1);
12123 if (sv)
12124 return sv;
12125 o = o->op_sibling;
d0063567 12126 }
bd81e77b 12127 break;
f9893866 12128 }
a0714e2c 12129 return NULL;
9f4817db
JH
12130}
12131
220e2d4e 12132
bd81e77b
NC
12133/*
12134=for apidoc report_uninit
68795e93 12135
bd81e77b 12136Print appropriate "Use of uninitialized variable" warning
220e2d4e 12137
bd81e77b
NC
12138=cut
12139*/
220e2d4e 12140
bd81e77b
NC
12141void
12142Perl_report_uninit(pTHX_ SV* uninit_sv)
220e2d4e 12143{
97aff369 12144 dVAR;
bd81e77b 12145 if (PL_op) {
a0714e2c 12146 SV* varname = NULL;
bd81e77b
NC
12147 if (uninit_sv) {
12148 varname = find_uninit_var(PL_op, uninit_sv,0);
12149 if (varname)
12150 sv_insert(varname, 0, 0, " ", 1);
12151 }
12152 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12153 varname ? SvPV_nolen_const(varname) : "",
12154 " in ", OP_DESC(PL_op));
220e2d4e 12155 }
a73e8557 12156 else
bd81e77b
NC
12157 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
12158 "", "", "");
220e2d4e 12159}
f9893866 12160
241d1a3b
NC
12161/*
12162 * Local variables:
12163 * c-indentation-style: bsd
12164 * c-basic-offset: 4
12165 * indent-tabs-mode: t
12166 * End:
12167 *
37442d52
RGS
12168 * ex: set ts=8 sts=4 sw=4 noet:
12169 */