This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
The logic to use SVt_NV or SVt_PVNV is in sv_upgrade, so no need to
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
241d1a3b 4 * 2000, 2001, 2002, 2003, 2004, 2005, 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
JH
32#ifdef PERL_UTF8_CACHE_ASSERT
33/* The cache element 0 is the Unicode offset;
34 * the cache element 1 is the byte offset of the element 0;
35 * the cache element 2 is the Unicode length of the substring;
36 * the cache element 3 is the byte length of the substring;
37 * The checking of the substring side would be good
38 * but substr() has enough code paths to make my head spin;
39 * if adding more checks watch out for the following tests:
40 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
41 * lib/utf8.t lib/Unicode/Collate/t/index.t
42 * --jhi
43 */
44#define ASSERT_UTF8_CACHE(cache) \
45 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); } } STMT_END
46#else
47#define ASSERT_UTF8_CACHE(cache) NOOP
48#endif
49
f8c7b90f 50#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 51#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 52#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 53/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 54 on-write. */
765f542d 55#endif
645c22ef
DM
56
57/* ============================================================================
58
59=head1 Allocation and deallocation of SVs.
60
5e045b90
AMS
61An SV (or AV, HV, etc.) is allocated in two parts: the head (struct sv,
62av, hv...) contains type and reference count information, as well as a
63pointer to the body (struct xrv, xpv, xpviv...), which contains fields
64specific to each type.
65
93e68bfb
JC
66In all but the most memory-paranoid configuations (ex: PURIFY), this
67allocation is done using arenas, which by default are approximately 4K
68chunks of memory parcelled up into N heads or bodies (of same size).
69Sv-bodies are allocated by their sv-type, guaranteeing size
70consistency needed to allocate safely from arrays.
71
72The first slot in each arena is reserved, and is used to hold a link
73to the next arena. In the case of heads, the unused first slot also
74contains some flags and a note of the number of slots. Snaked through
75each arena chain is a linked list of free items; when this becomes
76empty, an extra arena is allocated and divided up into N items which
77are threaded into the free list.
645c22ef
DM
78
79The following global variables are associated with arenas:
80
81 PL_sv_arenaroot pointer to list of SV arenas
82 PL_sv_root pointer to list of free SV structures
83
93e68bfb
JC
84 PL_body_arenaroots[] array of pointers to list of arenas, 1 per svtype
85 PL_body_roots[] array of pointers to list of free bodies of svtype
86 arrays are indexed by the svtype needed
645c22ef 87
93e68bfb
JC
88Note that some of the larger and more rarely used body types (eg
89xpvio) are not allocated using arenas, but are instead just
90malloc()/free()ed as required.
91
92In addition, a few SV heads are not allocated from an arena, but are
93instead directly created as static or auto variables, eg PL_sv_undef.
94The size of arenas can be changed from the default by setting
95PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
96
97The SV arena serves the secondary purpose of allowing still-live SVs
98to be located and destroyed during final cleanup.
99
100At the lowest level, the macros new_SV() and del_SV() grab and free
101an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
102to return the SV to the free list with error checking.) new_SV() calls
103more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
104SVs in the free list have their SvTYPE field set to all ones.
105
106Similarly, there are macros new_XIV()/del_XIV(), new_XNV()/del_XNV() etc
107that allocate and return individual body types. Normally these are mapped
ff276b08
RG
108to the arena-manipulating functions new_xiv()/del_xiv() etc, but may be
109instead mapped directly to malloc()/free() if PURIFY is defined. The
645c22ef
DM
110new/del functions remove from, or add to, the appropriate PL_foo_root
111list, and call more_xiv() etc to add a new arena if the list is empty.
112
ff276b08 113At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 114perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 115start of the interpreter.
645c22ef
DM
116
117Manipulation of any of the PL_*root pointers is protected by enclosing
118LOCK_SV_MUTEX; ... UNLOCK_SV_MUTEX calls which should Do the Right Thing
119if threads are enabled.
120
121The function visit() scans the SV arenas list, and calls a specified
122function for each SV it finds which is still live - ie which has an SvTYPE
123other than all 1's, and a non-zero SvREFCNT. visit() is used by the
124following functions (specified as [function that calls visit()] / [function
125called by visit() for each SV]):
126
127 sv_report_used() / do_report_used()
128 dump all remaining SVs (debugging aid)
129
130 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
131 Attempt to free all objects pointed to by RVs,
132 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
133 try to do the same for all objects indirectly
134 referenced by typeglobs too. Called once from
135 perl_destruct(), prior to calling sv_clean_all()
136 below.
137
138 sv_clean_all() / do_clean_all()
139 SvREFCNT_dec(sv) each remaining SV, possibly
140 triggering an sv_free(). It also sets the
141 SVf_BREAK flag on the SV to indicate that the
142 refcnt has been artificially lowered, and thus
143 stopping sv_free() from giving spurious warnings
144 about SVs which unexpectedly have a refcnt
145 of zero. called repeatedly from perl_destruct()
146 until there are no SVs left.
147
93e68bfb 148=head2 Arena allocator API Summary
645c22ef
DM
149
150Private API to rest of sv.c
151
152 new_SV(), del_SV(),
153
154 new_XIV(), del_XIV(),
155 new_XNV(), del_XNV(),
156 etc
157
158Public API:
159
8cf8f3d1 160 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef
DM
161
162
163=cut
164
165============================================================================ */
166
167
51371543 168
4561caa4
CS
169/*
170 * "A time to plant, and a time to uproot what was planted..."
171 */
172
77354fb4
NC
173/*
174 * nice_chunk and nice_chunk size need to be set
175 * and queried under the protection of sv_mutex
176 */
177void
178Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
179{
180 void *new_chunk;
181 U32 new_chunk_size;
182 LOCK_SV_MUTEX;
183 new_chunk = (void *)(chunk);
184 new_chunk_size = (chunk_size);
185 if (new_chunk_size > PL_nice_chunk_size) {
186 Safefree(PL_nice_chunk);
187 PL_nice_chunk = (char *) new_chunk;
188 PL_nice_chunk_size = new_chunk_size;
189 } else {
190 Safefree(chunk);
191 }
192 UNLOCK_SV_MUTEX;
193}
cac9b346 194
fd0854ff 195#ifdef DEBUG_LEAKING_SCALARS
22162ca8 196# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
197#else
198# define FREE_SV_DEBUG_FILE(sv)
199#endif
200
48614a46
NC
201#ifdef PERL_POISON
202# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
203/* Whilst I'd love to do this, it seems that things like to check on
204 unreferenced scalars
205# define POSION_SV_HEAD(sv) Poison(sv, 1, struct STRUCT_SV)
206*/
207# define POSION_SV_HEAD(sv) Poison(&SvANY(sv), 1, void *), \
208 Poison(&SvREFCNT(sv), 1, U32)
209#else
210# define SvARENA_CHAIN(sv) SvANY(sv)
211# define POSION_SV_HEAD(sv)
212#endif
213
053fc874
GS
214#define plant_SV(p) \
215 STMT_START { \
fd0854ff 216 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
217 POSION_SV_HEAD(p); \
218 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
219 SvFLAGS(p) = SVTYPEMASK; \
220 PL_sv_root = (p); \
221 --PL_sv_count; \
222 } STMT_END
a0d0e21e 223
fba3b22e 224/* sv_mutex must be held while calling uproot_SV() */
053fc874
GS
225#define uproot_SV(p) \
226 STMT_START { \
227 (p) = PL_sv_root; \
48614a46 228 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
229 ++PL_sv_count; \
230 } STMT_END
231
645c22ef 232
cac9b346
NC
233/* make some more SVs by adding another arena */
234
235/* sv_mutex must be held while calling more_sv() */
236STATIC SV*
237S_more_sv(pTHX)
238{
239 SV* sv;
240
241 if (PL_nice_chunk) {
242 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
243 PL_nice_chunk = Nullch;
244 PL_nice_chunk_size = 0;
245 }
246 else {
247 char *chunk; /* must use New here to match call to */
a02a5408 248 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 249 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
250 }
251 uproot_SV(sv);
252 return sv;
253}
254
645c22ef
DM
255/* new_SV(): return a new, empty SV head */
256
eba0f806
DM
257#ifdef DEBUG_LEAKING_SCALARS
258/* provide a real function for a debugger to play with */
259STATIC SV*
260S_new_SV(pTHX)
261{
262 SV* sv;
263
264 LOCK_SV_MUTEX;
265 if (PL_sv_root)
266 uproot_SV(sv);
267 else
cac9b346 268 sv = S_more_sv(aTHX);
eba0f806
DM
269 UNLOCK_SV_MUTEX;
270 SvANY(sv) = 0;
271 SvREFCNT(sv) = 1;
272 SvFLAGS(sv) = 0;
fd0854ff
DM
273 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
274 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
275 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_cloned = 0;
fd0854ff 278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 279
eba0f806
DM
280 return sv;
281}
282# define new_SV(p) (p)=S_new_SV(aTHX)
283
284#else
285# define new_SV(p) \
053fc874
GS
286 STMT_START { \
287 LOCK_SV_MUTEX; \
288 if (PL_sv_root) \
289 uproot_SV(p); \
290 else \
cac9b346 291 (p) = S_more_sv(aTHX); \
053fc874
GS
292 UNLOCK_SV_MUTEX; \
293 SvANY(p) = 0; \
294 SvREFCNT(p) = 1; \
295 SvFLAGS(p) = 0; \
296 } STMT_END
eba0f806 297#endif
463ee0b2 298
645c22ef
DM
299
300/* del_SV(): return an empty SV head to the free list */
301
a0d0e21e 302#ifdef DEBUGGING
4561caa4 303
053fc874
GS
304#define del_SV(p) \
305 STMT_START { \
306 LOCK_SV_MUTEX; \
aea4f609 307 if (DEBUG_D_TEST) \
053fc874
GS
308 del_sv(p); \
309 else \
310 plant_SV(p); \
311 UNLOCK_SV_MUTEX; \
312 } STMT_END
a0d0e21e 313
76e3520e 314STATIC void
cea2e8a9 315S_del_sv(pTHX_ SV *p)
463ee0b2 316{
aea4f609 317 if (DEBUG_D_TEST) {
4633a7c4 318 SV* sva;
a3b680e6 319 bool ok = 0;
3280af22 320 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
321 const SV * const sv = sva + 1;
322 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 323 if (p >= sv && p < svend) {
a0d0e21e 324 ok = 1;
c0ff570e
NC
325 break;
326 }
a0d0e21e
LW
327 }
328 if (!ok) {
0453d815 329 if (ckWARN_d(WARN_INTERNAL))
9014280d 330 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
331 "Attempt to free non-arena SV: 0x%"UVxf
332 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
333 return;
334 }
335 }
4561caa4 336 plant_SV(p);
463ee0b2 337}
a0d0e21e 338
4561caa4
CS
339#else /* ! DEBUGGING */
340
341#define del_SV(p) plant_SV(p)
342
343#endif /* DEBUGGING */
463ee0b2 344
645c22ef
DM
345
346/*
ccfc67b7
JH
347=head1 SV Manipulation Functions
348
645c22ef
DM
349=for apidoc sv_add_arena
350
351Given a chunk of memory, link it to the head of the list of arenas,
352and split it into a list of free SVs.
353
354=cut
355*/
356
4633a7c4 357void
864dbfa3 358Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 359{
0bd48802 360 SV* const sva = (SV*)ptr;
463ee0b2
LW
361 register SV* sv;
362 register SV* svend;
4633a7c4
LW
363
364 /* The first SV in an arena isn't an SV. */
3280af22 365 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
366 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
367 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
368
3280af22
NIS
369 PL_sv_arenaroot = sva;
370 PL_sv_root = sva + 1;
4633a7c4
LW
371
372 svend = &sva[SvREFCNT(sva) - 1];
373 sv = sva + 1;
463ee0b2 374 while (sv < svend) {
48614a46 375 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 376#ifdef DEBUGGING
978b032e 377 SvREFCNT(sv) = 0;
03e36789
NC
378#endif
379 /* Must always set typemask because it's awlays checked in on cleanup
380 when the arenas are walked looking for objects. */
8990e307 381 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
382 sv++;
383 }
48614a46 384 SvARENA_CHAIN(sv) = 0;
03e36789
NC
385#ifdef DEBUGGING
386 SvREFCNT(sv) = 0;
387#endif
4633a7c4
LW
388 SvFLAGS(sv) = SVTYPEMASK;
389}
390
055972dc
DM
391/* visit(): call the named function for each non-free SV in the arenas
392 * whose flags field matches the flags/mask args. */
645c22ef 393
5226ed68 394STATIC I32
055972dc 395S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 396{
4633a7c4 397 SV* sva;
5226ed68 398 I32 visited = 0;
8990e307 399
3280af22 400 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 401 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 402 register SV* sv;
4561caa4 403 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
404 if (SvTYPE(sv) != SVTYPEMASK
405 && (sv->sv_flags & mask) == flags
406 && SvREFCNT(sv))
407 {
acfe0abc 408 (FCALL)(aTHX_ sv);
5226ed68
JH
409 ++visited;
410 }
8990e307
LW
411 }
412 }
5226ed68 413 return visited;
8990e307
LW
414}
415
758a08c3
JH
416#ifdef DEBUGGING
417
645c22ef
DM
418/* called by sv_report_used() for each live SV */
419
420static void
acfe0abc 421do_report_used(pTHX_ SV *sv)
645c22ef
DM
422{
423 if (SvTYPE(sv) != SVTYPEMASK) {
424 PerlIO_printf(Perl_debug_log, "****\n");
425 sv_dump(sv);
426 }
427}
758a08c3 428#endif
645c22ef
DM
429
430/*
431=for apidoc sv_report_used
432
433Dump the contents of all SVs not yet freed. (Debugging aid).
434
435=cut
436*/
437
8990e307 438void
864dbfa3 439Perl_sv_report_used(pTHX)
4561caa4 440{
ff270d3a 441#ifdef DEBUGGING
055972dc 442 visit(do_report_used, 0, 0);
ff270d3a 443#endif
4561caa4
CS
444}
445
645c22ef
DM
446/* called by sv_clean_objs() for each live SV */
447
448static void
e15faf7d 449do_clean_objs(pTHX_ SV *ref)
645c22ef 450{
823a54a3
AL
451 if (SvROK(ref)) {
452 SV * const target = SvRV(ref);
453 if (SvOBJECT(target)) {
454 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
455 if (SvWEAKREF(ref)) {
456 sv_del_backref(target, ref);
457 SvWEAKREF_off(ref);
458 SvRV_set(ref, NULL);
459 } else {
460 SvROK_off(ref);
461 SvRV_set(ref, NULL);
462 SvREFCNT_dec(target);
463 }
645c22ef
DM
464 }
465 }
466
467 /* XXX Might want to check arrays, etc. */
468}
469
470/* called by sv_clean_objs() for each live SV */
471
472#ifndef DISABLE_DESTRUCTOR_KLUDGE
473static void
acfe0abc 474do_clean_named_objs(pTHX_ SV *sv)
645c22ef
DM
475{
476 if (SvTYPE(sv) == SVt_PVGV && GvGP(sv)) {
c69033f2
NC
477 if ((
478#ifdef PERL_DONT_CREATE_GVSV
479 GvSV(sv) &&
480#endif
481 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
482 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
483 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
484 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
485 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
486 {
487 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 488 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
489 SvREFCNT_dec(sv);
490 }
491 }
492}
493#endif
494
495/*
496=for apidoc sv_clean_objs
497
498Attempt to destroy all objects not yet freed
499
500=cut
501*/
502
4561caa4 503void
864dbfa3 504Perl_sv_clean_objs(pTHX)
4561caa4 505{
3280af22 506 PL_in_clean_objs = TRUE;
055972dc 507 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 508#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 509 /* some barnacles may yet remain, clinging to typeglobs */
055972dc 510 visit(do_clean_named_objs, SVt_PVGV, SVTYPEMASK);
4561caa4 511#endif
3280af22 512 PL_in_clean_objs = FALSE;
4561caa4
CS
513}
514
645c22ef
DM
515/* called by sv_clean_all() for each live SV */
516
517static void
acfe0abc 518do_clean_all(pTHX_ SV *sv)
645c22ef
DM
519{
520 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
521 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b
DM
522 if (PL_comppad == (AV*)sv) {
523 PL_comppad = Nullav;
524 PL_curpad = Null(SV**);
525 }
645c22ef
DM
526 SvREFCNT_dec(sv);
527}
528
529/*
530=for apidoc sv_clean_all
531
532Decrement the refcnt of each remaining SV, possibly triggering a
533cleanup. This function may have to be called multiple times to free
ff276b08 534SVs which are in complex self-referential hierarchies.
645c22ef
DM
535
536=cut
537*/
538
5226ed68 539I32
864dbfa3 540Perl_sv_clean_all(pTHX)
8990e307 541{
5226ed68 542 I32 cleaned;
3280af22 543 PL_in_clean_all = TRUE;
055972dc 544 cleaned = visit(do_clean_all, 0,0);
3280af22 545 PL_in_clean_all = FALSE;
5226ed68 546 return cleaned;
8990e307 547}
463ee0b2 548
7cfef17e
NC
549static void
550S_free_arena(pTHX_ void **root) {
551 while (root) {
1b6737cc 552 void ** const next = *(void **)root;
7cfef17e
NC
553 Safefree(root);
554 root = next;
555 }
556}
557
645c22ef
DM
558/*
559=for apidoc sv_free_arenas
560
561Deallocate the memory used by all arenas. Note that all the individual SV
562heads and bodies within the arenas must already have been freed.
563
564=cut
565*/
7cfef17e
NC
566#define free_arena(name) \
567 STMT_START { \
568 S_free_arena(aTHX_ (void**) PL_ ## name ## _arenaroot); \
569 PL_ ## name ## _arenaroot = 0; \
570 PL_ ## name ## _root = 0; \
571 } STMT_END
572
4633a7c4 573void
864dbfa3 574Perl_sv_free_arenas(pTHX)
4633a7c4
LW
575{
576 SV* sva;
577 SV* svanext;
93e68bfb 578 int i;
4633a7c4
LW
579
580 /* Free arenas here, but be careful about fake ones. (We assume
581 contiguity of the fake ones with the corresponding real ones.) */
582
3280af22 583 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
584 svanext = (SV*) SvANY(sva);
585 while (svanext && SvFAKE(svanext))
586 svanext = (SV*) SvANY(svanext);
587
588 if (!SvFAKE(sva))
1df70142 589 Safefree(sva);
4633a7c4 590 }
93e68bfb
JC
591
592 for (i=0; i<SVt_LAST; i++) {
593 S_free_arena(aTHX_ (void**) PL_body_arenaroots[i]);
594 PL_body_arenaroots[i] = 0;
595 PL_body_roots[i] = 0;
596 }
597
43c5f42d 598 Safefree(PL_nice_chunk);
3280af22
NIS
599 PL_nice_chunk = Nullch;
600 PL_nice_chunk_size = 0;
601 PL_sv_arenaroot = 0;
602 PL_sv_root = 0;
4633a7c4
LW
603}
604
bd81e77b
NC
605/*
606 Here are mid-level routines that manage the allocation of bodies out
607 of the various arenas. There are 5 kinds of arenas:
29489e7c 608
bd81e77b
NC
609 1. SV-head arenas, which are discussed and handled above
610 2. regular body arenas
611 3. arenas for reduced-size bodies
612 4. Hash-Entry arenas
613 5. pte arenas (thread related)
29489e7c 614
bd81e77b
NC
615 Arena types 2 & 3 are chained by body-type off an array of
616 arena-root pointers, which is indexed by svtype. Some of the
617 larger/less used body types are malloced singly, since a large
618 unused block of them is wasteful. Also, several svtypes dont have
619 bodies; the data fits into the sv-head itself. The arena-root
620 pointer thus has a few unused root-pointers (which may be hijacked
621 later for arena types 4,5)
29489e7c 622
bd81e77b
NC
623 3 differs from 2 as an optimization; some body types have several
624 unused fields in the front of the structure (which are kept in-place
625 for consistency). These bodies can be allocated in smaller chunks,
626 because the leading fields arent accessed. Pointers to such bodies
627 are decremented to point at the unused 'ghost' memory, knowing that
628 the pointers are used with offsets to the real memory.
29489e7c 629
bd81e77b
NC
630 HE, HEK arenas are managed separately, with separate code, but may
631 be merge-able later..
632
633 PTE arenas are not sv-bodies, but they share these mid-level
634 mechanics, so are considered here. The new mid-level mechanics rely
635 on the sv_type of the body being allocated, so we just reserve one
636 of the unused body-slots for PTEs, then use it in those (2) PTE
637 contexts below (line ~10k)
638*/
639
640STATIC void *
641S_more_bodies (pTHX_ size_t size, svtype sv_type)
29489e7c 642{
00b6aa41
AL
643 void ** const arena_root = &PL_body_arenaroots[sv_type];
644 void ** const root = &PL_body_roots[sv_type];
bd81e77b
NC
645 char *start;
646 const char *end;
647 const size_t count = PERL_ARENA_SIZE / size;
29489e7c 648
bd81e77b
NC
649 Newx(start, count*size, char);
650 *((void **) start) = *arena_root;
651 *arena_root = (void *)start;
29489e7c 652
bd81e77b 653 end = start + (count-1) * size;
29489e7c 654
bd81e77b
NC
655 /* The initial slot is used to link the arenas together, so it isn't to be
656 linked into the list of ready-to-use bodies. */
29489e7c 657
bd81e77b 658 start += size;
29489e7c 659
bd81e77b 660 *root = (void *)start;
29489e7c 661
bd81e77b
NC
662 while (start < end) {
663 char * const next = start + size;
664 *(void**) start = (void *)next;
665 start = next;
29489e7c 666 }
bd81e77b
NC
667 *(void **)start = 0;
668
669 return *root;
29489e7c
DM
670}
671
bd81e77b 672/* grab a new thing from the free list, allocating more if necessary */
29489e7c 673
bd81e77b 674/* 1st, the inline version */
29489e7c 675
bd81e77b
NC
676#define new_body_inline(xpv, size, sv_type) \
677 STMT_START { \
00b6aa41 678 void ** const r3wt = &PL_body_roots[sv_type]; \
bd81e77b
NC
679 LOCK_SV_MUTEX; \
680 xpv = *((void **)(r3wt)) \
681 ? *((void **)(r3wt)) : S_more_bodies(aTHX_ size, sv_type); \
682 *(r3wt) = *(void**)(xpv); \
683 UNLOCK_SV_MUTEX; \
684 } STMT_END
29489e7c 685
bd81e77b 686/* now use the inline version in the proper function */
29489e7c 687
bd81e77b 688#ifndef PURIFY
9393da09 689
bd81e77b
NC
690/* This isn't being used with -DPURIFY, so don't declare it. Otherwise
691 compilers issue warnings. */
9393da09 692
bd81e77b
NC
693STATIC void *
694S_new_body(pTHX_ size_t size, svtype sv_type)
695{
696 void *xpv;
697 new_body_inline(xpv, size, sv_type);
698 return xpv;
699}
9393da09 700
bd81e77b 701#endif
53c1dcc0 702
bd81e77b 703/* return a thing to the free list */
29489e7c 704
bd81e77b
NC
705#define del_body(thing, root) \
706 STMT_START { \
00b6aa41 707 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
708 LOCK_SV_MUTEX; \
709 *thing_copy = *root; \
710 *root = (void*)thing_copy; \
711 UNLOCK_SV_MUTEX; \
712 } STMT_END
29489e7c 713
bd81e77b
NC
714/*
715 Revisiting type 3 arenas, there are 4 body-types which have some
716 members that are never accessed. They are XPV, XPVIV, XPVAV,
717 XPVHV, which have corresponding types: xpv_allocated,
718 xpviv_allocated, xpvav_allocated, xpvhv_allocated,
29489e7c 719
bd81e77b
NC
720 For these types, the arenas are carved up into *_allocated size
721 chunks, we thus avoid wasted memory for those unaccessed members.
722 When bodies are allocated, we adjust the pointer back in memory by
723 the size of the bit not allocated, so it's as if we allocated the
724 full structure. (But things will all go boom if you write to the
725 part that is "not there", because you'll be overwriting the last
726 members of the preceding structure in memory.)
29489e7c 727
bd81e77b
NC
728 We calculate the correction using the STRUCT_OFFSET macro. For example, if
729 xpv_allocated is the same structure as XPV then the two OFFSETs sum to zero,
730 and the pointer is unchanged. If the allocated structure is smaller (no
731 initial NV actually allocated) then the net effect is to subtract the size
732 of the NV from the pointer, to return a new pointer as if an initial NV were
733 actually allocated.
29489e7c 734
bd81e77b
NC
735 This is the same trick as was used for NV and IV bodies. Ironically it
736 doesn't need to be used for NV bodies any more, because NV is now at the
737 start of the structure. IV bodies don't need it either, because they are
738 no longer allocated. */
29489e7c 739
bd81e77b
NC
740/* The following 2 arrays hide the above details in a pair of
741 lookup-tables, allowing us to be body-type agnostic.
29489e7c 742
bd81e77b
NC
743 size maps svtype to its body's allocated size.
744 offset maps svtype to the body-pointer adjustment needed
29489e7c 745
bd81e77b
NC
746 NB: elements in latter are 0 or <0, and are added during
747 allocation, and subtracted during deallocation. It may be clearer
748 to invert the values, and call it shrinkage_by_svtype.
29489e7c
DM
749*/
750
bd81e77b
NC
751struct body_details {
752 size_t size; /* Size to allocate */
753 size_t copy; /* Size of structure to copy (may be shorter) */
754 size_t offset;
755 bool cant_upgrade; /* Can upgrade this type */
756 bool zero_nv; /* zero the NV when upgrading from this */
757 bool arena; /* Allocated from an arena */
758};
29489e7c 759
bd81e77b
NC
760#define HADNV FALSE
761#define NONV TRUE
29489e7c 762
bd81e77b
NC
763#ifdef PURIFY
764/* With -DPURFIY we allocate everything directly, and don't use arenas.
765 This seems a rather elegant way to simplify some of the code below. */
766#define HASARENA FALSE
767#else
768#define HASARENA TRUE
769#endif
770#define NOARENA FALSE
29489e7c 771
bd81e77b 772/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 773
bd81e77b
NC
774typedef struct {
775 STRLEN xpv_cur;
776 STRLEN xpv_len;
777} xpv_allocated;
29489e7c 778
bd81e77b 779to make its members accessible via a pointer to (say)
29489e7c 780
bd81e77b
NC
781struct xpv {
782 NV xnv_nv;
783 STRLEN xpv_cur;
784 STRLEN xpv_len;
785};
29489e7c 786
bd81e77b 787*/
29489e7c 788
bd81e77b
NC
789#define relative_STRUCT_OFFSET(longer, shorter, member) \
790 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 791
bd81e77b
NC
792/* Calculate the length to copy. Specifically work out the length less any
793 final padding the compiler needed to add. See the comment in sv_upgrade
794 for why copying the padding proved to be a bug. */
29489e7c 795
bd81e77b
NC
796#define copy_length(type, last_member) \
797 STRUCT_OFFSET(type, last_member) \
798 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 799
bd81e77b
NC
800static const struct body_details bodies_by_type[] = {
801 {0, 0, 0, FALSE, NONV, NOARENA},
802 /* IVs are in the head, so the allocation size is 0 */
803 {0, sizeof(IV), STRUCT_OFFSET(XPVIV, xiv_iv), FALSE, NONV, NOARENA},
804 /* 8 bytes on most ILP32 with IEEE doubles */
805 {sizeof(NV), sizeof(NV), 0, FALSE, HADNV, HASARENA},
806 /* RVs are in the head now */
807 /* However, this slot is overloaded and used by the pte */
808 {0, 0, 0, FALSE, NONV, NOARENA},
809 /* 8 bytes on most ILP32 with IEEE doubles */
810 {sizeof(xpv_allocated),
811 copy_length(XPV, xpv_len)
812 + relative_STRUCT_OFFSET(XPV, xpv_allocated, xpv_cur),
813 - relative_STRUCT_OFFSET(XPV, xpv_allocated, xpv_cur),
814 FALSE, NONV, HASARENA},
815 /* 12 */
816 {sizeof(xpviv_allocated),
817 copy_length(XPVIV, xiv_u)
818 + relative_STRUCT_OFFSET(XPVIV, xpviv_allocated, xpv_cur),
819 - relative_STRUCT_OFFSET(XPVIV, xpviv_allocated, xpv_cur),
820 FALSE, NONV, HASARENA},
821 /* 20 */
822 {sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, FALSE, HADNV, HASARENA},
823 /* 28 */
824 {sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, FALSE, HADNV, HASARENA},
825 /* 36 */
826 {sizeof(XPVBM), sizeof(XPVBM), 0, TRUE, HADNV, HASARENA},
827 /* 48 */
828 {sizeof(XPVGV), sizeof(XPVGV), 0, TRUE, HADNV, HASARENA},
829 /* 64 */
830 {sizeof(XPVLV), sizeof(XPVLV), 0, TRUE, HADNV, HASARENA},
831 /* 20 */
832 {sizeof(xpvav_allocated),
833 copy_length(XPVAV, xmg_stash)
834 + relative_STRUCT_OFFSET(XPVAV, xpvav_allocated, xav_fill),
835 - relative_STRUCT_OFFSET(XPVAV, xpvav_allocated, xav_fill),
836 TRUE, HADNV, HASARENA},
837 /* 20 */
838 {sizeof(xpvhv_allocated),
839 copy_length(XPVHV, xmg_stash)
840 + relative_STRUCT_OFFSET(XPVHV, xpvhv_allocated, xhv_fill),
841 - relative_STRUCT_OFFSET(XPVHV, xpvhv_allocated, xhv_fill),
842 TRUE, HADNV, HASARENA},
843 /* 76 */
844 {sizeof(XPVCV), sizeof(XPVCV), 0, TRUE, HADNV, HASARENA},
845 /* 80 */
846 {sizeof(XPVFM), sizeof(XPVFM), 0, TRUE, HADNV, NOARENA},
847 /* 84 */
848 {sizeof(XPVIO), sizeof(XPVIO), 0, TRUE, HADNV, NOARENA}
849};
29489e7c 850
bd81e77b
NC
851#define new_body_type(sv_type) \
852 (void *)((char *)S_new_body(aTHX_ bodies_by_type[sv_type].size, sv_type)\
853 - bodies_by_type[sv_type].offset)
29489e7c 854
bd81e77b
NC
855#define del_body_type(p, sv_type) \
856 del_body(p, &PL_body_roots[sv_type])
29489e7c 857
29489e7c 858
bd81e77b
NC
859#define new_body_allocated(sv_type) \
860 (void *)((char *)S_new_body(aTHX_ bodies_by_type[sv_type].size, sv_type)\
861 - bodies_by_type[sv_type].offset)
29489e7c 862
bd81e77b
NC
863#define del_body_allocated(p, sv_type) \
864 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 865
29489e7c 866
bd81e77b
NC
867#define my_safemalloc(s) (void*)safemalloc(s)
868#define my_safecalloc(s) (void*)safecalloc(s, 1)
869#define my_safefree(p) safefree((char*)p)
29489e7c 870
bd81e77b 871#ifdef PURIFY
29489e7c 872
bd81e77b
NC
873#define new_XNV() my_safemalloc(sizeof(XPVNV))
874#define del_XNV(p) my_safefree(p)
29489e7c 875
bd81e77b
NC
876#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
877#define del_XPVNV(p) my_safefree(p)
29489e7c 878
bd81e77b
NC
879#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
880#define del_XPVAV(p) my_safefree(p)
29489e7c 881
bd81e77b
NC
882#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
883#define del_XPVHV(p) my_safefree(p)
29489e7c 884
bd81e77b
NC
885#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
886#define del_XPVMG(p) my_safefree(p)
29489e7c 887
bd81e77b
NC
888#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
889#define del_XPVGV(p) my_safefree(p)
29489e7c 890
bd81e77b 891#else /* !PURIFY */
29489e7c 892
bd81e77b
NC
893#define new_XNV() new_body_type(SVt_NV)
894#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 895
bd81e77b
NC
896#define new_XPVNV() new_body_type(SVt_PVNV)
897#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 898
bd81e77b
NC
899#define new_XPVAV() new_body_allocated(SVt_PVAV)
900#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 901
bd81e77b
NC
902#define new_XPVHV() new_body_allocated(SVt_PVHV)
903#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 904
bd81e77b
NC
905#define new_XPVMG() new_body_type(SVt_PVMG)
906#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 907
bd81e77b
NC
908#define new_XPVGV() new_body_type(SVt_PVGV)
909#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 910
bd81e77b 911#endif /* PURIFY */
93e68bfb 912
bd81e77b 913/* no arena for you! */
93e68bfb 914
bd81e77b
NC
915#define new_NOARENA(details) \
916 my_safemalloc((details)->size + (details)->offset)
917#define new_NOARENAZ(details) \
918 my_safecalloc((details)->size + (details)->offset)
93e68bfb 919
bd81e77b
NC
920/*
921=for apidoc sv_upgrade
93e68bfb 922
bd81e77b
NC
923Upgrade an SV to a more complex form. Generally adds a new body type to the
924SV, then copies across as much information as possible from the old body.
925You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 926
bd81e77b 927=cut
93e68bfb 928*/
93e68bfb 929
bd81e77b
NC
930void
931Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type)
cac9b346 932{
bd81e77b
NC
933 void* old_body;
934 void* new_body;
935 const U32 old_type = SvTYPE(sv);
936 const struct body_details *const old_type_details
937 = bodies_by_type + old_type;
938 const struct body_details *new_type_details = bodies_by_type + new_type;
cac9b346 939
bd81e77b
NC
940 if (new_type != SVt_PV && SvIsCOW(sv)) {
941 sv_force_normal_flags(sv, 0);
942 }
cac9b346 943
bd81e77b
NC
944 if (old_type == new_type)
945 return;
cac9b346 946
bd81e77b
NC
947 if (old_type > new_type)
948 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
949 (int)old_type, (int)new_type);
cac9b346 950
cac9b346 951
bd81e77b 952 old_body = SvANY(sv);
de042e1d 953
bd81e77b
NC
954 /* Copying structures onto other structures that have been neatly zeroed
955 has a subtle gotcha. Consider XPVMG
cac9b346 956
bd81e77b
NC
957 +------+------+------+------+------+-------+-------+
958 | NV | CUR | LEN | IV | MAGIC | STASH |
959 +------+------+------+------+------+-------+-------+
960 0 4 8 12 16 20 24 28
645c22ef 961
bd81e77b
NC
962 where NVs are aligned to 8 bytes, so that sizeof that structure is
963 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 964
bd81e77b
NC
965 +------+------+------+------+------+-------+-------+------+
966 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
967 +------+------+------+------+------+-------+-------+------+
968 0 4 8 12 16 20 24 28 32
08742458 969
bd81e77b 970 so what happens if you allocate memory for this structure:
30f9da9e 971
bd81e77b
NC
972 +------+------+------+------+------+-------+-------+------+------+...
973 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
974 +------+------+------+------+------+-------+-------+------+------+...
975 0 4 8 12 16 20 24 28 32 36
bfc44f79 976
bd81e77b
NC
977 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
978 expect, because you copy the area marked ??? onto GP. Now, ??? may have
979 started out as zero once, but it's quite possible that it isn't. So now,
980 rather than a nicely zeroed GP, you have it pointing somewhere random.
981 Bugs ensue.
bfc44f79 982
bd81e77b
NC
983 (In fact, GP ends up pointing at a previous GP structure, because the
984 principle cause of the padding in XPVMG getting garbage is a copy of
985 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 986
bd81e77b
NC
987 So we are careful and work out the size of used parts of all the
988 structures. */
bfc44f79 989
bd81e77b
NC
990 switch (old_type) {
991 case SVt_NULL:
992 break;
993 case SVt_IV:
994 if (new_type < SVt_PVIV) {
995 new_type = (new_type == SVt_NV)
996 ? SVt_PVNV : SVt_PVIV;
997 new_type_details = bodies_by_type + new_type;
998 }
999 break;
1000 case SVt_NV:
1001 if (new_type < SVt_PVNV) {
1002 new_type = SVt_PVNV;
1003 new_type_details = bodies_by_type + new_type;
1004 }
1005 break;
1006 case SVt_RV:
1007 break;
1008 case SVt_PV:
1009 assert(new_type > SVt_PV);
1010 assert(SVt_IV < SVt_PV);
1011 assert(SVt_NV < SVt_PV);
1012 break;
1013 case SVt_PVIV:
1014 break;
1015 case SVt_PVNV:
1016 break;
1017 case SVt_PVMG:
1018 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1019 there's no way that it can be safely upgraded, because perl.c
1020 expects to Safefree(SvANY(PL_mess_sv)) */
1021 assert(sv != PL_mess_sv);
1022 /* This flag bit is used to mean other things in other scalar types.
1023 Given that it only has meaning inside the pad, it shouldn't be set
1024 on anything that can get upgraded. */
1025 assert((SvFLAGS(sv) & SVpad_TYPED) == 0);
1026 break;
1027 default:
1028 if (old_type_details->cant_upgrade)
1029 Perl_croak(aTHX_ "Can't upgrade that kind of scalar");
1030 }
645c22ef 1031
bd81e77b
NC
1032 SvFLAGS(sv) &= ~SVTYPEMASK;
1033 SvFLAGS(sv) |= new_type;
932e9ff9 1034
bd81e77b
NC
1035 switch (new_type) {
1036 case SVt_NULL:
1037 Perl_croak(aTHX_ "Can't upgrade to undef");
1038 case SVt_IV:
1039 assert(old_type == SVt_NULL);
1040 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1041 SvIV_set(sv, 0);
1042 return;
1043 case SVt_NV:
1044 assert(old_type == SVt_NULL);
1045 SvANY(sv) = new_XNV();
1046 SvNV_set(sv, 0);
1047 return;
1048 case SVt_RV:
1049 assert(old_type == SVt_NULL);
1050 SvANY(sv) = &sv->sv_u.svu_rv;
1051 SvRV_set(sv, 0);
1052 return;
1053 case SVt_PVHV:
1054 SvANY(sv) = new_XPVHV();
1055 HvFILL(sv) = 0;
1056 HvMAX(sv) = 0;
1057 HvTOTALKEYS(sv) = 0;
645c22ef 1058
bd81e77b 1059 goto hv_av_common;
aeb18a1e 1060
bd81e77b
NC
1061 case SVt_PVAV:
1062 SvANY(sv) = new_XPVAV();
1063 AvMAX(sv) = -1;
1064 AvFILLp(sv) = -1;
1065 AvALLOC(sv) = 0;
1066 AvREAL_only(sv);
aeb18a1e 1067
bd81e77b
NC
1068 hv_av_common:
1069 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1070 The target created by newSVrv also is, and it can have magic.
1071 However, it never has SvPVX set.
1072 */
1073 if (old_type >= SVt_RV) {
1074 assert(SvPVX_const(sv) == 0);
1075 }
aeb18a1e 1076
bd81e77b
NC
1077 /* Could put this in the else clause below, as PVMG must have SvPVX
1078 0 already (the assertion above) */
1079 SvPV_set(sv, (char*)0);
93e68bfb 1080
bd81e77b
NC
1081 if (old_type >= SVt_PVMG) {
1082 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_magic);
1083 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1084 } else {
1085 SvMAGIC_set(sv, 0);
1086 SvSTASH_set(sv, 0);
1087 }
1088 break;
93e68bfb 1089
93e68bfb 1090
bd81e77b
NC
1091 case SVt_PVIV:
1092 /* XXX Is this still needed? Was it ever needed? Surely as there is
1093 no route from NV to PVIV, NOK can never be true */
1094 assert(!SvNOKp(sv));
1095 assert(!SvNOK(sv));
1096 case SVt_PVIO:
1097 case SVt_PVFM:
1098 case SVt_PVBM:
1099 case SVt_PVGV:
1100 case SVt_PVCV:
1101 case SVt_PVLV:
1102 case SVt_PVMG:
1103 case SVt_PVNV:
1104 case SVt_PV:
93e68bfb 1105
bd81e77b
NC
1106 assert(new_type_details->size);
1107 /* We always allocated the full length item with PURIFY. To do this
1108 we fake things so that arena is false for all 16 types.. */
1109 if(new_type_details->arena) {
1110 /* This points to the start of the allocated area. */
1111 new_body_inline(new_body, new_type_details->size, new_type);
1112 Zero(new_body, new_type_details->size, char);
1113 new_body = ((char *)new_body) - new_type_details->offset;
1114 } else {
1115 new_body = new_NOARENAZ(new_type_details);
1116 }
1117 SvANY(sv) = new_body;
5e2fc214 1118
bd81e77b
NC
1119 if (old_type_details->copy) {
1120 Copy((char *)old_body + old_type_details->offset,
1121 (char *)new_body + old_type_details->offset,
1122 old_type_details->copy, char);
1123 }
1124
1125#ifndef NV_ZERO_IS_ALLBITS_ZERO
1126 /* If NV 0.0 is store as all bits 0 then Zero() already creates a correct
1127 0.0 for us. */
1128 if (old_type_details->zero_nv)
1129 SvNV_set(sv, 0);
82048762 1130#endif
5e2fc214 1131
bd81e77b
NC
1132 if (new_type == SVt_PVIO)
1133 IoPAGE_LEN(sv) = 60;
1134 if (old_type < SVt_RV)
1135 SvPV_set(sv, 0);
1136 break;
1137 default:
1138 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu", new_type);
1139 }
73171d91 1140
bd81e77b
NC
1141 if (old_type_details->size) {
1142 /* If the old body had an allocated size, then we need to free it. */
1143#ifdef PURIFY
1144 my_safefree(old_body);
1145#else
1146 del_body((void*)((char*)old_body + old_type_details->offset),
1147 &PL_body_roots[old_type]);
1148#endif
1149 }
1150}
73171d91 1151
bd81e77b
NC
1152/*
1153=for apidoc sv_backoff
73171d91 1154
bd81e77b
NC
1155Remove any string offset. You should normally use the C<SvOOK_off> macro
1156wrapper instead.
73171d91 1157
bd81e77b 1158=cut
73171d91
NC
1159*/
1160
bd81e77b
NC
1161int
1162Perl_sv_backoff(pTHX_ register SV *sv)
1163{
1164 assert(SvOOK(sv));
1165 assert(SvTYPE(sv) != SVt_PVHV);
1166 assert(SvTYPE(sv) != SVt_PVAV);
1167 if (SvIVX(sv)) {
1168 const char * const s = SvPVX_const(sv);
1169 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1170 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1171 SvIV_set(sv, 0);
1172 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1173 }
1174 SvFLAGS(sv) &= ~SVf_OOK;
1175 return 0;
1176}
73171d91 1177
bd81e77b
NC
1178/*
1179=for apidoc sv_grow
73171d91 1180
bd81e77b
NC
1181Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1182upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1183Use the C<SvGROW> wrapper instead.
93e68bfb 1184
bd81e77b
NC
1185=cut
1186*/
93e68bfb 1187
bd81e77b
NC
1188char *
1189Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1190{
1191 register char *s;
93e68bfb 1192
bd81e77b
NC
1193#ifdef HAS_64K_LIMIT
1194 if (newlen >= 0x10000) {
1195 PerlIO_printf(Perl_debug_log,
1196 "Allocation too large: %"UVxf"\n", (UV)newlen);
1197 my_exit(1);
1198 }
1199#endif /* HAS_64K_LIMIT */
1200 if (SvROK(sv))
1201 sv_unref(sv);
1202 if (SvTYPE(sv) < SVt_PV) {
1203 sv_upgrade(sv, SVt_PV);
1204 s = SvPVX_mutable(sv);
1205 }
1206 else if (SvOOK(sv)) { /* pv is offset? */
1207 sv_backoff(sv);
1208 s = SvPVX_mutable(sv);
1209 if (newlen > SvLEN(sv))
1210 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1211#ifdef HAS_64K_LIMIT
1212 if (newlen >= 0x10000)
1213 newlen = 0xFFFF;
1214#endif
1215 }
1216 else
1217 s = SvPVX_mutable(sv);
aeb18a1e 1218
bd81e77b
NC
1219 if (newlen > SvLEN(sv)) { /* need more room? */
1220 newlen = PERL_STRLEN_ROUNDUP(newlen);
1221 if (SvLEN(sv) && s) {
1222#ifdef MYMALLOC
1223 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1224 if (newlen <= l) {
1225 SvLEN_set(sv, l);
1226 return s;
1227 } else
1228#endif
1229 s = saferealloc(s, newlen);
1230 }
1231 else {
1232 s = safemalloc(newlen);
1233 if (SvPVX_const(sv) && SvCUR(sv)) {
1234 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1235 }
1236 }
1237 SvPV_set(sv, s);
1238 SvLEN_set(sv, newlen);
1239 }
1240 return s;
1241}
aeb18a1e 1242
bd81e77b
NC
1243/*
1244=for apidoc sv_setiv
932e9ff9 1245
bd81e77b
NC
1246Copies an integer into the given SV, upgrading first if necessary.
1247Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1248
bd81e77b
NC
1249=cut
1250*/
463ee0b2 1251
bd81e77b
NC
1252void
1253Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1254{
1255 SV_CHECK_THINKFIRST_COW_DROP(sv);
1256 switch (SvTYPE(sv)) {
1257 case SVt_NULL:
1258 sv_upgrade(sv, SVt_IV);
1259 break;
1260 case SVt_NV:
1261 sv_upgrade(sv, SVt_PVNV);
1262 break;
1263 case SVt_RV:
1264 case SVt_PV:
1265 sv_upgrade(sv, SVt_PVIV);
1266 break;
463ee0b2 1267
bd81e77b
NC
1268 case SVt_PVGV:
1269 case SVt_PVAV:
1270 case SVt_PVHV:
1271 case SVt_PVCV:
1272 case SVt_PVFM:
1273 case SVt_PVIO:
1274 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1275 OP_DESC(PL_op));
1276 }
1277 (void)SvIOK_only(sv); /* validate number */
1278 SvIV_set(sv, i);
1279 SvTAINT(sv);
1280}
932e9ff9 1281
bd81e77b
NC
1282/*
1283=for apidoc sv_setiv_mg
d33b2eba 1284
bd81e77b 1285Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1286
bd81e77b
NC
1287=cut
1288*/
d33b2eba 1289
bd81e77b
NC
1290void
1291Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1292{
1293 sv_setiv(sv,i);
1294 SvSETMAGIC(sv);
1295}
727879eb 1296
bd81e77b
NC
1297/*
1298=for apidoc sv_setuv
d33b2eba 1299
bd81e77b
NC
1300Copies an unsigned integer into the given SV, upgrading first if necessary.
1301Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1302
bd81e77b
NC
1303=cut
1304*/
d33b2eba 1305
bd81e77b
NC
1306void
1307Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1308{
1309 /* With these two if statements:
1310 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1311
bd81e77b
NC
1312 without
1313 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1314
bd81e77b
NC
1315 If you wish to remove them, please benchmark to see what the effect is
1316 */
1317 if (u <= (UV)IV_MAX) {
1318 sv_setiv(sv, (IV)u);
1319 return;
1320 }
1321 sv_setiv(sv, 0);
1322 SvIsUV_on(sv);
1323 SvUV_set(sv, u);
1324}
d33b2eba 1325
bd81e77b
NC
1326/*
1327=for apidoc sv_setuv_mg
727879eb 1328
bd81e77b 1329Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1330
bd81e77b
NC
1331=cut
1332*/
5e2fc214 1333
bd81e77b
NC
1334void
1335Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1336{
1337 sv_setiv(sv, 0);
1338 SvIsUV_on(sv);
1339 sv_setuv(sv,u);
1340 SvSETMAGIC(sv);
1341}
5e2fc214 1342
954c1994 1343/*
bd81e77b 1344=for apidoc sv_setnv
954c1994 1345
bd81e77b
NC
1346Copies a double into the given SV, upgrading first if necessary.
1347Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1348
1349=cut
1350*/
1351
63f97190 1352void
bd81e77b 1353Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1354{
bd81e77b
NC
1355 SV_CHECK_THINKFIRST_COW_DROP(sv);
1356 switch (SvTYPE(sv)) {
79072805 1357 case SVt_NULL:
79072805 1358 case SVt_IV:
bd81e77b 1359 sv_upgrade(sv, SVt_NV);
79072805 1360 break;
ed6116ce 1361 case SVt_RV:
79072805 1362 case SVt_PV:
79072805 1363 case SVt_PVIV:
bd81e77b 1364 sv_upgrade(sv, SVt_PVNV);
79072805 1365 break;
bd4b1eb5 1366
bd4b1eb5 1367 case SVt_PVGV:
bd81e77b
NC
1368 case SVt_PVAV:
1369 case SVt_PVHV:
79072805 1370 case SVt_PVCV:
bd81e77b
NC
1371 case SVt_PVFM:
1372 case SVt_PVIO:
1373 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1374 OP_NAME(PL_op));
2068cd4d 1375 }
bd81e77b
NC
1376 SvNV_set(sv, num);
1377 (void)SvNOK_only(sv); /* validate number */
1378 SvTAINT(sv);
79072805
LW
1379}
1380
645c22ef 1381/*
bd81e77b 1382=for apidoc sv_setnv_mg
645c22ef 1383
bd81e77b 1384Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1385
1386=cut
1387*/
1388
bd81e77b
NC
1389void
1390Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1391{
bd81e77b
NC
1392 sv_setnv(sv,num);
1393 SvSETMAGIC(sv);
79072805
LW
1394}
1395
bd81e77b
NC
1396/* Print an "isn't numeric" warning, using a cleaned-up,
1397 * printable version of the offending string
1398 */
954c1994 1399
bd81e77b
NC
1400STATIC void
1401S_not_a_number(pTHX_ SV *sv)
79072805 1402{
bd81e77b
NC
1403 SV *dsv;
1404 char tmpbuf[64];
1405 const char *pv;
94463019
JH
1406
1407 if (DO_UTF8(sv)) {
d0043bd1 1408 dsv = sv_2mortal(newSVpvn("", 0));
94463019
JH
1409 pv = sv_uni_display(dsv, sv, 10, 0);
1410 } else {
1411 char *d = tmpbuf;
551405c4 1412 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1413 /* each *s can expand to 4 chars + "...\0",
1414 i.e. need room for 8 chars */
ecdeb87c 1415
00b6aa41
AL
1416 const char *s = SvPVX_const(sv);
1417 const char * const end = s + SvCUR(sv);
1418 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1419 int ch = *s & 0xFF;
1420 if (ch & 128 && !isPRINT_LC(ch)) {
1421 *d++ = 'M';
1422 *d++ = '-';
1423 ch &= 127;
1424 }
1425 if (ch == '\n') {
1426 *d++ = '\\';
1427 *d++ = 'n';
1428 }
1429 else if (ch == '\r') {
1430 *d++ = '\\';
1431 *d++ = 'r';
1432 }
1433 else if (ch == '\f') {
1434 *d++ = '\\';
1435 *d++ = 'f';
1436 }
1437 else if (ch == '\\') {
1438 *d++ = '\\';
1439 *d++ = '\\';
1440 }
1441 else if (ch == '\0') {
1442 *d++ = '\\';
1443 *d++ = '0';
1444 }
1445 else if (isPRINT_LC(ch))
1446 *d++ = ch;
1447 else {
1448 *d++ = '^';
1449 *d++ = toCTRL(ch);
1450 }
1451 }
1452 if (s < end) {
1453 *d++ = '.';
1454 *d++ = '.';
1455 *d++ = '.';
1456 }
1457 *d = '\0';
1458 pv = tmpbuf;
a0d0e21e 1459 }
a0d0e21e 1460
533c011a 1461 if (PL_op)
9014280d 1462 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1463 "Argument \"%s\" isn't numeric in %s", pv,
1464 OP_DESC(PL_op));
a0d0e21e 1465 else
9014280d 1466 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1467 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1468}
1469
c2988b20
NC
1470/*
1471=for apidoc looks_like_number
1472
645c22ef
DM
1473Test if the content of an SV looks like a number (or is a number).
1474C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1475non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1476
1477=cut
1478*/
1479
1480I32
1481Perl_looks_like_number(pTHX_ SV *sv)
1482{
a3b680e6 1483 register const char *sbegin;
c2988b20
NC
1484 STRLEN len;
1485
1486 if (SvPOK(sv)) {
3f7c398e 1487 sbegin = SvPVX_const(sv);
c2988b20
NC
1488 len = SvCUR(sv);
1489 }
1490 else if (SvPOKp(sv))
83003860 1491 sbegin = SvPV_const(sv, len);
c2988b20 1492 else
e0ab1c0e 1493 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1494 return grok_number(sbegin, len, NULL);
1495}
25da4f38
IZ
1496
1497/* Actually, ISO C leaves conversion of UV to IV undefined, but
1498 until proven guilty, assume that things are not that bad... */
1499
645c22ef
DM
1500/*
1501 NV_PRESERVES_UV:
1502
1503 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1504 an IV (an assumption perl has been based on to date) it becomes necessary
1505 to remove the assumption that the NV always carries enough precision to
1506 recreate the IV whenever needed, and that the NV is the canonical form.
1507 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1508 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1509 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1510 1) to distinguish between IV/UV/NV slots that have cached a valid
1511 conversion where precision was lost and IV/UV/NV slots that have a
1512 valid conversion which has lost no precision
645c22ef 1513 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1514 would lose precision, the precise conversion (or differently
1515 imprecise conversion) is also performed and cached, to prevent
1516 requests for different numeric formats on the same SV causing
1517 lossy conversion chains. (lossless conversion chains are perfectly
1518 acceptable (still))
1519
1520
1521 flags are used:
1522 SvIOKp is true if the IV slot contains a valid value
1523 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1524 SvNOKp is true if the NV slot contains a valid value
1525 SvNOK is true only if the NV value is accurate
1526
1527 so
645c22ef 1528 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1529 IV(or UV) would lose accuracy over a direct conversion from PV to
1530 IV(or UV). If it would, cache both conversions, return NV, but mark
1531 SV as IOK NOKp (ie not NOK).
1532
645c22ef 1533 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1534 NV would lose accuracy over a direct conversion from PV to NV. If it
1535 would, cache both conversions, flag similarly.
1536
1537 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1538 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1539 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1540 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1541 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1542
645c22ef
DM
1543 The benefit of this is that operations such as pp_add know that if
1544 SvIOK is true for both left and right operands, then integer addition
1545 can be used instead of floating point (for cases where the result won't
1546 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1547 loss of precision compared with integer addition.
1548
1549 * making IV and NV equal status should make maths accurate on 64 bit
1550 platforms
1551 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1552 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1553 looking for SvIOK and checking for overflow will not outweigh the
1554 fp to integer speedup)
1555 * will slow down integer operations (callers of SvIV) on "inaccurate"
1556 values, as the change from SvIOK to SvIOKp will cause a call into
1557 sv_2iv each time rather than a macro access direct to the IV slot
1558 * should speed up number->string conversion on integers as IV is
645c22ef 1559 favoured when IV and NV are equally accurate
28e5dec8
JH
1560
1561 ####################################################################
645c22ef
DM
1562 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1563 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1564 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1565 ####################################################################
1566
645c22ef 1567 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1568 performance ratio.
1569*/
1570
1571#ifndef NV_PRESERVES_UV
645c22ef
DM
1572# define IS_NUMBER_UNDERFLOW_IV 1
1573# define IS_NUMBER_UNDERFLOW_UV 2
1574# define IS_NUMBER_IV_AND_UV 2
1575# define IS_NUMBER_OVERFLOW_IV 4
1576# define IS_NUMBER_OVERFLOW_UV 5
1577
1578/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1579
1580/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1581STATIC int
645c22ef 1582S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1583{
3f7c398e 1584 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
1585 if (SvNVX(sv) < (NV)IV_MIN) {
1586 (void)SvIOKp_on(sv);
1587 (void)SvNOK_on(sv);
45977657 1588 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1589 return IS_NUMBER_UNDERFLOW_IV;
1590 }
1591 if (SvNVX(sv) > (NV)UV_MAX) {
1592 (void)SvIOKp_on(sv);
1593 (void)SvNOK_on(sv);
1594 SvIsUV_on(sv);
607fa7f2 1595 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1596 return IS_NUMBER_OVERFLOW_UV;
1597 }
c2988b20
NC
1598 (void)SvIOKp_on(sv);
1599 (void)SvNOK_on(sv);
1600 /* Can't use strtol etc to convert this string. (See truth table in
1601 sv_2iv */
1602 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1603 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1604 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1605 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1606 } else {
1607 /* Integer is imprecise. NOK, IOKp */
1608 }
1609 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1610 }
1611 SvIsUV_on(sv);
607fa7f2 1612 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1613 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1614 if (SvUVX(sv) == UV_MAX) {
1615 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1616 possibly be preserved by NV. Hence, it must be overflow.
1617 NOK, IOKp */
1618 return IS_NUMBER_OVERFLOW_UV;
1619 }
1620 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1621 } else {
1622 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1623 }
c2988b20 1624 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1625}
645c22ef
DM
1626#endif /* !NV_PRESERVES_UV*/
1627
af359546
NC
1628STATIC bool
1629S_sv_2iuv_common(pTHX_ SV *sv) {
1630 if (SvNOKp(sv)) {
28e5dec8
JH
1631 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1632 * without also getting a cached IV/UV from it at the same time
1633 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1634 * IV or UV at same time to avoid this. */
1635 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1636
1637 if (SvTYPE(sv) == SVt_NV)
1638 sv_upgrade(sv, SVt_PVNV);
1639
28e5dec8
JH
1640 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1641 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1642 certainly cast into the IV range at IV_MAX, whereas the correct
1643 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1644 cases go to UV */
1645 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1646 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1647 if (SvNVX(sv) == (NV) SvIVX(sv)
1648#ifndef NV_PRESERVES_UV
1649 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1650 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1651 /* Don't flag it as "accurately an integer" if the number
1652 came from a (by definition imprecise) NV operation, and
1653 we're outside the range of NV integer precision */
1654#endif
1655 ) {
1656 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1657 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1658 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1659 PTR2UV(sv),
1660 SvNVX(sv),
1661 SvIVX(sv)));
1662
1663 } else {
1664 /* IV not precise. No need to convert from PV, as NV
1665 conversion would already have cached IV if it detected
1666 that PV->IV would be better than PV->NV->IV
1667 flags already correct - don't set public IOK. */
1668 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1669 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1670 PTR2UV(sv),
1671 SvNVX(sv),
1672 SvIVX(sv)));
1673 }
1674 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1675 but the cast (NV)IV_MIN rounds to a the value less (more
1676 negative) than IV_MIN which happens to be equal to SvNVX ??
1677 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1678 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1679 (NV)UVX == NVX are both true, but the values differ. :-(
1680 Hopefully for 2s complement IV_MIN is something like
1681 0x8000000000000000 which will be exact. NWC */
d460ef45 1682 }
25da4f38 1683 else {
607fa7f2 1684 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1685 if (
1686 (SvNVX(sv) == (NV) SvUVX(sv))
1687#ifndef NV_PRESERVES_UV
1688 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1689 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1690 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1691 /* Don't flag it as "accurately an integer" if the number
1692 came from a (by definition imprecise) NV operation, and
1693 we're outside the range of NV integer precision */
1694#endif
1695 )
1696 SvIOK_on(sv);
25da4f38 1697 SvIsUV_on(sv);
1c846c1f 1698 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1699 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1700 PTR2UV(sv),
57def98f
JH
1701 SvUVX(sv),
1702 SvUVX(sv)));
25da4f38 1703 }
748a9306
LW
1704 }
1705 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1706 UV value;
504618e9 1707 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1708 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1709 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1710 the same as the direct translation of the initial string
1711 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1712 be careful to ensure that the value with the .456 is around if the
1713 NV value is requested in the future).
1c846c1f 1714
af359546 1715 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1716 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1717 cache the NV if we are sure it's not needed.
25da4f38 1718 */
16b7a9a4 1719
c2988b20
NC
1720 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1721 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1722 == IS_NUMBER_IN_UV) {
5e045b90 1723 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1724 if (SvTYPE(sv) < SVt_PVIV)
1725 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1726 (void)SvIOK_on(sv);
c2988b20
NC
1727 } else if (SvTYPE(sv) < SVt_PVNV)
1728 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1729
c2988b20
NC
1730 /* If NV preserves UV then we only use the UV value if we know that
1731 we aren't going to call atof() below. If NVs don't preserve UVs
1732 then the value returned may have more precision than atof() will
1733 return, even though value isn't perfectly accurate. */
1734 if ((numtype & (IS_NUMBER_IN_UV
1735#ifdef NV_PRESERVES_UV
1736 | IS_NUMBER_NOT_INT
1737#endif
1738 )) == IS_NUMBER_IN_UV) {
1739 /* This won't turn off the public IOK flag if it was set above */
1740 (void)SvIOKp_on(sv);
1741
1742 if (!(numtype & IS_NUMBER_NEG)) {
1743 /* positive */;
1744 if (value <= (UV)IV_MAX) {
45977657 1745 SvIV_set(sv, (IV)value);
c2988b20 1746 } else {
af359546 1747 /* it didn't overflow, and it was positive. */
607fa7f2 1748 SvUV_set(sv, value);
c2988b20
NC
1749 SvIsUV_on(sv);
1750 }
1751 } else {
1752 /* 2s complement assumption */
1753 if (value <= (UV)IV_MIN) {
45977657 1754 SvIV_set(sv, -(IV)value);
c2988b20
NC
1755 } else {
1756 /* Too negative for an IV. This is a double upgrade, but
d1be9408 1757 I'm assuming it will be rare. */
c2988b20
NC
1758 if (SvTYPE(sv) < SVt_PVNV)
1759 sv_upgrade(sv, SVt_PVNV);
1760 SvNOK_on(sv);
1761 SvIOK_off(sv);
1762 SvIOKp_on(sv);
9d6ce603 1763 SvNV_set(sv, -(NV)value);
45977657 1764 SvIV_set(sv, IV_MIN);
c2988b20
NC
1765 }
1766 }
1767 }
1768 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
1769 will be in the previous block to set the IV slot, and the next
1770 block to set the NV slot. So no else here. */
1771
1772 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1773 != IS_NUMBER_IN_UV) {
1774 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 1775 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 1776
c2988b20
NC
1777 if (! numtype && ckWARN(WARN_NUMERIC))
1778 not_a_number(sv);
28e5dec8 1779
65202027 1780#if defined(USE_LONG_DOUBLE)
c2988b20
NC
1781 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
1782 PTR2UV(sv), SvNVX(sv)));
65202027 1783#else
1779d84d 1784 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 1785 PTR2UV(sv), SvNVX(sv)));
65202027 1786#endif
28e5dec8 1787
28e5dec8 1788#ifdef NV_PRESERVES_UV
af359546
NC
1789 (void)SvIOKp_on(sv);
1790 (void)SvNOK_on(sv);
1791 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
1792 SvIV_set(sv, I_V(SvNVX(sv)));
1793 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1794 SvIOK_on(sv);
1795 } else {
1796 /* Integer is imprecise. NOK, IOKp */
1797 }
1798 /* UV will not work better than IV */
1799 } else {
1800 if (SvNVX(sv) > (NV)UV_MAX) {
1801 SvIsUV_on(sv);
1802 /* Integer is inaccurate. NOK, IOKp, is UV */
1803 SvUV_set(sv, UV_MAX);
af359546
NC
1804 } else {
1805 SvUV_set(sv, U_V(SvNVX(sv)));
1806 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
1807 NV preservse UV so can do correct comparison. */
1808 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1809 SvIOK_on(sv);
af359546
NC
1810 } else {
1811 /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
1812 }
1813 }
4b0c9573 1814 SvIsUV_on(sv);
af359546 1815 }
28e5dec8 1816#else /* NV_PRESERVES_UV */
c2988b20
NC
1817 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1818 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 1819 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
1820 grok_number above. The NV slot has just been set using
1821 Atof. */
560b0c46 1822 SvNOK_on(sv);
c2988b20
NC
1823 assert (SvIOKp(sv));
1824 } else {
1825 if (((UV)1 << NV_PRESERVES_UV_BITS) >
1826 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
1827 /* Small enough to preserve all bits. */
1828 (void)SvIOKp_on(sv);
1829 SvNOK_on(sv);
45977657 1830 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1831 if ((NV)(SvIVX(sv)) == SvNVX(sv))
1832 SvIOK_on(sv);
1833 /* Assumption: first non-preserved integer is < IV_MAX,
1834 this NV is in the preserved range, therefore: */
1835 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
1836 < (UV)IV_MAX)) {
32fdb065 1837 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
1838 }
1839 } else {
1840 /* IN_UV NOT_INT
1841 0 0 already failed to read UV.
1842 0 1 already failed to read UV.
1843 1 0 you won't get here in this case. IV/UV
1844 slot set, public IOK, Atof() unneeded.
1845 1 1 already read UV.
1846 so there's no point in sv_2iuv_non_preserve() attempting
1847 to use atol, strtol, strtoul etc. */
40a17c4c 1848 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
1849 }
1850 }
28e5dec8 1851#endif /* NV_PRESERVES_UV */
25da4f38 1852 }
af359546
NC
1853 }
1854 else {
1855 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
1856 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
1857 report_uninit(sv);
1858 }
25da4f38
IZ
1859 if (SvTYPE(sv) < SVt_IV)
1860 /* Typically the caller expects that sv_any is not NULL now. */
1861 sv_upgrade(sv, SVt_IV);
af359546
NC
1862 /* Return 0 from the caller. */
1863 return TRUE;
1864 }
1865 return FALSE;
1866}
1867
1868/*
1869=for apidoc sv_2iv_flags
1870
1871Return the integer value of an SV, doing any necessary string
1872conversion. If flags includes SV_GMAGIC, does an mg_get() first.
1873Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
1874
1875=cut
1876*/
1877
1878IV
1879Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
1880{
1881 if (!sv)
a0d0e21e 1882 return 0;
af359546
NC
1883 if (SvGMAGICAL(sv)) {
1884 if (flags & SV_GMAGIC)
1885 mg_get(sv);
1886 if (SvIOKp(sv))
1887 return SvIVX(sv);
1888 if (SvNOKp(sv)) {
1889 return I_V(SvNVX(sv));
1890 }
71c558c3
NC
1891 if (SvPOKp(sv) && SvLEN(sv)) {
1892 UV value;
1893 const int numtype
1894 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
1895
1896 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1897 == IS_NUMBER_IN_UV) {
1898 /* It's definitely an integer */
1899 if (numtype & IS_NUMBER_NEG) {
1900 if (value < (UV)IV_MIN)
1901 return -(IV)value;
1902 } else {
1903 if (value < (UV)IV_MAX)
1904 return (IV)value;
1905 }
1906 }
1907 if (!numtype) {
1908 if (ckWARN(WARN_NUMERIC))
1909 not_a_number(sv);
1910 }
1911 return I_V(Atof(SvPVX_const(sv)));
1912 }
af359546
NC
1913 if (!SvROK(sv)) {
1914 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
1915 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
1916 report_uninit(sv);
1917 }
1918 return 0;
1919 }
71c558c3
NC
1920 /* Else this will drop through into the SvROK case just below, which
1921 will return within the {} for all code paths. */
af359546
NC
1922 }
1923 if (SvTHINKFIRST(sv)) {
1924 if (SvROK(sv)) {
1925 if (SvAMAGIC(sv)) {
1926 SV * const tmpstr=AMG_CALLun(sv,numer);
1927 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
1928 return SvIV(tmpstr);
1929 }
1930 }
1931 return PTR2IV(SvRV(sv));
1932 }
1933 if (SvIsCOW(sv)) {
1934 sv_force_normal_flags(sv, 0);
1935 }
1936 if (SvREADONLY(sv) && !SvOK(sv)) {
1937 if (ckWARN(WARN_UNINITIALIZED))
1938 report_uninit(sv);
1939 return 0;
1940 }
1941 }
1942 if (!SvIOKp(sv)) {
1943 if (S_sv_2iuv_common(aTHX_ sv))
1944 return 0;
79072805 1945 }
1d7c1841
GS
1946 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
1947 PTR2UV(sv),SvIVX(sv)));
25da4f38 1948 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
1949}
1950
645c22ef 1951/*
891f9566 1952=for apidoc sv_2uv_flags
645c22ef
DM
1953
1954Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
1955conversion. If flags includes SV_GMAGIC, does an mg_get() first.
1956Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
1957
1958=cut
1959*/
1960
ff68c719 1961UV
891f9566 1962Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719 1963{
1964 if (!sv)
1965 return 0;
1966 if (SvGMAGICAL(sv)) {
891f9566
YST
1967 if (flags & SV_GMAGIC)
1968 mg_get(sv);
ff68c719 1969 if (SvIOKp(sv))
1970 return SvUVX(sv);
1971 if (SvNOKp(sv))
1972 return U_V(SvNVX(sv));
71c558c3
NC
1973 if (SvPOKp(sv) && SvLEN(sv)) {
1974 UV value;
1975 const int numtype
1976 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
1977
1978 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1979 == IS_NUMBER_IN_UV) {
1980 /* It's definitely an integer */
1981 if (!(numtype & IS_NUMBER_NEG))
1982 return value;
1983 }
1984 if (!numtype) {
1985 if (ckWARN(WARN_NUMERIC))
1986 not_a_number(sv);
1987 }
1988 return U_V(Atof(SvPVX_const(sv)));
1989 }
3fe9a6f1 1990 if (!SvROK(sv)) {
d008e5eb 1991 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 1992 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 1993 report_uninit(sv);
c6ee37c5 1994 }
36477c24 1995 return 0;
3fe9a6f1 1996 }
71c558c3
NC
1997 /* Else this will drop through into the SvROK case just below, which
1998 will return within the {} for all code paths. */
ff68c719 1999 }
2000 if (SvTHINKFIRST(sv)) {
2001 if (SvROK(sv)) {
ff68c719 2002 SV* tmpstr;
1554e226 2003 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,numer)) &&
b4b9a328 2004 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv))))
9e7bc3e8 2005 return SvUV(tmpstr);
56431972 2006 return PTR2UV(SvRV(sv));
ff68c719 2007 }
765f542d
NC
2008 if (SvIsCOW(sv)) {
2009 sv_force_normal_flags(sv, 0);
8a818333 2010 }
0336b60e 2011 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2012 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2013 report_uninit(sv);
ff68c719 2014 return 0;
2015 }
2016 }
af359546
NC
2017 if (!SvIOKp(sv)) {
2018 if (S_sv_2iuv_common(aTHX_ sv))
2019 return 0;
ff68c719 2020 }
25da4f38 2021
1d7c1841
GS
2022 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2023 PTR2UV(sv),SvUVX(sv)));
25da4f38 2024 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2025}
2026
645c22ef
DM
2027/*
2028=for apidoc sv_2nv
2029
2030Return the num value of an SV, doing any necessary string or integer
2031conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2032macros.
2033
2034=cut
2035*/
2036
65202027 2037NV
864dbfa3 2038Perl_sv_2nv(pTHX_ register SV *sv)
79072805
LW
2039{
2040 if (!sv)
2041 return 0.0;
8990e307 2042 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2043 mg_get(sv);
2044 if (SvNOKp(sv))
2045 return SvNVX(sv);
a0d0e21e 2046 if (SvPOKp(sv) && SvLEN(sv)) {
041457d9 2047 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2048 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2049 not_a_number(sv);
3f7c398e 2050 return Atof(SvPVX_const(sv));
a0d0e21e 2051 }
25da4f38 2052 if (SvIOKp(sv)) {
1c846c1f 2053 if (SvIsUV(sv))
65202027 2054 return (NV)SvUVX(sv);
25da4f38 2055 else
65202027 2056 return (NV)SvIVX(sv);
25da4f38 2057 }
16d20bd9 2058 if (!SvROK(sv)) {
d008e5eb 2059 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 2060 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 2061 report_uninit(sv);
c6ee37c5 2062 }
66a1b24b 2063 return (NV)0;
16d20bd9 2064 }
71c558c3
NC
2065 /* Else this will drop through into the SvROK case just below, which
2066 will return within the {} for all code paths. */
463ee0b2 2067 }
ed6116ce 2068 if (SvTHINKFIRST(sv)) {
a0d0e21e 2069 if (SvROK(sv)) {
a0d0e21e 2070 SV* tmpstr;
1554e226 2071 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,numer)) &&
b4b9a328 2072 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv))))
9e7bc3e8 2073 return SvNV(tmpstr);
56431972 2074 return PTR2NV(SvRV(sv));
a0d0e21e 2075 }
765f542d
NC
2076 if (SvIsCOW(sv)) {
2077 sv_force_normal_flags(sv, 0);
8a818333 2078 }
0336b60e 2079 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2080 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2081 report_uninit(sv);
ed6116ce
LW
2082 return 0.0;
2083 }
79072805
LW
2084 }
2085 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2086 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2087 sv_upgrade(sv, SVt_NV);
906f284f 2088#ifdef USE_LONG_DOUBLE
097ee67d 2089 DEBUG_c({
f93f4e46 2090 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2091 PerlIO_printf(Perl_debug_log,
2092 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2093 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2094 RESTORE_NUMERIC_LOCAL();
2095 });
65202027 2096#else
572bbb43 2097 DEBUG_c({
f93f4e46 2098 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2099 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2100 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2101 RESTORE_NUMERIC_LOCAL();
2102 });
572bbb43 2103#endif
79072805
LW
2104 }
2105 else if (SvTYPE(sv) < SVt_PVNV)
2106 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2107 if (SvNOKp(sv)) {
2108 return SvNVX(sv);
61604483 2109 }
59d8ce62 2110 if (SvIOKp(sv)) {
9d6ce603 2111 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2112#ifdef NV_PRESERVES_UV
2113 SvNOK_on(sv);
2114#else
2115 /* Only set the public NV OK flag if this NV preserves the IV */
2116 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2117 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2118 : (SvIVX(sv) == I_V(SvNVX(sv))))
2119 SvNOK_on(sv);
2120 else
2121 SvNOKp_on(sv);
2122#endif
93a17b20 2123 }
748a9306 2124 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2125 UV value;
3f7c398e 2126 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2127 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2128 not_a_number(sv);
28e5dec8 2129#ifdef NV_PRESERVES_UV
c2988b20
NC
2130 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2131 == IS_NUMBER_IN_UV) {
5e045b90 2132 /* It's definitely an integer */
9d6ce603 2133 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2134 } else
3f7c398e 2135 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2136 SvNOK_on(sv);
2137#else
3f7c398e 2138 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2139 /* Only set the public NV OK flag if this NV preserves the value in
2140 the PV at least as well as an IV/UV would.
2141 Not sure how to do this 100% reliably. */
2142 /* if that shift count is out of range then Configure's test is
2143 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2144 UV_BITS */
2145 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2146 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2147 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2148 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2149 /* Can't use strtol etc to convert this string, so don't try.
2150 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2151 SvNOK_on(sv);
2152 } else {
2153 /* value has been set. It may not be precise. */
2154 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2155 /* 2s complement assumption for (UV)IV_MIN */
2156 SvNOK_on(sv); /* Integer is too negative. */
2157 } else {
2158 SvNOKp_on(sv);
2159 SvIOKp_on(sv);
6fa402ec 2160
c2988b20 2161 if (numtype & IS_NUMBER_NEG) {
45977657 2162 SvIV_set(sv, -(IV)value);
c2988b20 2163 } else if (value <= (UV)IV_MAX) {
45977657 2164 SvIV_set(sv, (IV)value);
c2988b20 2165 } else {
607fa7f2 2166 SvUV_set(sv, value);
c2988b20
NC
2167 SvIsUV_on(sv);
2168 }
2169
2170 if (numtype & IS_NUMBER_NOT_INT) {
2171 /* I believe that even if the original PV had decimals,
2172 they are lost beyond the limit of the FP precision.
2173 However, neither is canonical, so both only get p
2174 flags. NWC, 2000/11/25 */
2175 /* Both already have p flags, so do nothing */
2176 } else {
66a1b24b 2177 const NV nv = SvNVX(sv);
c2988b20
NC
2178 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2179 if (SvIVX(sv) == I_V(nv)) {
2180 SvNOK_on(sv);
c2988b20 2181 } else {
c2988b20
NC
2182 /* It had no "." so it must be integer. */
2183 }
00b6aa41 2184 SvIOK_on(sv);
c2988b20
NC
2185 } else {
2186 /* between IV_MAX and NV(UV_MAX).
2187 Could be slightly > UV_MAX */
6fa402ec 2188
c2988b20
NC
2189 if (numtype & IS_NUMBER_NOT_INT) {
2190 /* UV and NV both imprecise. */
2191 } else {
66a1b24b 2192 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2193
2194 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2195 SvNOK_on(sv);
c2988b20 2196 }
00b6aa41 2197 SvIOK_on(sv);
c2988b20
NC
2198 }
2199 }
2200 }
2201 }
2202 }
28e5dec8 2203#endif /* NV_PRESERVES_UV */
93a17b20 2204 }
79072805 2205 else {
041457d9 2206 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2207 report_uninit(sv);
7e25a7e9
NC
2208 assert (SvTYPE(sv) >= SVt_NV);
2209 /* Typically the caller expects that sv_any is not NULL now. */
2210 /* XXX Ilya implies that this is a bug in callers that assume this
2211 and ideally should be fixed. */
a0d0e21e 2212 return 0.0;
79072805 2213 }
572bbb43 2214#if defined(USE_LONG_DOUBLE)
097ee67d 2215 DEBUG_c({
f93f4e46 2216 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2217 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2218 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2219 RESTORE_NUMERIC_LOCAL();
2220 });
65202027 2221#else
572bbb43 2222 DEBUG_c({
f93f4e46 2223 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2224 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2225 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2226 RESTORE_NUMERIC_LOCAL();
2227 });
572bbb43 2228#endif
463ee0b2 2229 return SvNVX(sv);
79072805
LW
2230}
2231
645c22ef
DM
2232/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2233 * UV as a string towards the end of buf, and return pointers to start and
2234 * end of it.
2235 *
2236 * We assume that buf is at least TYPE_CHARS(UV) long.
2237 */
2238
864dbfa3 2239static char *
aec46f14 2240S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2241{
25da4f38 2242 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2243 char * const ebuf = ptr;
25da4f38 2244 int sign;
25da4f38
IZ
2245
2246 if (is_uv)
2247 sign = 0;
2248 else if (iv >= 0) {
2249 uv = iv;
2250 sign = 0;
2251 } else {
2252 uv = -iv;
2253 sign = 1;
2254 }
2255 do {
eb160463 2256 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2257 } while (uv /= 10);
2258 if (sign)
2259 *--ptr = '-';
2260 *peob = ebuf;
2261 return ptr;
2262}
2263
9af30d34
NC
2264/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2265 * a regexp to its stringified form.
2266 */
2267
2268static char *
2269S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
00b6aa41 2270 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2271
2272 if (!mg->mg_ptr) {
2273 const char *fptr = "msix";
2274 char reflags[6];
2275 char ch;
2276 int left = 0;
2277 int right = 4;
00b6aa41 2278 bool need_newline = 0;
9af30d34
NC
2279 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2280
2281 while((ch = *fptr++)) {
2282 if(reganch & 1) {
2283 reflags[left++] = ch;
2284 }
2285 else {
2286 reflags[right--] = ch;
2287 }
2288 reganch >>= 1;
2289 }
2290 if(left != 4) {
2291 reflags[left] = '-';
2292 left = 5;
2293 }
2294
2295 mg->mg_len = re->prelen + 4 + left;
2296 /*
2297 * If /x was used, we have to worry about a regex ending with a
2298 * comment later being embedded within another regex. If so, we don't
2299 * want this regex's "commentization" to leak out to the right part of
2300 * the enclosing regex, we must cap it with a newline.
2301 *
2302 * So, if /x was used, we scan backwards from the end of the regex. If
2303 * we find a '#' before we find a newline, we need to add a newline
2304 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2305 * we don't need to add anything. -jfriedl
2306 */
2307 if (PMf_EXTENDED & re->reganch) {
2308 const char *endptr = re->precomp + re->prelen;
2309 while (endptr >= re->precomp) {
2310 const char c = *(endptr--);
2311 if (c == '\n')
2312 break; /* don't need another */
2313 if (c == '#') {
2314 /* we end while in a comment, so we need a newline */
2315 mg->mg_len++; /* save space for it */
2316 need_newline = 1; /* note to add it */
2317 break;
2318 }
2319 }
2320 }
2321
2322 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2323 mg->mg_ptr[0] = '(';
2324 mg->mg_ptr[1] = '?';
2325 Copy(reflags, mg->mg_ptr+2, left, char);
2326 *(mg->mg_ptr+left+2) = ':';
2327 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2328 if (need_newline)
2329 mg->mg_ptr[mg->mg_len - 2] = '\n';
2330 mg->mg_ptr[mg->mg_len - 1] = ')';
2331 mg->mg_ptr[mg->mg_len] = 0;
2332 }
2333 PL_reginterp_cnt += re->program[0].next_off;
2334
2335 if (re->reganch & ROPT_UTF8)
2336 SvUTF8_on(sv);
2337 else
2338 SvUTF8_off(sv);
2339 if (lp)
2340 *lp = mg->mg_len;
2341 return mg->mg_ptr;
2342}
2343
645c22ef
DM
2344/*
2345=for apidoc sv_2pv_flags
2346
ff276b08 2347Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2348If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2349if necessary.
2350Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2351usually end up here too.
2352
2353=cut
2354*/
2355
8d6d96c1
HS
2356char *
2357Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2358{
79072805
LW
2359 register char *s;
2360 int olderrno;
2361
463ee0b2 2362 if (!sv) {
cdb061a3
NC
2363 if (lp)
2364 *lp = 0;
73d840c0 2365 return (char *)"";
463ee0b2 2366 }
8990e307 2367 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2368 if (flags & SV_GMAGIC)
2369 mg_get(sv);
463ee0b2 2370 if (SvPOKp(sv)) {
cdb061a3
NC
2371 if (lp)
2372 *lp = SvCUR(sv);
10516c54
NC
2373 if (flags & SV_MUTABLE_RETURN)
2374 return SvPVX_mutable(sv);
4d84ee25
NC
2375 if (flags & SV_CONST_RETURN)
2376 return (char *)SvPVX_const(sv);
463ee0b2
LW
2377 return SvPVX(sv);
2378 }
75dfc8ec
NC
2379 if (SvIOKp(sv) || SvNOKp(sv)) {
2380 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2381 STRLEN len;
2382
2383 if (SvIOKp(sv)) {
e8ada2d0
NC
2384 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2385 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2386 } else {
e8ada2d0
NC
2387 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2388 len = strlen(tbuf);
75dfc8ec
NC
2389 }
2390 if (SvROK(sv)) { /* XXX Skip this when sv_pvn_force calls */
2391 /* Sneaky stuff here */
00b6aa41 2392 SV * const tsv = newSVpvn(tbuf, len);
75dfc8ec
NC
2393
2394 sv_2mortal(tsv);
2395 if (lp)
2396 *lp = SvCUR(tsv);
2397 return SvPVX(tsv);
2398 }
2399 else {
2400 dVAR;
2401
2402#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2403 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2404 tbuf[0] = '0';
2405 tbuf[1] = 0;
75dfc8ec
NC
2406 len = 1;
2407 }
2408#endif
2409 SvUPGRADE(sv, SVt_PV);
2410 if (lp)
2411 *lp = len;
2412 s = SvGROW_mutable(sv, len + 1);
2413 SvCUR_set(sv, len);
2414 SvPOKp_on(sv);
e8ada2d0 2415 return memcpy(s, tbuf, len + 1);
75dfc8ec 2416 }
463ee0b2 2417 }
16d20bd9 2418 if (!SvROK(sv)) {
d008e5eb 2419 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 2420 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 2421 report_uninit(sv);
c6ee37c5 2422 }
cdb061a3
NC
2423 if (lp)
2424 *lp = 0;
73d840c0 2425 return (char *)"";
16d20bd9 2426 }
71c558c3
NC
2427 /* Else this will drop through into the SvROK case just below, which
2428 will return within the {} for all code paths. */
463ee0b2 2429 }
ed6116ce
LW
2430 if (SvTHINKFIRST(sv)) {
2431 if (SvROK(sv)) {
a0d0e21e 2432 SV* tmpstr;
d8eae41e 2433
1554e226 2434 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,string)) &&
b4b9a328 2435 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
50adf7d2
NC
2436 /* Unwrap this: */
2437 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr); */
2438
2439 char *pv;
2440 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2441 if (flags & SV_CONST_RETURN) {
2442 pv = (char *) SvPVX_const(tmpstr);
2443 } else {
2444 pv = (flags & SV_MUTABLE_RETURN)
2445 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2446 }
2447 if (lp)
2448 *lp = SvCUR(tmpstr);
2449 } else {
2450 pv = sv_2pv_flags(tmpstr, lp, flags);
2451 }
446eaa42
YST
2452 if (SvUTF8(tmpstr))
2453 SvUTF8_on(sv);
2454 else
2455 SvUTF8_off(sv);
2456 return pv;
d8eae41e 2457 } else {
75dfc8ec 2458 SV *tsv;
f9277f47 2459 MAGIC *mg;
d8eae41e
NC
2460 const SV *const referent = (SV*)SvRV(sv);
2461
2462 if (!referent) {
042dae7a
NC
2463 tsv = sv_2mortal(newSVpvn("NULLREF", 7));
2464 } else if (SvTYPE(referent) == SVt_PVMG
2465 && ((SvFLAGS(referent) &
2466 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2467 == (SVs_OBJECT|SVs_SMG))
2468 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
9af30d34 2469 return S_stringify_regexp(aTHX_ sv, mg, lp);
d8eae41e
NC
2470 } else {
2471 const char *const typestr = sv_reftype(referent, 0);
2472
2473 tsv = sv_newmortal();
2474 if (SvOBJECT(referent)) {
2475 const char *const name = HvNAME_get(SvSTASH(referent));
2476 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2477 name ? name : "__ANON__" , typestr,
2478 PTR2UV(referent));
2479 }
2480 else
2481 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2482 PTR2UV(referent));
c080367d 2483 }
042dae7a
NC
2484 if (lp)
2485 *lp = SvCUR(tsv);
2486 return SvPVX(tsv);
463ee0b2 2487 }
79072805 2488 }
0336b60e 2489 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2490 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2491 report_uninit(sv);
cdb061a3
NC
2492 if (lp)
2493 *lp = 0;
73d840c0 2494 return (char *)"";
79072805 2495 }
79072805 2496 }
28e5dec8
JH
2497 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2498 /* I'm assuming that if both IV and NV are equally valid then
2499 converting the IV is going to be more efficient */
e1ec3a88
AL
2500 const U32 isIOK = SvIOK(sv);
2501 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2502 char buf[TYPE_CHARS(UV)];
2503 char *ebuf, *ptr;
2504
2505 if (SvTYPE(sv) < SVt_PVIV)
2506 sv_upgrade(sv, SVt_PVIV);
2507 if (isUIOK)
2508 ptr = uiv_2buf(buf, 0, SvUVX(sv), 1, &ebuf);
2509 else
2510 ptr = uiv_2buf(buf, SvIVX(sv), 0, 0, &ebuf);
5902b6a9
NC
2511 /* inlined from sv_setpvn */
2512 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2513 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2514 SvCUR_set(sv, ebuf - ptr);
2515 s = SvEND(sv);
2516 *s = '\0';
2517 if (isIOK)
2518 SvIOK_on(sv);
2519 else
2520 SvIOKp_on(sv);
2521 if (isUIOK)
2522 SvIsUV_on(sv);
2523 }
2524 else if (SvNOKp(sv)) {
79072805
LW
2525 if (SvTYPE(sv) < SVt_PVNV)
2526 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2527 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2528 s = SvGROW_mutable(sv, NV_DIG + 20);
79072805 2529 olderrno = errno; /* some Xenix systems wipe out errno here */
79072805 2530#ifdef apollo
463ee0b2 2531 if (SvNVX(sv) == 0.0)
79072805
LW
2532 (void)strcpy(s,"0");
2533 else
2534#endif /*apollo*/
bbce6d69 2535 {
2d4389e4 2536 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2537 }
79072805 2538 errno = olderrno;
a0d0e21e
LW
2539#ifdef FIXNEGATIVEZERO
2540 if (*s == '-' && s[1] == '0' && !s[2])
2541 strcpy(s,"0");
2542#endif
79072805
LW
2543 while (*s) s++;
2544#ifdef hcx
2545 if (s[-1] == '.')
46fc3d4c 2546 *--s = '\0';
79072805
LW
2547#endif
2548 }
79072805 2549 else {
041457d9 2550 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2551 report_uninit(sv);
cdb061a3 2552 if (lp)
00b6aa41 2553 *lp = 0;
25da4f38
IZ
2554 if (SvTYPE(sv) < SVt_PV)
2555 /* Typically the caller expects that sv_any is not NULL now. */
2556 sv_upgrade(sv, SVt_PV);
73d840c0 2557 return (char *)"";
79072805 2558 }
cdb061a3 2559 {
823a54a3 2560 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2561 if (lp)
2562 *lp = len;
2563 SvCUR_set(sv, len);
2564 }
79072805 2565 SvPOK_on(sv);
1d7c1841 2566 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2567 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2568 if (flags & SV_CONST_RETURN)
2569 return (char *)SvPVX_const(sv);
10516c54
NC
2570 if (flags & SV_MUTABLE_RETURN)
2571 return SvPVX_mutable(sv);
463ee0b2
LW
2572 return SvPVX(sv);
2573}
2574
645c22ef 2575/*
6050d10e
JP
2576=for apidoc sv_copypv
2577
2578Copies a stringified representation of the source SV into the
2579destination SV. Automatically performs any necessary mg_get and
54f0641b 2580coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2581UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2582sv_2pv[_flags] but operates directly on an SV instead of just the
2583string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2584would lose the UTF-8'ness of the PV.
2585
2586=cut
2587*/
2588
2589void
2590Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2591{
446eaa42 2592 STRLEN len;
53c1dcc0 2593 const char * const s = SvPV_const(ssv,len);
cb50f42d 2594 sv_setpvn(dsv,s,len);
446eaa42 2595 if (SvUTF8(ssv))
cb50f42d 2596 SvUTF8_on(dsv);
446eaa42 2597 else
cb50f42d 2598 SvUTF8_off(dsv);
6050d10e
JP
2599}
2600
2601/*
645c22ef
DM
2602=for apidoc sv_2pvbyte
2603
2604Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2605to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2606side-effect.
2607
2608Usually accessed via the C<SvPVbyte> macro.
2609
2610=cut
2611*/
2612
7340a771
GS
2613char *
2614Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2615{
0875d2fe 2616 sv_utf8_downgrade(sv,0);
97972285 2617 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2618}
2619
645c22ef 2620/*
035cbb0e
RGS
2621=for apidoc sv_2pvutf8
2622
2623Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2624to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2625
2626Usually accessed via the C<SvPVutf8> macro.
2627
2628=cut
2629*/
645c22ef 2630
7340a771
GS
2631char *
2632Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2633{
035cbb0e
RGS
2634 sv_utf8_upgrade(sv);
2635 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2636}
1c846c1f 2637
7ee2227d 2638
645c22ef
DM
2639/*
2640=for apidoc sv_2bool
2641
2642This function is only called on magical items, and is only used by
8cf8f3d1 2643sv_true() or its macro equivalent.
645c22ef
DM
2644
2645=cut
2646*/
2647
463ee0b2 2648bool
864dbfa3 2649Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2650{
5b295bef 2651 SvGETMAGIC(sv);
463ee0b2 2652
a0d0e21e
LW
2653 if (!SvOK(sv))
2654 return 0;
2655 if (SvROK(sv)) {
a0d0e21e 2656 SV* tmpsv;
1554e226 2657 if (SvAMAGIC(sv) && (tmpsv=AMG_CALLun(sv,bool_)) &&
9e3013b1 2658 (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
8a31060d 2659 return (bool)SvTRUE(tmpsv);
a0d0e21e
LW
2660 return SvRV(sv) != 0;
2661 }
463ee0b2 2662 if (SvPOKp(sv)) {
53c1dcc0
AL
2663 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2664 if (Xpvtmp &&
339049b0 2665 (*sv->sv_u.svu_pv > '0' ||
11343788 2666 Xpvtmp->xpv_cur > 1 ||
339049b0 2667 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2668 return 1;
2669 else
2670 return 0;
2671 }
2672 else {
2673 if (SvIOKp(sv))
2674 return SvIVX(sv) != 0;
2675 else {
2676 if (SvNOKp(sv))
2677 return SvNVX(sv) != 0.0;
2678 else
2679 return FALSE;
2680 }
2681 }
79072805
LW
2682}
2683
c461cf8f
JH
2684/*
2685=for apidoc sv_utf8_upgrade
2686
78ea37eb 2687Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2688Forces the SV to string form if it is not already.
4411f3b6
NIS
2689Always sets the SvUTF8 flag to avoid future validity checks even
2690if all the bytes have hibit clear.
c461cf8f 2691
13a6c0e0
JH
2692This is not as a general purpose byte encoding to Unicode interface:
2693use the Encode extension for that.
2694
8d6d96c1
HS
2695=for apidoc sv_utf8_upgrade_flags
2696
78ea37eb 2697Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2698Forces the SV to string form if it is not already.
8d6d96c1
HS
2699Always sets the SvUTF8 flag to avoid future validity checks even
2700if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2701will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2702C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2703
13a6c0e0
JH
2704This is not as a general purpose byte encoding to Unicode interface:
2705use the Encode extension for that.
2706
8d6d96c1
HS
2707=cut
2708*/
2709
2710STRLEN
2711Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2712{
808c356f
RGS
2713 if (sv == &PL_sv_undef)
2714 return 0;
e0e62c2a
NIS
2715 if (!SvPOK(sv)) {
2716 STRLEN len = 0;
d52b7888
NC
2717 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2718 (void) sv_2pv_flags(sv,&len, flags);
2719 if (SvUTF8(sv))
2720 return len;
2721 } else {
2722 (void) SvPV_force(sv,len);
2723 }
e0e62c2a 2724 }
4411f3b6 2725
f5cee72b 2726 if (SvUTF8(sv)) {
5fec3b1d 2727 return SvCUR(sv);
f5cee72b 2728 }
5fec3b1d 2729
765f542d
NC
2730 if (SvIsCOW(sv)) {
2731 sv_force_normal_flags(sv, 0);
db42d148
NIS
2732 }
2733
88632417 2734 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2735 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2736 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2737 /* This function could be much more efficient if we
2738 * had a FLAG in SVs to signal if there are any hibit
2739 * chars in the PV. Given that there isn't such a flag
2740 * make the loop as fast as possible. */
00b6aa41 2741 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 2742 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 2743 const U8 *t = s;
c4e7c712
NC
2744
2745 while (t < e) {
53c1dcc0 2746 const U8 ch = *t++;
00b6aa41
AL
2747 /* Check for hi bit */
2748 if (!NATIVE_IS_INVARIANT(ch)) {
2749 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
2750 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
2751
2752 SvPV_free(sv); /* No longer using what was there before. */
2753 SvPV_set(sv, (char*)recoded);
2754 SvCUR_set(sv, len - 1);
2755 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 2756 break;
00b6aa41 2757 }
c4e7c712
NC
2758 }
2759 /* Mark as UTF-8 even if no hibit - saves scanning loop */
2760 SvUTF8_on(sv);
560a288e 2761 }
4411f3b6 2762 return SvCUR(sv);
560a288e
GS
2763}
2764
c461cf8f
JH
2765/*
2766=for apidoc sv_utf8_downgrade
2767
78ea37eb
TS
2768Attempts to convert the PV of an SV from characters to bytes.
2769If the PV contains a character beyond byte, this conversion will fail;
2770in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
2771true, croaks.
2772
13a6c0e0
JH
2773This is not as a general purpose Unicode to byte encoding interface:
2774use the Encode extension for that.
2775
c461cf8f
JH
2776=cut
2777*/
2778
560a288e
GS
2779bool
2780Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
2781{
78ea37eb 2782 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 2783 if (SvCUR(sv)) {
03cfe0ae 2784 U8 *s;
652088fc 2785 STRLEN len;
fa301091 2786
765f542d
NC
2787 if (SvIsCOW(sv)) {
2788 sv_force_normal_flags(sv, 0);
2789 }
03cfe0ae
NIS
2790 s = (U8 *) SvPV(sv, len);
2791 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
2792 if (fail_ok)
2793 return FALSE;
2794 else {
2795 if (PL_op)
2796 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 2797 OP_DESC(PL_op));
fa301091
JH
2798 else
2799 Perl_croak(aTHX_ "Wide character");
2800 }
4b3603a4 2801 }
b162af07 2802 SvCUR_set(sv, len);
67e989fb 2803 }
560a288e 2804 }
ffebcc3e 2805 SvUTF8_off(sv);
560a288e
GS
2806 return TRUE;
2807}
2808
c461cf8f
JH
2809/*
2810=for apidoc sv_utf8_encode
2811
78ea37eb
TS
2812Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
2813flag off so that it looks like octets again.
c461cf8f
JH
2814
2815=cut
2816*/
2817
560a288e
GS
2818void
2819Perl_sv_utf8_encode(pTHX_ register SV *sv)
2820{
4411f3b6 2821 (void) sv_utf8_upgrade(sv);
4c94c214
NC
2822 if (SvIsCOW(sv)) {
2823 sv_force_normal_flags(sv, 0);
2824 }
2825 if (SvREADONLY(sv)) {
2826 Perl_croak(aTHX_ PL_no_modify);
2827 }
560a288e
GS
2828 SvUTF8_off(sv);
2829}
2830
4411f3b6
NIS
2831/*
2832=for apidoc sv_utf8_decode
2833
78ea37eb
TS
2834If the PV of the SV is an octet sequence in UTF-8
2835and contains a multiple-byte character, the C<SvUTF8> flag is turned on
2836so that it looks like a character. If the PV contains only single-byte
2837characters, the C<SvUTF8> flag stays being off.
2838Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
2839
2840=cut
2841*/
2842
560a288e
GS
2843bool
2844Perl_sv_utf8_decode(pTHX_ register SV *sv)
2845{
78ea37eb 2846 if (SvPOKp(sv)) {
93524f2b
NC
2847 const U8 *c;
2848 const U8 *e;
9cbac4c7 2849
645c22ef
DM
2850 /* The octets may have got themselves encoded - get them back as
2851 * bytes
2852 */
2853 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
2854 return FALSE;
2855
2856 /* it is actually just a matter of turning the utf8 flag on, but
2857 * we want to make sure everything inside is valid utf8 first.
2858 */
93524f2b 2859 c = (const U8 *) SvPVX_const(sv);
63cd0674 2860 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 2861 return FALSE;
93524f2b 2862 e = (const U8 *) SvEND(sv);
511c2ff0 2863 while (c < e) {
b64e5050 2864 const U8 ch = *c++;
c4d5f83a 2865 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
2866 SvUTF8_on(sv);
2867 break;
2868 }
560a288e 2869 }
560a288e
GS
2870 }
2871 return TRUE;
2872}
2873
954c1994
GS
2874/*
2875=for apidoc sv_setsv
2876
645c22ef
DM
2877Copies the contents of the source SV C<ssv> into the destination SV
2878C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
2879function if the source SV needs to be reused. Does not handle 'set' magic.
2880Loosely speaking, it performs a copy-by-value, obliterating any previous
2881content of the destination.
2882
2883You probably want to use one of the assortment of wrappers, such as
2884C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
2885C<SvSetMagicSV_nosteal>.
2886
8d6d96c1
HS
2887=for apidoc sv_setsv_flags
2888
645c22ef
DM
2889Copies the contents of the source SV C<ssv> into the destination SV
2890C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
2891function if the source SV needs to be reused. Does not handle 'set' magic.
2892Loosely speaking, it performs a copy-by-value, obliterating any previous
2893content of the destination.
2894If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
2895C<ssv> if appropriate, else not. If the C<flags> parameter has the
2896C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
2897and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
2898
2899You probably want to use one of the assortment of wrappers, such as
2900C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
2901C<SvSetMagicSV_nosteal>.
2902
2903This is the primary function for copying scalars, and most other
2904copy-ish functions and macros use this underneath.
8d6d96c1
HS
2905
2906=cut
2907*/
2908
2909void
2910Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
2911{
8990e307
LW
2912 register U32 sflags;
2913 register int dtype;
2914 register int stype;
463ee0b2 2915
79072805
LW
2916 if (sstr == dstr)
2917 return;
765f542d 2918 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 2919 if (!sstr)
3280af22 2920 sstr = &PL_sv_undef;
8990e307
LW
2921 stype = SvTYPE(sstr);
2922 dtype = SvTYPE(dstr);
79072805 2923
a0d0e21e 2924 SvAMAGIC_off(dstr);
7a5fa8a2 2925 if ( SvVOK(dstr) )
ece467f9
JP
2926 {
2927 /* need to nuke the magic */
2928 mg_free(dstr);
2929 SvRMAGICAL_off(dstr);
2930 }
9e7bc3e8 2931
463ee0b2 2932 /* There's a lot of redundancy below but we're going for speed here */
79072805 2933
8990e307 2934 switch (stype) {
79072805 2935 case SVt_NULL:
aece5585 2936 undef_sstr:
20408e3c
GS
2937 if (dtype != SVt_PVGV) {
2938 (void)SvOK_off(dstr);
2939 return;
2940 }
2941 break;
463ee0b2 2942 case SVt_IV:
aece5585
GA
2943 if (SvIOK(sstr)) {
2944 switch (dtype) {
2945 case SVt_NULL:
8990e307 2946 sv_upgrade(dstr, SVt_IV);
aece5585
GA
2947 break;
2948 case SVt_NV:
8990e307 2949 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
2950 break;
2951 case SVt_RV:
2952 case SVt_PV:
a0d0e21e 2953 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
2954 break;
2955 }
2956 (void)SvIOK_only(dstr);
45977657 2957 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
2958 if (SvIsUV(sstr))
2959 SvIsUV_on(dstr);
27c9684d
AP
2960 if (SvTAINTED(sstr))
2961 SvTAINT(dstr);
aece5585 2962 return;
8990e307 2963 }
aece5585
GA
2964 goto undef_sstr;
2965
463ee0b2 2966 case SVt_NV:
aece5585
GA
2967 if (SvNOK(sstr)) {
2968 switch (dtype) {
2969 case SVt_NULL:
2970 case SVt_IV:
8990e307 2971 sv_upgrade(dstr, SVt_NV);
aece5585
GA
2972 break;
2973 case SVt_RV:
2974 case SVt_PV:
2975 case SVt_PVIV:
a0d0e21e 2976 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
2977 break;
2978 }
9d6ce603 2979 SvNV_set(dstr, SvNVX(sstr));
aece5585 2980 (void)SvNOK_only(dstr);
27c9684d
AP
2981 if (SvTAINTED(sstr))
2982 SvTAINT(dstr);
aece5585 2983 return;
8990e307 2984 }
aece5585
GA
2985 goto undef_sstr;
2986
ed6116ce 2987 case SVt_RV:
8990e307 2988 if (dtype < SVt_RV)
ed6116ce 2989 sv_upgrade(dstr, SVt_RV);
c07a80fd 2990 else if (dtype == SVt_PVGV &&
23bb1b96 2991 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
c07a80fd 2992 sstr = SvRV(sstr);
a5f75d66 2993 if (sstr == dstr) {
1d7c1841
GS
2994 if (GvIMPORTED(dstr) != GVf_IMPORTED
2995 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
2996 {
a5f75d66 2997 GvIMPORTED_on(dstr);
1d7c1841 2998 }
a5f75d66
AD
2999 GvMULTI_on(dstr);
3000 return;
3001 }
c07a80fd 3002 goto glob_assign;
3003 }
ed6116ce 3004 break;
fc36a67e 3005 case SVt_PVFM:
f8c7b90f 3006#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3007 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3008 if (dtype < SVt_PVIV)
3009 sv_upgrade(dstr, SVt_PVIV);
3010 break;
3011 }
3012 /* Fall through */
3013#endif
3014 case SVt_PV:
8990e307 3015 if (dtype < SVt_PV)
463ee0b2 3016 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3017 break;
3018 case SVt_PVIV:
8990e307 3019 if (dtype < SVt_PVIV)
463ee0b2 3020 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3021 break;
3022 case SVt_PVNV:
8990e307 3023 if (dtype < SVt_PVNV)
463ee0b2 3024 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3025 break;
4633a7c4
LW
3026 case SVt_PVAV:
3027 case SVt_PVHV:
3028 case SVt_PVCV:
4633a7c4 3029 case SVt_PVIO:
a3b680e6
AL
3030 {
3031 const char * const type = sv_reftype(sstr,0);
533c011a 3032 if (PL_op)
a3b680e6 3033 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3034 else
a3b680e6
AL
3035 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3036 }
4633a7c4
LW
3037 break;
3038
79072805 3039 case SVt_PVGV:
8990e307 3040 if (dtype <= SVt_PVGV) {
c07a80fd 3041 glob_assign:
a5f75d66 3042 if (dtype != SVt_PVGV) {
a3b680e6
AL
3043 const char * const name = GvNAME(sstr);
3044 const STRLEN len = GvNAMELEN(sstr);
b76195c2
DM
3045 /* don't upgrade SVt_PVLV: it can hold a glob */
3046 if (dtype != SVt_PVLV)
3047 sv_upgrade(dstr, SVt_PVGV);
14befaf4 3048 sv_magic(dstr, dstr, PERL_MAGIC_glob, Nullch, 0);
e15faf7d
NC
3049 GvSTASH(dstr) = GvSTASH(sstr);
3050 if (GvSTASH(dstr))
3051 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
a0d0e21e
LW
3052 GvNAME(dstr) = savepvn(name, len);
3053 GvNAMELEN(dstr) = len;
3054 SvFAKE_on(dstr); /* can coerce to non-glob */
3055 }
5bd07a3d 3056
7fb37951
AMS
3057#ifdef GV_UNIQUE_CHECK
3058 if (GvUNIQUE((GV*)dstr)) {
5bd07a3d
DM
3059 Perl_croak(aTHX_ PL_no_modify);
3060 }
3061#endif
3062
a0d0e21e 3063 (void)SvOK_off(dstr);
a5f75d66 3064 GvINTRO_off(dstr); /* one-shot flag */
1edc1566 3065 gp_free((GV*)dstr);
79072805 3066 GvGP(dstr) = gp_ref(GvGP(sstr));
27c9684d
AP
3067 if (SvTAINTED(sstr))
3068 SvTAINT(dstr);
1d7c1841
GS
3069 if (GvIMPORTED(dstr) != GVf_IMPORTED
3070 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3071 {
a5f75d66 3072 GvIMPORTED_on(dstr);
1d7c1841 3073 }
a5f75d66 3074 GvMULTI_on(dstr);
79072805
LW
3075 return;
3076 }
3077 /* FALL THROUGH */
3078
3079 default:
8d6d96c1 3080 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3081 mg_get(sstr);
eb160463 3082 if ((int)SvTYPE(sstr) != stype) {
973f89ab
CS
3083 stype = SvTYPE(sstr);
3084 if (stype == SVt_PVGV && dtype <= SVt_PVGV)
3085 goto glob_assign;
3086 }
3087 }
ded42b9f 3088 if (stype == SVt_PVLV)
862a34c6 3089 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3090 else
862a34c6 3091 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3092 }
3093
8990e307
LW
3094 sflags = SvFLAGS(sstr);
3095
3096 if (sflags & SVf_ROK) {
3097 if (dtype >= SVt_PV) {
3098 if (dtype == SVt_PVGV) {
823a54a3 3099 SV * const sref = SvREFCNT_inc(SvRV(sstr));
8990e307 3100 SV *dref = 0;
a3b680e6 3101 const int intro = GvINTRO(dstr);
a0d0e21e 3102
7fb37951
AMS
3103#ifdef GV_UNIQUE_CHECK
3104 if (GvUNIQUE((GV*)dstr)) {
5bd07a3d
DM
3105 Perl_croak(aTHX_ PL_no_modify);
3106 }
3107#endif
3108
a0d0e21e 3109 if (intro) {
a5f75d66 3110 GvINTRO_off(dstr); /* one-shot flag */
1d7c1841 3111 GvLINE(dstr) = CopLINE(PL_curcop);
1edc1566 3112 GvEGV(dstr) = (GV*)dstr;
a0d0e21e 3113 }
a5f75d66 3114 GvMULTI_on(dstr);
8990e307
LW
3115 switch (SvTYPE(sref)) {
3116 case SVt_PVAV:
a0d0e21e 3117 if (intro)
890ed176 3118 SAVEGENERICSV(GvAV(dstr));
a0d0e21e
LW
3119 else
3120 dref = (SV*)GvAV(dstr);
8990e307 3121 GvAV(dstr) = (AV*)sref;
39bac7f7 3122 if (!GvIMPORTED_AV(dstr)
1d7c1841
GS
3123 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3124 {
a5f75d66 3125 GvIMPORTED_AV_on(dstr);
1d7c1841 3126 }
8990e307
LW
3127 break;
3128 case SVt_PVHV:
a0d0e21e 3129 if (intro)
890ed176 3130 SAVEGENERICSV(GvHV(dstr));
a0d0e21e
LW
3131 else
3132 dref = (SV*)GvHV(dstr);
8990e307 3133 GvHV(dstr) = (HV*)sref;
39bac7f7 3134 if (!GvIMPORTED_HV(dstr)
1d7c1841
GS
3135 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3136 {
a5f75d66 3137 GvIMPORTED_HV_on(dstr);
1d7c1841 3138 }
8990e307
LW
3139 break;
3140 case SVt_PVCV:
8ebc5c01 3141 if (intro) {
3142 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3143 SvREFCNT_dec(GvCV(dstr));
3144 GvCV(dstr) = Nullcv;
68dc0745 3145 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3280af22 3146 PL_sub_generation++;
8ebc5c01 3147 }
890ed176 3148 SAVEGENERICSV(GvCV(dstr));
8ebc5c01 3149 }
68dc0745 3150 else
3151 dref = (SV*)GvCV(dstr);
3152 if (GvCV(dstr) != (CV*)sref) {
823a54a3 3153 CV* const cv = GvCV(dstr);
4633a7c4 3154 if (cv) {
68dc0745 3155 if (!GvCVGEN((GV*)dstr) &&
3156 (CvROOT(cv) || CvXSUB(cv)))
3157 {
beab0874
JT
3158 /* Redefining a sub - warning is mandatory if
3159 it was a const and its value changed. */
3160 if (ckWARN(WARN_REDEFINE)
3161 || (CvCONST(cv)
3162 && (!CvCONST((CV*)sref)
3163 || sv_cmp(cv_const_sv(cv),
3164 cv_const_sv((CV*)sref)))))
3165 {
9014280d 3166 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
beab0874 3167 CvCONST(cv)
910764e6
RGS
3168 ? "Constant subroutine %s::%s redefined"
3169 : "Subroutine %s::%s redefined",
bfcb3514 3170 HvNAME_get(GvSTASH((GV*)dstr)),
beab0874
JT
3171 GvENAME((GV*)dstr));
3172 }
9607fc9c 3173 }
fb24441d
RGS
3174 if (!intro)
3175 cv_ckproto(cv, (GV*)dstr,
93524f2b
NC
3176 SvPOK(sref)
3177 ? SvPVX_const(sref) : Nullch);
4633a7c4 3178 }
a5f75d66 3179 GvCV(dstr) = (CV*)sref;
7a4c00b4 3180 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
a5f75d66 3181 GvASSUMECV_on(dstr);
3280af22 3182 PL_sub_generation++;
a5f75d66 3183 }
39bac7f7 3184 if (!GvIMPORTED_CV(dstr)
1d7c1841
GS
3185 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3186 {
a5f75d66 3187 GvIMPORTED_CV_on(dstr);
1d7c1841 3188 }
8990e307 3189 break;
91bba347
LW
3190 case SVt_PVIO:
3191 if (intro)
890ed176 3192 SAVEGENERICSV(GvIOp(dstr));
91bba347
LW
3193 else
3194 dref = (SV*)GvIOp(dstr);
3195 GvIOp(dstr) = (IO*)sref;
3196 break;
f4d13ee9
JH
3197 case SVt_PVFM:
3198 if (intro)
890ed176 3199 SAVEGENERICSV(GvFORM(dstr));
f4d13ee9
JH
3200 else
3201 dref = (SV*)GvFORM(dstr);
3202 GvFORM(dstr) = (CV*)sref;
3203 break;
8990e307 3204 default:
a0d0e21e 3205 if (intro)
890ed176 3206 SAVEGENERICSV(GvSV(dstr));
a0d0e21e
LW
3207 else
3208 dref = (SV*)GvSV(dstr);
8990e307 3209 GvSV(dstr) = sref;
39bac7f7 3210 if (!GvIMPORTED_SV(dstr)
1d7c1841
GS
3211 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3212 {
a5f75d66 3213 GvIMPORTED_SV_on(dstr);
1d7c1841 3214 }
8990e307
LW
3215 break;
3216 }
3217 if (dref)
3218 SvREFCNT_dec(dref);
27c9684d
AP
3219 if (SvTAINTED(sstr))
3220 SvTAINT(dstr);
8990e307
LW
3221 return;
3222 }
3f7c398e 3223 if (SvPVX_const(dstr)) {
8bd4d4c5 3224 SvPV_free(dstr);
b162af07
SP
3225 SvLEN_set(dstr, 0);
3226 SvCUR_set(dstr, 0);
a0d0e21e 3227 }
8990e307 3228 }
a0d0e21e 3229 (void)SvOK_off(dstr);
b162af07 3230 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
ed6116ce 3231 SvROK_on(dstr);
8990e307 3232 if (sflags & SVp_NOK) {
3332b3c1
JH
3233 SvNOKp_on(dstr);
3234 /* Only set the public OK flag if the source has public OK. */
3235 if (sflags & SVf_NOK)
3236 SvFLAGS(dstr) |= SVf_NOK;
9d6ce603 3237 SvNV_set(dstr, SvNVX(sstr));
ed6116ce 3238 }
8990e307 3239 if (sflags & SVp_IOK) {
3332b3c1
JH
3240 (void)SvIOKp_on(dstr);
3241 if (sflags & SVf_IOK)
3242 SvFLAGS(dstr) |= SVf_IOK;
2b1c7e3e 3243 if (sflags & SVf_IVisUV)
25da4f38 3244 SvIsUV_on(dstr);
45977657 3245 SvIV_set(dstr, SvIVX(sstr));
ed6116ce 3246 }
a0d0e21e
LW
3247 if (SvAMAGIC(sstr)) {
3248 SvAMAGIC_on(dstr);
3249 }
ed6116ce 3250 }
8990e307 3251 else if (sflags & SVp_POK) {
765f542d 3252 bool isSwipe = 0;
79072805
LW
3253
3254 /*
3255 * Check to see if we can just swipe the string. If so, it's a
3256 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3257 * It might even be a win on short strings if SvPVX_const(dstr)
3258 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3259 */
3260
120fac95
NC
3261 /* Whichever path we take through the next code, we want this true,
3262 and doing it now facilitates the COW check. */
3263 (void)SvPOK_only(dstr);
3264
765f542d 3265 if (
b8f9541a
NC
3266 /* We're not already COW */
3267 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3268#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3269 /* or we are, but dstr isn't a suitable target. */
3270 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3271#endif
3272 )
765f542d 3273 &&
765f542d
NC
3274 !(isSwipe =
3275 (sflags & SVs_TEMP) && /* slated for free anyway? */
3276 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3277 (!(flags & SV_NOSTEAL)) &&
3278 /* and we're allowed to steal temps */
765f542d
NC
3279 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3280 SvLEN(sstr) && /* and really is a string */
645c22ef 3281 /* and won't be needed again, potentially */
765f542d 3282 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3283#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3284 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3285 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3286 && SvTYPE(sstr) >= SVt_PVIV)
3287#endif
3288 ) {
3289 /* Failed the swipe test, and it's not a shared hash key either.
3290 Have to copy the string. */
3291 STRLEN len = SvCUR(sstr);
3292 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3293 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3294 SvCUR_set(dstr, len);
3295 *SvEND(dstr) = '\0';
765f542d 3296 } else {
f8c7b90f 3297 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3298 be true in here. */
765f542d
NC
3299 /* Either it's a shared hash key, or it's suitable for
3300 copy-on-write or we can swipe the string. */
46187eeb 3301 if (DEBUG_C_TEST) {
ed252734 3302 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3303 sv_dump(sstr);
3304 sv_dump(dstr);
46187eeb 3305 }
f8c7b90f 3306#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3307 if (!isSwipe) {
3308 /* I believe I should acquire a global SV mutex if
3309 it's a COW sv (not a shared hash key) to stop
3310 it going un copy-on-write.
3311 If the source SV has gone un copy on write between up there
3312 and down here, then (assert() that) it is of the correct
3313 form to make it copy on write again */
3314 if ((sflags & (SVf_FAKE | SVf_READONLY))
3315 != (SVf_FAKE | SVf_READONLY)) {
3316 SvREADONLY_on(sstr);
3317 SvFAKE_on(sstr);
3318 /* Make the source SV into a loop of 1.
3319 (about to become 2) */
a29f6d03 3320 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3321 }
3322 }
3323#endif
3324 /* Initial code is common. */
94010e71
NC
3325 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3326 SvPV_free(dstr);
79072805 3327 }
765f542d 3328
765f542d
NC
3329 if (!isSwipe) {
3330 /* making another shared SV. */
3331 STRLEN cur = SvCUR(sstr);
3332 STRLEN len = SvLEN(sstr);
f8c7b90f 3333#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3334 if (len) {
b8f9541a 3335 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3336 /* SvIsCOW_normal */
3337 /* splice us in between source and next-after-source. */
a29f6d03
NC
3338 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3339 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3340 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3341 } else
3342#endif
3343 {
765f542d 3344 /* SvIsCOW_shared_hash */
46187eeb
NC
3345 DEBUG_C(PerlIO_printf(Perl_debug_log,
3346 "Copy on write: Sharing hash\n"));
b8f9541a 3347
bdd68bc3 3348 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3349 SvPV_set(dstr,
d1db91c6 3350 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3351 }
87a1ef3d
SP
3352 SvLEN_set(dstr, len);
3353 SvCUR_set(dstr, cur);
765f542d
NC
3354 SvREADONLY_on(dstr);
3355 SvFAKE_on(dstr);
3356 /* Relesase a global SV mutex. */
3357 }
3358 else
765f542d 3359 { /* Passes the swipe test. */
78d1e721 3360 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3361 SvLEN_set(dstr, SvLEN(sstr));
3362 SvCUR_set(dstr, SvCUR(sstr));
3363
3364 SvTEMP_off(dstr);
3365 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
3366 SvPV_set(sstr, Nullch);
3367 SvLEN_set(sstr, 0);
3368 SvCUR_set(sstr, 0);
3369 SvTEMP_off(sstr);
3370 }
3371 }
9aa983d2 3372 if (sflags & SVf_UTF8)
a7cb1f99 3373 SvUTF8_on(dstr);
8990e307 3374 if (sflags & SVp_NOK) {
3332b3c1
JH
3375 SvNOKp_on(dstr);
3376 if (sflags & SVf_NOK)
3377 SvFLAGS(dstr) |= SVf_NOK;
9d6ce603 3378 SvNV_set(dstr, SvNVX(sstr));
79072805 3379 }
8990e307 3380 if (sflags & SVp_IOK) {
3332b3c1
JH
3381 (void)SvIOKp_on(dstr);
3382 if (sflags & SVf_IOK)
3383 SvFLAGS(dstr) |= SVf_IOK;
2b1c7e3e 3384 if (sflags & SVf_IVisUV)
25da4f38 3385 SvIsUV_on(dstr);
45977657 3386 SvIV_set(dstr, SvIVX(sstr));
79072805 3387 }
92f0c265 3388 if (SvVOK(sstr)) {
00b6aa41 3389 const MAGIC * const smg = mg_find(sstr,PERL_MAGIC_vstring);
ece467f9
JP
3390 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3391 smg->mg_ptr, smg->mg_len);
439cb1c4 3392 SvRMAGICAL_on(dstr);
7a5fa8a2 3393 }
79072805 3394 }
8990e307 3395 else if (sflags & SVp_IOK) {
3332b3c1
JH
3396 if (sflags & SVf_IOK)
3397 (void)SvIOK_only(dstr);
3398 else {
9cbac4c7
DM
3399 (void)SvOK_off(dstr);
3400 (void)SvIOKp_on(dstr);
3332b3c1
JH
3401 }
3402 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
2b1c7e3e 3403 if (sflags & SVf_IVisUV)
25da4f38 3404 SvIsUV_on(dstr);
45977657 3405 SvIV_set(dstr, SvIVX(sstr));
3332b3c1
JH
3406 if (sflags & SVp_NOK) {
3407 if (sflags & SVf_NOK)
3408 (void)SvNOK_on(dstr);
3409 else
3410 (void)SvNOKp_on(dstr);
9d6ce603 3411 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3412 }
3413 }
3414 else if (sflags & SVp_NOK) {
3415 if (sflags & SVf_NOK)
3416 (void)SvNOK_only(dstr);
3417 else {
9cbac4c7 3418 (void)SvOK_off(dstr);
3332b3c1
JH
3419 SvNOKp_on(dstr);
3420 }
9d6ce603 3421 SvNV_set(dstr, SvNVX(sstr));
79072805
LW
3422 }
3423 else {
20408e3c 3424 if (dtype == SVt_PVGV) {
e476b1b5 3425 if (ckWARN(WARN_MISC))
9014280d 3426 Perl_warner(aTHX_ packWARN(WARN_MISC), "Undefined value assigned to typeglob");
20408e3c
GS
3427 }
3428 else
3429 (void)SvOK_off(dstr);
a0d0e21e 3430 }
27c9684d
AP
3431 if (SvTAINTED(sstr))
3432 SvTAINT(dstr);
79072805
LW
3433}
3434
954c1994
GS
3435/*
3436=for apidoc sv_setsv_mg
3437
3438Like C<sv_setsv>, but also handles 'set' magic.
3439
3440=cut
3441*/
3442
79072805 3443void
864dbfa3 3444Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3445{
3446 sv_setsv(dstr,sstr);
3447 SvSETMAGIC(dstr);
3448}
3449
f8c7b90f 3450#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3451SV *
3452Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3453{
3454 STRLEN cur = SvCUR(sstr);
3455 STRLEN len = SvLEN(sstr);
3456 register char *new_pv;
3457
3458 if (DEBUG_C_TEST) {
3459 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3460 sstr, dstr);
3461 sv_dump(sstr);
3462 if (dstr)
3463 sv_dump(dstr);
3464 }
3465
3466 if (dstr) {
3467 if (SvTHINKFIRST(dstr))
3468 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3469 else if (SvPVX_const(dstr))
3470 Safefree(SvPVX_const(dstr));
ed252734
NC
3471 }
3472 else
3473 new_SV(dstr);
862a34c6 3474 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3475
3476 assert (SvPOK(sstr));
3477 assert (SvPOKp(sstr));
3478 assert (!SvIOK(sstr));
3479 assert (!SvIOKp(sstr));
3480 assert (!SvNOK(sstr));
3481 assert (!SvNOKp(sstr));
3482
3483 if (SvIsCOW(sstr)) {
3484
3485 if (SvLEN(sstr) == 0) {
3486 /* source is a COW shared hash key. */
ed252734
NC
3487 DEBUG_C(PerlIO_printf(Perl_debug_log,
3488 "Fast copy on write: Sharing hash\n"));
d1db91c6 3489 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3490 goto common_exit;
3491 }
3492 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3493 } else {
3494 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3495 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3496 SvREADONLY_on(sstr);
3497 SvFAKE_on(sstr);
3498 DEBUG_C(PerlIO_printf(Perl_debug_log,
3499 "Fast copy on write: Converting sstr to COW\n"));
3500 SV_COW_NEXT_SV_SET(dstr, sstr);
3501 }
3502 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3503 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3504
3505 common_exit:
3506 SvPV_set(dstr, new_pv);
3507 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3508 if (SvUTF8(sstr))
3509 SvUTF8_on(dstr);
87a1ef3d
SP
3510 SvLEN_set(dstr, len);
3511 SvCUR_set(dstr, cur);
ed252734
NC
3512 if (DEBUG_C_TEST) {
3513 sv_dump(dstr);
3514 }
3515 return dstr;
3516}
3517#endif
3518
954c1994
GS
3519/*
3520=for apidoc sv_setpvn
3521
3522Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3523bytes to be copied. If the C<ptr> argument is NULL the SV will become
3524undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3525
3526=cut
3527*/
3528
ef50df4b 3529void
864dbfa3 3530Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3531{
c6f8c383 3532 register char *dptr;
22c522df 3533
765f542d 3534 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3535 if (!ptr) {
a0d0e21e 3536 (void)SvOK_off(sv);
463ee0b2
LW
3537 return;
3538 }
22c522df
JH
3539 else {
3540 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3541 const IV iv = len;
9c5ffd7c
JH
3542 if (iv < 0)
3543 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3544 }
862a34c6 3545 SvUPGRADE(sv, SVt_PV);
c6f8c383 3546
5902b6a9 3547 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3548 Move(ptr,dptr,len,char);
3549 dptr[len] = '\0';
79072805 3550 SvCUR_set(sv, len);
1aa99e6b 3551 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3552 SvTAINT(sv);
79072805
LW
3553}
3554
954c1994
GS
3555/*
3556=for apidoc sv_setpvn_mg
3557
3558Like C<sv_setpvn>, but also handles 'set' magic.
3559
3560=cut
3561*/
3562
79072805 3563void
864dbfa3 3564Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3565{
3566 sv_setpvn(sv,ptr,len);
3567 SvSETMAGIC(sv);
3568}
3569
954c1994
GS
3570/*
3571=for apidoc sv_setpv
3572
3573Copies a string into an SV. The string must be null-terminated. Does not
3574handle 'set' magic. See C<sv_setpv_mg>.
3575
3576=cut
3577*/
3578
ef50df4b 3579void
864dbfa3 3580Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
3581{
3582 register STRLEN len;
3583
765f542d 3584 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3585 if (!ptr) {
a0d0e21e 3586 (void)SvOK_off(sv);
463ee0b2
LW
3587 return;
3588 }
79072805 3589 len = strlen(ptr);
862a34c6 3590 SvUPGRADE(sv, SVt_PV);
c6f8c383 3591
79072805 3592 SvGROW(sv, len + 1);
463ee0b2 3593 Move(ptr,SvPVX(sv),len+1,char);
79072805 3594 SvCUR_set(sv, len);
1aa99e6b 3595 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3596 SvTAINT(sv);
3597}
3598
954c1994
GS
3599/*
3600=for apidoc sv_setpv_mg
3601
3602Like C<sv_setpv>, but also handles 'set' magic.
3603
3604=cut
3605*/
3606
463ee0b2 3607void
864dbfa3 3608Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3609{
3610 sv_setpv(sv,ptr);
3611 SvSETMAGIC(sv);
3612}
3613
954c1994
GS
3614/*
3615=for apidoc sv_usepvn
3616
3617Tells an SV to use C<ptr> to find its string value. Normally the string is
1c846c1f 3618stored inside the SV but sv_usepvn allows the SV to use an outside string.
954c1994
GS
3619The C<ptr> should point to memory that was allocated by C<malloc>. The
3620string length, C<len>, must be supplied. This function will realloc the
3621memory pointed to by C<ptr>, so that pointer should not be freed or used by
3622the programmer after giving it to sv_usepvn. Does not handle 'set' magic.
3623See C<sv_usepvn_mg>.
3624
3625=cut
3626*/
3627
ef50df4b 3628void
864dbfa3 3629Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
463ee0b2 3630{
1936d2a7 3631 STRLEN allocate;
765f542d 3632 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3633 SvUPGRADE(sv, SVt_PV);
463ee0b2 3634 if (!ptr) {
a0d0e21e 3635 (void)SvOK_off(sv);
463ee0b2
LW
3636 return;
3637 }
3f7c398e 3638 if (SvPVX_const(sv))
8bd4d4c5 3639 SvPV_free(sv);
1936d2a7
NC
3640
3641 allocate = PERL_STRLEN_ROUNDUP(len + 1);
7a9b70e9 3642 ptr = saferealloc (ptr, allocate);
f880fe2f 3643 SvPV_set(sv, ptr);
463ee0b2 3644 SvCUR_set(sv, len);
1936d2a7 3645 SvLEN_set(sv, allocate);
463ee0b2 3646 *SvEND(sv) = '\0';
1aa99e6b 3647 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3648 SvTAINT(sv);
79072805
LW
3649}
3650
954c1994
GS
3651/*
3652=for apidoc sv_usepvn_mg
3653
3654Like C<sv_usepvn>, but also handles 'set' magic.
3655
3656=cut
3657*/
3658
ef50df4b 3659void
864dbfa3 3660Perl_sv_usepvn_mg(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
ef50df4b 3661{
51c1089b 3662 sv_usepvn(sv,ptr,len);
ef50df4b
GS
3663 SvSETMAGIC(sv);
3664}
3665
f8c7b90f 3666#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3667/* Need to do this *after* making the SV normal, as we need the buffer
3668 pointer to remain valid until after we've copied it. If we let go too early,
3669 another thread could invalidate it by unsharing last of the same hash key
3670 (which it can do by means other than releasing copy-on-write Svs)
3671 or by changing the other copy-on-write SVs in the loop. */
3672STATIC void
bdd68bc3 3673S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3674{
3675 if (len) { /* this SV was SvIsCOW_normal(sv) */
3676 /* we need to find the SV pointing to us. */
aec46f14 3677 SV * const current = SV_COW_NEXT_SV(after);
7a5fa8a2 3678
765f542d
NC
3679 if (current == sv) {
3680 /* The SV we point to points back to us (there were only two of us
3681 in the loop.)
3682 Hence other SV is no longer copy on write either. */
3683 SvFAKE_off(after);
3684 SvREADONLY_off(after);
3685 } else {
3686 /* We need to follow the pointers around the loop. */
3687 SV *next;
3688 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3689 assert (next);
3690 current = next;
3691 /* don't loop forever if the structure is bust, and we have
3692 a pointer into a closed loop. */
3693 assert (current != after);
3f7c398e 3694 assert (SvPVX_const(current) == pvx);
765f542d
NC
3695 }
3696 /* Make the SV before us point to the SV after us. */
a29f6d03 3697 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3698 }
3699 } else {
bdd68bc3 3700 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3701 }
3702}
3703
3704int
3705Perl_sv_release_IVX(pTHX_ register SV *sv)
3706{
3707 if (SvIsCOW(sv))
3708 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
3709 SvOOK_off(sv);
3710 return 0;
765f542d
NC
3711}
3712#endif
645c22ef
DM
3713/*
3714=for apidoc sv_force_normal_flags
3715
3716Undo various types of fakery on an SV: if the PV is a shared string, make
3717a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
3718an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
3719we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
3720then a copy-on-write scalar drops its PV buffer (if any) and becomes
3721SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 3722set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
3723C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
3724with flags set to 0.
645c22ef
DM
3725
3726=cut
3727*/