This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add SV allocation tracing to -Dm and PERL_MEM_LOG
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
7272f7c1 4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137 32#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 33/* if adding more checks watch out for the following tests:
e23c8137
JH
34 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
35 * lib/utf8.t lib/Unicode/Collate/t/index.t
36 * --jhi
37 */
6f207bd3 38# define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
39 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
40 assert((cache)[2] <= (cache)[3]); \
41 assert((cache)[3] <= (cache)[1]);} \
42 } STMT_END
e23c8137 43#else
6f207bd3 44# define ASSERT_UTF8_CACHE(cache) NOOP
e23c8137
JH
45#endif
46
f8c7b90f 47#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 48#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 49#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 50/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 51 on-write. */
765f542d 52#endif
645c22ef
DM
53
54/* ============================================================================
55
56=head1 Allocation and deallocation of SVs.
57
d2a0f284
JC
58An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
59sv, av, hv...) contains type and reference count information, and for
60many types, a pointer to the body (struct xrv, xpv, xpviv...), which
61contains fields specific to each type. Some types store all they need
62in the head, so don't have a body.
63
64In all but the most memory-paranoid configuations (ex: PURIFY), heads
65and bodies are allocated out of arenas, which by default are
66approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
67Sv-bodies are allocated by their sv-type, guaranteeing size
68consistency needed to allocate safely from arrays.
69
d2a0f284
JC
70For SV-heads, the first slot in each arena is reserved, and holds a
71link to the next arena, some flags, and a note of the number of slots.
72Snaked through each arena chain is a linked list of free items; when
73this becomes empty, an extra arena is allocated and divided up into N
74items which are threaded into the free list.
75
76SV-bodies are similar, but they use arena-sets by default, which
77separate the link and info from the arena itself, and reclaim the 1st
78slot in the arena. SV-bodies are further described later.
645c22ef
DM
79
80The following global variables are associated with arenas:
81
82 PL_sv_arenaroot pointer to list of SV arenas
83 PL_sv_root pointer to list of free SV structures
84
d2a0f284
JC
85 PL_body_arenas head of linked-list of body arenas
86 PL_body_roots[] array of pointers to list of free bodies of svtype
87 arrays are indexed by the svtype needed
93e68bfb 88
d2a0f284
JC
89A few special SV heads are not allocated from an arena, but are
90instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
91The size of arenas can be changed from the default by setting
92PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
93
94The SV arena serves the secondary purpose of allowing still-live SVs
95to be located and destroyed during final cleanup.
96
97At the lowest level, the macros new_SV() and del_SV() grab and free
98an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
99to return the SV to the free list with error checking.) new_SV() calls
100more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
101SVs in the free list have their SvTYPE field set to all ones.
102
ff276b08 103At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 104perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 105start of the interpreter.
645c22ef 106
645c22ef
DM
107The function visit() scans the SV arenas list, and calls a specified
108function for each SV it finds which is still live - ie which has an SvTYPE
109other than all 1's, and a non-zero SvREFCNT. visit() is used by the
110following functions (specified as [function that calls visit()] / [function
111called by visit() for each SV]):
112
113 sv_report_used() / do_report_used()
f2524eef 114 dump all remaining SVs (debugging aid)
645c22ef
DM
115
116 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
117 Attempt to free all objects pointed to by RVs,
118 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
119 try to do the same for all objects indirectly
120 referenced by typeglobs too. Called once from
121 perl_destruct(), prior to calling sv_clean_all()
122 below.
123
124 sv_clean_all() / do_clean_all()
125 SvREFCNT_dec(sv) each remaining SV, possibly
126 triggering an sv_free(). It also sets the
127 SVf_BREAK flag on the SV to indicate that the
128 refcnt has been artificially lowered, and thus
129 stopping sv_free() from giving spurious warnings
130 about SVs which unexpectedly have a refcnt
131 of zero. called repeatedly from perl_destruct()
132 until there are no SVs left.
133
93e68bfb 134=head2 Arena allocator API Summary
645c22ef
DM
135
136Private API to rest of sv.c
137
138 new_SV(), del_SV(),
139
140 new_XIV(), del_XIV(),
141 new_XNV(), del_XNV(),
142 etc
143
144Public API:
145
8cf8f3d1 146 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 147
645c22ef
DM
148=cut
149
150============================================================================ */
151
4561caa4
CS
152/*
153 * "A time to plant, and a time to uproot what was planted..."
154 */
155
77354fb4 156void
de37a194 157Perl_offer_nice_chunk(pTHX_ void *const chunk, const U32 chunk_size)
77354fb4 158{
97aff369 159 dVAR;
77354fb4
NC
160 void *new_chunk;
161 U32 new_chunk_size;
7918f24d
NC
162
163 PERL_ARGS_ASSERT_OFFER_NICE_CHUNK;
164
77354fb4
NC
165 new_chunk = (void *)(chunk);
166 new_chunk_size = (chunk_size);
167 if (new_chunk_size > PL_nice_chunk_size) {
168 Safefree(PL_nice_chunk);
169 PL_nice_chunk = (char *) new_chunk;
170 PL_nice_chunk_size = new_chunk_size;
171 } else {
172 Safefree(chunk);
173 }
77354fb4 174}
cac9b346 175
d7a2c63c
MHM
176#ifdef PERL_MEM_LOG
177# define MEM_LOG_NEW_SV(sv, file, line, func) \
178 Perl_mem_log_new_sv(sv, file, line, func)
179# define MEM_LOG_DEL_SV(sv, file, line, func) \
180 Perl_mem_log_del_sv(sv, file, line, func)
181#else
182# define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
183# define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184#endif
185
fd0854ff 186#ifdef DEBUG_LEAKING_SCALARS
22162ca8 187# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
d7a2c63c
MHM
188# define DEBUG_SV_SERIAL(sv) \
189 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
190 PTR2UV(sv), (long)(sv)->sv_debug_serial))
fd0854ff
DM
191#else
192# define FREE_SV_DEBUG_FILE(sv)
d7a2c63c 193# define DEBUG_SV_SERIAL(sv) NOOP
fd0854ff
DM
194#endif
195
48614a46
NC
196#ifdef PERL_POISON
197# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
198/* Whilst I'd love to do this, it seems that things like to check on
199 unreferenced scalars
7e337ee0 200# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 201*/
7e337ee0
JH
202# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
203 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
204#else
205# define SvARENA_CHAIN(sv) SvANY(sv)
206# define POSION_SV_HEAD(sv)
207#endif
208
990198f0
DM
209/* Mark an SV head as unused, and add to free list.
210 *
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
215
053fc874
GS
216#define plant_SV(p) \
217 STMT_START { \
990198f0 218 const U32 old_flags = SvFLAGS(p); \
d7a2c63c
MHM
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
fd0854ff 221 FREE_SV_DEBUG_FILE(p); \
48614a46 222 POSION_SV_HEAD(p); \
053fc874 223 SvFLAGS(p) = SVTYPEMASK; \
990198f0
DM
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
226 PL_sv_root = (p); \
227 } \
053fc874
GS
228 --PL_sv_count; \
229 } STMT_END
a0d0e21e 230
053fc874
GS
231#define uproot_SV(p) \
232 STMT_START { \
233 (p) = PL_sv_root; \
bb7bbd9c 234 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
235 ++PL_sv_count; \
236 } STMT_END
237
645c22ef 238
cac9b346
NC
239/* make some more SVs by adding another arena */
240
cac9b346
NC
241STATIC SV*
242S_more_sv(pTHX)
243{
97aff369 244 dVAR;
cac9b346
NC
245 SV* sv;
246
247 if (PL_nice_chunk) {
248 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 249 PL_nice_chunk = NULL;
cac9b346
NC
250 PL_nice_chunk_size = 0;
251 }
252 else {
253 char *chunk; /* must use New here to match call to */
d2a0f284 254 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 255 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
256 }
257 uproot_SV(sv);
258 return sv;
259}
260
645c22ef
DM
261/* new_SV(): return a new, empty SV head */
262
eba0f806
DM
263#ifdef DEBUG_LEAKING_SCALARS
264/* provide a real function for a debugger to play with */
265STATIC SV*
d7a2c63c 266S_new_SV(pTHX_ const char *file, int line, const char *func)
eba0f806
DM
267{
268 SV* sv;
269
eba0f806
DM
270 if (PL_sv_root)
271 uproot_SV(sv);
272 else
cac9b346 273 sv = S_more_sv(aTHX);
eba0f806
DM
274 SvANY(sv) = 0;
275 SvREFCNT(sv) = 1;
276 SvFLAGS(sv) = 0;
fd0854ff 277 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
e385c3bf
DM
278 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
279 ? PL_parser->copline
280 : PL_curcop
f24aceb1
DM
281 ? CopLINE(PL_curcop)
282 : 0
e385c3bf 283 );
fd0854ff
DM
284 sv->sv_debug_inpad = 0;
285 sv->sv_debug_cloned = 0;
fd0854ff 286 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
d7a2c63c
MHM
287
288 sv->sv_debug_serial = PL_sv_serial++;
289
290 MEM_LOG_NEW_SV(sv, file, line, func);
291 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
292 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
293
eba0f806
DM
294 return sv;
295}
d7a2c63c 296# define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
eba0f806
DM
297
298#else
299# define new_SV(p) \
053fc874 300 STMT_START { \
053fc874
GS
301 if (PL_sv_root) \
302 uproot_SV(p); \
303 else \
cac9b346 304 (p) = S_more_sv(aTHX); \
053fc874
GS
305 SvANY(p) = 0; \
306 SvREFCNT(p) = 1; \
307 SvFLAGS(p) = 0; \
d7a2c63c 308 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
053fc874 309 } STMT_END
eba0f806 310#endif
463ee0b2 311
645c22ef
DM
312
313/* del_SV(): return an empty SV head to the free list */
314
a0d0e21e 315#ifdef DEBUGGING
4561caa4 316
053fc874
GS
317#define del_SV(p) \
318 STMT_START { \
aea4f609 319 if (DEBUG_D_TEST) \
053fc874
GS
320 del_sv(p); \
321 else \
322 plant_SV(p); \
053fc874 323 } STMT_END
a0d0e21e 324
76e3520e 325STATIC void
cea2e8a9 326S_del_sv(pTHX_ SV *p)
463ee0b2 327{
97aff369 328 dVAR;
7918f24d
NC
329
330 PERL_ARGS_ASSERT_DEL_SV;
331
aea4f609 332 if (DEBUG_D_TEST) {
4633a7c4 333 SV* sva;
a3b680e6 334 bool ok = 0;
3280af22 335 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
336 const SV * const sv = sva + 1;
337 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 338 if (p >= sv && p < svend) {
a0d0e21e 339 ok = 1;
c0ff570e
NC
340 break;
341 }
a0d0e21e
LW
342 }
343 if (!ok) {
0453d815 344 if (ckWARN_d(WARN_INTERNAL))
9014280d 345 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
346 "Attempt to free non-arena SV: 0x%"UVxf
347 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
348 return;
349 }
350 }
4561caa4 351 plant_SV(p);
463ee0b2 352}
a0d0e21e 353
4561caa4
CS
354#else /* ! DEBUGGING */
355
356#define del_SV(p) plant_SV(p)
357
358#endif /* DEBUGGING */
463ee0b2 359
645c22ef
DM
360
361/*
ccfc67b7
JH
362=head1 SV Manipulation Functions
363
645c22ef
DM
364=for apidoc sv_add_arena
365
366Given a chunk of memory, link it to the head of the list of arenas,
367and split it into a list of free SVs.
368
369=cut
370*/
371
4633a7c4 372void
de37a194 373Perl_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
463ee0b2 374{
97aff369 375 dVAR;
0bd48802 376 SV* const sva = (SV*)ptr;
463ee0b2
LW
377 register SV* sv;
378 register SV* svend;
4633a7c4 379
7918f24d
NC
380 PERL_ARGS_ASSERT_SV_ADD_ARENA;
381
4633a7c4 382 /* The first SV in an arena isn't an SV. */
3280af22 383 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
384 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
385 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
386
3280af22
NIS
387 PL_sv_arenaroot = sva;
388 PL_sv_root = sva + 1;
4633a7c4
LW
389
390 svend = &sva[SvREFCNT(sva) - 1];
391 sv = sva + 1;
463ee0b2 392 while (sv < svend) {
48614a46 393 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 394#ifdef DEBUGGING
978b032e 395 SvREFCNT(sv) = 0;
03e36789 396#endif
4b69cbe3 397 /* Must always set typemask because it's always checked in on cleanup
03e36789 398 when the arenas are walked looking for objects. */
8990e307 399 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
400 sv++;
401 }
48614a46 402 SvARENA_CHAIN(sv) = 0;
03e36789
NC
403#ifdef DEBUGGING
404 SvREFCNT(sv) = 0;
405#endif
4633a7c4
LW
406 SvFLAGS(sv) = SVTYPEMASK;
407}
408
055972dc
DM
409/* visit(): call the named function for each non-free SV in the arenas
410 * whose flags field matches the flags/mask args. */
645c22ef 411
5226ed68 412STATIC I32
de37a194 413S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
8990e307 414{
97aff369 415 dVAR;
4633a7c4 416 SV* sva;
5226ed68 417 I32 visited = 0;
8990e307 418
7918f24d
NC
419 PERL_ARGS_ASSERT_VISIT;
420
3280af22 421 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 422 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 423 register SV* sv;
4561caa4 424 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
425 if (SvTYPE(sv) != SVTYPEMASK
426 && (sv->sv_flags & mask) == flags
427 && SvREFCNT(sv))
428 {
acfe0abc 429 (FCALL)(aTHX_ sv);
5226ed68
JH
430 ++visited;
431 }
8990e307
LW
432 }
433 }
5226ed68 434 return visited;
8990e307
LW
435}
436
758a08c3
JH
437#ifdef DEBUGGING
438
645c22ef
DM
439/* called by sv_report_used() for each live SV */
440
441static void
5fa45a31 442do_report_used(pTHX_ SV *const sv)
645c22ef
DM
443{
444 if (SvTYPE(sv) != SVTYPEMASK) {
445 PerlIO_printf(Perl_debug_log, "****\n");
446 sv_dump(sv);
447 }
448}
758a08c3 449#endif
645c22ef
DM
450
451/*
452=for apidoc sv_report_used
453
454Dump the contents of all SVs not yet freed. (Debugging aid).
455
456=cut
457*/
458
8990e307 459void
864dbfa3 460Perl_sv_report_used(pTHX)
4561caa4 461{
ff270d3a 462#ifdef DEBUGGING
055972dc 463 visit(do_report_used, 0, 0);
96a5add6
AL
464#else
465 PERL_UNUSED_CONTEXT;
ff270d3a 466#endif
4561caa4
CS
467}
468
645c22ef
DM
469/* called by sv_clean_objs() for each live SV */
470
471static void
de37a194 472do_clean_objs(pTHX_ SV *const ref)
645c22ef 473{
97aff369 474 dVAR;
ea724faa
NC
475 assert (SvROK(ref));
476 {
823a54a3
AL
477 SV * const target = SvRV(ref);
478 if (SvOBJECT(target)) {
479 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
480 if (SvWEAKREF(ref)) {
481 sv_del_backref(target, ref);
482 SvWEAKREF_off(ref);
483 SvRV_set(ref, NULL);
484 } else {
485 SvROK_off(ref);
486 SvRV_set(ref, NULL);
487 SvREFCNT_dec(target);
488 }
645c22ef
DM
489 }
490 }
491
492 /* XXX Might want to check arrays, etc. */
493}
494
495/* called by sv_clean_objs() for each live SV */
496
497#ifndef DISABLE_DESTRUCTOR_KLUDGE
498static void
f30de749 499do_clean_named_objs(pTHX_ SV *const sv)
645c22ef 500{
97aff369 501 dVAR;
ea724faa 502 assert(SvTYPE(sv) == SVt_PVGV);
d011219a
NC
503 assert(isGV_with_GP(sv));
504 if (GvGP(sv)) {
c69033f2
NC
505 if ((
506#ifdef PERL_DONT_CREATE_GVSV
507 GvSV(sv) &&
508#endif
509 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
510 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
511 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
9c12f1e5
RGS
512 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
513 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
645c22ef
DM
514 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
515 {
516 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 517 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
518 SvREFCNT_dec(sv);
519 }
520 }
521}
522#endif
523
524/*
525=for apidoc sv_clean_objs
526
527Attempt to destroy all objects not yet freed
528
529=cut
530*/
531
4561caa4 532void
864dbfa3 533Perl_sv_clean_objs(pTHX)
4561caa4 534{
97aff369 535 dVAR;
3280af22 536 PL_in_clean_objs = TRUE;
055972dc 537 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 538#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 539 /* some barnacles may yet remain, clinging to typeglobs */
d011219a 540 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
4561caa4 541#endif
3280af22 542 PL_in_clean_objs = FALSE;
4561caa4
CS
543}
544
645c22ef
DM
545/* called by sv_clean_all() for each live SV */
546
547static void
de37a194 548do_clean_all(pTHX_ SV *const sv)
645c22ef 549{
97aff369 550 dVAR;
cddfcddc
NC
551 if (sv == (SV*) PL_fdpid || sv == (SV *)PL_strtab) {
552 /* don't clean pid table and strtab */
d17ea597 553 return;
cddfcddc 554 }
645c22ef
DM
555 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
556 SvFLAGS(sv) |= SVf_BREAK;
557 SvREFCNT_dec(sv);
558}
559
560/*
561=for apidoc sv_clean_all
562
563Decrement the refcnt of each remaining SV, possibly triggering a
564cleanup. This function may have to be called multiple times to free
ff276b08 565SVs which are in complex self-referential hierarchies.
645c22ef
DM
566
567=cut
568*/
569
5226ed68 570I32
864dbfa3 571Perl_sv_clean_all(pTHX)
8990e307 572{
97aff369 573 dVAR;
5226ed68 574 I32 cleaned;
3280af22 575 PL_in_clean_all = TRUE;
055972dc 576 cleaned = visit(do_clean_all, 0,0);
3280af22 577 PL_in_clean_all = FALSE;
5226ed68 578 return cleaned;
8990e307 579}
463ee0b2 580
5e258f8c
JC
581/*
582 ARENASETS: a meta-arena implementation which separates arena-info
583 into struct arena_set, which contains an array of struct
584 arena_descs, each holding info for a single arena. By separating
585 the meta-info from the arena, we recover the 1st slot, formerly
586 borrowed for list management. The arena_set is about the size of an
39244528 587 arena, avoiding the needless malloc overhead of a naive linked-list.
5e258f8c
JC
588
589 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
590 memory in the last arena-set (1/2 on average). In trade, we get
591 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284 592 smaller types). The recovery of the wasted space allows use of
e15dad31
JC
593 small arenas for large, rare body types, by changing array* fields
594 in body_details_by_type[] below.
5e258f8c 595*/
5e258f8c 596struct arena_desc {
398c677b
NC
597 char *arena; /* the raw storage, allocated aligned */
598 size_t size; /* its size ~4k typ */
0a848332 599 U32 misc; /* type, and in future other things. */
5e258f8c
JC
600};
601
e6148039
NC
602struct arena_set;
603
604/* Get the maximum number of elements in set[] such that struct arena_set
e15dad31 605 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
e6148039
NC
606 therefore likely to be 1 aligned memory page. */
607
608#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
609 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
610
611struct arena_set {
612 struct arena_set* next;
0a848332
NC
613 unsigned int set_size; /* ie ARENAS_PER_SET */
614 unsigned int curr; /* index of next available arena-desc */
5e258f8c
JC
615 struct arena_desc set[ARENAS_PER_SET];
616};
617
645c22ef
DM
618/*
619=for apidoc sv_free_arenas
620
621Deallocate the memory used by all arenas. Note that all the individual SV
622heads and bodies within the arenas must already have been freed.
623
624=cut
625*/
4633a7c4 626void
864dbfa3 627Perl_sv_free_arenas(pTHX)
4633a7c4 628{
97aff369 629 dVAR;
4633a7c4
LW
630 SV* sva;
631 SV* svanext;
0a848332 632 unsigned int i;
4633a7c4
LW
633
634 /* Free arenas here, but be careful about fake ones. (We assume
635 contiguity of the fake ones with the corresponding real ones.) */
636
3280af22 637 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
638 svanext = (SV*) SvANY(sva);
639 while (svanext && SvFAKE(svanext))
640 svanext = (SV*) SvANY(svanext);
641
642 if (!SvFAKE(sva))
1df70142 643 Safefree(sva);
4633a7c4 644 }
93e68bfb 645
5e258f8c 646 {
0a848332
NC
647 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
648
649 while (aroot) {
650 struct arena_set *current = aroot;
651 i = aroot->curr;
652 while (i--) {
5e258f8c
JC
653 assert(aroot->set[i].arena);
654 Safefree(aroot->set[i].arena);
655 }
0a848332
NC
656 aroot = aroot->next;
657 Safefree(current);
5e258f8c
JC
658 }
659 }
dc8220bf 660 PL_body_arenas = 0;
fdda85ca 661
0a848332
NC
662 i = PERL_ARENA_ROOTS_SIZE;
663 while (i--)
93e68bfb 664 PL_body_roots[i] = 0;
93e68bfb 665
43c5f42d 666 Safefree(PL_nice_chunk);
bd61b366 667 PL_nice_chunk = NULL;
3280af22
NIS
668 PL_nice_chunk_size = 0;
669 PL_sv_arenaroot = 0;
670 PL_sv_root = 0;
4633a7c4
LW
671}
672
bd81e77b
NC
673/*
674 Here are mid-level routines that manage the allocation of bodies out
675 of the various arenas. There are 5 kinds of arenas:
29489e7c 676
bd81e77b
NC
677 1. SV-head arenas, which are discussed and handled above
678 2. regular body arenas
679 3. arenas for reduced-size bodies
680 4. Hash-Entry arenas
681 5. pte arenas (thread related)
29489e7c 682
bd81e77b
NC
683 Arena types 2 & 3 are chained by body-type off an array of
684 arena-root pointers, which is indexed by svtype. Some of the
685 larger/less used body types are malloced singly, since a large
686 unused block of them is wasteful. Also, several svtypes dont have
687 bodies; the data fits into the sv-head itself. The arena-root
688 pointer thus has a few unused root-pointers (which may be hijacked
689 later for arena types 4,5)
29489e7c 690
bd81e77b
NC
691 3 differs from 2 as an optimization; some body types have several
692 unused fields in the front of the structure (which are kept in-place
693 for consistency). These bodies can be allocated in smaller chunks,
694 because the leading fields arent accessed. Pointers to such bodies
695 are decremented to point at the unused 'ghost' memory, knowing that
696 the pointers are used with offsets to the real memory.
29489e7c 697
bd81e77b
NC
698 HE, HEK arenas are managed separately, with separate code, but may
699 be merge-able later..
700
701 PTE arenas are not sv-bodies, but they share these mid-level
702 mechanics, so are considered here. The new mid-level mechanics rely
703 on the sv_type of the body being allocated, so we just reserve one
704 of the unused body-slots for PTEs, then use it in those (2) PTE
705 contexts below (line ~10k)
706*/
707
bd26d9a3 708/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
709 TBD: export properly for hv.c: S_more_he().
710*/
711void*
de37a194 712Perl_get_arena(pTHX_ const size_t arena_size, const U32 misc)
5e258f8c 713{
7a89be66 714 dVAR;
5e258f8c 715 struct arena_desc* adesc;
39244528 716 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
0a848332 717 unsigned int curr;
5e258f8c 718
476a1e16
JC
719 /* shouldnt need this
720 if (!arena_size) arena_size = PERL_ARENA_SIZE;
721 */
5e258f8c
JC
722
723 /* may need new arena-set to hold new arena */
39244528
NC
724 if (!aroot || aroot->curr >= aroot->set_size) {
725 struct arena_set *newroot;
5e258f8c
JC
726 Newxz(newroot, 1, struct arena_set);
727 newroot->set_size = ARENAS_PER_SET;
39244528
NC
728 newroot->next = aroot;
729 aroot = newroot;
730 PL_body_arenas = (void *) newroot;
52944de8 731 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
5e258f8c
JC
732 }
733
734 /* ok, now have arena-set with at least 1 empty/available arena-desc */
39244528
NC
735 curr = aroot->curr++;
736 adesc = &(aroot->set[curr]);
5e258f8c
JC
737 assert(!adesc->arena);
738
89086707 739 Newx(adesc->arena, arena_size, char);
5e258f8c 740 adesc->size = arena_size;
0a848332 741 adesc->misc = misc;
d67b3c53
JH
742 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
743 curr, (void*)adesc->arena, (UV)arena_size));
5e258f8c
JC
744
745 return adesc->arena;
5e258f8c
JC
746}
747
53c1dcc0 748
bd81e77b 749/* return a thing to the free list */
29489e7c 750
bd81e77b
NC
751#define del_body(thing, root) \
752 STMT_START { \
00b6aa41 753 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
754 *thing_copy = *root; \
755 *root = (void*)thing_copy; \
bd81e77b 756 } STMT_END
29489e7c 757
bd81e77b 758/*
d2a0f284
JC
759
760=head1 SV-Body Allocation
761
762Allocation of SV-bodies is similar to SV-heads, differing as follows;
763the allocation mechanism is used for many body types, so is somewhat
764more complicated, it uses arena-sets, and has no need for still-live
765SV detection.
766
767At the outermost level, (new|del)_X*V macros return bodies of the
768appropriate type. These macros call either (new|del)_body_type or
769(new|del)_body_allocated macro pairs, depending on specifics of the
770type. Most body types use the former pair, the latter pair is used to
771allocate body types with "ghost fields".
772
773"ghost fields" are fields that are unused in certain types, and
774consequently dont need to actually exist. They are declared because
775they're part of a "base type", which allows use of functions as
776methods. The simplest examples are AVs and HVs, 2 aggregate types
777which don't use the fields which support SCALAR semantics.
778
779For these types, the arenas are carved up into *_allocated size
780chunks, we thus avoid wasted memory for those unaccessed members.
781When bodies are allocated, we adjust the pointer back in memory by the
782size of the bit not allocated, so it's as if we allocated the full
783structure. (But things will all go boom if you write to the part that
784is "not there", because you'll be overwriting the last members of the
785preceding structure in memory.)
786
787We calculate the correction using the STRUCT_OFFSET macro. For
788example, if xpv_allocated is the same structure as XPV then the two
789OFFSETs sum to zero, and the pointer is unchanged. If the allocated
790structure is smaller (no initial NV actually allocated) then the net
791effect is to subtract the size of the NV from the pointer, to return a
792new pointer as if an initial NV were actually allocated.
793
794This is the same trick as was used for NV and IV bodies. Ironically it
795doesn't need to be used for NV bodies any more, because NV is now at
796the start of the structure. IV bodies don't need it either, because
797they are no longer allocated.
798
799In turn, the new_body_* allocators call S_new_body(), which invokes
800new_body_inline macro, which takes a lock, and takes a body off the
801linked list at PL_body_roots[sv_type], calling S_more_bodies() if
802necessary to refresh an empty list. Then the lock is released, and
803the body is returned.
804
805S_more_bodies calls get_arena(), and carves it up into an array of N
806bodies, which it strings into a linked list. It looks up arena-size
807and body-size from the body_details table described below, thus
808supporting the multiple body-types.
809
810If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
811the (new|del)_X*V macros are mapped directly to malloc/free.
812
813*/
814
815/*
816
817For each sv-type, struct body_details bodies_by_type[] carries
818parameters which control these aspects of SV handling:
819
820Arena_size determines whether arenas are used for this body type, and if
821so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
822zero, forcing individual mallocs and frees.
823
824Body_size determines how big a body is, and therefore how many fit into
825each arena. Offset carries the body-pointer adjustment needed for
826*_allocated body types, and is used in *_allocated macros.
827
828But its main purpose is to parameterize info needed in
829Perl_sv_upgrade(). The info here dramatically simplifies the function
830vs the implementation in 5.8.7, making it table-driven. All fields
831are used for this, except for arena_size.
832
833For the sv-types that have no bodies, arenas are not used, so those
834PL_body_roots[sv_type] are unused, and can be overloaded. In
835something of a special case, SVt_NULL is borrowed for HE arenas;
c6f8b1d0 836PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
d2a0f284 837bodies_by_type[SVt_NULL] slot is not used, as the table is not
c6f8b1d0 838available in hv.c.
d2a0f284 839
c6f8b1d0
JC
840PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
841they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
842just use the same allocation semantics. At first, PTEs were also
843overloaded to a non-body sv-type, but this yielded hard-to-find malloc
844bugs, so was simplified by claiming a new slot. This choice has no
845consequence at this time.
d2a0f284 846
29489e7c
DM
847*/
848
bd81e77b 849struct body_details {
0fb58b32 850 U8 body_size; /* Size to allocate */
10666ae3 851 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 852 U8 offset;
10666ae3
NC
853 unsigned int type : 4; /* We have space for a sanity check. */
854 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
855 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
856 unsigned int arena : 1; /* Allocated from an arena */
857 size_t arena_size; /* Size of arena to allocate */
bd81e77b 858};
29489e7c 859
bd81e77b
NC
860#define HADNV FALSE
861#define NONV TRUE
29489e7c 862
d2a0f284 863
bd81e77b
NC
864#ifdef PURIFY
865/* With -DPURFIY we allocate everything directly, and don't use arenas.
866 This seems a rather elegant way to simplify some of the code below. */
867#define HASARENA FALSE
868#else
869#define HASARENA TRUE
870#endif
871#define NOARENA FALSE
29489e7c 872
d2a0f284
JC
873/* Size the arenas to exactly fit a given number of bodies. A count
874 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
875 simplifying the default. If count > 0, the arena is sized to fit
876 only that many bodies, allowing arenas to be used for large, rare
877 bodies (XPVFM, XPVIO) without undue waste. The arena size is
878 limited by PERL_ARENA_SIZE, so we can safely oversize the
879 declarations.
880 */
95db5f15
MB
881#define FIT_ARENA0(body_size) \
882 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
883#define FIT_ARENAn(count,body_size) \
884 ( count * body_size <= PERL_ARENA_SIZE) \
885 ? count * body_size \
886 : FIT_ARENA0 (body_size)
887#define FIT_ARENA(count,body_size) \
888 count \
889 ? FIT_ARENAn (count, body_size) \
890 : FIT_ARENA0 (body_size)
d2a0f284 891
bd81e77b 892/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 893
bd81e77b
NC
894typedef struct {
895 STRLEN xpv_cur;
896 STRLEN xpv_len;
897} xpv_allocated;
29489e7c 898
bd81e77b 899to make its members accessible via a pointer to (say)
29489e7c 900
bd81e77b
NC
901struct xpv {
902 NV xnv_nv;
903 STRLEN xpv_cur;
904 STRLEN xpv_len;
905};
29489e7c 906
bd81e77b 907*/
29489e7c 908
bd81e77b
NC
909#define relative_STRUCT_OFFSET(longer, shorter, member) \
910 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 911
bd81e77b
NC
912/* Calculate the length to copy. Specifically work out the length less any
913 final padding the compiler needed to add. See the comment in sv_upgrade
914 for why copying the padding proved to be a bug. */
29489e7c 915
bd81e77b
NC
916#define copy_length(type, last_member) \
917 STRUCT_OFFSET(type, last_member) \
918 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 919
bd81e77b 920static const struct body_details bodies_by_type[] = {
10666ae3
NC
921 { sizeof(HE), 0, 0, SVt_NULL,
922 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284 923
1cb9cd50 924 /* The bind placeholder pretends to be an RV for now.
c6f8b1d0 925 Also it's marked as "can't upgrade" to stop anyone using it before it's
1cb9cd50
NC
926 implemented. */
927 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
928
d2a0f284
JC
929 /* IVs are in the head, so the allocation size is 0.
930 However, the slot is overloaded for PTEs. */
931 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
932 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 933 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
934 NOARENA /* IVS don't need an arena */,
935 /* But PTEs need to know the size of their arena */
936 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
937 },
938
bd81e77b 939 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 940 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
941 FIT_ARENA(0, sizeof(NV)) },
942
bd81e77b 943 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
944 { sizeof(xpv_allocated),
945 copy_length(XPV, xpv_len)
946 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
947 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 948 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 949
bd81e77b 950 /* 12 */
d2a0f284
JC
951 { sizeof(xpviv_allocated),
952 copy_length(XPVIV, xiv_u)
953 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
954 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 955 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 956
bd81e77b 957 /* 20 */
10666ae3 958 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
959 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
960
bd81e77b 961 /* 28 */
10666ae3 962 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284 963 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
4df7f6af 964
288b8c02 965 /* something big */
08e44740
NC
966 { sizeof(struct regexp_allocated), sizeof(struct regexp_allocated),
967 + relative_STRUCT_OFFSET(struct regexp_allocated, regexp, xpv_cur),
968 SVt_REGEXP, FALSE, NONV, HASARENA,
969 FIT_ARENA(0, sizeof(struct regexp_allocated))
5c35adbb 970 },
4df7f6af 971
bd81e77b 972 /* 48 */
10666ae3 973 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
974 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
975
bd81e77b 976 /* 64 */
10666ae3 977 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
978 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
979
980 { sizeof(xpvav_allocated),
981 copy_length(XPVAV, xmg_stash)
982 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
983 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
9c59bb28 984 SVt_PVAV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
985
986 { sizeof(xpvhv_allocated),
987 copy_length(XPVHV, xmg_stash)
988 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
989 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
9c59bb28 990 SVt_PVHV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 991
c84c4652 992 /* 56 */
4115f141 993 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 994 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 995 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 996
4115f141 997 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 998 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 999 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
1000
1001 /* XPVIO is 84 bytes, fits 48x */
167f2c4d
NC
1002 { sizeof(xpvio_allocated), sizeof(xpvio_allocated),
1003 + relative_STRUCT_OFFSET(xpvio_allocated, XPVIO, xpv_cur),
1004 SVt_PVIO, TRUE, NONV, HASARENA, FIT_ARENA(24, sizeof(xpvio_allocated)) },
bd81e77b 1005};
29489e7c 1006
d2a0f284
JC
1007#define new_body_type(sv_type) \
1008 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 1009
bd81e77b
NC
1010#define del_body_type(p, sv_type) \
1011 del_body(p, &PL_body_roots[sv_type])
29489e7c 1012
29489e7c 1013
bd81e77b 1014#define new_body_allocated(sv_type) \
d2a0f284 1015 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 1016 - bodies_by_type[sv_type].offset)
29489e7c 1017
bd81e77b
NC
1018#define del_body_allocated(p, sv_type) \
1019 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 1020
29489e7c 1021
bd81e77b
NC
1022#define my_safemalloc(s) (void*)safemalloc(s)
1023#define my_safecalloc(s) (void*)safecalloc(s, 1)
1024#define my_safefree(p) safefree((char*)p)
29489e7c 1025
bd81e77b 1026#ifdef PURIFY
29489e7c 1027
bd81e77b
NC
1028#define new_XNV() my_safemalloc(sizeof(XPVNV))
1029#define del_XNV(p) my_safefree(p)
29489e7c 1030
bd81e77b
NC
1031#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1032#define del_XPVNV(p) my_safefree(p)
29489e7c 1033
bd81e77b
NC
1034#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1035#define del_XPVAV(p) my_safefree(p)
29489e7c 1036
bd81e77b
NC
1037#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1038#define del_XPVHV(p) my_safefree(p)
29489e7c 1039
bd81e77b
NC
1040#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1041#define del_XPVMG(p) my_safefree(p)
29489e7c 1042
bd81e77b
NC
1043#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1044#define del_XPVGV(p) my_safefree(p)
29489e7c 1045
bd81e77b 1046#else /* !PURIFY */
29489e7c 1047
bd81e77b
NC
1048#define new_XNV() new_body_type(SVt_NV)
1049#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1050
bd81e77b
NC
1051#define new_XPVNV() new_body_type(SVt_PVNV)
1052#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1053
bd81e77b
NC
1054#define new_XPVAV() new_body_allocated(SVt_PVAV)
1055#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1056
bd81e77b
NC
1057#define new_XPVHV() new_body_allocated(SVt_PVHV)
1058#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1059
bd81e77b
NC
1060#define new_XPVMG() new_body_type(SVt_PVMG)
1061#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1062
bd81e77b
NC
1063#define new_XPVGV() new_body_type(SVt_PVGV)
1064#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1065
bd81e77b 1066#endif /* PURIFY */
93e68bfb 1067
bd81e77b 1068/* no arena for you! */
93e68bfb 1069
bd81e77b 1070#define new_NOARENA(details) \
d2a0f284 1071 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1072#define new_NOARENAZ(details) \
d2a0f284
JC
1073 my_safecalloc((details)->body_size + (details)->offset)
1074
1075STATIC void *
de37a194 1076S_more_bodies (pTHX_ const svtype sv_type)
d2a0f284
JC
1077{
1078 dVAR;
1079 void ** const root = &PL_body_roots[sv_type];
96a5add6 1080 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1081 const size_t body_size = bdp->body_size;
1082 char *start;
1083 const char *end;
d8fca402 1084 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
0b2d3faa 1085#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
23e9d66c
NC
1086 static bool done_sanity_check;
1087
0b2d3faa
JH
1088 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1089 * variables like done_sanity_check. */
10666ae3 1090 if (!done_sanity_check) {
ea471437 1091 unsigned int i = SVt_LAST;
10666ae3
NC
1092
1093 done_sanity_check = TRUE;
1094
1095 while (i--)
1096 assert (bodies_by_type[i].type == i);
1097 }
1098#endif
1099
23e9d66c
NC
1100 assert(bdp->arena_size);
1101
d8fca402 1102 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
d2a0f284 1103
d8fca402 1104 end = start + arena_size - 2 * body_size;
d2a0f284 1105
d2a0f284 1106 /* computed count doesnt reflect the 1st slot reservation */
d8fca402
NC
1107#if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1108 DEBUG_m(PerlIO_printf(Perl_debug_log,
1109 "arena %p end %p arena-size %d (from %d) type %d "
1110 "size %d ct %d\n",
1111 (void*)start, (void*)end, (int)arena_size,
1112 (int)bdp->arena_size, sv_type, (int)body_size,
1113 (int)arena_size / (int)body_size));
1114#else
d2a0f284
JC
1115 DEBUG_m(PerlIO_printf(Perl_debug_log,
1116 "arena %p end %p arena-size %d type %d size %d ct %d\n",
6c9570dc 1117 (void*)start, (void*)end,
0e84aef4
JH
1118 (int)bdp->arena_size, sv_type, (int)body_size,
1119 (int)bdp->arena_size / (int)body_size));
d8fca402 1120#endif
d2a0f284
JC
1121 *root = (void *)start;
1122
d8fca402 1123 while (start <= end) {
d2a0f284
JC
1124 char * const next = start + body_size;
1125 *(void**) start = (void *)next;
1126 start = next;
1127 }
1128 *(void **)start = 0;
1129
1130 return *root;
1131}
1132
1133/* grab a new thing from the free list, allocating more if necessary.
1134 The inline version is used for speed in hot routines, and the
1135 function using it serves the rest (unless PURIFY).
1136*/
1137#define new_body_inline(xpv, sv_type) \
1138 STMT_START { \
1139 void ** const r3wt = &PL_body_roots[sv_type]; \
11b79775
DD
1140 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1141 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284 1142 *(r3wt) = *(void**)(xpv); \
d2a0f284
JC
1143 } STMT_END
1144
1145#ifndef PURIFY
1146
1147STATIC void *
de37a194 1148S_new_body(pTHX_ const svtype sv_type)
d2a0f284
JC
1149{
1150 dVAR;
1151 void *xpv;
1152 new_body_inline(xpv, sv_type);
1153 return xpv;
1154}
1155
1156#endif
93e68bfb 1157
238b27b3
NC
1158static const struct body_details fake_rv =
1159 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1160
bd81e77b
NC
1161/*
1162=for apidoc sv_upgrade
93e68bfb 1163
bd81e77b
NC
1164Upgrade an SV to a more complex form. Generally adds a new body type to the
1165SV, then copies across as much information as possible from the old body.
1166You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1167
bd81e77b 1168=cut
93e68bfb 1169*/
93e68bfb 1170
bd81e77b 1171void
aad570aa 1172Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
cac9b346 1173{
97aff369 1174 dVAR;
bd81e77b
NC
1175 void* old_body;
1176 void* new_body;
42d0e0b7 1177 const svtype old_type = SvTYPE(sv);
d2a0f284 1178 const struct body_details *new_type_details;
238b27b3 1179 const struct body_details *old_type_details
bd81e77b 1180 = bodies_by_type + old_type;
4df7f6af 1181 SV *referant = NULL;
cac9b346 1182
7918f24d
NC
1183 PERL_ARGS_ASSERT_SV_UPGRADE;
1184
bd81e77b
NC
1185 if (new_type != SVt_PV && SvIsCOW(sv)) {
1186 sv_force_normal_flags(sv, 0);
1187 }
cac9b346 1188
bd81e77b
NC
1189 if (old_type == new_type)
1190 return;
cac9b346 1191
bd81e77b 1192 old_body = SvANY(sv);
de042e1d 1193
bd81e77b
NC
1194 /* Copying structures onto other structures that have been neatly zeroed
1195 has a subtle gotcha. Consider XPVMG
cac9b346 1196
bd81e77b
NC
1197 +------+------+------+------+------+-------+-------+
1198 | NV | CUR | LEN | IV | MAGIC | STASH |
1199 +------+------+------+------+------+-------+-------+
1200 0 4 8 12 16 20 24 28
645c22ef 1201
bd81e77b
NC
1202 where NVs are aligned to 8 bytes, so that sizeof that structure is
1203 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1204
bd81e77b
NC
1205 +------+------+------+------+------+-------+-------+------+
1206 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1207 +------+------+------+------+------+-------+-------+------+
1208 0 4 8 12 16 20 24 28 32
08742458 1209
bd81e77b 1210 so what happens if you allocate memory for this structure:
30f9da9e 1211
bd81e77b
NC
1212 +------+------+------+------+------+-------+-------+------+------+...
1213 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1214 +------+------+------+------+------+-------+-------+------+------+...
1215 0 4 8 12 16 20 24 28 32 36
bfc44f79 1216
bd81e77b
NC
1217 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1218 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1219 started out as zero once, but it's quite possible that it isn't. So now,
1220 rather than a nicely zeroed GP, you have it pointing somewhere random.
1221 Bugs ensue.
bfc44f79 1222
bd81e77b
NC
1223 (In fact, GP ends up pointing at a previous GP structure, because the
1224 principle cause of the padding in XPVMG getting garbage is a copy of
6c9e42f7
NC
1225 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1226 this happens to be moot because XPVGV has been re-ordered, with GP
1227 no longer after STASH)
30f9da9e 1228
bd81e77b
NC
1229 So we are careful and work out the size of used parts of all the
1230 structures. */
bfc44f79 1231
bd81e77b
NC
1232 switch (old_type) {
1233 case SVt_NULL:
1234 break;
1235 case SVt_IV:
4df7f6af
NC
1236 if (SvROK(sv)) {
1237 referant = SvRV(sv);
238b27b3
NC
1238 old_type_details = &fake_rv;
1239 if (new_type == SVt_NV)
1240 new_type = SVt_PVNV;
4df7f6af
NC
1241 } else {
1242 if (new_type < SVt_PVIV) {
1243 new_type = (new_type == SVt_NV)
1244 ? SVt_PVNV : SVt_PVIV;
1245 }
bd81e77b
NC
1246 }
1247 break;
1248 case SVt_NV:
1249 if (new_type < SVt_PVNV) {
1250 new_type = SVt_PVNV;
bd81e77b
NC
1251 }
1252 break;
bd81e77b
NC
1253 case SVt_PV:
1254 assert(new_type > SVt_PV);
1255 assert(SVt_IV < SVt_PV);
1256 assert(SVt_NV < SVt_PV);
1257 break;
1258 case SVt_PVIV:
1259 break;
1260 case SVt_PVNV:
1261 break;
1262 case SVt_PVMG:
1263 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1264 there's no way that it can be safely upgraded, because perl.c
1265 expects to Safefree(SvANY(PL_mess_sv)) */
1266 assert(sv != PL_mess_sv);
1267 /* This flag bit is used to mean other things in other scalar types.
1268 Given that it only has meaning inside the pad, it shouldn't be set
1269 on anything that can get upgraded. */
00b1698f 1270 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1271 break;
1272 default:
1273 if (old_type_details->cant_upgrade)
c81225bc
NC
1274 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1275 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1276 }
3376de98
NC
1277
1278 if (old_type > new_type)
1279 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1280 (int)old_type, (int)new_type);
1281
2fa1109b 1282 new_type_details = bodies_by_type + new_type;
645c22ef 1283
bd81e77b
NC
1284 SvFLAGS(sv) &= ~SVTYPEMASK;
1285 SvFLAGS(sv) |= new_type;
932e9ff9 1286
ab4416c0
NC
1287 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1288 the return statements above will have triggered. */
1289 assert (new_type != SVt_NULL);
bd81e77b 1290 switch (new_type) {
bd81e77b
NC
1291 case SVt_IV:
1292 assert(old_type == SVt_NULL);
1293 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1294 SvIV_set(sv, 0);
1295 return;
1296 case SVt_NV:
1297 assert(old_type == SVt_NULL);
1298 SvANY(sv) = new_XNV();
1299 SvNV_set(sv, 0);
1300 return;
bd81e77b 1301 case SVt_PVHV:
bd81e77b 1302 case SVt_PVAV:
d2a0f284 1303 assert(new_type_details->body_size);
c1ae03ae
NC
1304
1305#ifndef PURIFY
1306 assert(new_type_details->arena);
d2a0f284 1307 assert(new_type_details->arena_size);
c1ae03ae 1308 /* This points to the start of the allocated area. */
d2a0f284
JC
1309 new_body_inline(new_body, new_type);
1310 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1311 new_body = ((char *)new_body) - new_type_details->offset;
1312#else
1313 /* We always allocated the full length item with PURIFY. To do this
1314 we fake things so that arena is false for all 16 types.. */
1315 new_body = new_NOARENAZ(new_type_details);
1316#endif
1317 SvANY(sv) = new_body;
1318 if (new_type == SVt_PVAV) {
1319 AvMAX(sv) = -1;
1320 AvFILLp(sv) = -1;
1321 AvREAL_only(sv);
64484faa 1322 if (old_type_details->body_size) {
ac572bf4
NC
1323 AvALLOC(sv) = 0;
1324 } else {
1325 /* It will have been zeroed when the new body was allocated.
1326 Lets not write to it, in case it confuses a write-back
1327 cache. */
1328 }
78ac7dd9
NC
1329 } else {
1330 assert(!SvOK(sv));
1331 SvOK_off(sv);
1332#ifndef NODEFAULT_SHAREKEYS
1333 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1334#endif
1335 HvMAX(sv) = 7; /* (start with 8 buckets) */
64484faa 1336 if (old_type_details->body_size) {
78ac7dd9
NC
1337 HvFILL(sv) = 0;
1338 } else {
1339 /* It will have been zeroed when the new body was allocated.
1340 Lets not write to it, in case it confuses a write-back
1341 cache. */
1342 }
c1ae03ae 1343 }
aeb18a1e 1344
bd81e77b
NC
1345 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1346 The target created by newSVrv also is, and it can have magic.
1347 However, it never has SvPVX set.
1348 */
4df7f6af
NC
1349 if (old_type == SVt_IV) {
1350 assert(!SvROK(sv));
1351 } else if (old_type >= SVt_PV) {
bd81e77b
NC
1352 assert(SvPVX_const(sv) == 0);
1353 }
aeb18a1e 1354
bd81e77b 1355 if (old_type >= SVt_PVMG) {
e736a858 1356 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1357 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1358 } else {
1359 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1360 }
1361 break;
93e68bfb 1362
93e68bfb 1363
bd81e77b
NC
1364 case SVt_PVIV:
1365 /* XXX Is this still needed? Was it ever needed? Surely as there is
1366 no route from NV to PVIV, NOK can never be true */
1367 assert(!SvNOKp(sv));
1368 assert(!SvNOK(sv));
1369 case SVt_PVIO:
1370 case SVt_PVFM:
bd81e77b
NC
1371 case SVt_PVGV:
1372 case SVt_PVCV:
1373 case SVt_PVLV:
5c35adbb 1374 case SVt_REGEXP:
bd81e77b
NC
1375 case SVt_PVMG:
1376 case SVt_PVNV:
1377 case SVt_PV:
93e68bfb 1378
d2a0f284 1379 assert(new_type_details->body_size);
bd81e77b
NC
1380 /* We always allocated the full length item with PURIFY. To do this
1381 we fake things so that arena is false for all 16 types.. */
1382 if(new_type_details->arena) {
1383 /* This points to the start of the allocated area. */
d2a0f284
JC
1384 new_body_inline(new_body, new_type);
1385 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1386 new_body = ((char *)new_body) - new_type_details->offset;
1387 } else {
1388 new_body = new_NOARENAZ(new_type_details);
1389 }
1390 SvANY(sv) = new_body;
5e2fc214 1391
bd81e77b 1392 if (old_type_details->copy) {
f9ba3d20
NC
1393 /* There is now the potential for an upgrade from something without
1394 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1395 int offset = old_type_details->offset;
1396 int length = old_type_details->copy;
1397
1398 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1399 const int difference
f9ba3d20
NC
1400 = new_type_details->offset - old_type_details->offset;
1401 offset += difference;
1402 length -= difference;
1403 }
1404 assert (length >= 0);
1405
1406 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1407 char);
bd81e77b
NC
1408 }
1409
1410#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1411 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1412 * correct 0.0 for us. Otherwise, if the old body didn't have an
1413 * NV slot, but the new one does, then we need to initialise the
1414 * freshly created NV slot with whatever the correct bit pattern is
1415 * for 0.0 */
e22a937e
NC
1416 if (old_type_details->zero_nv && !new_type_details->zero_nv
1417 && !isGV_with_GP(sv))
bd81e77b 1418 SvNV_set(sv, 0);
82048762 1419#endif
5e2fc214 1420
bd81e77b 1421 if (new_type == SVt_PVIO)
f2524eef 1422 IoPAGE_LEN(sv) = 60;
4df7f6af
NC
1423 if (old_type < SVt_PV) {
1424 /* referant will be NULL unless the old type was SVt_IV emulating
1425 SVt_RV */
1426 sv->sv_u.svu_rv = referant;
1427 }
bd81e77b
NC
1428 break;
1429 default:
afd78fd5
JH
1430 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1431 (unsigned long)new_type);
bd81e77b 1432 }
73171d91 1433
d2a0f284
JC
1434 if (old_type_details->arena) {
1435 /* If there was an old body, then we need to free it.
1436 Note that there is an assumption that all bodies of types that
1437 can be upgraded came from arenas. Only the more complex non-
1438 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1439#ifdef PURIFY
1440 my_safefree(old_body);
1441#else
1442 del_body((void*)((char*)old_body + old_type_details->offset),
1443 &PL_body_roots[old_type]);
1444#endif
1445 }
1446}
73171d91 1447
bd81e77b
NC
1448/*
1449=for apidoc sv_backoff
73171d91 1450
bd81e77b
NC
1451Remove any string offset. You should normally use the C<SvOOK_off> macro
1452wrapper instead.
73171d91 1453
bd81e77b 1454=cut
73171d91
NC
1455*/
1456
bd81e77b 1457int
aad570aa 1458Perl_sv_backoff(pTHX_ register SV *const sv)
bd81e77b 1459{
69240efd 1460 STRLEN delta;
7a4bba22 1461 const char * const s = SvPVX_const(sv);
7918f24d
NC
1462
1463 PERL_ARGS_ASSERT_SV_BACKOFF;
96a5add6 1464 PERL_UNUSED_CONTEXT;
7918f24d 1465
bd81e77b
NC
1466 assert(SvOOK(sv));
1467 assert(SvTYPE(sv) != SVt_PVHV);
1468 assert(SvTYPE(sv) != SVt_PVAV);
7a4bba22 1469
69240efd
NC
1470 SvOOK_offset(sv, delta);
1471
7a4bba22
NC
1472 SvLEN_set(sv, SvLEN(sv) + delta);
1473 SvPV_set(sv, SvPVX(sv) - delta);
1474 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
bd81e77b
NC
1475 SvFLAGS(sv) &= ~SVf_OOK;
1476 return 0;
1477}
73171d91 1478
bd81e77b
NC
1479/*
1480=for apidoc sv_grow
73171d91 1481
bd81e77b
NC
1482Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1483upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1484Use the C<SvGROW> wrapper instead.
93e68bfb 1485
bd81e77b
NC
1486=cut
1487*/
93e68bfb 1488
bd81e77b 1489char *
aad570aa 1490Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
bd81e77b
NC
1491{
1492 register char *s;
93e68bfb 1493
7918f24d
NC
1494 PERL_ARGS_ASSERT_SV_GROW;
1495
5db06880
NC
1496 if (PL_madskills && newlen >= 0x100000) {
1497 PerlIO_printf(Perl_debug_log,
1498 "Allocation too large: %"UVxf"\n", (UV)newlen);
1499 }
bd81e77b
NC
1500#ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000) {
1502 PerlIO_printf(Perl_debug_log,
1503 "Allocation too large: %"UVxf"\n", (UV)newlen);
1504 my_exit(1);
1505 }
1506#endif /* HAS_64K_LIMIT */
1507 if (SvROK(sv))
1508 sv_unref(sv);
1509 if (SvTYPE(sv) < SVt_PV) {
1510 sv_upgrade(sv, SVt_PV);
1511 s = SvPVX_mutable(sv);
1512 }
1513 else if (SvOOK(sv)) { /* pv is offset? */
1514 sv_backoff(sv);
1515 s = SvPVX_mutable(sv);
1516 if (newlen > SvLEN(sv))
1517 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1518#ifdef HAS_64K_LIMIT
1519 if (newlen >= 0x10000)
1520 newlen = 0xFFFF;
1521#endif
1522 }
1523 else
1524 s = SvPVX_mutable(sv);
aeb18a1e 1525
bd81e77b 1526 if (newlen > SvLEN(sv)) { /* need more room? */
aedff202 1527#ifndef Perl_safesysmalloc_size
bd81e77b 1528 newlen = PERL_STRLEN_ROUNDUP(newlen);
bd81e77b 1529#endif
98653f18 1530 if (SvLEN(sv) && s) {
10edeb5d 1531 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1532 }
1533 else {
10edeb5d 1534 s = (char*)safemalloc(newlen);
bd81e77b
NC
1535 if (SvPVX_const(sv) && SvCUR(sv)) {
1536 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1537 }
1538 }
1539 SvPV_set(sv, s);
ca7c1a29 1540#ifdef Perl_safesysmalloc_size
98653f18
NC
1541 /* Do this here, do it once, do it right, and then we will never get
1542 called back into sv_grow() unless there really is some growing
1543 needed. */
ca7c1a29 1544 SvLEN_set(sv, Perl_safesysmalloc_size(s));
98653f18 1545#else
bd81e77b 1546 SvLEN_set(sv, newlen);
98653f18 1547#endif
bd81e77b
NC
1548 }
1549 return s;
1550}
aeb18a1e 1551
bd81e77b
NC
1552/*
1553=for apidoc sv_setiv
932e9ff9 1554
bd81e77b
NC
1555Copies an integer into the given SV, upgrading first if necessary.
1556Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1557
bd81e77b
NC
1558=cut
1559*/
463ee0b2 1560
bd81e77b 1561void
aad570aa 1562Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
bd81e77b 1563{
97aff369 1564 dVAR;
7918f24d
NC
1565
1566 PERL_ARGS_ASSERT_SV_SETIV;
1567
bd81e77b
NC
1568 SV_CHECK_THINKFIRST_COW_DROP(sv);
1569 switch (SvTYPE(sv)) {
1570 case SVt_NULL:
bd81e77b 1571 case SVt_NV:
3376de98 1572 sv_upgrade(sv, SVt_IV);
bd81e77b 1573 break;
bd81e77b
NC
1574 case SVt_PV:
1575 sv_upgrade(sv, SVt_PVIV);
1576 break;
463ee0b2 1577
bd81e77b 1578 case SVt_PVGV:
6e592b3a
BM
1579 if (!isGV_with_GP(sv))
1580 break;
bd81e77b
NC
1581 case SVt_PVAV:
1582 case SVt_PVHV:
1583 case SVt_PVCV:
1584 case SVt_PVFM:
1585 case SVt_PVIO:
1586 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1587 OP_DESC(PL_op));
42d0e0b7 1588 default: NOOP;
bd81e77b
NC
1589 }
1590 (void)SvIOK_only(sv); /* validate number */
1591 SvIV_set(sv, i);
1592 SvTAINT(sv);
1593}
932e9ff9 1594
bd81e77b
NC
1595/*
1596=for apidoc sv_setiv_mg
d33b2eba 1597
bd81e77b 1598Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1599
bd81e77b
NC
1600=cut
1601*/
d33b2eba 1602
bd81e77b 1603void
aad570aa 1604Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
bd81e77b 1605{
7918f24d
NC
1606 PERL_ARGS_ASSERT_SV_SETIV_MG;
1607
bd81e77b
NC
1608 sv_setiv(sv,i);
1609 SvSETMAGIC(sv);
1610}
727879eb 1611
bd81e77b
NC
1612/*
1613=for apidoc sv_setuv
d33b2eba 1614
bd81e77b
NC
1615Copies an unsigned integer into the given SV, upgrading first if necessary.
1616Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1617
bd81e77b
NC
1618=cut
1619*/
d33b2eba 1620
bd81e77b 1621void
aad570aa 1622Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
bd81e77b 1623{
7918f24d
NC
1624 PERL_ARGS_ASSERT_SV_SETUV;
1625
bd81e77b
NC
1626 /* With these two if statements:
1627 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1628
bd81e77b
NC
1629 without
1630 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1631
bd81e77b
NC
1632 If you wish to remove them, please benchmark to see what the effect is
1633 */
1634 if (u <= (UV)IV_MAX) {
1635 sv_setiv(sv, (IV)u);
1636 return;
1637 }
1638 sv_setiv(sv, 0);
1639 SvIsUV_on(sv);
1640 SvUV_set(sv, u);
1641}
d33b2eba 1642
bd81e77b
NC
1643/*
1644=for apidoc sv_setuv_mg
727879eb 1645
bd81e77b 1646Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1647
bd81e77b
NC
1648=cut
1649*/
5e2fc214 1650
bd81e77b 1651void
aad570aa 1652Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
bd81e77b 1653{
7918f24d
NC
1654 PERL_ARGS_ASSERT_SV_SETUV_MG;
1655
bd81e77b
NC
1656 sv_setuv(sv,u);
1657 SvSETMAGIC(sv);
1658}
5e2fc214 1659
954c1994 1660/*
bd81e77b 1661=for apidoc sv_setnv
954c1994 1662
bd81e77b
NC
1663Copies a double into the given SV, upgrading first if necessary.
1664Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1665
1666=cut
1667*/
1668
63f97190 1669void
aad570aa 1670Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
79072805 1671{
97aff369 1672 dVAR;
7918f24d
NC
1673
1674 PERL_ARGS_ASSERT_SV_SETNV;
1675
bd81e77b
NC
1676 SV_CHECK_THINKFIRST_COW_DROP(sv);
1677 switch (SvTYPE(sv)) {
79072805 1678 case SVt_NULL:
79072805 1679 case SVt_IV:
bd81e77b 1680 sv_upgrade(sv, SVt_NV);
79072805
LW
1681 break;
1682 case SVt_PV:
79072805 1683 case SVt_PVIV:
bd81e77b 1684 sv_upgrade(sv, SVt_PVNV);
79072805 1685 break;
bd4b1eb5 1686
bd4b1eb5 1687 case SVt_PVGV:
6e592b3a
BM
1688 if (!isGV_with_GP(sv))
1689 break;
bd81e77b
NC
1690 case SVt_PVAV:
1691 case SVt_PVHV:
79072805 1692 case SVt_PVCV:
bd81e77b
NC
1693 case SVt_PVFM:
1694 case SVt_PVIO:
1695 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1696 OP_NAME(PL_op));
42d0e0b7 1697 default: NOOP;
2068cd4d 1698 }
bd81e77b
NC
1699 SvNV_set(sv, num);
1700 (void)SvNOK_only(sv); /* validate number */
1701 SvTAINT(sv);
79072805
LW
1702}
1703
645c22ef 1704/*
bd81e77b 1705=for apidoc sv_setnv_mg
645c22ef 1706
bd81e77b 1707Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1708
1709=cut
1710*/
1711
bd81e77b 1712void
aad570aa 1713Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
79072805 1714{
7918f24d
NC
1715 PERL_ARGS_ASSERT_SV_SETNV_MG;
1716
bd81e77b
NC
1717 sv_setnv(sv,num);
1718 SvSETMAGIC(sv);
79072805
LW
1719}
1720
bd81e77b
NC
1721/* Print an "isn't numeric" warning, using a cleaned-up,
1722 * printable version of the offending string
1723 */
954c1994 1724
bd81e77b 1725STATIC void
aad570aa 1726S_not_a_number(pTHX_ SV *const sv)
79072805 1727{
97aff369 1728 dVAR;
bd81e77b
NC
1729 SV *dsv;
1730 char tmpbuf[64];
1731 const char *pv;
94463019 1732
7918f24d
NC
1733 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1734
94463019 1735 if (DO_UTF8(sv)) {
84bafc02 1736 dsv = newSVpvs_flags("", SVs_TEMP);
94463019
JH
1737 pv = sv_uni_display(dsv, sv, 10, 0);
1738 } else {
1739 char *d = tmpbuf;
551405c4 1740 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1741 /* each *s can expand to 4 chars + "...\0",
1742 i.e. need room for 8 chars */
ecdeb87c 1743
00b6aa41
AL
1744 const char *s = SvPVX_const(sv);
1745 const char * const end = s + SvCUR(sv);
1746 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1747 int ch = *s & 0xFF;
1748 if (ch & 128 && !isPRINT_LC(ch)) {
1749 *d++ = 'M';
1750 *d++ = '-';
1751 ch &= 127;
1752 }
1753 if (ch == '\n') {
1754 *d++ = '\\';
1755 *d++ = 'n';
1756 }
1757 else if (ch == '\r') {
1758 *d++ = '\\';
1759 *d++ = 'r';
1760 }
1761 else if (ch == '\f') {
1762 *d++ = '\\';
1763 *d++ = 'f';
1764 }
1765 else if (ch == '\\') {
1766 *d++ = '\\';
1767 *d++ = '\\';
1768 }
1769 else if (ch == '\0') {
1770 *d++ = '\\';
1771 *d++ = '0';
1772 }
1773 else if (isPRINT_LC(ch))
1774 *d++ = ch;
1775 else {
1776 *d++ = '^';
1777 *d++ = toCTRL(ch);
1778 }
1779 }
1780 if (s < end) {
1781 *d++ = '.';
1782 *d++ = '.';
1783 *d++ = '.';
1784 }
1785 *d = '\0';
1786 pv = tmpbuf;
a0d0e21e 1787 }
a0d0e21e 1788
533c011a 1789 if (PL_op)
9014280d 1790 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1791 "Argument \"%s\" isn't numeric in %s", pv,
1792 OP_DESC(PL_op));
a0d0e21e 1793 else
9014280d 1794 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1795 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1796}
1797
c2988b20
NC
1798/*
1799=for apidoc looks_like_number
1800
645c22ef
DM
1801Test if the content of an SV looks like a number (or is a number).
1802C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1803non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1804
1805=cut
1806*/
1807
1808I32
aad570aa 1809Perl_looks_like_number(pTHX_ SV *const sv)
c2988b20 1810{
a3b680e6 1811 register const char *sbegin;
c2988b20
NC
1812 STRLEN len;
1813
7918f24d
NC
1814 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1815
c2988b20 1816 if (SvPOK(sv)) {
3f7c398e 1817 sbegin = SvPVX_const(sv);
c2988b20
NC
1818 len = SvCUR(sv);
1819 }
1820 else if (SvPOKp(sv))
83003860 1821 sbegin = SvPV_const(sv, len);
c2988b20 1822 else
e0ab1c0e 1823 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1824 return grok_number(sbegin, len, NULL);
1825}
25da4f38 1826
19f6321d
NC
1827STATIC bool
1828S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1829{
1830 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1831 SV *const buffer = sv_newmortal();
1832
7918f24d
NC
1833 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1834
180488f8
NC
1835 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1836 is on. */
1837 SvFAKE_off(gv);
1838 gv_efullname3(buffer, gv, "*");
1839 SvFLAGS(gv) |= wasfake;
1840
675c862f
AL
1841 /* We know that all GVs stringify to something that is not-a-number,
1842 so no need to test that. */
1843 if (ckWARN(WARN_NUMERIC))
1844 not_a_number(buffer);
1845 /* We just want something true to return, so that S_sv_2iuv_common
1846 can tail call us and return true. */
19f6321d 1847 return TRUE;
675c862f
AL
1848}
1849
1850STATIC char *
19f6321d 1851S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1852{
1853 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1854 SV *const buffer = sv_newmortal();
1855
7918f24d
NC
1856 PERL_ARGS_ASSERT_GLOB_2PV;
1857
675c862f
AL
1858 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1859 is on. */
1860 SvFAKE_off(gv);
1861 gv_efullname3(buffer, gv, "*");
1862 SvFLAGS(gv) |= wasfake;
1863
1864 assert(SvPOK(buffer));
a6d61a6c
NC
1865 if (len) {
1866 *len = SvCUR(buffer);
1867 }
675c862f 1868 return SvPVX(buffer);
180488f8
NC
1869}
1870
25da4f38
IZ
1871/* Actually, ISO C leaves conversion of UV to IV undefined, but
1872 until proven guilty, assume that things are not that bad... */
1873
645c22ef
DM
1874/*
1875 NV_PRESERVES_UV:
1876
1877 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1878 an IV (an assumption perl has been based on to date) it becomes necessary
1879 to remove the assumption that the NV always carries enough precision to
1880 recreate the IV whenever needed, and that the NV is the canonical form.
1881 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1882 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1883 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1884 1) to distinguish between IV/UV/NV slots that have cached a valid
1885 conversion where precision was lost and IV/UV/NV slots that have a
1886 valid conversion which has lost no precision
645c22ef 1887 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1888 would lose precision, the precise conversion (or differently
1889 imprecise conversion) is also performed and cached, to prevent
1890 requests for different numeric formats on the same SV causing
1891 lossy conversion chains. (lossless conversion chains are perfectly
1892 acceptable (still))
1893
1894
1895 flags are used:
1896 SvIOKp is true if the IV slot contains a valid value
1897 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1898 SvNOKp is true if the NV slot contains a valid value
1899 SvNOK is true only if the NV value is accurate
1900
1901 so
645c22ef 1902 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1903 IV(or UV) would lose accuracy over a direct conversion from PV to
1904 IV(or UV). If it would, cache both conversions, return NV, but mark
1905 SV as IOK NOKp (ie not NOK).
1906
645c22ef 1907 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1908 NV would lose accuracy over a direct conversion from PV to NV. If it
1909 would, cache both conversions, flag similarly.
1910
1911 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1912 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1913 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1914 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1915 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1916
645c22ef
DM
1917 The benefit of this is that operations such as pp_add know that if
1918 SvIOK is true for both left and right operands, then integer addition
1919 can be used instead of floating point (for cases where the result won't
1920 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1921 loss of precision compared with integer addition.
1922
1923 * making IV and NV equal status should make maths accurate on 64 bit
1924 platforms
1925 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1926 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1927 looking for SvIOK and checking for overflow will not outweigh the
1928 fp to integer speedup)
1929 * will slow down integer operations (callers of SvIV) on "inaccurate"
1930 values, as the change from SvIOK to SvIOKp will cause a call into
1931 sv_2iv each time rather than a macro access direct to the IV slot
1932 * should speed up number->string conversion on integers as IV is
645c22ef 1933 favoured when IV and NV are equally accurate
28e5dec8
JH
1934
1935 ####################################################################
645c22ef
DM
1936 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1937 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1938 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1939 ####################################################################
1940
645c22ef 1941 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1942 performance ratio.
1943*/
1944
1945#ifndef NV_PRESERVES_UV
645c22ef
DM
1946# define IS_NUMBER_UNDERFLOW_IV 1
1947# define IS_NUMBER_UNDERFLOW_UV 2
1948# define IS_NUMBER_IV_AND_UV 2
1949# define IS_NUMBER_OVERFLOW_IV 4
1950# define IS_NUMBER_OVERFLOW_UV 5
1951
1952/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1953
1954/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1955STATIC int
5de3775c 1956S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
47031da6
NC
1957# ifdef DEBUGGING
1958 , I32 numtype
1959# endif
1960 )
28e5dec8 1961{
97aff369 1962 dVAR;
7918f24d
NC
1963
1964 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1965
3f7c398e 1966 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
1967 if (SvNVX(sv) < (NV)IV_MIN) {
1968 (void)SvIOKp_on(sv);
1969 (void)SvNOK_on(sv);
45977657 1970 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1971 return IS_NUMBER_UNDERFLOW_IV;
1972 }
1973 if (SvNVX(sv) > (NV)UV_MAX) {
1974 (void)SvIOKp_on(sv);
1975 (void)SvNOK_on(sv);
1976 SvIsUV_on(sv);
607fa7f2 1977 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1978 return IS_NUMBER_OVERFLOW_UV;
1979 }
c2988b20
NC
1980 (void)SvIOKp_on(sv);
1981 (void)SvNOK_on(sv);
1982 /* Can't use strtol etc to convert this string. (See truth table in
1983 sv_2iv */
1984 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1985 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1986 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1987 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1988 } else {
1989 /* Integer is imprecise. NOK, IOKp */
1990 }
1991 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1992 }
1993 SvIsUV_on(sv);
607fa7f2 1994 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1995 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1996 if (SvUVX(sv) == UV_MAX) {
1997 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1998 possibly be preserved by NV. Hence, it must be overflow.
1999 NOK, IOKp */
2000 return IS_NUMBER_OVERFLOW_UV;
2001 }
2002 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2003 } else {
2004 /* Integer is imprecise. NOK, IOKp */
28e5dec8 2005 }
c2988b20 2006 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 2007}
645c22ef
DM
2008#endif /* !NV_PRESERVES_UV*/
2009
af359546 2010STATIC bool
7918f24d
NC
2011S_sv_2iuv_common(pTHX_ SV *const sv)
2012{
97aff369 2013 dVAR;
7918f24d
NC
2014
2015 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2016
af359546 2017 if (SvNOKp(sv)) {
28e5dec8
JH
2018 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2019 * without also getting a cached IV/UV from it at the same time
2020 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
2021 * IV or UV at same time to avoid this. */
2022 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
2023
2024 if (SvTYPE(sv) == SVt_NV)
2025 sv_upgrade(sv, SVt_PVNV);
2026
28e5dec8
JH
2027 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2028 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2029 certainly cast into the IV range at IV_MAX, whereas the correct
2030 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2031 cases go to UV */
cab190d4
JD
2032#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2033 if (Perl_isnan(SvNVX(sv))) {
2034 SvUV_set(sv, 0);
2035 SvIsUV_on(sv);
fdbe6d7c 2036 return FALSE;
cab190d4 2037 }
cab190d4 2038#endif
28e5dec8 2039 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 2040 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
2041 if (SvNVX(sv) == (NV) SvIVX(sv)
2042#ifndef NV_PRESERVES_UV
2043 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2044 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2045 /* Don't flag it as "accurately an integer" if the number
2046 came from a (by definition imprecise) NV operation, and
2047 we're outside the range of NV integer precision */
2048#endif
2049 ) {
a43d94f2
NC
2050 if (SvNOK(sv))
2051 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2052 else {
2053 /* scalar has trailing garbage, eg "42a" */
2054 }
28e5dec8 2055 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2056 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
2057 PTR2UV(sv),
2058 SvNVX(sv),
2059 SvIVX(sv)));
2060
2061 } else {
2062 /* IV not precise. No need to convert from PV, as NV
2063 conversion would already have cached IV if it detected
2064 that PV->IV would be better than PV->NV->IV
2065 flags already correct - don't set public IOK. */
2066 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2067 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
2068 PTR2UV(sv),
2069 SvNVX(sv),
2070 SvIVX(sv)));
2071 }
2072 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2073 but the cast (NV)IV_MIN rounds to a the value less (more
2074 negative) than IV_MIN which happens to be equal to SvNVX ??
2075 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2076 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2077 (NV)UVX == NVX are both true, but the values differ. :-(
2078 Hopefully for 2s complement IV_MIN is something like
2079 0x8000000000000000 which will be exact. NWC */
d460ef45 2080 }
25da4f38 2081 else {
607fa7f2 2082 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
2083 if (
2084 (SvNVX(sv) == (NV) SvUVX(sv))
2085#ifndef NV_PRESERVES_UV
2086 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2087 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2088 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2089 /* Don't flag it as "accurately an integer" if the number
2090 came from a (by definition imprecise) NV operation, and
2091 we're outside the range of NV integer precision */
2092#endif
a43d94f2 2093 && SvNOK(sv)
28e5dec8
JH
2094 )
2095 SvIOK_on(sv);
25da4f38 2096 SvIsUV_on(sv);
1c846c1f 2097 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 2098 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 2099 PTR2UV(sv),
57def98f
JH
2100 SvUVX(sv),
2101 SvUVX(sv)));
25da4f38 2102 }
748a9306
LW
2103 }
2104 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2105 UV value;
504618e9 2106 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 2107 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 2108 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
2109 the same as the direct translation of the initial string
2110 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2111 be careful to ensure that the value with the .456 is around if the
2112 NV value is requested in the future).
1c846c1f 2113
af359546 2114 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 2115 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 2116 cache the NV if we are sure it's not needed.
25da4f38 2117 */
16b7a9a4 2118
c2988b20
NC
2119 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2120 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2121 == IS_NUMBER_IN_UV) {
5e045b90 2122 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
2123 if (SvTYPE(sv) < SVt_PVIV)
2124 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 2125 (void)SvIOK_on(sv);
c2988b20
NC
2126 } else if (SvTYPE(sv) < SVt_PVNV)
2127 sv_upgrade(sv, SVt_PVNV);
28e5dec8 2128
f2524eef 2129 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2130 we aren't going to call atof() below. If NVs don't preserve UVs
2131 then the value returned may have more precision than atof() will
2132 return, even though value isn't perfectly accurate. */
2133 if ((numtype & (IS_NUMBER_IN_UV
2134#ifdef NV_PRESERVES_UV
2135 | IS_NUMBER_NOT_INT
2136#endif
2137 )) == IS_NUMBER_IN_UV) {
2138 /* This won't turn off the public IOK flag if it was set above */
2139 (void)SvIOKp_on(sv);
2140
2141 if (!(numtype & IS_NUMBER_NEG)) {
2142 /* positive */;
2143 if (value <= (UV)IV_MAX) {
45977657 2144 SvIV_set(sv, (IV)value);
c2988b20 2145 } else {
af359546 2146 /* it didn't overflow, and it was positive. */
607fa7f2 2147 SvUV_set(sv, value);
c2988b20
NC
2148 SvIsUV_on(sv);
2149 }
2150 } else {
2151 /* 2s complement assumption */
2152 if (value <= (UV)IV_MIN) {
45977657 2153 SvIV_set(sv, -(IV)value);
c2988b20
NC
2154 } else {
2155 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2156 I'm assuming it will be rare. */
c2988b20
NC
2157 if (SvTYPE(sv) < SVt_PVNV)
2158 sv_upgrade(sv, SVt_PVNV);
2159 SvNOK_on(sv);
2160 SvIOK_off(sv);
2161 SvIOKp_on(sv);
9d6ce603 2162 SvNV_set(sv, -(NV)value);
45977657 2163 SvIV_set(sv, IV_MIN);
c2988b20
NC
2164 }
2165 }
2166 }
2167 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2168 will be in the previous block to set the IV slot, and the next
2169 block to set the NV slot. So no else here. */
2170
2171 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2172 != IS_NUMBER_IN_UV) {
2173 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2174 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2175
c2988b20
NC
2176 if (! numtype && ckWARN(WARN_NUMERIC))
2177 not_a_number(sv);
28e5dec8 2178
65202027 2179#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2180 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2181 PTR2UV(sv), SvNVX(sv)));
65202027 2182#else
1779d84d 2183 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2184 PTR2UV(sv), SvNVX(sv)));
65202027 2185#endif
28e5dec8 2186
28e5dec8 2187#ifdef NV_PRESERVES_UV
af359546
NC
2188 (void)SvIOKp_on(sv);
2189 (void)SvNOK_on(sv);
2190 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2193 SvIOK_on(sv);
2194 } else {
6f207bd3 2195 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2196 }
2197 /* UV will not work better than IV */
2198 } else {
2199 if (SvNVX(sv) > (NV)UV_MAX) {
2200 SvIsUV_on(sv);
2201 /* Integer is inaccurate. NOK, IOKp, is UV */
2202 SvUV_set(sv, UV_MAX);
af359546
NC
2203 } else {
2204 SvUV_set(sv, U_V(SvNVX(sv)));
2205 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2206 NV preservse UV so can do correct comparison. */
2207 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2208 SvIOK_on(sv);
af359546 2209 } else {
6f207bd3 2210 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2211 }
2212 }
4b0c9573 2213 SvIsUV_on(sv);
af359546 2214 }
28e5dec8 2215#else /* NV_PRESERVES_UV */
c2988b20
NC
2216 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2217 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2218 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2219 grok_number above. The NV slot has just been set using
2220 Atof. */
560b0c46 2221 SvNOK_on(sv);
c2988b20
NC
2222 assert (SvIOKp(sv));
2223 } else {
2224 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2225 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2226 /* Small enough to preserve all bits. */
2227 (void)SvIOKp_on(sv);
2228 SvNOK_on(sv);
45977657 2229 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2230 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2231 SvIOK_on(sv);
2232 /* Assumption: first non-preserved integer is < IV_MAX,
2233 this NV is in the preserved range, therefore: */
2234 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2235 < (UV)IV_MAX)) {
32fdb065 2236 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
2237 }
2238 } else {
2239 /* IN_UV NOT_INT
2240 0 0 already failed to read UV.
2241 0 1 already failed to read UV.
2242 1 0 you won't get here in this case. IV/UV
2243 slot set, public IOK, Atof() unneeded.
2244 1 1 already read UV.
2245 so there's no point in sv_2iuv_non_preserve() attempting
2246 to use atol, strtol, strtoul etc. */
47031da6 2247# ifdef DEBUGGING
40a17c4c 2248 sv_2iuv_non_preserve (sv, numtype);
47031da6
NC
2249# else
2250 sv_2iuv_non_preserve (sv);
2251# endif
c2988b20
NC
2252 }
2253 }
28e5dec8 2254#endif /* NV_PRESERVES_UV */
a43d94f2
NC
2255 /* It might be more code efficient to go through the entire logic above
2256 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2257 gets complex and potentially buggy, so more programmer efficient
2258 to do it this way, by turning off the public flags: */
2259 if (!numtype)
2260 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
25da4f38 2261 }
af359546
NC
2262 }
2263 else {
675c862f 2264 if (isGV_with_GP(sv))
a0933d07 2265 return glob_2number((GV *)sv);
180488f8 2266
af359546
NC
2267 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2268 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2269 report_uninit(sv);
2270 }
25da4f38
IZ
2271 if (SvTYPE(sv) < SVt_IV)
2272 /* Typically the caller expects that sv_any is not NULL now. */
2273 sv_upgrade(sv, SVt_IV);
af359546
NC
2274 /* Return 0 from the caller. */
2275 return TRUE;
2276 }
2277 return FALSE;
2278}
2279
2280/*
2281=for apidoc sv_2iv_flags
2282
2283Return the integer value of an SV, doing any necessary string
2284conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2285Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2286
2287=cut
2288*/
2289
2290IV
5de3775c 2291Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
af359546 2292{
97aff369 2293 dVAR;
af359546 2294 if (!sv)
a0d0e21e 2295 return 0;
cecf5685
NC
2296 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2297 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e
NC
2298 cache IVs just in case. In practice it seems that they never
2299 actually anywhere accessible by user Perl code, let alone get used
2300 in anything other than a string context. */
af359546
NC
2301 if (flags & SV_GMAGIC)
2302 mg_get(sv);
2303 if (SvIOKp(sv))
2304 return SvIVX(sv);
2305 if (SvNOKp(sv)) {
2306 return I_V(SvNVX(sv));
2307 }
71c558c3
NC
2308 if (SvPOKp(sv) && SvLEN(sv)) {
2309 UV value;
2310 const int numtype
2311 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2312
2313 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2314 == IS_NUMBER_IN_UV) {
2315 /* It's definitely an integer */
2316 if (numtype & IS_NUMBER_NEG) {
2317 if (value < (UV)IV_MIN)
2318 return -(IV)value;
2319 } else {
2320 if (value < (UV)IV_MAX)
2321 return (IV)value;
2322 }
2323 }
2324 if (!numtype) {
2325 if (ckWARN(WARN_NUMERIC))
2326 not_a_number(sv);
2327 }
2328 return I_V(Atof(SvPVX_const(sv)));
2329 }
1c7ff15e
NC
2330 if (SvROK(sv)) {
2331 goto return_rok;
af359546 2332 }
1c7ff15e
NC
2333 assert(SvTYPE(sv) >= SVt_PVMG);
2334 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2335 } else if (SvTHINKFIRST(sv)) {
af359546 2336 if (SvROK(sv)) {
1c7ff15e 2337 return_rok:
af359546
NC
2338 if (SvAMAGIC(sv)) {
2339 SV * const tmpstr=AMG_CALLun(sv,numer);
2340 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2341 return SvIV(tmpstr);
2342 }
2343 }
2344 return PTR2IV(SvRV(sv));
2345 }
2346 if (SvIsCOW(sv)) {
2347 sv_force_normal_flags(sv, 0);
2348 }
2349 if (SvREADONLY(sv) && !SvOK(sv)) {
2350 if (ckWARN(WARN_UNINITIALIZED))
2351 report_uninit(sv);
2352 return 0;
2353 }
2354 }
2355 if (!SvIOKp(sv)) {
2356 if (S_sv_2iuv_common(aTHX_ sv))
2357 return 0;
79072805 2358 }
1d7c1841
GS
2359 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2360 PTR2UV(sv),SvIVX(sv)));
25da4f38 2361 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2362}
2363
645c22ef 2364/*
891f9566 2365=for apidoc sv_2uv_flags
645c22ef
DM
2366
2367Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2368conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2369Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2370
2371=cut
2372*/
2373
ff68c719 2374UV
5de3775c 2375Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
ff68c719 2376{
97aff369 2377 dVAR;
ff68c719
PP
2378 if (!sv)
2379 return 0;
cecf5685
NC
2380 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2381 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2382 cache IVs just in case. */
891f9566
YST
2383 if (flags & SV_GMAGIC)
2384 mg_get(sv);
ff68c719
PP
2385 if (SvIOKp(sv))
2386 return SvUVX(sv);
2387 if (SvNOKp(sv))
2388 return U_V(SvNVX(sv));
71c558c3
NC
2389 if (SvPOKp(sv) && SvLEN(sv)) {
2390 UV value;
2391 const int numtype
2392 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2393
2394 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2395 == IS_NUMBER_IN_UV) {
2396 /* It's definitely an integer */
2397 if (!(numtype & IS_NUMBER_NEG))
2398 return value;
2399 }
2400 if (!numtype) {
2401 if (ckWARN(WARN_NUMERIC))
2402 not_a_number(sv);
2403 }
2404 return U_V(Atof(SvPVX_const(sv)));
2405 }
1c7ff15e
NC
2406 if (SvROK(sv)) {
2407 goto return_rok;
3fe9a6f1 2408 }
1c7ff15e
NC
2409 assert(SvTYPE(sv) >= SVt_PVMG);
2410 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2411 } else if (SvTHINKFIRST(sv)) {
ff68c719 2412 if (SvROK(sv)) {
1c7ff15e 2413 return_rok:
deb46114
NC
2414 if (SvAMAGIC(sv)) {
2415 SV *const tmpstr = AMG_CALLun(sv,numer);
2416 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2417 return SvUV(tmpstr);
2418 }
2419 }
2420 return PTR2UV(SvRV(sv));
ff68c719 2421 }
765f542d
NC
2422 if (SvIsCOW(sv)) {
2423 sv_force_normal_flags(sv, 0);
8a818333 2424 }
0336b60e 2425 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2426 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2427 report_uninit(sv);
ff68c719
PP
2428 return 0;
2429 }
2430 }
af359546
NC
2431 if (!SvIOKp(sv)) {
2432 if (S_sv_2iuv_common(aTHX_ sv))
2433 return 0;
ff68c719 2434 }
25da4f38 2435
1d7c1841
GS
2436 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2437 PTR2UV(sv),SvUVX(sv)));
25da4f38 2438 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719
PP
2439}
2440
645c22ef
DM
2441/*
2442=for apidoc sv_2nv
2443
2444Return the num value of an SV, doing any necessary string or integer
2445conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2446macros.
2447
2448=cut
2449*/
2450
65202027 2451NV
5de3775c 2452Perl_sv_2nv(pTHX_ register SV *const sv)
79072805 2453{
97aff369 2454 dVAR;
79072805
LW
2455 if (!sv)
2456 return 0.0;
cecf5685
NC
2457 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2458 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2459 cache IVs just in case. */
463ee0b2
LW
2460 mg_get(sv);
2461 if (SvNOKp(sv))
2462 return SvNVX(sv);
0aa395f8 2463 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2464 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2465 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2466 not_a_number(sv);
3f7c398e 2467 return Atof(SvPVX_const(sv));
a0d0e21e 2468 }
25da4f38 2469 if (SvIOKp(sv)) {
1c846c1f 2470 if (SvIsUV(sv))
65202027 2471 return (NV)SvUVX(sv);
25da4f38 2472 else
65202027 2473 return (NV)SvIVX(sv);
47a72cb8
NC
2474 }
2475 if (SvROK(sv)) {
2476 goto return_rok;
2477 }
2478 assert(SvTYPE(sv) >= SVt_PVMG);
2479 /* This falls through to the report_uninit near the end of the
2480 function. */
2481 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2482 if (SvROK(sv)) {
47a72cb8 2483 return_rok:
deb46114
NC
2484 if (SvAMAGIC(sv)) {
2485 SV *const tmpstr = AMG_CALLun(sv,numer);
2486 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2487 return SvNV(tmpstr);
2488 }
2489 }
2490 return PTR2NV(SvRV(sv));
a0d0e21e 2491 }
765f542d
NC
2492 if (SvIsCOW(sv)) {
2493 sv_force_normal_flags(sv, 0);
8a818333 2494 }
0336b60e 2495 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2496 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2497 report_uninit(sv);
ed6116ce
LW
2498 return 0.0;
2499 }
79072805
LW
2500 }
2501 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2502 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2503 sv_upgrade(sv, SVt_NV);
906f284f 2504#ifdef USE_LONG_DOUBLE
097ee67d 2505 DEBUG_c({
f93f4e46 2506 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2507 PerlIO_printf(Perl_debug_log,
2508 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2509 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2510 RESTORE_NUMERIC_LOCAL();
2511 });
65202027 2512#else
572bbb43 2513 DEBUG_c({
f93f4e46 2514 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2515 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2516 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2517 RESTORE_NUMERIC_LOCAL();
2518 });
572bbb43 2519#endif
79072805
LW
2520 }
2521 else if (SvTYPE(sv) < SVt_PVNV)
2522 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2523 if (SvNOKp(sv)) {
2524 return SvNVX(sv);
61604483 2525 }
59d8ce62 2526 if (SvIOKp(sv)) {
9d6ce603 2527 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8 2528#ifdef NV_PRESERVES_UV
a43d94f2
NC
2529 if (SvIOK(sv))
2530 SvNOK_on(sv);
2531 else
2532 SvNOKp_on(sv);
28e5dec8
JH
2533#else
2534 /* Only set the public NV OK flag if this NV preserves the IV */
2535 /* Check it's not 0xFFFFFFFFFFFFFFFF */
a43d94f2
NC
2536 if (SvIOK(sv) &&
2537 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
28e5dec8
JH
2538 : (SvIVX(sv) == I_V(SvNVX(sv))))
2539 SvNOK_on(sv);
2540 else
2541 SvNOKp_on(sv);
2542#endif
93a17b20 2543 }
748a9306 2544 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2545 UV value;
3f7c398e 2546 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2547 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2548 not_a_number(sv);
28e5dec8 2549#ifdef NV_PRESERVES_UV
c2988b20
NC
2550 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2551 == IS_NUMBER_IN_UV) {
5e045b90 2552 /* It's definitely an integer */
9d6ce603 2553 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2554 } else
3f7c398e 2555 SvNV_set(sv, Atof(SvPVX_const(sv)));
a43d94f2
NC
2556 if (numtype)
2557 SvNOK_on(sv);
2558 else
2559 SvNOKp_on(sv);
28e5dec8 2560#else
3f7c398e 2561 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2562 /* Only set the public NV OK flag if this NV preserves the value in
2563 the PV at least as well as an IV/UV would.
2564 Not sure how to do this 100% reliably. */
2565 /* if that shift count is out of range then Configure's test is
2566 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2567 UV_BITS */
2568 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2569 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2570 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2571 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2572 /* Can't use strtol etc to convert this string, so don't try.
2573 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2574 SvNOK_on(sv);
2575 } else {
2576 /* value has been set. It may not be precise. */
2577 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2578 /* 2s complement assumption for (UV)IV_MIN */
2579 SvNOK_on(sv); /* Integer is too negative. */
2580 } else {
2581 SvNOKp_on(sv);
2582 SvIOKp_on(sv);
6fa402ec 2583
c2988b20 2584 if (numtype & IS_NUMBER_NEG) {
45977657 2585 SvIV_set(sv, -(IV)value);
c2988b20 2586 } else if (value <= (UV)IV_MAX) {
45977657 2587 SvIV_set(sv, (IV)value);
c2988b20 2588 } else {
607fa7f2 2589 SvUV_set(sv, value);
c2988b20
NC
2590 SvIsUV_on(sv);
2591 }
2592
2593 if (numtype & IS_NUMBER_NOT_INT) {
2594 /* I believe that even if the original PV had decimals,
2595 they are lost beyond the limit of the FP precision.
2596 However, neither is canonical, so both only get p
2597 flags. NWC, 2000/11/25 */
2598 /* Both already have p flags, so do nothing */
2599 } else {
66a1b24b 2600 const NV nv = SvNVX(sv);
c2988b20
NC
2601 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2602 if (SvIVX(sv) == I_V(nv)) {
2603 SvNOK_on(sv);
c2988b20 2604 } else {
c2988b20
NC
2605 /* It had no "." so it must be integer. */
2606 }
00b6aa41 2607 SvIOK_on(sv);
c2988b20
NC
2608 } else {
2609 /* between IV_MAX and NV(UV_MAX).
2610 Could be slightly > UV_MAX */
6fa402ec 2611
c2988b20
NC
2612 if (numtype & IS_NUMBER_NOT_INT) {
2613 /* UV and NV both imprecise. */
2614 } else {
66a1b24b 2615 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2616
2617 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2618 SvNOK_on(sv);
c2988b20 2619 }
00b6aa41 2620 SvIOK_on(sv);
c2988b20
NC
2621 }
2622 }
2623 }
2624 }
2625 }
a43d94f2
NC
2626 /* It might be more code efficient to go through the entire logic above
2627 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2628 gets complex and potentially buggy, so more programmer efficient
2629 to do it this way, by turning off the public flags: */
2630 if (!numtype)
2631 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
28e5dec8 2632#endif /* NV_PRESERVES_UV */
93a17b20 2633 }
79072805 2634 else {
f7877b28 2635 if (isGV_with_GP(sv)) {
19f6321d 2636 glob_2number((GV *)sv);
180488f8
NC
2637 return 0.0;
2638 }
2639
041457d9 2640 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2641 report_uninit(sv);
7e25a7e9
NC
2642 assert (SvTYPE(sv) >= SVt_NV);
2643 /* Typically the caller expects that sv_any is not NULL now. */
2644 /* XXX Ilya implies that this is a bug in callers that assume this
2645 and ideally should be fixed. */
a0d0e21e 2646 return 0.0;
79072805 2647 }
572bbb43 2648#if defined(USE_LONG_DOUBLE)
097ee67d 2649 DEBUG_c({
f93f4e46 2650 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2651 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2652 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2653 RESTORE_NUMERIC_LOCAL();
2654 });
65202027 2655#else
572bbb43 2656 DEBUG_c({
f93f4e46 2657 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2658 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2659 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2660 RESTORE_NUMERIC_LOCAL();
2661 });
572bbb43 2662#endif
463ee0b2 2663 return SvNVX(sv);
79072805
LW
2664}
2665
800401ee
JH
2666/*
2667=for apidoc sv_2num
2668
2669Return an SV with the numeric value of the source SV, doing any necessary
a196a5fa
JH
2670reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2671access this function.
800401ee
JH
2672
2673=cut
2674*/
2675
2676SV *
5de3775c 2677Perl_sv_2num(pTHX_ register SV *const sv)
800401ee 2678{
7918f24d
NC
2679 PERL_ARGS_ASSERT_SV_2NUM;
2680
b9ee0594
RGS
2681 if (!SvROK(sv))
2682 return sv;
800401ee
JH
2683 if (SvAMAGIC(sv)) {
2684 SV * const tmpsv = AMG_CALLun(sv,numer);
2685 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2686 return sv_2num(tmpsv);
2687 }
2688 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2689}
2690
645c22ef
DM
2691/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2692 * UV as a string towards the end of buf, and return pointers to start and
2693 * end of it.
2694 *
2695 * We assume that buf is at least TYPE_CHARS(UV) long.
2696 */
2697
864dbfa3 2698static char *
5de3775c 2699S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
25da4f38 2700{
25da4f38 2701 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2702 char * const ebuf = ptr;
25da4f38 2703 int sign;
25da4f38 2704
7918f24d
NC
2705 PERL_ARGS_ASSERT_UIV_2BUF;
2706
25da4f38
IZ
2707 if (is_uv)
2708 sign = 0;
2709 else if (iv >= 0) {
2710 uv = iv;
2711 sign = 0;
2712 } else {
2713 uv = -iv;
2714 sign = 1;
2715 }
2716 do {
eb160463 2717 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2718 } while (uv /= 10);
2719 if (sign)
2720 *--ptr = '-';
2721 *peob = ebuf;
2722 return ptr;
2723}
2724
645c22ef
DM
2725/*
2726=for apidoc sv_2pv_flags
2727
ff276b08 2728Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2729If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2730if necessary.
2731Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2732usually end up here too.
2733
2734=cut
2735*/
2736
8d6d96c1 2737char *
5de3775c 2738Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
8d6d96c1 2739{
97aff369 2740 dVAR;
79072805 2741 register char *s;
79072805 2742
463ee0b2 2743 if (!sv) {
cdb061a3
NC
2744 if (lp)
2745 *lp = 0;
73d840c0 2746 return (char *)"";
463ee0b2 2747 }
8990e307 2748 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2749 if (flags & SV_GMAGIC)
2750 mg_get(sv);
463ee0b2 2751 if (SvPOKp(sv)) {
cdb061a3
NC
2752 if (lp)
2753 *lp = SvCUR(sv);
10516c54
NC
2754 if (flags & SV_MUTABLE_RETURN)
2755 return SvPVX_mutable(sv);
4d84ee25
NC
2756 if (flags & SV_CONST_RETURN)
2757 return (char *)SvPVX_const(sv);
463ee0b2
LW
2758 return SvPVX(sv);
2759 }
75dfc8ec
NC
2760 if (SvIOKp(sv) || SvNOKp(sv)) {
2761 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2762 STRLEN len;
2763
2764 if (SvIOKp(sv)) {
e80fed9d 2765 len = SvIsUV(sv)
d9fad198
JH
2766 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2767 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2768 } else {
e8ada2d0
NC
2769 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2770 len = strlen(tbuf);
75dfc8ec 2771 }
b5b886f0
NC
2772 assert(!SvROK(sv));
2773 {
75dfc8ec
NC
2774 dVAR;
2775
2776#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2777 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2778 tbuf[0] = '0';
2779 tbuf[1] = 0;
75dfc8ec
NC
2780 len = 1;
2781 }
2782#endif
2783 SvUPGRADE(sv, SVt_PV);
2784 if (lp)
2785 *lp = len;
2786 s = SvGROW_mutable(sv, len + 1);
2787 SvCUR_set(sv, len);
2788 SvPOKp_on(sv);
10edeb5d 2789 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2790 }
463ee0b2 2791 }
1c7ff15e
NC
2792 if (SvROK(sv)) {
2793 goto return_rok;
2794 }
2795 assert(SvTYPE(sv) >= SVt_PVMG);
2796 /* This falls through to the report_uninit near the end of the
2797 function. */
2798 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2799 if (SvROK(sv)) {
1c7ff15e 2800 return_rok:
deb46114
NC
2801 if (SvAMAGIC(sv)) {
2802 SV *const tmpstr = AMG_CALLun(sv,string);
2803 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2804 /* Unwrap this: */
2805 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2806 */
2807
2808 char *pv;
2809 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2810 if (flags & SV_CONST_RETURN) {
2811 pv = (char *) SvPVX_const(tmpstr);
2812 } else {
2813 pv = (flags & SV_MUTABLE_RETURN)
2814 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2815 }
2816 if (lp)
2817 *lp = SvCUR(tmpstr);
50adf7d2 2818 } else {
deb46114 2819 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2820 }
deb46114
NC
2821 if (SvUTF8(tmpstr))
2822 SvUTF8_on(sv);
2823 else
2824 SvUTF8_off(sv);
2825 return pv;
50adf7d2 2826 }
deb46114
NC
2827 }
2828 {
fafee734
NC
2829 STRLEN len;
2830 char *retval;
2831 char *buffer;
d8eae41e
NC
2832 const SV *const referent = (SV*)SvRV(sv);
2833
2834 if (!referent) {
fafee734
NC
2835 len = 7;
2836 retval = buffer = savepvn("NULLREF", len);
5c35adbb 2837 } else if (SvTYPE(referent) == SVt_REGEXP) {
67d2d14d
AB
2838 const REGEXP * const re = (REGEXP *)referent;
2839 I32 seen_evals = 0;
2840
2841 assert(re);
2842
2843 /* If the regex is UTF-8 we want the containing scalar to
2844 have an UTF-8 flag too */
2845 if (RX_UTF8(re))
2846 SvUTF8_on(sv);
2847 else
2848 SvUTF8_off(sv);
2849
2850 if ((seen_evals = RX_SEEN_EVALS(re)))
2851 PL_reginterp_cnt += seen_evals;
2852
2853 if (lp)
2854 *lp = RX_WRAPLEN(re);
2855
2856 return RX_WRAPPED(re);
d8eae41e
NC
2857 } else {
2858 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2859 const STRLEN typelen = strlen(typestr);
2860 UV addr = PTR2UV(referent);
2861 const char *stashname = NULL;
2862 STRLEN stashnamelen = 0; /* hush, gcc */
2863 const char *buffer_end;
d8eae41e 2864
d8eae41e 2865 if (SvOBJECT(referent)) {
fafee734
NC
2866 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2867
2868 if (name) {
2869 stashname = HEK_KEY(name);
2870 stashnamelen = HEK_LEN(name);
2871
2872 if (HEK_UTF8(name)) {
2873 SvUTF8_on(sv);
2874 } else {
2875 SvUTF8_off(sv);
2876 }
2877 } else {
2878 stashname = "__ANON__";
2879 stashnamelen = 8;
2880 }
2881 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2882 + 2 * sizeof(UV) + 2 /* )\0 */;
2883 } else {
2884 len = typelen + 3 /* (0x */
2885 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2886 }
fafee734
NC
2887
2888 Newx(buffer, len, char);
2889 buffer_end = retval = buffer + len;
2890
2891 /* Working backwards */
2892 *--retval = '\0';
2893 *--retval = ')';
2894 do {
2895 *--retval = PL_hexdigit[addr & 15];
2896 } while (addr >>= 4);
2897 *--retval = 'x';
2898 *--retval = '0';
2899 *--retval = '(';
2900
2901 retval -= typelen;
2902 memcpy(retval, typestr, typelen);
2903
2904 if (stashname) {
2905 *--retval = '=';
2906 retval -= stashnamelen;
2907 memcpy(retval, stashname, stashnamelen);
2908 }
2909 /* retval may not neccesarily have reached the start of the
2910 buffer here. */
2911 assert (retval >= buffer);
2912
2913 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2914 }
042dae7a 2915 if (lp)
fafee734
NC
2916 *lp = len;
2917 SAVEFREEPV(buffer);
2918 return retval;
463ee0b2 2919 }
79072805 2920 }
0336b60e 2921 if (SvREADONLY(sv) && !SvOK(sv)) {
cdb061a3
NC
2922 if (lp)
2923 *lp = 0;
9f621bb0
NC
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2925 return NULL;
2926 if (ckWARN(WARN_UNINITIALIZED))
2927 report_uninit(sv);
73d840c0 2928 return (char *)"";
79072805 2929 }
79072805 2930 }
28e5dec8
JH
2931 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2932 /* I'm assuming that if both IV and NV are equally valid then
2933 converting the IV is going to be more efficient */
e1ec3a88 2934 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2935 char buf[TYPE_CHARS(UV)];
2936 char *ebuf, *ptr;
97a130b8 2937 STRLEN len;
28e5dec8
JH
2938
2939 if (SvTYPE(sv) < SVt_PVIV)
2940 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2941 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
97a130b8 2942 len = ebuf - ptr;
5902b6a9 2943 /* inlined from sv_setpvn */
97a130b8
NC
2944 s = SvGROW_mutable(sv, len + 1);
2945 Move(ptr, s, len, char);
2946 s += len;
28e5dec8 2947 *s = '\0';
28e5dec8
JH
2948 }
2949 else if (SvNOKp(sv)) {
c81271c3 2950 const int olderrno = errno;
79072805
LW
2951 if (SvTYPE(sv) < SVt_PVNV)
2952 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2953 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2954 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2955 /* some Xenix systems wipe out errno here */
79072805 2956#ifdef apollo
463ee0b2 2957 if (SvNVX(sv) == 0.0)
d1307786 2958 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2959 else
2960#endif /*apollo*/
bbce6d69 2961 {
2d4389e4 2962 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2963 }
79072805 2964 errno = olderrno;
a0d0e21e 2965#ifdef FIXNEGATIVEZERO
20773dcd
NC
2966 if (*s == '-' && s[1] == '0' && !s[2]) {
2967 s[0] = '0';
2968 s[1] = 0;
2969 }
a0d0e21e 2970#endif
79072805
LW
2971 while (*s) s++;
2972#ifdef hcx
2973 if (s[-1] == '.')
46fc3d4c 2974 *--s = '\0';
79072805
LW
2975#endif
2976 }
79072805 2977 else {
675c862f 2978 if (isGV_with_GP(sv))
19f6321d 2979 return glob_2pv((GV *)sv, lp);
180488f8 2980
cdb061a3 2981 if (lp)
00b6aa41 2982 *lp = 0;
9f621bb0
NC
2983 if (flags & SV_UNDEF_RETURNS_NULL)
2984 return NULL;
2985 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2986 report_uninit(sv);
25da4f38
IZ
2987 if (SvTYPE(sv) < SVt_PV)
2988 /* Typically the caller expects that sv_any is not NULL now. */
2989 sv_upgrade(sv, SVt_PV);
73d840c0 2990 return (char *)"";
79072805 2991 }
cdb061a3 2992 {
823a54a3 2993 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2994 if (lp)
2995 *lp = len;
2996 SvCUR_set(sv, len);
2997 }
79072805 2998 SvPOK_on(sv);
1d7c1841 2999 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 3000 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
3001 if (flags & SV_CONST_RETURN)
3002 return (char *)SvPVX_const(sv);
10516c54
NC
3003 if (flags & SV_MUTABLE_RETURN)
3004 return SvPVX_mutable(sv);
463ee0b2
LW
3005 return SvPVX(sv);
3006}
3007
645c22ef 3008/*
6050d10e
JP
3009=for apidoc sv_copypv
3010
3011Copies a stringified representation of the source SV into the
3012destination SV. Automatically performs any necessary mg_get and
54f0641b 3013coercion of numeric values into strings. Guaranteed to preserve
2575c402 3014UTF8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
3015sv_2pv[_flags] but operates directly on an SV instead of just the
3016string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
3017would lose the UTF-8'ness of the PV.
3018
3019=cut
3020*/
3021
3022void
5de3775c 3023Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
6050d10e 3024{
446eaa42 3025 STRLEN len;
53c1dcc0 3026 const char * const s = SvPV_const(ssv,len);
7918f24d
NC
3027
3028 PERL_ARGS_ASSERT_SV_COPYPV;
3029
cb50f42d 3030 sv_setpvn(dsv,s,len);
446eaa42 3031 if (SvUTF8(ssv))
cb50f42d 3032 SvUTF8_on(dsv);
446eaa42 3033 else
cb50f42d 3034 SvUTF8_off(dsv);
6050d10e
JP
3035}
3036
3037/*
645c22ef
DM
3038=for apidoc sv_2pvbyte
3039
3040Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 3041to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
3042side-effect.
3043
3044Usually accessed via the C<SvPVbyte> macro.
3045
3046=cut
3047*/
3048
7340a771 3049char *
5de3775c 3050Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3051{
7918f24d
NC
3052 PERL_ARGS_ASSERT_SV_2PVBYTE;
3053
0875d2fe 3054 sv_utf8_downgrade(sv,0);
97972285 3055 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
3056}
3057
645c22ef 3058/*
035cbb0e
RGS
3059=for apidoc sv_2pvutf8
3060
3061Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3062to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3063
3064Usually accessed via the C<SvPVutf8> macro.
3065
3066=cut
3067*/
645c22ef 3068
7340a771 3069char *
7bc54cea 3070Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3071{
7918f24d
NC
3072 PERL_ARGS_ASSERT_SV_2PVUTF8;
3073
035cbb0e
RGS
3074 sv_utf8_upgrade(sv);
3075 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 3076}
1c846c1f 3077
7ee2227d 3078
645c22ef
DM
3079/*
3080=for apidoc sv_2bool
3081
3082This function is only called on magical items, and is only used by
8cf8f3d1 3083sv_true() or its macro equivalent.
645c22ef
DM
3084
3085=cut
3086*/
3087
463ee0b2 3088bool
7bc54cea 3089Perl_sv_2bool(pTHX_ register SV *const sv)
463ee0b2 3090{
97aff369 3091 dVAR;
7918f24d
NC
3092
3093 PERL_ARGS_ASSERT_SV_2BOOL;
3094
5b295bef 3095 SvGETMAGIC(sv);
463ee0b2 3096
a0d0e21e
LW
3097 if (!SvOK(sv))
3098 return 0;
3099 if (SvROK(sv)) {
fabdb6c0
AL
3100 if (SvAMAGIC(sv)) {
3101 SV * const tmpsv = AMG_CALLun(sv,bool_);
3102 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3103 return (bool)SvTRUE(tmpsv);
3104 }
3105 return SvRV(sv) != 0;
a0d0e21e 3106 }
463ee0b2 3107 if (SvPOKp(sv)) {
53c1dcc0
AL
3108 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3109 if (Xpvtmp &&
339049b0 3110 (*sv->sv_u.svu_pv > '0' ||
11343788 3111 Xpvtmp->xpv_cur > 1 ||
339049b0 3112 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
3113 return 1;
3114 else
3115 return 0;
3116 }
3117 else {
3118 if (SvIOKp(sv))
3119 return SvIVX(sv) != 0;
3120 else {
3121 if (SvNOKp(sv))
3122 return SvNVX(sv) != 0.0;
180488f8 3123 else {
f7877b28 3124 if (isGV_with_GP(sv))
180488f8
NC
3125 return TRUE;
3126 else
3127 return FALSE;
3128 }
463ee0b2
LW
3129 }
3130 }
79072805
LW
3131}
3132
c461cf8f
JH
3133/*
3134=for apidoc sv_utf8_upgrade
3135
78ea37eb 3136Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3137Forces the SV to string form if it is not already.
4411f3b6
NIS
3138Always sets the SvUTF8 flag to avoid future validity checks even
3139if all the bytes have hibit clear.
c461cf8f 3140
13a6c0e0
JH
3141This is not as a general purpose byte encoding to Unicode interface:
3142use the Encode extension for that.
3143
8d6d96c1
HS
3144=for apidoc sv_utf8_upgrade_flags
3145
78ea37eb 3146Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3147Forces the SV to string form if it is not already.
8d6d96c1
HS
3148Always sets the SvUTF8 flag to avoid future validity checks even
3149if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
3150will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
3151C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3152
13a6c0e0
JH
3153This is not as a general purpose byte encoding to Unicode interface:
3154use the Encode extension for that.
3155
8d6d96c1
HS
3156=cut
3157*/
3158
3159STRLEN
7bc54cea 3160Perl_sv_utf8_upgrade_flags(pTHX_ register SV *const sv, const I32 flags)
8d6d96c1 3161{
97aff369 3162 dVAR;
7918f24d
NC
3163
3164 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS;
3165
808c356f
RGS
3166 if (sv == &PL_sv_undef)
3167 return 0;
e0e62c2a
NIS
3168 if (!SvPOK(sv)) {
3169 STRLEN len = 0;
d52b7888
NC
3170 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3171 (void) sv_2pv_flags(sv,&len, flags);
3172 if (SvUTF8(sv))
3173 return len;
3174 } else {
3175 (void) SvPV_force(sv,len);
3176 }
e0e62c2a 3177 }
4411f3b6 3178
f5cee72b 3179 if (SvUTF8(sv)) {
5fec3b1d 3180 return SvCUR(sv);
f5cee72b 3181 }
5fec3b1d 3182
765f542d
NC
3183 if (SvIsCOW(sv)) {
3184 sv_force_normal_flags(sv, 0);
db42d148
NIS
3185 }
3186
88632417 3187 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 3188 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 3189 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
3190 /* This function could be much more efficient if we
3191 * had a FLAG in SVs to signal if there are any hibit
3192 * chars in the PV. Given that there isn't such a flag
3193 * make the loop as fast as possible. */
00b6aa41 3194 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3195 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3196 const U8 *t = s;
c4e7c712
NC
3197
3198 while (t < e) {
53c1dcc0 3199 const U8 ch = *t++;
00b6aa41
AL
3200 /* Check for hi bit */
3201 if (!NATIVE_IS_INVARIANT(ch)) {
4612962a
NC
3202 STRLEN len = SvCUR(sv);
3203 /* *Currently* bytes_to_utf8() adds a '\0' after every string
3204 it converts. This isn't documented. It's not clear if it's
3205 a bad thing to be doing, and should be changed to do exactly
3206 what the documentation says. If so, this code will have to
3207 be changed.
3208 As is, we mustn't rely on our incoming SV being well formed
3209 and having a trailing '\0', as certain code in pp_formline
3210 can send us partially built SVs. */
00b6aa41
AL
3211 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3212
3213 SvPV_free(sv); /* No longer using what was there before. */
3214 SvPV_set(sv, (char*)recoded);
4612962a
NC
3215 SvCUR_set(sv, len);
3216 SvLEN_set(sv, len + 1); /* No longer know the real size. */
c4e7c712 3217 break;
00b6aa41 3218 }
c4e7c712
NC
3219 }
3220 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3221 SvUTF8_on(sv);
560a288e 3222 }
4411f3b6 3223 return SvCUR(sv);
560a288e
GS
3224}
3225
c461cf8f
JH
3226/*
3227=for apidoc sv_utf8_downgrade
3228
78ea37eb
ST
3229Attempts to convert the PV of an SV from characters to bytes.
3230If the PV contains a character beyond byte, this conversion will fail;
3231in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3232true, croaks.
3233
13a6c0e0
JH
3234This is not as a general purpose Unicode to byte encoding interface:
3235use the Encode extension for that.
3236
c461cf8f
JH
3237=cut
3238*/
3239
560a288e 3240bool
7bc54cea 3241Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
560a288e 3242{
97aff369 3243 dVAR;
7918f24d
NC
3244
3245 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3246
78ea37eb 3247 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3248 if (SvCUR(sv)) {
03cfe0ae 3249 U8 *s;
652088fc 3250 STRLEN len;
fa301091 3251
765f542d
NC
3252 if (SvIsCOW(sv)) {
3253 sv_force_normal_flags(sv, 0);
3254 }
03cfe0ae
NIS
3255 s = (U8 *) SvPV(sv, len);
3256 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3257 if (fail_ok)
3258 return FALSE;
3259 else {
3260 if (PL_op)
3261 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3262 OP_DESC(PL_op));
fa301091
JH
3263 else
3264 Perl_croak(aTHX_ "Wide character");
3265 }
4b3603a4 3266 }
b162af07 3267 SvCUR_set(sv, len);
67e989fb 3268 }
560a288e 3269 }
ffebcc3e 3270 SvUTF8_off(sv);
560a288e
GS
3271 return TRUE;
3272}
3273
c461cf8f
JH
3274/*
3275=for apidoc sv_utf8_encode
3276
78ea37eb
ST
3277Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3278flag off so that it looks like octets again.
c461cf8f
JH
3279
3280=cut
3281*/
3282
560a288e 3283void
7bc54cea 3284Perl_sv_utf8_encode(pTHX_ register SV *const sv)
560a288e 3285{
7918f24d
NC
3286 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3287
4c94c214
NC
3288 if (SvIsCOW(sv)) {
3289 sv_force_normal_flags(sv, 0);
3290 }
3291 if (SvREADONLY(sv)) {
3292 Perl_croak(aTHX_ PL_no_modify);
3293 }
a5f5288a 3294 (void) sv_utf8_upgrade(sv);
560a288e
GS
3295 SvUTF8_off(sv);
3296}
3297
4411f3b6
NIS
3298/*
3299=for apidoc sv_utf8_decode
3300
78ea37eb
ST
3301If the PV of the SV is an octet sequence in UTF-8
3302and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3303so that it looks like a character. If the PV contains only single-byte
3304characters, the C<SvUTF8> flag stays being off.
3305Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3306
3307=cut
3308*/
3309
560a288e 3310bool
7bc54cea 3311Perl_sv_utf8_decode(pTHX_ register SV *const sv)
560a288e 3312{
7918f24d
NC
3313 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3314
78ea37eb 3315 if (SvPOKp(sv)) {
93524f2b
NC
3316 const U8 *c;
3317 const U8 *e;
9cbac4c7 3318
645c22ef
DM
3319 /* The octets may have got themselves encoded - get them back as
3320 * bytes
3321 */
3322 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3323 return FALSE;
3324
3325 /* it is actually just a matter of turning the utf8 flag on, but
3326 * we want to make sure everything inside is valid utf8 first.
3327 */
93524f2b 3328 c = (const U8 *) SvPVX_const(sv);
63cd0674 3329 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3330 return FALSE;
93524f2b 3331 e = (const U8 *) SvEND(sv);
511c2ff0 3332 while (c < e) {
b64e5050 3333 const U8 ch = *c++;
c4d5f83a 3334 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3335 SvUTF8_on(sv);
3336 break;
3337 }
560a288e 3338 }
560a288e
GS
3339 }
3340 return TRUE;
3341}
3342
954c1994
GS
3343/*
3344=for apidoc sv_setsv
3345
645c22ef
DM
3346Copies the contents of the source SV C<ssv> into the destination SV
3347C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3348function if the source SV needs to be reused. Does not handle 'set' magic.
3349Loosely speaking, it performs a copy-by-value, obliterating any previous
3350content of the destination.
3351
3352You probably want to use one of the assortment of wrappers, such as
3353C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3354C<SvSetMagicSV_nosteal>.
3355
8d6d96c1
HS
3356=for apidoc sv_setsv_flags
3357
645c22ef
DM
3358Copies the contents of the source SV C<ssv> into the destination SV
3359C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3360function if the source SV needs to be reused. Does not handle 'set' magic.
3361Loosely speaking, it performs a copy-by-value, obliterating any previous
3362content of the destination.
3363If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3364C<ssv> if appropriate, else not. If the C<flags> parameter has the
3365C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3366and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3367
3368You probably want to use one of the assortment of wrappers, such as
3369C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3370C<SvSetMagicSV_nosteal>.
3371
3372This is the primary function for copying scalars, and most other
3373copy-ish functions and macros use this underneath.
8d6d96c1
HS
3374
3375=cut
3376*/
3377
5d0301b7 3378static void
7bc54cea 3379S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
5d0301b7 3380{
70cd14a1 3381 I32 mro_changes = 0; /* 1 = method, 2 = isa */
dd69841b 3382
7918f24d
NC
3383 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3384
5d0301b7
NC
3385 if (dtype != SVt_PVGV) {
3386 const char * const name = GvNAME(sstr);
3387 const STRLEN len = GvNAMELEN(sstr);
0d092c36 3388 {
f7877b28
NC
3389 if (dtype >= SVt_PV) {
3390 SvPV_free(dstr);
3391 SvPV_set(dstr, 0);
3392 SvLEN_set(dstr, 0);
3393 SvCUR_set(dstr, 0);
3394 }
0d092c36 3395 SvUPGRADE(dstr, SVt_PVGV);
dedf8e73 3396 (void)SvOK_off(dstr);
2e5b91de
NC
3397 /* FIXME - why are we doing this, then turning it off and on again
3398 below? */
3399 isGV_with_GP_on(dstr);
f7877b28 3400 }
5d0301b7
NC
3401 GvSTASH(dstr) = GvSTASH(sstr);
3402 if (GvSTASH(dstr))
3403 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3404 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3405 SvFAKE_on(dstr); /* can coerce to non-glob */
3406 }
3407
3408#ifdef GV_UNIQUE_CHECK
3409 if (GvUNIQUE((GV*)dstr)) {
3410 Perl_croak(aTHX_ PL_no_modify);
3411 }
3412#endif
3413
dd69841b
BB
3414 if(GvGP((GV*)sstr)) {
3415 /* If source has method cache entry, clear it */
3416 if(GvCVGEN(sstr)) {
3417 SvREFCNT_dec(GvCV(sstr));
3418 GvCV(sstr) = NULL;
3419 GvCVGEN(sstr) = 0;
3420 }
3421 /* If source has a real method, then a method is
3422 going to change */
3423 else if(GvCV((GV*)sstr)) {
70cd14a1 3424 mro_changes = 1;
dd69841b
BB
3425 }
3426 }
3427
3428 /* If dest already had a real method, that's a change as well */
70cd14a1
CB
3429 if(!mro_changes && GvGP((GV*)dstr) && GvCVu((GV*)dstr)) {
3430 mro_changes = 1;
dd69841b
BB
3431 }
3432
70cd14a1
CB
3433 if(strEQ(GvNAME((GV*)dstr),"ISA"))
3434 mro_changes = 2;
3435
f7877b28 3436 gp_free((GV*)dstr);
2e5b91de 3437 isGV_with_GP_off(dstr);
5d0301b7 3438 (void)SvOK_off(dstr);
2e5b91de 3439 isGV_with_GP_on(dstr);
dedf8e73 3440 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3441 GvGP(dstr) = gp_ref(GvGP(sstr));
3442 if (SvTAINTED(sstr))
3443 SvTAINT(dstr);
3444 if (GvIMPORTED(dstr) != GVf_IMPORTED
3445 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3446 {
3447 GvIMPORTED_on(dstr);
3448 }
3449 GvMULTI_on(dstr);
70cd14a1
CB
3450 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3451 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
5d0301b7
NC
3452 return;
3453}
3454
b8473700 3455static void
7bc54cea 3456S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
7918f24d 3457{
b8473700
NC
3458 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3459 SV *dref = NULL;
3460 const int intro = GvINTRO(dstr);
2440974c 3461 SV **location;
3386d083 3462 U8 import_flag = 0;
27242d61
NC
3463 const U32 stype = SvTYPE(sref);
3464
7918f24d 3465 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
b8473700
NC
3466
3467#ifdef GV_UNIQUE_CHECK
3468 if (GvUNIQUE((GV*)dstr)) {
3469 Perl_croak(aTHX_ PL_no_modify);
3470 }
3471#endif
3472
3473 if (intro) {
3474 GvINTRO_off(dstr); /* one-shot flag */
3475 GvLINE(dstr) = CopLINE(PL_curcop);
3476 GvEGV(dstr) = (GV*)dstr;
3477 }
3478 GvMULTI_on(dstr);
27242d61 3479 switch (stype) {
b8473700 3480 case SVt_PVCV:
27242d61
NC
3481 location = (SV **) &GvCV(dstr);
3482 import_flag = GVf_IMPORTED_CV;
3483 goto common;
3484 case SVt_PVHV:
3485 location = (SV **) &GvHV(dstr);
3486 import_flag = GVf_IMPORTED_HV;
3487 goto common;
3488 case SVt_PVAV:
3489 location = (SV **) &GvAV(dstr);
3490 import_flag = GVf_IMPORTED_AV;
3491 goto common;
3492 case SVt_PVIO:
3493 location = (SV **) &GvIOp(dstr);
3494 goto common;
3495 case SVt_PVFM:
3496 location = (SV **) &GvFORM(dstr);
3497 default:
3498 location = &GvSV(dstr);
3499 import_flag = GVf_IMPORTED_SV;
3500 common:
b8473700 3501 if (intro) {
27242d61 3502 if (stype == SVt_PVCV) {
5f2fca8a
BB
3503 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (CV*)sref || GvCVGEN(dstr))) {*/
3504 if (GvCVGEN(dstr)) {
27242d61
NC
3505 SvREFCNT_dec(GvCV(dstr));
3506 GvCV(dstr) = NULL;
3507 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
27242d61 3508 }
b8473700 3509 }
27242d61 3510 SAVEGENERICSV(*location);
b8473700
NC
3511 }
3512 else
27242d61 3513 dref = *location;
5f2fca8a 3514 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
27242d61 3515 CV* const cv = (CV*)*location;
b8473700
NC
3516 if (cv) {
3517 if (!GvCVGEN((GV*)dstr) &&
3518 (CvROOT(cv) || CvXSUB(cv)))
3519 {
3520 /* Redefining a sub - warning is mandatory if
3521 it was a const and its value changed. */
3522 if (CvCONST(cv) && CvCONST((CV*)sref)
3523 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3524 NOOP;
b8473700
NC
3525 /* They are 2 constant subroutines generated from
3526 the same constant. This probably means that
3527 they are really the "same" proxy subroutine
3528 instantiated in 2 places. Most likely this is
3529 when a constant is exported twice. Don't warn.
3530 */
3531 }
3532 else if (ckWARN(WARN_REDEFINE)
3533 || (CvCONST(cv)
3534 && (!CvCONST((CV*)sref)
3535 || sv_cmp(cv_const_sv(cv),
3536 cv_const_sv((CV*)sref))))) {
3537 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3538 (const char *)
3539 (CvCONST(cv)
3540 ? "Constant subroutine %s::%s redefined"
3541 : "Subroutine %s::%s redefined"),
b8473700
NC
3542 HvNAME_get(GvSTASH((GV*)dstr)),
3543 GvENAME((GV*)dstr));
3544 }
3545 }
3546 if (!intro)
cbf82dd0
NC
3547 cv_ckproto_len(cv, (GV*)dstr,
3548 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3549 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3550 }
b8473700
NC
3551 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3552 GvASSUMECV_on(dstr);
dd69841b 3553 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
b8473700 3554 }
2440974c 3555 *location = sref;
3386d083
NC
3556 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3557 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3558 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3559 }
3560 break;
3561 }
b37c2d43 3562 SvREFCNT_dec(dref);
b8473700
NC
3563 if (SvTAINTED(sstr))
3564 SvTAINT(dstr);
3565 return;
3566}
3567
8d6d96c1 3568void
7bc54cea 3569Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
8d6d96c1 3570{
97aff369 3571 dVAR;
8990e307
LW
3572 register U32 sflags;
3573 register int dtype;
42d0e0b7 3574 register svtype stype;
463ee0b2 3575
7918f24d
NC
3576 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3577
79072805
LW
3578 if (sstr == dstr)
3579 return;
29f4f0ab
NC
3580
3581 if (SvIS_FREED(dstr)) {
3582 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
be2597df 3583 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
29f4f0ab 3584 }
765f542d 3585 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3586 if (!sstr)
3280af22 3587 sstr = &PL_sv_undef;
29f4f0ab 3588 if (SvIS_FREED(sstr)) {
6c9570dc
MHM
3589 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3590 (void*)sstr, (void*)dstr);
29f4f0ab 3591 }
8990e307
LW
3592 stype = SvTYPE(sstr);
3593 dtype = SvTYPE(dstr);
79072805 3594
52944de8 3595 (void)SvAMAGIC_off(dstr);
7a5fa8a2 3596 if ( SvVOK(dstr) )
ece467f9
JP
3597 {
3598 /* need to nuke the magic */
3599 mg_free(dstr);
ece467f9 3600 }
9e7bc3e8 3601
463ee0b2 3602 /* There's a lot of redundancy below but we're going for speed here */
79072805 3603
8990e307 3604 switch (stype) {
79072805 3605 case SVt_NULL:
aece5585 3606 undef_sstr:
20408e3c
GS
3607 if (dtype != SVt_PVGV) {
3608 (void)SvOK_off(dstr);
3609 return;
3610 }
3611 break;
463ee0b2 3612 case SVt_IV:
aece5585
GA
3613 if (SvIOK(sstr)) {
3614 switch (dtype) {
3615 case SVt_NULL:
8990e307 3616 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3617 break;
3618 case SVt_NV:
aece5585 3619 case SVt_PV:
a0d0e21e 3620 sv_upgrade(dstr, SVt_PVIV);
aece5585 3621 break;
010be86b
NC
3622 case SVt_PVGV:
3623 goto end_of_first_switch;
aece5585
GA
3624 }
3625 (void)SvIOK_only(dstr);
45977657 3626 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3627 if (SvIsUV(sstr))
3628 SvIsUV_on(dstr);
37c25af0
NC
3629 /* SvTAINTED can only be true if the SV has taint magic, which in
3630 turn means that the SV type is PVMG (or greater). This is the
3631 case statement for SVt_IV, so this cannot be true (whatever gcov
3632 may say). */
3633 assert(!SvTAINTED(sstr));
aece5585 3634 return;
8990e307 3635 }
4df7f6af
NC
3636 if (!SvROK(sstr))
3637 goto undef_sstr;
3638 if (dtype < SVt_PV && dtype != SVt_IV)
3639 sv_upgrade(dstr, SVt_IV);
3640 break;
aece5585 3641
463ee0b2 3642