This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
In S_mro_get_linear_isa_dfs(), save copying by making a shared hash
[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
fd0854ff 176#ifdef DEBUG_LEAKING_SCALARS
22162ca8 177# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
178#else
179# define FREE_SV_DEBUG_FILE(sv)
180#endif
181
48614a46
NC
182#ifdef PERL_POISON
183# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
184/* Whilst I'd love to do this, it seems that things like to check on
185 unreferenced scalars
7e337ee0 186# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 187*/
7e337ee0
JH
188# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
189 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
190#else
191# define SvARENA_CHAIN(sv) SvANY(sv)
192# define POSION_SV_HEAD(sv)
193#endif
194
053fc874
GS
195#define plant_SV(p) \
196 STMT_START { \
fd0854ff 197 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
198 POSION_SV_HEAD(p); \
199 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
200 SvFLAGS(p) = SVTYPEMASK; \
201 PL_sv_root = (p); \
202 --PL_sv_count; \
203 } STMT_END
a0d0e21e 204
053fc874
GS
205#define uproot_SV(p) \
206 STMT_START { \
207 (p) = PL_sv_root; \
bb7bbd9c 208 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
209 ++PL_sv_count; \
210 } STMT_END
211
645c22ef 212
cac9b346
NC
213/* make some more SVs by adding another arena */
214
cac9b346
NC
215STATIC SV*
216S_more_sv(pTHX)
217{
97aff369 218 dVAR;
cac9b346
NC
219 SV* sv;
220
221 if (PL_nice_chunk) {
222 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 223 PL_nice_chunk = NULL;
cac9b346
NC
224 PL_nice_chunk_size = 0;
225 }
226 else {
227 char *chunk; /* must use New here to match call to */
d2a0f284 228 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 229 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
230 }
231 uproot_SV(sv);
232 return sv;
233}
234
645c22ef
DM
235/* new_SV(): return a new, empty SV head */
236
eba0f806
DM
237#ifdef DEBUG_LEAKING_SCALARS
238/* provide a real function for a debugger to play with */
239STATIC SV*
240S_new_SV(pTHX)
241{
242 SV* sv;
243
eba0f806
DM
244 if (PL_sv_root)
245 uproot_SV(sv);
246 else
cac9b346 247 sv = S_more_sv(aTHX);
eba0f806
DM
248 SvANY(sv) = 0;
249 SvREFCNT(sv) = 1;
250 SvFLAGS(sv) = 0;
fd0854ff 251 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
e385c3bf
DM
252 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
253 ? PL_parser->copline
254 : PL_curcop
f24aceb1
DM
255 ? CopLINE(PL_curcop)
256 : 0
e385c3bf 257 );
fd0854ff
DM
258 sv->sv_debug_inpad = 0;
259 sv->sv_debug_cloned = 0;
fd0854ff 260 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 261
eba0f806
DM
262 return sv;
263}
264# define new_SV(p) (p)=S_new_SV(aTHX)
265
266#else
267# define new_SV(p) \
053fc874 268 STMT_START { \
053fc874
GS
269 if (PL_sv_root) \
270 uproot_SV(p); \
271 else \
cac9b346 272 (p) = S_more_sv(aTHX); \
053fc874
GS
273 SvANY(p) = 0; \
274 SvREFCNT(p) = 1; \
275 SvFLAGS(p) = 0; \
276 } STMT_END
eba0f806 277#endif
463ee0b2 278
645c22ef
DM
279
280/* del_SV(): return an empty SV head to the free list */
281
a0d0e21e 282#ifdef DEBUGGING
4561caa4 283
053fc874
GS
284#define del_SV(p) \
285 STMT_START { \
aea4f609 286 if (DEBUG_D_TEST) \
053fc874
GS
287 del_sv(p); \
288 else \
289 plant_SV(p); \
053fc874 290 } STMT_END
a0d0e21e 291
76e3520e 292STATIC void
cea2e8a9 293S_del_sv(pTHX_ SV *p)
463ee0b2 294{
97aff369 295 dVAR;
7918f24d
NC
296
297 PERL_ARGS_ASSERT_DEL_SV;
298
aea4f609 299 if (DEBUG_D_TEST) {
4633a7c4 300 SV* sva;
a3b680e6 301 bool ok = 0;
3280af22 302 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
303 const SV * const sv = sva + 1;
304 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 305 if (p >= sv && p < svend) {
a0d0e21e 306 ok = 1;
c0ff570e
NC
307 break;
308 }
a0d0e21e
LW
309 }
310 if (!ok) {
0453d815 311 if (ckWARN_d(WARN_INTERNAL))
9014280d 312 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
313 "Attempt to free non-arena SV: 0x%"UVxf
314 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
315 return;
316 }
317 }
4561caa4 318 plant_SV(p);
463ee0b2 319}
a0d0e21e 320
4561caa4
CS
321#else /* ! DEBUGGING */
322
323#define del_SV(p) plant_SV(p)
324
325#endif /* DEBUGGING */
463ee0b2 326
645c22ef
DM
327
328/*
ccfc67b7
JH
329=head1 SV Manipulation Functions
330
645c22ef
DM
331=for apidoc sv_add_arena
332
333Given a chunk of memory, link it to the head of the list of arenas,
334and split it into a list of free SVs.
335
336=cut
337*/
338
4633a7c4 339void
de37a194 340Perl_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
463ee0b2 341{
97aff369 342 dVAR;
0bd48802 343 SV* const sva = (SV*)ptr;
463ee0b2
LW
344 register SV* sv;
345 register SV* svend;
4633a7c4 346
7918f24d
NC
347 PERL_ARGS_ASSERT_SV_ADD_ARENA;
348
4633a7c4 349 /* The first SV in an arena isn't an SV. */
3280af22 350 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
351 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
352 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
353
3280af22
NIS
354 PL_sv_arenaroot = sva;
355 PL_sv_root = sva + 1;
4633a7c4
LW
356
357 svend = &sva[SvREFCNT(sva) - 1];
358 sv = sva + 1;
463ee0b2 359 while (sv < svend) {
48614a46 360 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 361#ifdef DEBUGGING
978b032e 362 SvREFCNT(sv) = 0;
03e36789 363#endif
4b69cbe3 364 /* Must always set typemask because it's always checked in on cleanup
03e36789 365 when the arenas are walked looking for objects. */
8990e307 366 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
367 sv++;
368 }
48614a46 369 SvARENA_CHAIN(sv) = 0;
03e36789
NC
370#ifdef DEBUGGING
371 SvREFCNT(sv) = 0;
372#endif
4633a7c4
LW
373 SvFLAGS(sv) = SVTYPEMASK;
374}
375
055972dc
DM
376/* visit(): call the named function for each non-free SV in the arenas
377 * whose flags field matches the flags/mask args. */
645c22ef 378
5226ed68 379STATIC I32
de37a194 380S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
8990e307 381{
97aff369 382 dVAR;
4633a7c4 383 SV* sva;
5226ed68 384 I32 visited = 0;
8990e307 385
7918f24d
NC
386 PERL_ARGS_ASSERT_VISIT;
387
3280af22 388 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 389 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 390 register SV* sv;
4561caa4 391 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
392 if (SvTYPE(sv) != SVTYPEMASK
393 && (sv->sv_flags & mask) == flags
394 && SvREFCNT(sv))
395 {
acfe0abc 396 (FCALL)(aTHX_ sv);
5226ed68
JH
397 ++visited;
398 }
8990e307
LW
399 }
400 }
5226ed68 401 return visited;
8990e307
LW
402}
403
758a08c3
JH
404#ifdef DEBUGGING
405
645c22ef
DM
406/* called by sv_report_used() for each live SV */
407
408static void
5fa45a31 409do_report_used(pTHX_ SV *const sv)
645c22ef
DM
410{
411 if (SvTYPE(sv) != SVTYPEMASK) {
412 PerlIO_printf(Perl_debug_log, "****\n");
413 sv_dump(sv);
414 }
415}
758a08c3 416#endif
645c22ef
DM
417
418/*
419=for apidoc sv_report_used
420
421Dump the contents of all SVs not yet freed. (Debugging aid).
422
423=cut
424*/
425
8990e307 426void
864dbfa3 427Perl_sv_report_used(pTHX)
4561caa4 428{
ff270d3a 429#ifdef DEBUGGING
055972dc 430 visit(do_report_used, 0, 0);
96a5add6
AL
431#else
432 PERL_UNUSED_CONTEXT;
ff270d3a 433#endif
4561caa4
CS
434}
435
645c22ef
DM
436/* called by sv_clean_objs() for each live SV */
437
438static void
de37a194 439do_clean_objs(pTHX_ SV *const ref)
645c22ef 440{
97aff369 441 dVAR;
ea724faa
NC
442 assert (SvROK(ref));
443 {
823a54a3
AL
444 SV * const target = SvRV(ref);
445 if (SvOBJECT(target)) {
446 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
447 if (SvWEAKREF(ref)) {
448 sv_del_backref(target, ref);
449 SvWEAKREF_off(ref);
450 SvRV_set(ref, NULL);
451 } else {
452 SvROK_off(ref);
453 SvRV_set(ref, NULL);
454 SvREFCNT_dec(target);
455 }
645c22ef
DM
456 }
457 }
458
459 /* XXX Might want to check arrays, etc. */
460}
461
462/* called by sv_clean_objs() for each live SV */
463
464#ifndef DISABLE_DESTRUCTOR_KLUDGE
465static void
f30de749 466do_clean_named_objs(pTHX_ SV *const sv)
645c22ef 467{
97aff369 468 dVAR;
ea724faa 469 assert(SvTYPE(sv) == SVt_PVGV);
d011219a
NC
470 assert(isGV_with_GP(sv));
471 if (GvGP(sv)) {
c69033f2
NC
472 if ((
473#ifdef PERL_DONT_CREATE_GVSV
474 GvSV(sv) &&
475#endif
476 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
477 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
478 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
9c12f1e5
RGS
479 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
480 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
645c22ef
DM
481 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
482 {
483 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 484 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
485 SvREFCNT_dec(sv);
486 }
487 }
488}
489#endif
490
491/*
492=for apidoc sv_clean_objs
493
494Attempt to destroy all objects not yet freed
495
496=cut
497*/
498
4561caa4 499void
864dbfa3 500Perl_sv_clean_objs(pTHX)
4561caa4 501{
97aff369 502 dVAR;
3280af22 503 PL_in_clean_objs = TRUE;
055972dc 504 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 505#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 506 /* some barnacles may yet remain, clinging to typeglobs */
d011219a 507 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
4561caa4 508#endif
3280af22 509 PL_in_clean_objs = FALSE;
4561caa4
CS
510}
511
645c22ef
DM
512/* called by sv_clean_all() for each live SV */
513
514static void
de37a194 515do_clean_all(pTHX_ SV *const sv)
645c22ef 516{
97aff369 517 dVAR;
cddfcddc
NC
518 if (sv == (SV*) PL_fdpid || sv == (SV *)PL_strtab) {
519 /* don't clean pid table and strtab */
d17ea597 520 return;
cddfcddc 521 }
645c22ef
DM
522 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
523 SvFLAGS(sv) |= SVf_BREAK;
524 SvREFCNT_dec(sv);
525}
526
527/*
528=for apidoc sv_clean_all
529
530Decrement the refcnt of each remaining SV, possibly triggering a
531cleanup. This function may have to be called multiple times to free
ff276b08 532SVs which are in complex self-referential hierarchies.
645c22ef
DM
533
534=cut
535*/
536
5226ed68 537I32
864dbfa3 538Perl_sv_clean_all(pTHX)
8990e307 539{
97aff369 540 dVAR;
5226ed68 541 I32 cleaned;
3280af22 542 PL_in_clean_all = TRUE;
055972dc 543 cleaned = visit(do_clean_all, 0,0);
3280af22 544 PL_in_clean_all = FALSE;
5226ed68 545 return cleaned;
8990e307 546}
463ee0b2 547
5e258f8c
JC
548/*
549 ARENASETS: a meta-arena implementation which separates arena-info
550 into struct arena_set, which contains an array of struct
551 arena_descs, each holding info for a single arena. By separating
552 the meta-info from the arena, we recover the 1st slot, formerly
553 borrowed for list management. The arena_set is about the size of an
39244528 554 arena, avoiding the needless malloc overhead of a naive linked-list.
5e258f8c
JC
555
556 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
557 memory in the last arena-set (1/2 on average). In trade, we get
558 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284 559 smaller types). The recovery of the wasted space allows use of
e15dad31
JC
560 small arenas for large, rare body types, by changing array* fields
561 in body_details_by_type[] below.
5e258f8c 562*/
5e258f8c 563struct arena_desc {
398c677b
NC
564 char *arena; /* the raw storage, allocated aligned */
565 size_t size; /* its size ~4k typ */
0a848332 566 U32 misc; /* type, and in future other things. */
5e258f8c
JC
567};
568
e6148039
NC
569struct arena_set;
570
571/* Get the maximum number of elements in set[] such that struct arena_set
e15dad31 572 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
e6148039
NC
573 therefore likely to be 1 aligned memory page. */
574
575#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
576 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
577
578struct arena_set {
579 struct arena_set* next;
0a848332
NC
580 unsigned int set_size; /* ie ARENAS_PER_SET */
581 unsigned int curr; /* index of next available arena-desc */
5e258f8c
JC
582 struct arena_desc set[ARENAS_PER_SET];
583};
584
645c22ef
DM
585/*
586=for apidoc sv_free_arenas
587
588Deallocate the memory used by all arenas. Note that all the individual SV
589heads and bodies within the arenas must already have been freed.
590
591=cut
592*/
4633a7c4 593void
864dbfa3 594Perl_sv_free_arenas(pTHX)
4633a7c4 595{
97aff369 596 dVAR;
4633a7c4
LW
597 SV* sva;
598 SV* svanext;
0a848332 599 unsigned int i;
4633a7c4
LW
600
601 /* Free arenas here, but be careful about fake ones. (We assume
602 contiguity of the fake ones with the corresponding real ones.) */
603
3280af22 604 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
605 svanext = (SV*) SvANY(sva);
606 while (svanext && SvFAKE(svanext))
607 svanext = (SV*) SvANY(svanext);
608
609 if (!SvFAKE(sva))
1df70142 610 Safefree(sva);
4633a7c4 611 }
93e68bfb 612
5e258f8c 613 {
0a848332
NC
614 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
615
616 while (aroot) {
617 struct arena_set *current = aroot;
618 i = aroot->curr;
619 while (i--) {
5e258f8c
JC
620 assert(aroot->set[i].arena);
621 Safefree(aroot->set[i].arena);
622 }
0a848332
NC
623 aroot = aroot->next;
624 Safefree(current);
5e258f8c
JC
625 }
626 }
dc8220bf 627 PL_body_arenas = 0;
fdda85ca 628
0a848332
NC
629 i = PERL_ARENA_ROOTS_SIZE;
630 while (i--)
93e68bfb 631 PL_body_roots[i] = 0;
93e68bfb 632
43c5f42d 633 Safefree(PL_nice_chunk);
bd61b366 634 PL_nice_chunk = NULL;
3280af22
NIS
635 PL_nice_chunk_size = 0;
636 PL_sv_arenaroot = 0;
637 PL_sv_root = 0;
4633a7c4
LW
638}
639
bd81e77b
NC
640/*
641 Here are mid-level routines that manage the allocation of bodies out
642 of the various arenas. There are 5 kinds of arenas:
29489e7c 643
bd81e77b
NC
644 1. SV-head arenas, which are discussed and handled above
645 2. regular body arenas
646 3. arenas for reduced-size bodies
647 4. Hash-Entry arenas
648 5. pte arenas (thread related)
29489e7c 649
bd81e77b
NC
650 Arena types 2 & 3 are chained by body-type off an array of
651 arena-root pointers, which is indexed by svtype. Some of the
652 larger/less used body types are malloced singly, since a large
653 unused block of them is wasteful. Also, several svtypes dont have
654 bodies; the data fits into the sv-head itself. The arena-root
655 pointer thus has a few unused root-pointers (which may be hijacked
656 later for arena types 4,5)
29489e7c 657
bd81e77b
NC
658 3 differs from 2 as an optimization; some body types have several
659 unused fields in the front of the structure (which are kept in-place
660 for consistency). These bodies can be allocated in smaller chunks,
661 because the leading fields arent accessed. Pointers to such bodies
662 are decremented to point at the unused 'ghost' memory, knowing that
663 the pointers are used with offsets to the real memory.
29489e7c 664
bd81e77b
NC
665 HE, HEK arenas are managed separately, with separate code, but may
666 be merge-able later..
667
668 PTE arenas are not sv-bodies, but they share these mid-level
669 mechanics, so are considered here. The new mid-level mechanics rely
670 on the sv_type of the body being allocated, so we just reserve one
671 of the unused body-slots for PTEs, then use it in those (2) PTE
672 contexts below (line ~10k)
673*/
674
bd26d9a3 675/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
676 TBD: export properly for hv.c: S_more_he().
677*/
678void*
de37a194 679Perl_get_arena(pTHX_ const size_t arena_size, const U32 misc)
5e258f8c 680{
7a89be66 681 dVAR;
5e258f8c 682 struct arena_desc* adesc;
39244528 683 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
0a848332 684 unsigned int curr;
5e258f8c 685
476a1e16
JC
686 /* shouldnt need this
687 if (!arena_size) arena_size = PERL_ARENA_SIZE;
688 */
5e258f8c
JC
689
690 /* may need new arena-set to hold new arena */
39244528
NC
691 if (!aroot || aroot->curr >= aroot->set_size) {
692 struct arena_set *newroot;
5e258f8c
JC
693 Newxz(newroot, 1, struct arena_set);
694 newroot->set_size = ARENAS_PER_SET;
39244528
NC
695 newroot->next = aroot;
696 aroot = newroot;
697 PL_body_arenas = (void *) newroot;
52944de8 698 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
5e258f8c
JC
699 }
700
701 /* ok, now have arena-set with at least 1 empty/available arena-desc */
39244528
NC
702 curr = aroot->curr++;
703 adesc = &(aroot->set[curr]);
5e258f8c
JC
704 assert(!adesc->arena);
705
89086707 706 Newx(adesc->arena, arena_size, char);
5e258f8c 707 adesc->size = arena_size;
0a848332 708 adesc->misc = misc;
d67b3c53
JH
709 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
710 curr, (void*)adesc->arena, (UV)arena_size));
5e258f8c
JC
711
712 return adesc->arena;
5e258f8c
JC
713}
714
53c1dcc0 715
bd81e77b 716/* return a thing to the free list */
29489e7c 717
bd81e77b
NC
718#define del_body(thing, root) \
719 STMT_START { \
00b6aa41 720 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
721 *thing_copy = *root; \
722 *root = (void*)thing_copy; \
bd81e77b 723 } STMT_END
29489e7c 724
bd81e77b 725/*
d2a0f284
JC
726
727=head1 SV-Body Allocation
728
729Allocation of SV-bodies is similar to SV-heads, differing as follows;
730the allocation mechanism is used for many body types, so is somewhat
731more complicated, it uses arena-sets, and has no need for still-live
732SV detection.
733
734At the outermost level, (new|del)_X*V macros return bodies of the
735appropriate type. These macros call either (new|del)_body_type or
736(new|del)_body_allocated macro pairs, depending on specifics of the
737type. Most body types use the former pair, the latter pair is used to
738allocate body types with "ghost fields".
739
740"ghost fields" are fields that are unused in certain types, and
741consequently dont need to actually exist. They are declared because
742they're part of a "base type", which allows use of functions as
743methods. The simplest examples are AVs and HVs, 2 aggregate types
744which don't use the fields which support SCALAR semantics.
745
746For these types, the arenas are carved up into *_allocated size
747chunks, we thus avoid wasted memory for those unaccessed members.
748When bodies are allocated, we adjust the pointer back in memory by the
749size of the bit not allocated, so it's as if we allocated the full
750structure. (But things will all go boom if you write to the part that
751is "not there", because you'll be overwriting the last members of the
752preceding structure in memory.)
753
754We calculate the correction using the STRUCT_OFFSET macro. For
755example, if xpv_allocated is the same structure as XPV then the two
756OFFSETs sum to zero, and the pointer is unchanged. If the allocated
757structure is smaller (no initial NV actually allocated) then the net
758effect is to subtract the size of the NV from the pointer, to return a
759new pointer as if an initial NV were actually allocated.
760
761This is the same trick as was used for NV and IV bodies. Ironically it
762doesn't need to be used for NV bodies any more, because NV is now at
763the start of the structure. IV bodies don't need it either, because
764they are no longer allocated.
765
766In turn, the new_body_* allocators call S_new_body(), which invokes
767new_body_inline macro, which takes a lock, and takes a body off the
768linked list at PL_body_roots[sv_type], calling S_more_bodies() if
769necessary to refresh an empty list. Then the lock is released, and
770the body is returned.
771
772S_more_bodies calls get_arena(), and carves it up into an array of N
773bodies, which it strings into a linked list. It looks up arena-size
774and body-size from the body_details table described below, thus
775supporting the multiple body-types.
776
777If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
778the (new|del)_X*V macros are mapped directly to malloc/free.
779
780*/
781
782/*
783
784For each sv-type, struct body_details bodies_by_type[] carries
785parameters which control these aspects of SV handling:
786
787Arena_size determines whether arenas are used for this body type, and if
788so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
789zero, forcing individual mallocs and frees.
790
791Body_size determines how big a body is, and therefore how many fit into
792each arena. Offset carries the body-pointer adjustment needed for
793*_allocated body types, and is used in *_allocated macros.
794
795But its main purpose is to parameterize info needed in
796Perl_sv_upgrade(). The info here dramatically simplifies the function
797vs the implementation in 5.8.7, making it table-driven. All fields
798are used for this, except for arena_size.
799
800For the sv-types that have no bodies, arenas are not used, so those
801PL_body_roots[sv_type] are unused, and can be overloaded. In
802something of a special case, SVt_NULL is borrowed for HE arenas;
c6f8b1d0 803PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
d2a0f284 804bodies_by_type[SVt_NULL] slot is not used, as the table is not
c6f8b1d0 805available in hv.c.
d2a0f284 806
c6f8b1d0
JC
807PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
808they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
809just use the same allocation semantics. At first, PTEs were also
810overloaded to a non-body sv-type, but this yielded hard-to-find malloc
811bugs, so was simplified by claiming a new slot. This choice has no
812consequence at this time.
d2a0f284 813
29489e7c
DM
814*/
815
bd81e77b 816struct body_details {
0fb58b32 817 U8 body_size; /* Size to allocate */
10666ae3 818 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 819 U8 offset;
10666ae3
NC
820 unsigned int type : 4; /* We have space for a sanity check. */
821 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
822 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
823 unsigned int arena : 1; /* Allocated from an arena */
824 size_t arena_size; /* Size of arena to allocate */
bd81e77b 825};
29489e7c 826
bd81e77b
NC
827#define HADNV FALSE
828#define NONV TRUE
29489e7c 829
d2a0f284 830
bd81e77b
NC
831#ifdef PURIFY
832/* With -DPURFIY we allocate everything directly, and don't use arenas.
833 This seems a rather elegant way to simplify some of the code below. */
834#define HASARENA FALSE
835#else
836#define HASARENA TRUE
837#endif
838#define NOARENA FALSE
29489e7c 839
d2a0f284
JC
840/* Size the arenas to exactly fit a given number of bodies. A count
841 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
842 simplifying the default. If count > 0, the arena is sized to fit
843 only that many bodies, allowing arenas to be used for large, rare
844 bodies (XPVFM, XPVIO) without undue waste. The arena size is
845 limited by PERL_ARENA_SIZE, so we can safely oversize the
846 declarations.
847 */
95db5f15
MB
848#define FIT_ARENA0(body_size) \
849 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
850#define FIT_ARENAn(count,body_size) \
851 ( count * body_size <= PERL_ARENA_SIZE) \
852 ? count * body_size \
853 : FIT_ARENA0 (body_size)
854#define FIT_ARENA(count,body_size) \
855 count \
856 ? FIT_ARENAn (count, body_size) \
857 : FIT_ARENA0 (body_size)
d2a0f284 858
bd81e77b 859/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 860
bd81e77b
NC
861typedef struct {
862 STRLEN xpv_cur;
863 STRLEN xpv_len;
864} xpv_allocated;
29489e7c 865
bd81e77b 866to make its members accessible via a pointer to (say)
29489e7c 867
bd81e77b
NC
868struct xpv {
869 NV xnv_nv;
870 STRLEN xpv_cur;
871 STRLEN xpv_len;
872};
29489e7c 873
bd81e77b 874*/
29489e7c 875
bd81e77b
NC
876#define relative_STRUCT_OFFSET(longer, shorter, member) \
877 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 878
bd81e77b
NC
879/* Calculate the length to copy. Specifically work out the length less any
880 final padding the compiler needed to add. See the comment in sv_upgrade
881 for why copying the padding proved to be a bug. */
29489e7c 882
bd81e77b
NC
883#define copy_length(type, last_member) \
884 STRUCT_OFFSET(type, last_member) \
885 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 886
bd81e77b 887static const struct body_details bodies_by_type[] = {
10666ae3
NC
888 { sizeof(HE), 0, 0, SVt_NULL,
889 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284 890
1cb9cd50 891 /* The bind placeholder pretends to be an RV for now.
c6f8b1d0 892 Also it's marked as "can't upgrade" to stop anyone using it before it's
1cb9cd50
NC
893 implemented. */
894 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
895
d2a0f284
JC
896 /* IVs are in the head, so the allocation size is 0.
897 However, the slot is overloaded for PTEs. */
898 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
899 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 900 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
901 NOARENA /* IVS don't need an arena */,
902 /* But PTEs need to know the size of their arena */
903 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
904 },
905
bd81e77b 906 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 907 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
908 FIT_ARENA(0, sizeof(NV)) },
909
bd81e77b 910 /* 8 bytes on most ILP32 with IEEE doubles */
d2a0f284
JC
911 { sizeof(xpv_allocated),
912 copy_length(XPV, xpv_len)
913 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
914 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
10666ae3 915 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
d2a0f284 916
bd81e77b 917 /* 12 */
d2a0f284
JC
918 { sizeof(xpviv_allocated),
919 copy_length(XPVIV, xiv_u)
920 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
921 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
10666ae3 922 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
d2a0f284 923
bd81e77b 924 /* 20 */
10666ae3 925 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
926 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
927
bd81e77b 928 /* 28 */
10666ae3 929 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284 930 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
4df7f6af 931
288b8c02 932 /* something big */
08e44740
NC
933 { sizeof(struct regexp_allocated), sizeof(struct regexp_allocated),
934 + relative_STRUCT_OFFSET(struct regexp_allocated, regexp, xpv_cur),
935 SVt_REGEXP, FALSE, NONV, HASARENA,
936 FIT_ARENA(0, sizeof(struct regexp_allocated))
5c35adbb 937 },
4df7f6af 938
bd81e77b 939 /* 48 */
10666ae3 940 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
941 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
942
bd81e77b 943 /* 64 */
10666ae3 944 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
945 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
946
947 { sizeof(xpvav_allocated),
948 copy_length(XPVAV, xmg_stash)
949 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
950 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
9c59bb28 951 SVt_PVAV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
d2a0f284
JC
952
953 { sizeof(xpvhv_allocated),
954 copy_length(XPVHV, xmg_stash)
955 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
956 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
9c59bb28 957 SVt_PVHV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
d2a0f284 958
c84c4652 959 /* 56 */
4115f141 960 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
c84c4652 961 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
10666ae3 962 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
d2a0f284 963
4115f141 964 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
3038937b 965 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
10666ae3 966 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
d2a0f284
JC
967
968 /* XPVIO is 84 bytes, fits 48x */
167f2c4d
NC
969 { sizeof(xpvio_allocated), sizeof(xpvio_allocated),
970 + relative_STRUCT_OFFSET(xpvio_allocated, XPVIO, xpv_cur),
971 SVt_PVIO, TRUE, NONV, HASARENA, FIT_ARENA(24, sizeof(xpvio_allocated)) },
bd81e77b 972};
29489e7c 973
d2a0f284
JC
974#define new_body_type(sv_type) \
975 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 976
bd81e77b
NC
977#define del_body_type(p, sv_type) \
978 del_body(p, &PL_body_roots[sv_type])
29489e7c 979
29489e7c 980
bd81e77b 981#define new_body_allocated(sv_type) \
d2a0f284 982 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 983 - bodies_by_type[sv_type].offset)
29489e7c 984
bd81e77b
NC
985#define del_body_allocated(p, sv_type) \
986 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 987
29489e7c 988
bd81e77b
NC
989#define my_safemalloc(s) (void*)safemalloc(s)
990#define my_safecalloc(s) (void*)safecalloc(s, 1)
991#define my_safefree(p) safefree((char*)p)
29489e7c 992
bd81e77b 993#ifdef PURIFY
29489e7c 994
bd81e77b
NC
995#define new_XNV() my_safemalloc(sizeof(XPVNV))
996#define del_XNV(p) my_safefree(p)
29489e7c 997
bd81e77b
NC
998#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
999#define del_XPVNV(p) my_safefree(p)
29489e7c 1000
bd81e77b
NC
1001#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1002#define del_XPVAV(p) my_safefree(p)
29489e7c 1003
bd81e77b
NC
1004#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1005#define del_XPVHV(p) my_safefree(p)
29489e7c 1006
bd81e77b
NC
1007#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1008#define del_XPVMG(p) my_safefree(p)
29489e7c 1009
bd81e77b
NC
1010#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1011#define del_XPVGV(p) my_safefree(p)
29489e7c 1012
bd81e77b 1013#else /* !PURIFY */
29489e7c 1014
bd81e77b
NC
1015#define new_XNV() new_body_type(SVt_NV)
1016#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1017
bd81e77b
NC
1018#define new_XPVNV() new_body_type(SVt_PVNV)
1019#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1020
bd81e77b
NC
1021#define new_XPVAV() new_body_allocated(SVt_PVAV)
1022#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1023
bd81e77b
NC
1024#define new_XPVHV() new_body_allocated(SVt_PVHV)
1025#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1026
bd81e77b
NC
1027#define new_XPVMG() new_body_type(SVt_PVMG)
1028#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1029
bd81e77b
NC
1030#define new_XPVGV() new_body_type(SVt_PVGV)
1031#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1032
bd81e77b 1033#endif /* PURIFY */
93e68bfb 1034
bd81e77b 1035/* no arena for you! */
93e68bfb 1036
bd81e77b 1037#define new_NOARENA(details) \
d2a0f284 1038 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1039#define new_NOARENAZ(details) \
d2a0f284
JC
1040 my_safecalloc((details)->body_size + (details)->offset)
1041
1042STATIC void *
de37a194 1043S_more_bodies (pTHX_ const svtype sv_type)
d2a0f284
JC
1044{
1045 dVAR;
1046 void ** const root = &PL_body_roots[sv_type];
96a5add6 1047 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1048 const size_t body_size = bdp->body_size;
1049 char *start;
1050 const char *end;
d8fca402 1051 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
0b2d3faa 1052#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
23e9d66c
NC
1053 static bool done_sanity_check;
1054
0b2d3faa
JH
1055 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1056 * variables like done_sanity_check. */
10666ae3 1057 if (!done_sanity_check) {
ea471437 1058 unsigned int i = SVt_LAST;
10666ae3
NC
1059
1060 done_sanity_check = TRUE;
1061
1062 while (i--)
1063 assert (bodies_by_type[i].type == i);
1064 }
1065#endif
1066
23e9d66c
NC
1067 assert(bdp->arena_size);
1068
d8fca402 1069 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
d2a0f284 1070
d8fca402 1071 end = start + arena_size - 2 * body_size;
d2a0f284 1072
d2a0f284 1073 /* computed count doesnt reflect the 1st slot reservation */
d8fca402
NC
1074#if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1075 DEBUG_m(PerlIO_printf(Perl_debug_log,
1076 "arena %p end %p arena-size %d (from %d) type %d "
1077 "size %d ct %d\n",
1078 (void*)start, (void*)end, (int)arena_size,
1079 (int)bdp->arena_size, sv_type, (int)body_size,
1080 (int)arena_size / (int)body_size));
1081#else
d2a0f284
JC
1082 DEBUG_m(PerlIO_printf(Perl_debug_log,
1083 "arena %p end %p arena-size %d type %d size %d ct %d\n",
6c9570dc 1084 (void*)start, (void*)end,
0e84aef4
JH
1085 (int)bdp->arena_size, sv_type, (int)body_size,
1086 (int)bdp->arena_size / (int)body_size));
d8fca402 1087#endif
d2a0f284
JC
1088 *root = (void *)start;
1089
d8fca402 1090 while (start <= end) {
d2a0f284
JC
1091 char * const next = start + body_size;
1092 *(void**) start = (void *)next;
1093 start = next;
1094 }
1095 *(void **)start = 0;
1096
1097 return *root;
1098}
1099
1100/* grab a new thing from the free list, allocating more if necessary.
1101 The inline version is used for speed in hot routines, and the
1102 function using it serves the rest (unless PURIFY).
1103*/
1104#define new_body_inline(xpv, sv_type) \
1105 STMT_START { \
1106 void ** const r3wt = &PL_body_roots[sv_type]; \
11b79775
DD
1107 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1108 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284 1109 *(r3wt) = *(void**)(xpv); \
d2a0f284
JC
1110 } STMT_END
1111
1112#ifndef PURIFY
1113
1114STATIC void *
de37a194 1115S_new_body(pTHX_ const svtype sv_type)
d2a0f284
JC
1116{
1117 dVAR;
1118 void *xpv;
1119 new_body_inline(xpv, sv_type);
1120 return xpv;
1121}
1122
1123#endif
93e68bfb 1124
238b27b3
NC
1125static const struct body_details fake_rv =
1126 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1127
bd81e77b
NC
1128/*
1129=for apidoc sv_upgrade
93e68bfb 1130
bd81e77b
NC
1131Upgrade an SV to a more complex form. Generally adds a new body type to the
1132SV, then copies across as much information as possible from the old body.
1133You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1134
bd81e77b 1135=cut
93e68bfb 1136*/
93e68bfb 1137
bd81e77b 1138void
aad570aa 1139Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
cac9b346 1140{
97aff369 1141 dVAR;
bd81e77b
NC
1142 void* old_body;
1143 void* new_body;
42d0e0b7 1144 const svtype old_type = SvTYPE(sv);
d2a0f284 1145 const struct body_details *new_type_details;
238b27b3 1146 const struct body_details *old_type_details
bd81e77b 1147 = bodies_by_type + old_type;
4df7f6af 1148 SV *referant = NULL;
cac9b346 1149
7918f24d
NC
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1151
bd81e77b
NC
1152 if (new_type != SVt_PV && SvIsCOW(sv)) {
1153 sv_force_normal_flags(sv, 0);
1154 }
cac9b346 1155
bd81e77b
NC
1156 if (old_type == new_type)
1157 return;
cac9b346 1158
bd81e77b 1159 old_body = SvANY(sv);
de042e1d 1160
bd81e77b
NC
1161 /* Copying structures onto other structures that have been neatly zeroed
1162 has a subtle gotcha. Consider XPVMG
cac9b346 1163
bd81e77b
NC
1164 +------+------+------+------+------+-------+-------+
1165 | NV | CUR | LEN | IV | MAGIC | STASH |
1166 +------+------+------+------+------+-------+-------+
1167 0 4 8 12 16 20 24 28
645c22ef 1168
bd81e77b
NC
1169 where NVs are aligned to 8 bytes, so that sizeof that structure is
1170 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1171
bd81e77b
NC
1172 +------+------+------+------+------+-------+-------+------+
1173 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1174 +------+------+------+------+------+-------+-------+------+
1175 0 4 8 12 16 20 24 28 32
08742458 1176
bd81e77b 1177 so what happens if you allocate memory for this structure:
30f9da9e 1178
bd81e77b
NC
1179 +------+------+------+------+------+-------+-------+------+------+...
1180 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1181 +------+------+------+------+------+-------+-------+------+------+...
1182 0 4 8 12 16 20 24 28 32 36
bfc44f79 1183
bd81e77b
NC
1184 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1185 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1186 started out as zero once, but it's quite possible that it isn't. So now,
1187 rather than a nicely zeroed GP, you have it pointing somewhere random.
1188 Bugs ensue.
bfc44f79 1189
bd81e77b
NC
1190 (In fact, GP ends up pointing at a previous GP structure, because the
1191 principle cause of the padding in XPVMG getting garbage is a copy of
6c9e42f7
NC
1192 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1193 this happens to be moot because XPVGV has been re-ordered, with GP
1194 no longer after STASH)
30f9da9e 1195
bd81e77b
NC
1196 So we are careful and work out the size of used parts of all the
1197 structures. */
bfc44f79 1198
bd81e77b
NC
1199 switch (old_type) {
1200 case SVt_NULL:
1201 break;
1202 case SVt_IV:
4df7f6af
NC
1203 if (SvROK(sv)) {
1204 referant = SvRV(sv);
238b27b3
NC
1205 old_type_details = &fake_rv;
1206 if (new_type == SVt_NV)
1207 new_type = SVt_PVNV;
4df7f6af
NC
1208 } else {
1209 if (new_type < SVt_PVIV) {
1210 new_type = (new_type == SVt_NV)
1211 ? SVt_PVNV : SVt_PVIV;
1212 }
bd81e77b
NC
1213 }
1214 break;
1215 case SVt_NV:
1216 if (new_type < SVt_PVNV) {
1217 new_type = SVt_PVNV;
bd81e77b
NC
1218 }
1219 break;
bd81e77b
NC
1220 case SVt_PV:
1221 assert(new_type > SVt_PV);
1222 assert(SVt_IV < SVt_PV);
1223 assert(SVt_NV < SVt_PV);
1224 break;
1225 case SVt_PVIV:
1226 break;
1227 case SVt_PVNV:
1228 break;
1229 case SVt_PVMG:
1230 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1231 there's no way that it can be safely upgraded, because perl.c
1232 expects to Safefree(SvANY(PL_mess_sv)) */
1233 assert(sv != PL_mess_sv);
1234 /* This flag bit is used to mean other things in other scalar types.
1235 Given that it only has meaning inside the pad, it shouldn't be set
1236 on anything that can get upgraded. */
00b1698f 1237 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1238 break;
1239 default:
1240 if (old_type_details->cant_upgrade)
c81225bc
NC
1241 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1242 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1243 }
3376de98
NC
1244
1245 if (old_type > new_type)
1246 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1247 (int)old_type, (int)new_type);
1248
2fa1109b 1249 new_type_details = bodies_by_type + new_type;
645c22ef 1250
bd81e77b
NC
1251 SvFLAGS(sv) &= ~SVTYPEMASK;
1252 SvFLAGS(sv) |= new_type;
932e9ff9 1253
ab4416c0
NC
1254 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1255 the return statements above will have triggered. */
1256 assert (new_type != SVt_NULL);
bd81e77b 1257 switch (new_type) {
bd81e77b
NC
1258 case SVt_IV:
1259 assert(old_type == SVt_NULL);
1260 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1261 SvIV_set(sv, 0);
1262 return;
1263 case SVt_NV:
1264 assert(old_type == SVt_NULL);
1265 SvANY(sv) = new_XNV();
1266 SvNV_set(sv, 0);
1267 return;
bd81e77b 1268 case SVt_PVHV:
bd81e77b 1269 case SVt_PVAV:
d2a0f284 1270 assert(new_type_details->body_size);
c1ae03ae
NC
1271
1272#ifndef PURIFY
1273 assert(new_type_details->arena);
d2a0f284 1274 assert(new_type_details->arena_size);
c1ae03ae 1275 /* This points to the start of the allocated area. */
d2a0f284
JC
1276 new_body_inline(new_body, new_type);
1277 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1278 new_body = ((char *)new_body) - new_type_details->offset;
1279#else
1280 /* We always allocated the full length item with PURIFY. To do this
1281 we fake things so that arena is false for all 16 types.. */
1282 new_body = new_NOARENAZ(new_type_details);
1283#endif
1284 SvANY(sv) = new_body;
1285 if (new_type == SVt_PVAV) {
1286 AvMAX(sv) = -1;
1287 AvFILLp(sv) = -1;
1288 AvREAL_only(sv);
64484faa 1289 if (old_type_details->body_size) {
ac572bf4
NC
1290 AvALLOC(sv) = 0;
1291 } else {
1292 /* It will have been zeroed when the new body was allocated.
1293 Lets not write to it, in case it confuses a write-back
1294 cache. */
1295 }
78ac7dd9
NC
1296 } else {
1297 assert(!SvOK(sv));
1298 SvOK_off(sv);
1299#ifndef NODEFAULT_SHAREKEYS
1300 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1301#endif
1302 HvMAX(sv) = 7; /* (start with 8 buckets) */
64484faa 1303 if (old_type_details->body_size) {
78ac7dd9
NC
1304 HvFILL(sv) = 0;
1305 } else {
1306 /* It will have been zeroed when the new body was allocated.
1307 Lets not write to it, in case it confuses a write-back
1308 cache. */
1309 }
c1ae03ae 1310 }
aeb18a1e 1311
bd81e77b
NC
1312 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1313 The target created by newSVrv also is, and it can have magic.
1314 However, it never has SvPVX set.
1315 */
4df7f6af
NC
1316 if (old_type == SVt_IV) {
1317 assert(!SvROK(sv));
1318 } else if (old_type >= SVt_PV) {
bd81e77b
NC
1319 assert(SvPVX_const(sv) == 0);
1320 }
aeb18a1e 1321
bd81e77b 1322 if (old_type >= SVt_PVMG) {
e736a858 1323 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1324 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1325 } else {
1326 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1327 }
1328 break;
93e68bfb 1329
93e68bfb 1330
bd81e77b
NC
1331 case SVt_PVIV:
1332 /* XXX Is this still needed? Was it ever needed? Surely as there is
1333 no route from NV to PVIV, NOK can never be true */
1334 assert(!SvNOKp(sv));
1335 assert(!SvNOK(sv));
1336 case SVt_PVIO:
1337 case SVt_PVFM:
bd81e77b
NC
1338 case SVt_PVGV:
1339 case SVt_PVCV:
1340 case SVt_PVLV:
5c35adbb 1341 case SVt_REGEXP:
bd81e77b
NC
1342 case SVt_PVMG:
1343 case SVt_PVNV:
1344 case SVt_PV:
93e68bfb 1345
d2a0f284 1346 assert(new_type_details->body_size);
bd81e77b
NC
1347 /* We always allocated the full length item with PURIFY. To do this
1348 we fake things so that arena is false for all 16 types.. */
1349 if(new_type_details->arena) {
1350 /* This points to the start of the allocated area. */
d2a0f284
JC
1351 new_body_inline(new_body, new_type);
1352 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1353 new_body = ((char *)new_body) - new_type_details->offset;
1354 } else {
1355 new_body = new_NOARENAZ(new_type_details);
1356 }
1357 SvANY(sv) = new_body;
5e2fc214 1358
bd81e77b 1359 if (old_type_details->copy) {
f9ba3d20
NC
1360 /* There is now the potential for an upgrade from something without
1361 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1362 int offset = old_type_details->offset;
1363 int length = old_type_details->copy;
1364
1365 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1366 const int difference
f9ba3d20
NC
1367 = new_type_details->offset - old_type_details->offset;
1368 offset += difference;
1369 length -= difference;
1370 }
1371 assert (length >= 0);
1372
1373 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1374 char);
bd81e77b
NC
1375 }
1376
1377#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1378 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1379 * correct 0.0 for us. Otherwise, if the old body didn't have an
1380 * NV slot, but the new one does, then we need to initialise the
1381 * freshly created NV slot with whatever the correct bit pattern is
1382 * for 0.0 */
e22a937e
NC
1383 if (old_type_details->zero_nv && !new_type_details->zero_nv
1384 && !isGV_with_GP(sv))
bd81e77b 1385 SvNV_set(sv, 0);
82048762 1386#endif
5e2fc214 1387
bd81e77b 1388 if (new_type == SVt_PVIO)
f2524eef 1389 IoPAGE_LEN(sv) = 60;
4df7f6af
NC
1390 if (old_type < SVt_PV) {
1391 /* referant will be NULL unless the old type was SVt_IV emulating
1392 SVt_RV */
1393 sv->sv_u.svu_rv = referant;
1394 }
bd81e77b
NC
1395 break;
1396 default:
afd78fd5
JH
1397 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1398 (unsigned long)new_type);
bd81e77b 1399 }
73171d91 1400
d2a0f284
JC
1401 if (old_type_details->arena) {
1402 /* If there was an old body, then we need to free it.
1403 Note that there is an assumption that all bodies of types that
1404 can be upgraded came from arenas. Only the more complex non-
1405 upgradable types are allowed to be directly malloc()ed. */
bd81e77b
NC
1406#ifdef PURIFY
1407 my_safefree(old_body);
1408#else
1409 del_body((void*)((char*)old_body + old_type_details->offset),
1410 &PL_body_roots[old_type]);
1411#endif
1412 }
1413}
73171d91 1414
bd81e77b
NC
1415/*
1416=for apidoc sv_backoff
73171d91 1417
bd81e77b
NC
1418Remove any string offset. You should normally use the C<SvOOK_off> macro
1419wrapper instead.
73171d91 1420
bd81e77b 1421=cut
73171d91
NC
1422*/
1423
bd81e77b 1424int
aad570aa 1425Perl_sv_backoff(pTHX_ register SV *const sv)
bd81e77b 1426{
69240efd 1427 STRLEN delta;
7a4bba22 1428 const char * const s = SvPVX_const(sv);
7918f24d
NC
1429
1430 PERL_ARGS_ASSERT_SV_BACKOFF;
96a5add6 1431 PERL_UNUSED_CONTEXT;
7918f24d 1432
bd81e77b
NC
1433 assert(SvOOK(sv));
1434 assert(SvTYPE(sv) != SVt_PVHV);
1435 assert(SvTYPE(sv) != SVt_PVAV);
7a4bba22 1436
69240efd
NC
1437 SvOOK_offset(sv, delta);
1438
7a4bba22
NC
1439 SvLEN_set(sv, SvLEN(sv) + delta);
1440 SvPV_set(sv, SvPVX(sv) - delta);
1441 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
bd81e77b
NC
1442 SvFLAGS(sv) &= ~SVf_OOK;
1443 return 0;
1444}
73171d91 1445
bd81e77b
NC
1446/*
1447=for apidoc sv_grow
73171d91 1448
bd81e77b
NC
1449Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1450upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1451Use the C<SvGROW> wrapper instead.
93e68bfb 1452
bd81e77b
NC
1453=cut
1454*/
93e68bfb 1455
bd81e77b 1456char *
aad570aa 1457Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
bd81e77b
NC
1458{
1459 register char *s;
93e68bfb 1460
7918f24d
NC
1461 PERL_ARGS_ASSERT_SV_GROW;
1462
5db06880
NC
1463 if (PL_madskills && newlen >= 0x100000) {
1464 PerlIO_printf(Perl_debug_log,
1465 "Allocation too large: %"UVxf"\n", (UV)newlen);
1466 }
bd81e77b
NC
1467#ifdef HAS_64K_LIMIT
1468 if (newlen >= 0x10000) {
1469 PerlIO_printf(Perl_debug_log,
1470 "Allocation too large: %"UVxf"\n", (UV)newlen);
1471 my_exit(1);
1472 }
1473#endif /* HAS_64K_LIMIT */
1474 if (SvROK(sv))
1475 sv_unref(sv);
1476 if (SvTYPE(sv) < SVt_PV) {
1477 sv_upgrade(sv, SVt_PV);
1478 s = SvPVX_mutable(sv);
1479 }
1480 else if (SvOOK(sv)) { /* pv is offset? */
1481 sv_backoff(sv);
1482 s = SvPVX_mutable(sv);
1483 if (newlen > SvLEN(sv))
1484 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1485#ifdef HAS_64K_LIMIT
1486 if (newlen >= 0x10000)
1487 newlen = 0xFFFF;
1488#endif
1489 }
1490 else
1491 s = SvPVX_mutable(sv);
aeb18a1e 1492
bd81e77b 1493 if (newlen > SvLEN(sv)) { /* need more room? */
aedff202 1494#ifndef Perl_safesysmalloc_size
bd81e77b 1495 newlen = PERL_STRLEN_ROUNDUP(newlen);
bd81e77b 1496#endif
98653f18 1497 if (SvLEN(sv) && s) {
10edeb5d 1498 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1499 }
1500 else {
10edeb5d 1501 s = (char*)safemalloc(newlen);
bd81e77b
NC
1502 if (SvPVX_const(sv) && SvCUR(sv)) {
1503 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1504 }
1505 }
1506 SvPV_set(sv, s);
ca7c1a29 1507#ifdef Perl_safesysmalloc_size
98653f18
NC
1508 /* Do this here, do it once, do it right, and then we will never get
1509 called back into sv_grow() unless there really is some growing
1510 needed. */
ca7c1a29 1511 SvLEN_set(sv, Perl_safesysmalloc_size(s));
98653f18 1512#else
bd81e77b 1513 SvLEN_set(sv, newlen);
98653f18 1514#endif
bd81e77b
NC
1515 }
1516 return s;
1517}
aeb18a1e 1518
bd81e77b
NC
1519/*
1520=for apidoc sv_setiv
932e9ff9 1521
bd81e77b
NC
1522Copies an integer into the given SV, upgrading first if necessary.
1523Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1524
bd81e77b
NC
1525=cut
1526*/
463ee0b2 1527
bd81e77b 1528void
aad570aa 1529Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
bd81e77b 1530{
97aff369 1531 dVAR;
7918f24d
NC
1532
1533 PERL_ARGS_ASSERT_SV_SETIV;
1534
bd81e77b
NC
1535 SV_CHECK_THINKFIRST_COW_DROP(sv);
1536 switch (SvTYPE(sv)) {
1537 case SVt_NULL:
bd81e77b 1538 case SVt_NV:
3376de98 1539 sv_upgrade(sv, SVt_IV);
bd81e77b 1540 break;
bd81e77b
NC
1541 case SVt_PV:
1542 sv_upgrade(sv, SVt_PVIV);
1543 break;
463ee0b2 1544
bd81e77b 1545 case SVt_PVGV:
6e592b3a
BM
1546 if (!isGV_with_GP(sv))
1547 break;
bd81e77b
NC
1548 case SVt_PVAV:
1549 case SVt_PVHV:
1550 case SVt_PVCV:
1551 case SVt_PVFM:
1552 case SVt_PVIO:
1553 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1554 OP_DESC(PL_op));
42d0e0b7 1555 default: NOOP;
bd81e77b
NC
1556 }
1557 (void)SvIOK_only(sv); /* validate number */
1558 SvIV_set(sv, i);
1559 SvTAINT(sv);
1560}
932e9ff9 1561
bd81e77b
NC
1562/*
1563=for apidoc sv_setiv_mg
d33b2eba 1564
bd81e77b 1565Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1566
bd81e77b
NC
1567=cut
1568*/
d33b2eba 1569
bd81e77b 1570void
aad570aa 1571Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
bd81e77b 1572{
7918f24d
NC
1573 PERL_ARGS_ASSERT_SV_SETIV_MG;
1574
bd81e77b
NC
1575 sv_setiv(sv,i);
1576 SvSETMAGIC(sv);
1577}
727879eb 1578
bd81e77b
NC
1579/*
1580=for apidoc sv_setuv
d33b2eba 1581
bd81e77b
NC
1582Copies an unsigned integer into the given SV, upgrading first if necessary.
1583Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1584
bd81e77b
NC
1585=cut
1586*/
d33b2eba 1587
bd81e77b 1588void
aad570aa 1589Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
bd81e77b 1590{
7918f24d
NC
1591 PERL_ARGS_ASSERT_SV_SETUV;
1592
bd81e77b
NC
1593 /* With these two if statements:
1594 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1595
bd81e77b
NC
1596 without
1597 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1598
bd81e77b
NC
1599 If you wish to remove them, please benchmark to see what the effect is
1600 */
1601 if (u <= (UV)IV_MAX) {
1602 sv_setiv(sv, (IV)u);
1603 return;
1604 }
1605 sv_setiv(sv, 0);
1606 SvIsUV_on(sv);
1607 SvUV_set(sv, u);
1608}
d33b2eba 1609
bd81e77b
NC
1610/*
1611=for apidoc sv_setuv_mg
727879eb 1612
bd81e77b 1613Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1614
bd81e77b
NC
1615=cut
1616*/
5e2fc214 1617
bd81e77b 1618void
aad570aa 1619Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
bd81e77b 1620{
7918f24d
NC
1621 PERL_ARGS_ASSERT_SV_SETUV_MG;
1622
bd81e77b
NC
1623 sv_setuv(sv,u);
1624 SvSETMAGIC(sv);
1625}
5e2fc214 1626
954c1994 1627/*
bd81e77b 1628=for apidoc sv_setnv
954c1994 1629
bd81e77b
NC
1630Copies a double into the given SV, upgrading first if necessary.
1631Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1632
1633=cut
1634*/
1635
63f97190 1636void
aad570aa 1637Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
79072805 1638{
97aff369 1639 dVAR;
7918f24d
NC
1640
1641 PERL_ARGS_ASSERT_SV_SETNV;
1642
bd81e77b
NC
1643 SV_CHECK_THINKFIRST_COW_DROP(sv);
1644 switch (SvTYPE(sv)) {
79072805 1645 case SVt_NULL:
79072805 1646 case SVt_IV:
bd81e77b 1647 sv_upgrade(sv, SVt_NV);
79072805
LW
1648 break;
1649 case SVt_PV:
79072805 1650 case SVt_PVIV:
bd81e77b 1651 sv_upgrade(sv, SVt_PVNV);
79072805 1652 break;
bd4b1eb5 1653
bd4b1eb5 1654 case SVt_PVGV:
6e592b3a
BM
1655 if (!isGV_with_GP(sv))
1656 break;
bd81e77b
NC
1657 case SVt_PVAV:
1658 case SVt_PVHV:
79072805 1659 case SVt_PVCV:
bd81e77b
NC
1660 case SVt_PVFM:
1661 case SVt_PVIO:
1662 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1663 OP_NAME(PL_op));
42d0e0b7 1664 default: NOOP;
2068cd4d 1665 }
bd81e77b
NC
1666 SvNV_set(sv, num);
1667 (void)SvNOK_only(sv); /* validate number */
1668 SvTAINT(sv);
79072805
LW
1669}
1670
645c22ef 1671/*
bd81e77b 1672=for apidoc sv_setnv_mg
645c22ef 1673
bd81e77b 1674Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1675
1676=cut
1677*/
1678
bd81e77b 1679void
aad570aa 1680Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
79072805 1681{
7918f24d
NC
1682 PERL_ARGS_ASSERT_SV_SETNV_MG;
1683
bd81e77b
NC
1684 sv_setnv(sv,num);
1685 SvSETMAGIC(sv);
79072805
LW
1686}
1687
bd81e77b
NC
1688/* Print an "isn't numeric" warning, using a cleaned-up,
1689 * printable version of the offending string
1690 */
954c1994 1691
bd81e77b 1692STATIC void
aad570aa 1693S_not_a_number(pTHX_ SV *const sv)
79072805 1694{
97aff369 1695 dVAR;
bd81e77b
NC
1696 SV *dsv;
1697 char tmpbuf[64];
1698 const char *pv;
94463019 1699
7918f24d
NC
1700 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1701
94463019 1702 if (DO_UTF8(sv)) {
84bafc02 1703 dsv = newSVpvs_flags("", SVs_TEMP);
94463019
JH
1704 pv = sv_uni_display(dsv, sv, 10, 0);
1705 } else {
1706 char *d = tmpbuf;
551405c4 1707 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1708 /* each *s can expand to 4 chars + "...\0",
1709 i.e. need room for 8 chars */
ecdeb87c 1710
00b6aa41
AL
1711 const char *s = SvPVX_const(sv);
1712 const char * const end = s + SvCUR(sv);
1713 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1714 int ch = *s & 0xFF;
1715 if (ch & 128 && !isPRINT_LC(ch)) {
1716 *d++ = 'M';
1717 *d++ = '-';
1718 ch &= 127;
1719 }
1720 if (ch == '\n') {
1721 *d++ = '\\';
1722 *d++ = 'n';
1723 }
1724 else if (ch == '\r') {
1725 *d++ = '\\';
1726 *d++ = 'r';
1727 }
1728 else if (ch == '\f') {
1729 *d++ = '\\';
1730 *d++ = 'f';
1731 }
1732 else if (ch == '\\') {
1733 *d++ = '\\';
1734 *d++ = '\\';
1735 }
1736 else if (ch == '\0') {
1737 *d++ = '\\';
1738 *d++ = '0';
1739 }
1740 else if (isPRINT_LC(ch))
1741 *d++ = ch;
1742 else {
1743 *d++ = '^';
1744 *d++ = toCTRL(ch);
1745 }
1746 }
1747 if (s < end) {
1748 *d++ = '.';
1749 *d++ = '.';
1750 *d++ = '.';
1751 }
1752 *d = '\0';
1753 pv = tmpbuf;
a0d0e21e 1754 }
a0d0e21e 1755
533c011a 1756 if (PL_op)
9014280d 1757 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1758 "Argument \"%s\" isn't numeric in %s", pv,
1759 OP_DESC(PL_op));
a0d0e21e 1760 else
9014280d 1761 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1762 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1763}
1764
c2988b20
NC
1765/*
1766=for apidoc looks_like_number
1767
645c22ef
DM
1768Test if the content of an SV looks like a number (or is a number).
1769C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1770non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1771
1772=cut
1773*/
1774
1775I32
aad570aa 1776Perl_looks_like_number(pTHX_ SV *const sv)
c2988b20 1777{
a3b680e6 1778 register const char *sbegin;
c2988b20
NC
1779 STRLEN len;
1780
7918f24d
NC
1781 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1782
c2988b20 1783 if (SvPOK(sv)) {
3f7c398e 1784 sbegin = SvPVX_const(sv);
c2988b20
NC
1785 len = SvCUR(sv);
1786 }
1787 else if (SvPOKp(sv))
83003860 1788 sbegin = SvPV_const(sv, len);
c2988b20 1789 else
e0ab1c0e 1790 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1791 return grok_number(sbegin, len, NULL);
1792}
25da4f38 1793
19f6321d
NC
1794STATIC bool
1795S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1796{
1797 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1798 SV *const buffer = sv_newmortal();
1799
7918f24d
NC
1800 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1801
180488f8
NC
1802 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1803 is on. */
1804 SvFAKE_off(gv);
1805 gv_efullname3(buffer, gv, "*");
1806 SvFLAGS(gv) |= wasfake;
1807
675c862f
AL
1808 /* We know that all GVs stringify to something that is not-a-number,
1809 so no need to test that. */
1810 if (ckWARN(WARN_NUMERIC))
1811 not_a_number(buffer);
1812 /* We just want something true to return, so that S_sv_2iuv_common
1813 can tail call us and return true. */
19f6321d 1814 return TRUE;
675c862f
AL
1815}
1816
1817STATIC char *
19f6321d 1818S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
675c862f
AL
1819{
1820 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1821 SV *const buffer = sv_newmortal();
1822
7918f24d
NC
1823 PERL_ARGS_ASSERT_GLOB_2PV;
1824
675c862f
AL
1825 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1826 is on. */
1827 SvFAKE_off(gv);
1828 gv_efullname3(buffer, gv, "*");
1829 SvFLAGS(gv) |= wasfake;
1830
1831 assert(SvPOK(buffer));
a6d61a6c
NC
1832 if (len) {
1833 *len = SvCUR(buffer);
1834 }
675c862f 1835 return SvPVX(buffer);
180488f8
NC
1836}
1837
25da4f38
IZ
1838/* Actually, ISO C leaves conversion of UV to IV undefined, but
1839 until proven guilty, assume that things are not that bad... */
1840
645c22ef
DM
1841/*
1842 NV_PRESERVES_UV:
1843
1844 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1845 an IV (an assumption perl has been based on to date) it becomes necessary
1846 to remove the assumption that the NV always carries enough precision to
1847 recreate the IV whenever needed, and that the NV is the canonical form.
1848 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1849 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1850 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1851 1) to distinguish between IV/UV/NV slots that have cached a valid
1852 conversion where precision was lost and IV/UV/NV slots that have a
1853 valid conversion which has lost no precision
645c22ef 1854 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1855 would lose precision, the precise conversion (or differently
1856 imprecise conversion) is also performed and cached, to prevent
1857 requests for different numeric formats on the same SV causing
1858 lossy conversion chains. (lossless conversion chains are perfectly
1859 acceptable (still))
1860
1861
1862 flags are used:
1863 SvIOKp is true if the IV slot contains a valid value
1864 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1865 SvNOKp is true if the NV slot contains a valid value
1866 SvNOK is true only if the NV value is accurate
1867
1868 so
645c22ef 1869 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1870 IV(or UV) would lose accuracy over a direct conversion from PV to
1871 IV(or UV). If it would, cache both conversions, return NV, but mark
1872 SV as IOK NOKp (ie not NOK).
1873
645c22ef 1874 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1875 NV would lose accuracy over a direct conversion from PV to NV. If it
1876 would, cache both conversions, flag similarly.
1877
1878 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1879 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1880 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1881 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1882 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1883
645c22ef
DM
1884 The benefit of this is that operations such as pp_add know that if
1885 SvIOK is true for both left and right operands, then integer addition
1886 can be used instead of floating point (for cases where the result won't
1887 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1888 loss of precision compared with integer addition.
1889
1890 * making IV and NV equal status should make maths accurate on 64 bit
1891 platforms
1892 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1893 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1894 looking for SvIOK and checking for overflow will not outweigh the
1895 fp to integer speedup)
1896 * will slow down integer operations (callers of SvIV) on "inaccurate"
1897 values, as the change from SvIOK to SvIOKp will cause a call into
1898 sv_2iv each time rather than a macro access direct to the IV slot
1899 * should speed up number->string conversion on integers as IV is
645c22ef 1900 favoured when IV and NV are equally accurate
28e5dec8
JH
1901
1902 ####################################################################
645c22ef
DM
1903 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1904 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1905 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1906 ####################################################################
1907
645c22ef 1908 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1909 performance ratio.
1910*/
1911
1912#ifndef NV_PRESERVES_UV
645c22ef
DM
1913# define IS_NUMBER_UNDERFLOW_IV 1
1914# define IS_NUMBER_UNDERFLOW_UV 2
1915# define IS_NUMBER_IV_AND_UV 2
1916# define IS_NUMBER_OVERFLOW_IV 4
1917# define IS_NUMBER_OVERFLOW_UV 5
1918
1919/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1920
1921/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1922STATIC int
5de3775c 1923S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
47031da6
NC
1924# ifdef DEBUGGING
1925 , I32 numtype
1926# endif
1927 )
28e5dec8 1928{
97aff369 1929 dVAR;
7918f24d
NC
1930
1931 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1932
3f7c398e 1933 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
1934 if (SvNVX(sv) < (NV)IV_MIN) {
1935 (void)SvIOKp_on(sv);
1936 (void)SvNOK_on(sv);
45977657 1937 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1938 return IS_NUMBER_UNDERFLOW_IV;
1939 }
1940 if (SvNVX(sv) > (NV)UV_MAX) {
1941 (void)SvIOKp_on(sv);
1942 (void)SvNOK_on(sv);
1943 SvIsUV_on(sv);
607fa7f2 1944 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1945 return IS_NUMBER_OVERFLOW_UV;
1946 }
c2988b20
NC
1947 (void)SvIOKp_on(sv);
1948 (void)SvNOK_on(sv);
1949 /* Can't use strtol etc to convert this string. (See truth table in
1950 sv_2iv */
1951 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1952 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1953 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1954 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1955 } else {
1956 /* Integer is imprecise. NOK, IOKp */
1957 }
1958 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1959 }
1960 SvIsUV_on(sv);
607fa7f2 1961 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1962 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1963 if (SvUVX(sv) == UV_MAX) {
1964 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1965 possibly be preserved by NV. Hence, it must be overflow.
1966 NOK, IOKp */
1967 return IS_NUMBER_OVERFLOW_UV;
1968 }
1969 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1970 } else {
1971 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1972 }
c2988b20 1973 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1974}
645c22ef
DM
1975#endif /* !NV_PRESERVES_UV*/
1976
af359546 1977STATIC bool
7918f24d
NC
1978S_sv_2iuv_common(pTHX_ SV *const sv)
1979{
97aff369 1980 dVAR;
7918f24d
NC
1981
1982 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1983
af359546 1984 if (SvNOKp(sv)) {
28e5dec8
JH
1985 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1986 * without also getting a cached IV/UV from it at the same time
1987 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1988 * IV or UV at same time to avoid this. */
1989 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1990
1991 if (SvTYPE(sv) == SVt_NV)
1992 sv_upgrade(sv, SVt_PVNV);
1993
28e5dec8
JH
1994 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1995 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1996 certainly cast into the IV range at IV_MAX, whereas the correct
1997 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1998 cases go to UV */
cab190d4
JD
1999#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2000 if (Perl_isnan(SvNVX(sv))) {
2001 SvUV_set(sv, 0);
2002 SvIsUV_on(sv);
fdbe6d7c 2003 return FALSE;
cab190d4 2004 }
cab190d4 2005#endif
28e5dec8 2006 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 2007 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
2008 if (SvNVX(sv) == (NV) SvIVX(sv)
2009#ifndef NV_PRESERVES_UV
2010 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2011 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2012 /* Don't flag it as "accurately an integer" if the number
2013 came from a (by definition imprecise) NV operation, and
2014 we're outside the range of NV integer precision */
2015#endif
2016 ) {
a43d94f2
NC
2017 if (SvNOK(sv))
2018 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2019 else {
2020 /* scalar has trailing garbage, eg "42a" */
2021 }
28e5dec8 2022 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2023 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
2024 PTR2UV(sv),
2025 SvNVX(sv),
2026 SvIVX(sv)));
2027
2028 } else {
2029 /* IV not precise. No need to convert from PV, as NV
2030 conversion would already have cached IV if it detected
2031 that PV->IV would be better than PV->NV->IV
2032 flags already correct - don't set public IOK. */
2033 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2034 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
2035 PTR2UV(sv),
2036 SvNVX(sv),
2037 SvIVX(sv)));
2038 }
2039 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2040 but the cast (NV)IV_MIN rounds to a the value less (more
2041 negative) than IV_MIN which happens to be equal to SvNVX ??
2042 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2043 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2044 (NV)UVX == NVX are both true, but the values differ. :-(
2045 Hopefully for 2s complement IV_MIN is something like
2046 0x8000000000000000 which will be exact. NWC */
d460ef45 2047 }
25da4f38 2048 else {
607fa7f2 2049 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
2050 if (
2051 (SvNVX(sv) == (NV) SvUVX(sv))
2052#ifndef NV_PRESERVES_UV
2053 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2054 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2055 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2056 /* Don't flag it as "accurately an integer" if the number
2057 came from a (by definition imprecise) NV operation, and
2058 we're outside the range of NV integer precision */
2059#endif
a43d94f2 2060 && SvNOK(sv)
28e5dec8
JH
2061 )
2062 SvIOK_on(sv);
25da4f38 2063 SvIsUV_on(sv);
1c846c1f 2064 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 2065 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 2066 PTR2UV(sv),
57def98f
JH
2067 SvUVX(sv),
2068 SvUVX(sv)));
25da4f38 2069 }
748a9306
LW
2070 }
2071 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2072 UV value;
504618e9 2073 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 2074 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 2075 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
2076 the same as the direct translation of the initial string
2077 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2078 be careful to ensure that the value with the .456 is around if the
2079 NV value is requested in the future).
1c846c1f 2080
af359546 2081 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 2082 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 2083 cache the NV if we are sure it's not needed.
25da4f38 2084 */
16b7a9a4 2085
c2988b20
NC
2086 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2087 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2088 == IS_NUMBER_IN_UV) {
5e045b90 2089 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
2090 if (SvTYPE(sv) < SVt_PVIV)
2091 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 2092 (void)SvIOK_on(sv);
c2988b20
NC
2093 } else if (SvTYPE(sv) < SVt_PVNV)
2094 sv_upgrade(sv, SVt_PVNV);
28e5dec8 2095
f2524eef 2096 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2097 we aren't going to call atof() below. If NVs don't preserve UVs
2098 then the value returned may have more precision than atof() will
2099 return, even though value isn't perfectly accurate. */
2100 if ((numtype & (IS_NUMBER_IN_UV
2101#ifdef NV_PRESERVES_UV
2102 | IS_NUMBER_NOT_INT
2103#endif
2104 )) == IS_NUMBER_IN_UV) {
2105 /* This won't turn off the public IOK flag if it was set above */
2106 (void)SvIOKp_on(sv);
2107
2108 if (!(numtype & IS_NUMBER_NEG)) {
2109 /* positive */;
2110 if (value <= (UV)IV_MAX) {
45977657 2111 SvIV_set(sv, (IV)value);
c2988b20 2112 } else {
af359546 2113 /* it didn't overflow, and it was positive. */
607fa7f2 2114 SvUV_set(sv, value);
c2988b20
NC
2115 SvIsUV_on(sv);
2116 }
2117 } else {
2118 /* 2s complement assumption */
2119 if (value <= (UV)IV_MIN) {
45977657 2120 SvIV_set(sv, -(IV)value);
c2988b20
NC
2121 } else {
2122 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2123 I'm assuming it will be rare. */
c2988b20
NC
2124 if (SvTYPE(sv) < SVt_PVNV)
2125 sv_upgrade(sv, SVt_PVNV);
2126 SvNOK_on(sv);
2127 SvIOK_off(sv);
2128 SvIOKp_on(sv);
9d6ce603 2129 SvNV_set(sv, -(NV)value);
45977657 2130 SvIV_set(sv, IV_MIN);
c2988b20
NC
2131 }
2132 }
2133 }
2134 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2135 will be in the previous block to set the IV slot, and the next
2136 block to set the NV slot. So no else here. */
2137
2138 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2139 != IS_NUMBER_IN_UV) {
2140 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2141 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2142
c2988b20
NC
2143 if (! numtype && ckWARN(WARN_NUMERIC))
2144 not_a_number(sv);
28e5dec8 2145
65202027 2146#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2147 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2148 PTR2UV(sv), SvNVX(sv)));
65202027 2149#else
1779d84d 2150 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2151 PTR2UV(sv), SvNVX(sv)));
65202027 2152#endif
28e5dec8 2153
28e5dec8 2154#ifdef NV_PRESERVES_UV
af359546
NC
2155 (void)SvIOKp_on(sv);
2156 (void)SvNOK_on(sv);
2157 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2158 SvIV_set(sv, I_V(SvNVX(sv)));
2159 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2160 SvIOK_on(sv);
2161 } else {
6f207bd3 2162 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2163 }
2164 /* UV will not work better than IV */
2165 } else {
2166 if (SvNVX(sv) > (NV)UV_MAX) {
2167 SvIsUV_on(sv);
2168 /* Integer is inaccurate. NOK, IOKp, is UV */
2169 SvUV_set(sv, UV_MAX);
af359546
NC
2170 } else {
2171 SvUV_set(sv, U_V(SvNVX(sv)));
2172 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2173 NV preservse UV so can do correct comparison. */
2174 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2175 SvIOK_on(sv);
af359546 2176 } else {
6f207bd3 2177 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2178 }
2179 }
4b0c9573 2180 SvIsUV_on(sv);
af359546 2181 }
28e5dec8 2182#else /* NV_PRESERVES_UV */
c2988b20
NC
2183 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2184 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2185 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2186 grok_number above. The NV slot has just been set using
2187 Atof. */
560b0c46 2188 SvNOK_on(sv);
c2988b20
NC
2189 assert (SvIOKp(sv));
2190 } else {
2191 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2192 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2193 /* Small enough to preserve all bits. */
2194 (void)SvIOKp_on(sv);
2195 SvNOK_on(sv);
45977657 2196 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2197 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2198 SvIOK_on(sv);
2199 /* Assumption: first non-preserved integer is < IV_MAX,
2200 this NV is in the preserved range, therefore: */
2201 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2202 < (UV)IV_MAX)) {
32fdb065 2203 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
2204 }
2205 } else {
2206 /* IN_UV NOT_INT
2207 0 0 already failed to read UV.
2208 0 1 already failed to read UV.
2209 1 0 you won't get here in this case. IV/UV
2210 slot set, public IOK, Atof() unneeded.
2211 1 1 already read UV.
2212 so there's no point in sv_2iuv_non_preserve() attempting
2213 to use atol, strtol, strtoul etc. */
47031da6 2214# ifdef DEBUGGING
40a17c4c 2215 sv_2iuv_non_preserve (sv, numtype);
47031da6
NC
2216# else
2217 sv_2iuv_non_preserve (sv);
2218# endif
c2988b20
NC
2219 }
2220 }
28e5dec8 2221#endif /* NV_PRESERVES_UV */
a43d94f2
NC
2222 /* It might be more code efficient to go through the entire logic above
2223 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2224 gets complex and potentially buggy, so more programmer efficient
2225 to do it this way, by turning off the public flags: */
2226 if (!numtype)
2227 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
25da4f38 2228 }
af359546
NC
2229 }
2230 else {
675c862f 2231 if (isGV_with_GP(sv))
a0933d07 2232 return glob_2number((GV *)sv);
180488f8 2233
af359546
NC
2234 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2235 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2236 report_uninit(sv);
2237 }
25da4f38
IZ
2238 if (SvTYPE(sv) < SVt_IV)
2239 /* Typically the caller expects that sv_any is not NULL now. */
2240 sv_upgrade(sv, SVt_IV);
af359546
NC
2241 /* Return 0 from the caller. */
2242 return TRUE;
2243 }
2244 return FALSE;
2245}
2246
2247/*
2248=for apidoc sv_2iv_flags
2249
2250Return the integer value of an SV, doing any necessary string
2251conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2252Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2253
2254=cut
2255*/
2256
2257IV
5de3775c 2258Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
af359546 2259{
97aff369 2260 dVAR;
af359546 2261 if (!sv)
a0d0e21e 2262 return 0;
cecf5685
NC
2263 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2264 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e
NC
2265 cache IVs just in case. In practice it seems that they never
2266 actually anywhere accessible by user Perl code, let alone get used
2267 in anything other than a string context. */
af359546
NC
2268 if (flags & SV_GMAGIC)
2269 mg_get(sv);
2270 if (SvIOKp(sv))
2271 return SvIVX(sv);
2272 if (SvNOKp(sv)) {
2273 return I_V(SvNVX(sv));
2274 }
71c558c3
NC
2275 if (SvPOKp(sv) && SvLEN(sv)) {
2276 UV value;
2277 const int numtype
2278 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2279
2280 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2281 == IS_NUMBER_IN_UV) {
2282 /* It's definitely an integer */
2283 if (numtype & IS_NUMBER_NEG) {
2284 if (value < (UV)IV_MIN)
2285 return -(IV)value;
2286 } else {
2287 if (value < (UV)IV_MAX)
2288 return (IV)value;
2289 }
2290 }
2291 if (!numtype) {
2292 if (ckWARN(WARN_NUMERIC))
2293 not_a_number(sv);
2294 }
2295 return I_V(Atof(SvPVX_const(sv)));
2296 }
1c7ff15e
NC
2297 if (SvROK(sv)) {
2298 goto return_rok;
af359546 2299 }
1c7ff15e
NC
2300 assert(SvTYPE(sv) >= SVt_PVMG);
2301 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2302 } else if (SvTHINKFIRST(sv)) {
af359546 2303 if (SvROK(sv)) {
1c7ff15e 2304 return_rok:
af359546
NC
2305 if (SvAMAGIC(sv)) {
2306 SV * const tmpstr=AMG_CALLun(sv,numer);
2307 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2308 return SvIV(tmpstr);
2309 }
2310 }
2311 return PTR2IV(SvRV(sv));
2312 }
2313 if (SvIsCOW(sv)) {
2314 sv_force_normal_flags(sv, 0);
2315 }
2316 if (SvREADONLY(sv) && !SvOK(sv)) {
2317 if (ckWARN(WARN_UNINITIALIZED))
2318 report_uninit(sv);
2319 return 0;
2320 }
2321 }
2322 if (!SvIOKp(sv)) {
2323 if (S_sv_2iuv_common(aTHX_ sv))
2324 return 0;
79072805 2325 }
1d7c1841
GS
2326 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2327 PTR2UV(sv),SvIVX(sv)));
25da4f38 2328 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2329}
2330
645c22ef 2331/*
891f9566 2332=for apidoc sv_2uv_flags
645c22ef
DM
2333
2334Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2335conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2336Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2337
2338=cut
2339*/
2340
ff68c719 2341UV
5de3775c 2342Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
ff68c719 2343{
97aff369 2344 dVAR;
ff68c719
PP
2345 if (!sv)
2346 return 0;
cecf5685
NC
2347 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2348 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2349 cache IVs just in case. */
891f9566
YST
2350 if (flags & SV_GMAGIC)
2351 mg_get(sv);
ff68c719
PP
2352 if (SvIOKp(sv))
2353 return SvUVX(sv);
2354 if (SvNOKp(sv))
2355 return U_V(SvNVX(sv));
71c558c3
NC
2356 if (SvPOKp(sv) && SvLEN(sv)) {
2357 UV value;
2358 const int numtype
2359 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2360
2361 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2362 == IS_NUMBER_IN_UV) {
2363 /* It's definitely an integer */
2364 if (!(numtype & IS_NUMBER_NEG))
2365 return value;
2366 }
2367 if (!numtype) {
2368 if (ckWARN(WARN_NUMERIC))
2369 not_a_number(sv);
2370 }
2371 return U_V(Atof(SvPVX_const(sv)));
2372 }
1c7ff15e
NC
2373 if (SvROK(sv)) {
2374 goto return_rok;
3fe9a6f1 2375 }
1c7ff15e
NC
2376 assert(SvTYPE(sv) >= SVt_PVMG);
2377 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2378 } else if (SvTHINKFIRST(sv)) {
ff68c719 2379 if (SvROK(sv)) {
1c7ff15e 2380 return_rok:
deb46114
NC
2381 if (SvAMAGIC(sv)) {
2382 SV *const tmpstr = AMG_CALLun(sv,numer);
2383 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2384 return SvUV(tmpstr);
2385 }
2386 }
2387 return PTR2UV(SvRV(sv));
ff68c719 2388 }
765f542d
NC
2389 if (SvIsCOW(sv)) {
2390 sv_force_normal_flags(sv, 0);
8a818333 2391 }
0336b60e 2392 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2393 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2394 report_uninit(sv);
ff68c719
PP
2395 return 0;
2396 }
2397 }
af359546
NC
2398 if (!SvIOKp(sv)) {
2399 if (S_sv_2iuv_common(aTHX_ sv))
2400 return 0;
ff68c719 2401 }
25da4f38 2402
1d7c1841
GS
2403 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2404 PTR2UV(sv),SvUVX(sv)));
25da4f38 2405 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719
PP
2406}
2407
645c22ef
DM
2408/*
2409=for apidoc sv_2nv
2410
2411Return the num value of an SV, doing any necessary string or integer
2412conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2413macros.
2414
2415=cut
2416*/
2417
65202027 2418NV
5de3775c 2419Perl_sv_2nv(pTHX_ register SV *const sv)
79072805 2420{
97aff369 2421 dVAR;
79072805
LW
2422 if (!sv)
2423 return 0.0;
cecf5685
NC
2424 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2425 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2426 cache IVs just in case. */
463ee0b2
LW
2427 mg_get(sv);
2428 if (SvNOKp(sv))
2429 return SvNVX(sv);
0aa395f8 2430 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2431 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2432 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2433 not_a_number(sv);
3f7c398e 2434 return Atof(SvPVX_const(sv));
a0d0e21e 2435 }
25da4f38 2436 if (SvIOKp(sv)) {
1c846c1f 2437 if (SvIsUV(sv))
65202027 2438 return (NV)SvUVX(sv);
25da4f38 2439 else
65202027 2440 return (NV)SvIVX(sv);
47a72cb8
NC
2441 }
2442 if (SvROK(sv)) {
2443 goto return_rok;
2444 }
2445 assert(SvTYPE(sv) >= SVt_PVMG);
2446 /* This falls through to the report_uninit near the end of the
2447 function. */
2448 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2449 if (SvROK(sv)) {
47a72cb8 2450 return_rok:
deb46114
NC
2451 if (SvAMAGIC(sv)) {
2452 SV *const tmpstr = AMG_CALLun(sv,numer);
2453 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2454 return SvNV(tmpstr);
2455 }
2456 }
2457 return PTR2NV(SvRV(sv));
a0d0e21e 2458 }
765f542d
NC
2459 if (SvIsCOW(sv)) {
2460 sv_force_normal_flags(sv, 0);
8a818333 2461 }
0336b60e 2462 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2463 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2464 report_uninit(sv);
ed6116ce
LW
2465 return 0.0;
2466 }
79072805
LW
2467 }
2468 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2469 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2470 sv_upgrade(sv, SVt_NV);
906f284f 2471#ifdef USE_LONG_DOUBLE
097ee67d 2472 DEBUG_c({
f93f4e46 2473 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2474 PerlIO_printf(Perl_debug_log,
2475 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2476 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2477 RESTORE_NUMERIC_LOCAL();
2478 });
65202027 2479#else
572bbb43 2480 DEBUG_c({
f93f4e46 2481 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2482 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2483 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2484 RESTORE_NUMERIC_LOCAL();
2485 });
572bbb43 2486#endif
79072805
LW
2487 }
2488 else if (SvTYPE(sv) < SVt_PVNV)
2489 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2490 if (SvNOKp(sv)) {
2491 return SvNVX(sv);
61604483 2492 }
59d8ce62 2493 if (SvIOKp(sv)) {
9d6ce603 2494 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8 2495#ifdef NV_PRESERVES_UV
a43d94f2
NC
2496 if (SvIOK(sv))
2497 SvNOK_on(sv);
2498 else
2499 SvNOKp_on(sv);
28e5dec8
JH
2500#else
2501 /* Only set the public NV OK flag if this NV preserves the IV */
2502 /* Check it's not 0xFFFFFFFFFFFFFFFF */
a43d94f2
NC
2503 if (SvIOK(sv) &&
2504 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
28e5dec8
JH
2505 : (SvIVX(sv) == I_V(SvNVX(sv))))
2506 SvNOK_on(sv);
2507 else
2508 SvNOKp_on(sv);
2509#endif
93a17b20 2510 }
748a9306 2511 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2512 UV value;
3f7c398e 2513 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2514 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2515 not_a_number(sv);
28e5dec8 2516#ifdef NV_PRESERVES_UV
c2988b20
NC
2517 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2518 == IS_NUMBER_IN_UV) {
5e045b90 2519 /* It's definitely an integer */
9d6ce603 2520 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2521 } else
3f7c398e 2522 SvNV_set(sv, Atof(SvPVX_const(sv)));
a43d94f2
NC
2523 if (numtype)
2524 SvNOK_on(sv);
2525 else
2526 SvNOKp_on(sv);
28e5dec8 2527#else
3f7c398e 2528 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2529 /* Only set the public NV OK flag if this NV preserves the value in
2530 the PV at least as well as an IV/UV would.
2531 Not sure how to do this 100% reliably. */
2532 /* if that shift count is out of range then Configure's test is
2533 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2534 UV_BITS */
2535 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2536 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2537 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2538 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2539 /* Can't use strtol etc to convert this string, so don't try.
2540 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2541 SvNOK_on(sv);
2542 } else {
2543 /* value has been set. It may not be precise. */
2544 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2545 /* 2s complement assumption for (UV)IV_MIN */
2546 SvNOK_on(sv); /* Integer is too negative. */
2547 } else {
2548 SvNOKp_on(sv);
2549 SvIOKp_on(sv);
6fa402ec 2550
c2988b20 2551 if (numtype & IS_NUMBER_NEG) {
45977657 2552 SvIV_set(sv, -(IV)value);
c2988b20 2553 } else if (value <= (UV)IV_MAX) {
45977657 2554 SvIV_set(sv, (IV)value);
c2988b20 2555 } else {
607fa7f2 2556 SvUV_set(sv, value);
c2988b20
NC
2557 SvIsUV_on(sv);
2558 }
2559
2560 if (numtype & IS_NUMBER_NOT_INT) {
2561 /* I believe that even if the original PV had decimals,
2562 they are lost beyond the limit of the FP precision.
2563 However, neither is canonical, so both only get p
2564 flags. NWC, 2000/11/25 */
2565 /* Both already have p flags, so do nothing */
2566 } else {
66a1b24b 2567 const NV nv = SvNVX(sv);
c2988b20
NC
2568 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2569 if (SvIVX(sv) == I_V(nv)) {
2570 SvNOK_on(sv);
c2988b20 2571 } else {
c2988b20
NC
2572 /* It had no "." so it must be integer. */
2573 }
00b6aa41 2574 SvIOK_on(sv);
c2988b20
NC
2575 } else {
2576 /* between IV_MAX and NV(UV_MAX).
2577 Could be slightly > UV_MAX */
6fa402ec 2578
c2988b20
NC
2579 if (numtype & IS_NUMBER_NOT_INT) {
2580 /* UV and NV both imprecise. */
2581 } else {
66a1b24b 2582 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2583
2584 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2585 SvNOK_on(sv);
c2988b20 2586 }
00b6aa41 2587 SvIOK_on(sv);
c2988b20
NC
2588 }
2589 }
2590 }
2591 }
2592 }
a43d94f2
NC
2593 /* It might be more code efficient to go through the entire logic above
2594 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2595 gets complex and potentially buggy, so more programmer efficient
2596 to do it this way, by turning off the public flags: */
2597 if (!numtype)
2598 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
28e5dec8 2599#endif /* NV_PRESERVES_UV */
93a17b20 2600 }
79072805 2601 else {
f7877b28 2602 if (isGV_with_GP(sv)) {
19f6321d 2603 glob_2number((GV *)sv);
180488f8
NC
2604 return 0.0;
2605 }
2606
041457d9 2607 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2608 report_uninit(sv);
7e25a7e9
NC
2609 assert (SvTYPE(sv) >= SVt_NV);
2610 /* Typically the caller expects that sv_any is not NULL now. */
2611 /* XXX Ilya implies that this is a bug in callers that assume this
2612 and ideally should be fixed. */
a0d0e21e 2613 return 0.0;
79072805 2614 }
572bbb43 2615#if defined(USE_LONG_DOUBLE)
097ee67d 2616 DEBUG_c({
f93f4e46 2617 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2618 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2619 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2620 RESTORE_NUMERIC_LOCAL();
2621 });
65202027 2622#else
572bbb43 2623 DEBUG_c({
f93f4e46 2624 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2625 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2626 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2627 RESTORE_NUMERIC_LOCAL();
2628 });
572bbb43 2629#endif
463ee0b2 2630 return SvNVX(sv);
79072805
LW
2631}
2632
800401ee
JH
2633/*
2634=for apidoc sv_2num
2635
2636Return an SV with the numeric value of the source SV, doing any necessary
a196a5fa
JH
2637reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2638access this function.
800401ee
JH
2639
2640=cut
2641*/
2642
2643SV *
5de3775c 2644Perl_sv_2num(pTHX_ register SV *const sv)
800401ee 2645{
7918f24d
NC
2646 PERL_ARGS_ASSERT_SV_2NUM;
2647
b9ee0594
RGS
2648 if (!SvROK(sv))
2649 return sv;
800401ee
JH
2650 if (SvAMAGIC(sv)) {
2651 SV * const tmpsv = AMG_CALLun(sv,numer);
2652 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2653 return sv_2num(tmpsv);
2654 }
2655 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2656}
2657
645c22ef
DM
2658/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2659 * UV as a string towards the end of buf, and return pointers to start and
2660 * end of it.
2661 *
2662 * We assume that buf is at least TYPE_CHARS(UV) long.
2663 */
2664
864dbfa3 2665static char *
5de3775c 2666S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
25da4f38 2667{
25da4f38 2668 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2669 char * const ebuf = ptr;
25da4f38 2670 int sign;
25da4f38 2671
7918f24d
NC
2672 PERL_ARGS_ASSERT_UIV_2BUF;
2673
25da4f38
IZ
2674 if (is_uv)
2675 sign = 0;
2676 else if (iv >= 0) {
2677 uv = iv;
2678 sign = 0;
2679 } else {
2680 uv = -iv;
2681 sign = 1;
2682 }
2683 do {
eb160463 2684 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2685 } while (uv /= 10);
2686 if (sign)
2687 *--ptr = '-';
2688 *peob = ebuf;
2689 return ptr;
2690}
2691
645c22ef
DM
2692/*
2693=for apidoc sv_2pv_flags
2694
ff276b08 2695Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2696If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2697if necessary.
2698Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2699usually end up here too.
2700
2701=cut
2702*/
2703
8d6d96c1 2704char *
5de3775c 2705Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
8d6d96c1 2706{
97aff369 2707 dVAR;
79072805 2708 register char *s;
79072805 2709
463ee0b2 2710 if (!sv) {
cdb061a3
NC
2711 if (lp)
2712 *lp = 0;
73d840c0 2713 return (char *)"";
463ee0b2 2714 }
8990e307 2715 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2716 if (flags & SV_GMAGIC)
2717 mg_get(sv);
463ee0b2 2718 if (SvPOKp(sv)) {
cdb061a3
NC
2719 if (lp)
2720 *lp = SvCUR(sv);
10516c54
NC
2721 if (flags & SV_MUTABLE_RETURN)
2722 return SvPVX_mutable(sv);
4d84ee25
NC
2723 if (flags & SV_CONST_RETURN)
2724 return (char *)SvPVX_const(sv);
463ee0b2
LW
2725 return SvPVX(sv);
2726 }
75dfc8ec
NC
2727 if (SvIOKp(sv) || SvNOKp(sv)) {
2728 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2729 STRLEN len;
2730
2731 if (SvIOKp(sv)) {
e80fed9d 2732 len = SvIsUV(sv)
d9fad198
JH
2733 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2734 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2735 } else {
e8ada2d0
NC
2736 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2737 len = strlen(tbuf);
75dfc8ec 2738 }
b5b886f0
NC
2739 assert(!SvROK(sv));
2740 {
75dfc8ec
NC
2741 dVAR;
2742
2743#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2744 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2745 tbuf[0] = '0';
2746 tbuf[1] = 0;
75dfc8ec
NC
2747 len = 1;
2748 }
2749#endif
2750 SvUPGRADE(sv, SVt_PV);
2751 if (lp)
2752 *lp = len;
2753 s = SvGROW_mutable(sv, len + 1);
2754 SvCUR_set(sv, len);
2755 SvPOKp_on(sv);
10edeb5d 2756 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2757 }
463ee0b2 2758 }
1c7ff15e
NC
2759 if (SvROK(sv)) {
2760 goto return_rok;
2761 }
2762 assert(SvTYPE(sv) >= SVt_PVMG);
2763 /* This falls through to the report_uninit near the end of the
2764 function. */
2765 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2766 if (SvROK(sv)) {
1c7ff15e 2767 return_rok:
deb46114
NC
2768 if (SvAMAGIC(sv)) {
2769 SV *const tmpstr = AMG_CALLun(sv,string);
2770 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2771 /* Unwrap this: */
2772 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2773 */
2774
2775 char *pv;
2776 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2777 if (flags & SV_CONST_RETURN) {
2778 pv = (char *) SvPVX_const(tmpstr);
2779 } else {
2780 pv = (flags & SV_MUTABLE_RETURN)
2781 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2782 }
2783 if (lp)
2784 *lp = SvCUR(tmpstr);
50adf7d2 2785 } else {
deb46114 2786 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2787 }
deb46114
NC
2788 if (SvUTF8(tmpstr))
2789 SvUTF8_on(sv);
2790 else
2791 SvUTF8_off(sv);
2792 return pv;
50adf7d2 2793 }
deb46114
NC
2794 }
2795 {
fafee734
NC
2796 STRLEN len;
2797 char *retval;
2798 char *buffer;
d8eae41e
NC
2799 const SV *const referent = (SV*)SvRV(sv);
2800
2801 if (!referent) {
fafee734
NC
2802 len = 7;
2803 retval = buffer = savepvn("NULLREF", len);
5c35adbb 2804 } else if (SvTYPE(referent) == SVt_REGEXP) {
67d2d14d
AB
2805 const REGEXP * const re = (REGEXP *)referent;
2806 I32 seen_evals = 0;
2807
2808 assert(re);
2809
2810 /* If the regex is UTF-8 we want the containing scalar to
2811 have an UTF-8 flag too */
2812 if (RX_UTF8(re))
2813 SvUTF8_on(sv);
2814 else
2815 SvUTF8_off(sv);
2816
2817 if ((seen_evals = RX_SEEN_EVALS(re)))
2818 PL_reginterp_cnt += seen_evals;
2819
2820 if (lp)
2821 *lp = RX_WRAPLEN(re);
2822
2823 return RX_WRAPPED(re);
d8eae41e
NC
2824 } else {
2825 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2826 const STRLEN typelen = strlen(typestr);
2827 UV addr = PTR2UV(referent);
2828 const char *stashname = NULL;
2829 STRLEN stashnamelen = 0; /* hush, gcc */
2830 const char *buffer_end;
d8eae41e 2831
d8eae41e 2832 if (SvOBJECT(referent)) {
fafee734
NC
2833 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2834
2835 if (name) {
2836 stashname = HEK_KEY(name);
2837 stashnamelen = HEK_LEN(name);
2838
2839 if (HEK_UTF8(name)) {
2840 SvUTF8_on(sv);
2841 } else {
2842 SvUTF8_off(sv);
2843 }
2844 } else {
2845 stashname = "__ANON__";
2846 stashnamelen = 8;
2847 }
2848 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2849 + 2 * sizeof(UV) + 2 /* )\0 */;
2850 } else {
2851 len = typelen + 3 /* (0x */
2852 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2853 }
fafee734
NC
2854
2855 Newx(buffer, len, char);
2856 buffer_end = retval = buffer + len;
2857
2858 /* Working backwards */
2859 *--retval = '\0';
2860 *--retval = ')';
2861 do {
2862 *--retval = PL_hexdigit[addr & 15];
2863 } while (addr >>= 4);
2864 *--retval = 'x';
2865 *--retval = '0';
2866 *--retval = '(';
2867
2868 retval -= typelen;
2869 memcpy(retval, typestr, typelen);
2870
2871 if (stashname) {
2872 *--retval = '=';
2873 retval -= stashnamelen;
2874 memcpy(retval, stashname, stashnamelen);
2875 }
2876 /* retval may not neccesarily have reached the start of the
2877 buffer here. */
2878 assert (retval >= buffer);
2879
2880 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2881 }
042dae7a 2882 if (lp)
fafee734
NC
2883 *lp = len;
2884 SAVEFREEPV(buffer);
2885 return retval;
463ee0b2 2886 }
79072805 2887 }
0336b60e 2888 if (SvREADONLY(sv) && !SvOK(sv)) {
cdb061a3
NC
2889 if (lp)
2890 *lp = 0;
9f621bb0
NC
2891 if (flags & SV_UNDEF_RETURNS_NULL)
2892 return NULL;
2893 if (ckWARN(WARN_UNINITIALIZED))
2894 report_uninit(sv);
73d840c0 2895 return (char *)"";
79072805 2896 }
79072805 2897 }
28e5dec8
JH
2898 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2899 /* I'm assuming that if both IV and NV are equally valid then
2900 converting the IV is going to be more efficient */
e1ec3a88 2901 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2902 char buf[TYPE_CHARS(UV)];
2903 char *ebuf, *ptr;
97a130b8 2904 STRLEN len;
28e5dec8
JH
2905
2906 if (SvTYPE(sv) < SVt_PVIV)
2907 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2908 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
97a130b8 2909 len = ebuf - ptr;
5902b6a9 2910 /* inlined from sv_setpvn */
97a130b8
NC
2911 s = SvGROW_mutable(sv, len + 1);
2912 Move(ptr, s, len, char);
2913 s += len;
28e5dec8 2914 *s = '\0';
28e5dec8
JH
2915 }
2916 else if (SvNOKp(sv)) {
c81271c3 2917 const int olderrno = errno;
79072805
LW
2918 if (SvTYPE(sv) < SVt_PVNV)
2919 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2920 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2921 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2922 /* some Xenix systems wipe out errno here */
79072805 2923#ifdef apollo
463ee0b2 2924 if (SvNVX(sv) == 0.0)
d1307786 2925 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2926 else
2927#endif /*apollo*/
bbce6d69 2928 {
2d4389e4 2929 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2930 }
79072805 2931 errno = olderrno;
a0d0e21e 2932#ifdef FIXNEGATIVEZERO
20773dcd
NC
2933 if (*s == '-' && s[1] == '0' && !s[2]) {
2934 s[0] = '0';
2935 s[1] = 0;
2936 }
a0d0e21e 2937#endif
79072805
LW
2938 while (*s) s++;
2939#ifdef hcx
2940 if (s[-1] == '.')
46fc3d4c 2941 *--s = '\0';
79072805
LW
2942#endif
2943 }
79072805 2944 else {
675c862f 2945 if (isGV_with_GP(sv))
19f6321d 2946 return glob_2pv((GV *)sv, lp);
180488f8 2947
cdb061a3 2948 if (lp)
00b6aa41 2949 *lp = 0;
9f621bb0
NC
2950 if (flags & SV_UNDEF_RETURNS_NULL)
2951 return NULL;
2952 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2953 report_uninit(sv);
25da4f38
IZ
2954 if (SvTYPE(sv) < SVt_PV)
2955 /* Typically the caller expects that sv_any is not NULL now. */
2956 sv_upgrade(sv, SVt_PV);
73d840c0 2957 return (char *)"";
79072805 2958 }
cdb061a3 2959 {
823a54a3 2960 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2961 if (lp)
2962 *lp = len;
2963 SvCUR_set(sv, len);
2964 }
79072805 2965 SvPOK_on(sv);
1d7c1841 2966 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2967 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2968 if (flags & SV_CONST_RETURN)
2969 return (char *)SvPVX_const(sv);
10516c54
NC
2970 if (flags & SV_MUTABLE_RETURN)
2971 return SvPVX_mutable(sv);
463ee0b2
LW
2972 return SvPVX(sv);
2973}
2974
645c22ef 2975/*
6050d10e
JP
2976=for apidoc sv_copypv
2977
2978Copies a stringified representation of the source SV into the
2979destination SV. Automatically performs any necessary mg_get and
54f0641b 2980coercion of numeric values into strings. Guaranteed to preserve
2575c402 2981UTF8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2982sv_2pv[_flags] but operates directly on an SV instead of just the
2983string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2984would lose the UTF-8'ness of the PV.
2985
2986=cut
2987*/
2988
2989void
5de3775c 2990Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
6050d10e 2991{
446eaa42 2992 STRLEN len;
53c1dcc0 2993 const char * const s = SvPV_const(ssv,len);
7918f24d
NC
2994
2995 PERL_ARGS_ASSERT_SV_COPYPV;
2996
cb50f42d 2997 sv_setpvn(dsv,s,len);
446eaa42 2998 if (SvUTF8(ssv))
cb50f42d 2999 SvUTF8_on(dsv);
446eaa42 3000 else
cb50f42d 3001 SvUTF8_off(dsv);
6050d10e
JP
3002}
3003
3004/*
645c22ef
DM
3005=for apidoc sv_2pvbyte
3006
3007Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 3008to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
3009side-effect.
3010
3011Usually accessed via the C<SvPVbyte> macro.
3012
3013=cut
3014*/
3015
7340a771 3016char *
5de3775c 3017Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3018{
7918f24d
NC
3019 PERL_ARGS_ASSERT_SV_2PVBYTE;
3020
0875d2fe 3021 sv_utf8_downgrade(sv,0);
97972285 3022 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
3023}
3024
645c22ef 3025/*
035cbb0e
RGS
3026=for apidoc sv_2pvutf8
3027
3028Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3029to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3030
3031Usually accessed via the C<SvPVutf8> macro.
3032
3033=cut
3034*/
645c22ef 3035
7340a771 3036char *
7bc54cea 3037Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3038{
7918f24d
NC
3039 PERL_ARGS_ASSERT_SV_2PVUTF8;
3040
035cbb0e
RGS
3041 sv_utf8_upgrade(sv);
3042 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 3043}
1c846c1f 3044
7ee2227d 3045
645c22ef
DM
3046/*
3047=for apidoc sv_2bool
3048
3049This function is only called on magical items, and is only used by
8cf8f3d1 3050sv_true() or its macro equivalent.
645c22ef
DM
3051
3052=cut
3053*/
3054
463ee0b2 3055bool
7bc54cea 3056Perl_sv_2bool(pTHX_ register SV *const sv)
463ee0b2 3057{
97aff369 3058 dVAR;
7918f24d
NC
3059
3060 PERL_ARGS_ASSERT_SV_2BOOL;
3061
5b295bef 3062 SvGETMAGIC(sv);
463ee0b2 3063
a0d0e21e
LW
3064 if (!SvOK(sv))
3065 return 0;
3066 if (SvROK(sv)) {
fabdb6c0
AL
3067 if (SvAMAGIC(sv)) {
3068 SV * const tmpsv = AMG_CALLun(sv,bool_);
3069 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3070 return (bool)SvTRUE(tmpsv);
3071 }
3072 return SvRV(sv) != 0;
a0d0e21e 3073 }
463ee0b2 3074 if (SvPOKp(sv)) {
53c1dcc0
AL
3075 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3076 if (Xpvtmp &&
339049b0 3077 (*sv->sv_u.svu_pv > '0' ||
11343788 3078 Xpvtmp->xpv_cur > 1 ||
339049b0 3079 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
3080 return 1;
3081 else
3082 return 0;
3083 }
3084 else {
3085 if (SvIOKp(sv))
3086 return SvIVX(sv) != 0;
3087 else {
3088 if (SvNOKp(sv))
3089 return SvNVX(sv) != 0.0;
180488f8 3090 else {
f7877b28 3091 if (isGV_with_GP(sv))
180488f8
NC
3092 return TRUE;
3093 else
3094 return FALSE;
3095 }
463ee0b2
LW
3096 }
3097 }
79072805
LW
3098}
3099
c461cf8f
JH
3100/*
3101=for apidoc sv_utf8_upgrade
3102
78ea37eb 3103Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3104Forces the SV to string form if it is not already.
4411f3b6
NIS
3105Always sets the SvUTF8 flag to avoid future validity checks even
3106if all the bytes have hibit clear.
c461cf8f 3107
13a6c0e0
JH
3108This is not as a general purpose byte encoding to Unicode interface:
3109use the Encode extension for that.
3110
8d6d96c1
HS
3111=for apidoc sv_utf8_upgrade_flags
3112
78ea37eb 3113Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3114Forces the SV to string form if it is not already.
8d6d96c1
HS
3115Always sets the SvUTF8 flag to avoid future validity checks even
3116if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
3117will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
3118C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3119
13a6c0e0
JH
3120This is not as a general purpose byte encoding to Unicode interface:
3121use the Encode extension for that.
3122
8d6d96c1
HS
3123=cut
3124*/
3125
3126STRLEN
7bc54cea 3127Perl_sv_utf8_upgrade_flags(pTHX_ register SV *const sv, const I32 flags)
8d6d96c1 3128{
97aff369 3129 dVAR;
7918f24d
NC
3130
3131 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS;
3132
808c356f
RGS
3133 if (sv == &PL_sv_undef)
3134 return 0;
e0e62c2a
NIS
3135 if (!SvPOK(sv)) {
3136 STRLEN len = 0;
d52b7888
NC
3137 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3138 (void) sv_2pv_flags(sv,&len, flags);
3139 if (SvUTF8(sv))
3140 return len;
3141 } else {
3142 (void) SvPV_force(sv,len);
3143 }
e0e62c2a 3144 }
4411f3b6 3145
f5cee72b 3146 if (SvUTF8(sv)) {
5fec3b1d 3147 return SvCUR(sv);
f5cee72b 3148 }
5fec3b1d 3149
765f542d
NC
3150 if (SvIsCOW(sv)) {
3151 sv_force_normal_flags(sv, 0);
db42d148
NIS
3152 }
3153
88632417 3154 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 3155 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 3156 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
3157 /* This function could be much more efficient if we
3158 * had a FLAG in SVs to signal if there are any hibit
3159 * chars in the PV. Given that there isn't such a flag
3160 * make the loop as fast as possible. */
00b6aa41 3161 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 3162 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 3163 const U8 *t = s;
c4e7c712
NC
3164
3165 while (t < e) {
53c1dcc0 3166 const U8 ch = *t++;
00b6aa41
AL
3167 /* Check for hi bit */
3168 if (!NATIVE_IS_INVARIANT(ch)) {
4612962a
NC
3169 STRLEN len = SvCUR(sv);
3170 /* *Currently* bytes_to_utf8() adds a '\0' after every string
3171 it converts. This isn't documented. It's not clear if it's
3172 a bad thing to be doing, and should be changed to do exactly
3173 what the documentation says. If so, this code will have to
3174 be changed.
3175 As is, we mustn't rely on our incoming SV being well formed
3176 and having a trailing '\0', as certain code in pp_formline
3177 can send us partially built SVs. */
00b6aa41
AL
3178 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
3179
3180 SvPV_free(sv); /* No longer using what was there before. */
3181 SvPV_set(sv, (char*)recoded);
4612962a
NC
3182 SvCUR_set(sv, len);
3183 SvLEN_set(sv, len + 1); /* No longer know the real size. */
c4e7c712 3184 break;
00b6aa41 3185 }
c4e7c712
NC
3186 }
3187 /* Mark as UTF-8 even if no hibit - saves scanning loop */
3188 SvUTF8_on(sv);
560a288e 3189 }
4411f3b6 3190 return SvCUR(sv);
560a288e
GS
3191}
3192
c461cf8f
JH
3193/*
3194=for apidoc sv_utf8_downgrade
3195
78ea37eb
ST
3196Attempts to convert the PV of an SV from characters to bytes.
3197If the PV contains a character beyond byte, this conversion will fail;
3198in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3199true, croaks.
3200
13a6c0e0
JH
3201This is not as a general purpose Unicode to byte encoding interface:
3202use the Encode extension for that.
3203
c461cf8f
JH
3204=cut
3205*/
3206
560a288e 3207bool
7bc54cea 3208Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
560a288e 3209{
97aff369 3210 dVAR;
7918f24d
NC
3211
3212 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3213
78ea37eb 3214 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3215 if (SvCUR(sv)) {
03cfe0ae 3216 U8 *s;
652088fc 3217 STRLEN len;
fa301091 3218
765f542d
NC
3219 if (SvIsCOW(sv)) {
3220 sv_force_normal_flags(sv, 0);
3221 }
03cfe0ae
NIS
3222 s = (U8 *) SvPV(sv, len);
3223 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3224 if (fail_ok)
3225 return FALSE;
3226 else {
3227 if (PL_op)
3228 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3229 OP_DESC(PL_op));
fa301091
JH
3230 else
3231 Perl_croak(aTHX_ "Wide character");
3232 }
4b3603a4 3233 }
b162af07 3234 SvCUR_set(sv, len);
67e989fb 3235 }
560a288e 3236 }
ffebcc3e 3237 SvUTF8_off(sv);
560a288e
GS
3238 return TRUE;
3239}
3240
c461cf8f
JH
3241/*
3242=for apidoc sv_utf8_encode
3243
78ea37eb
ST
3244Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3245flag off so that it looks like octets again.
c461cf8f
JH
3246
3247=cut
3248*/
3249
560a288e 3250void
7bc54cea 3251Perl_sv_utf8_encode(pTHX_ register SV *const sv)
560a288e 3252{
7918f24d
NC
3253 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3254
4c94c214
NC
3255 if (SvIsCOW(sv)) {
3256 sv_force_normal_flags(sv, 0);
3257 }
3258 if (SvREADONLY(sv)) {
3259 Perl_croak(aTHX_ PL_no_modify);
3260 }
a5f5288a 3261 (void) sv_utf8_upgrade(sv);
560a288e
GS
3262 SvUTF8_off(sv);
3263}
3264
4411f3b6
NIS
3265/*
3266=for apidoc sv_utf8_decode
3267
78ea37eb
ST
3268If the PV of the SV is an octet sequence in UTF-8
3269and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3270so that it looks like a character. If the PV contains only single-byte
3271characters, the C<SvUTF8> flag stays being off.
3272Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3273
3274=cut
3275*/
3276
560a288e 3277bool
7bc54cea 3278Perl_sv_utf8_decode(pTHX_ register SV *const sv)
560a288e 3279{
7918f24d
NC
3280 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3281
78ea37eb 3282 if (SvPOKp(sv)) {
93524f2b
NC
3283 const U8 *c;
3284 const U8 *e;
9cbac4c7 3285
645c22ef
DM
3286 /* The octets may have got themselves encoded - get them back as
3287 * bytes
3288 */
3289 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3290 return FALSE;
3291
3292 /* it is actually just a matter of turning the utf8 flag on, but
3293 * we want to make sure everything inside is valid utf8 first.
3294 */
93524f2b 3295 c = (const U8 *) SvPVX_const(sv);
63cd0674 3296 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3297 return FALSE;
93524f2b 3298 e = (const U8 *) SvEND(sv);
511c2ff0 3299 while (c < e) {
b64e5050 3300 const U8 ch = *c++;
c4d5f83a 3301 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3302 SvUTF8_on(sv);
3303 break;
3304 }
560a288e 3305 }
560a288e
GS
3306 }
3307 return TRUE;
3308}
3309
954c1994
GS
3310/*
3311=for apidoc sv_setsv
3312
645c22ef
DM
3313Copies the contents of the source SV C<ssv> into the destination SV
3314C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3315function if the source SV needs to be reused. Does not handle 'set' magic.
3316Loosely speaking, it performs a copy-by-value, obliterating any previous
3317content of the destination.
3318
3319You probably want to use one of the assortment of wrappers, such as
3320C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3321C<SvSetMagicSV_nosteal>.
3322
8d6d96c1
HS
3323=for apidoc sv_setsv_flags
3324
645c22ef
DM
3325Copies the contents of the source SV C<ssv> into the destination SV
3326C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3327function if the source SV needs to be reused. Does not handle 'set' magic.
3328Loosely speaking, it performs a copy-by-value, obliterating any previous
3329content of the destination.
3330If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3331C<ssv> if appropriate, else not. If the C<flags> parameter has the
3332C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3333and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3334
3335You probably want to use one of the assortment of wrappers, such as
3336C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3337C<SvSetMagicSV_nosteal>.
3338
3339This is the primary function for copying scalars, and most other
3340copy-ish functions and macros use this underneath.
8d6d96c1
HS
3341
3342=cut
3343*/
3344
5d0301b7 3345static void
7bc54cea 3346S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
5d0301b7 3347{
70cd14a1 3348 I32 mro_changes = 0; /* 1 = method, 2 = isa */
dd69841b 3349
7918f24d
NC
3350 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3351
5d0301b7
NC
3352 if (dtype != SVt_PVGV) {
3353 const char * const name = GvNAME(sstr);
3354 const STRLEN len = GvNAMELEN(sstr);
0d092c36 3355 {
f7877b28
NC
3356 if (dtype >= SVt_PV) {
3357 SvPV_free(dstr);
3358 SvPV_set(dstr, 0);
3359 SvLEN_set(dstr, 0);
3360 SvCUR_set(dstr, 0);
3361 }
0d092c36 3362 SvUPGRADE(dstr, SVt_PVGV);
dedf8e73 3363 (void)SvOK_off(dstr);
2e5b91de
NC
3364 /* FIXME - why are we doing this, then turning it off and on again
3365 below? */
3366 isGV_with_GP_on(dstr);
f7877b28 3367 }
5d0301b7
NC
3368 GvSTASH(dstr) = GvSTASH(sstr);
3369 if (GvSTASH(dstr))
3370 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
ae8cc45f 3371 gv_name_set((GV *)dstr, name, len, GV_ADD);
5d0301b7
NC
3372 SvFAKE_on(dstr); /* can coerce to non-glob */
3373 }
3374
3375#ifdef GV_UNIQUE_CHECK
3376 if (GvUNIQUE((GV*)dstr)) {
3377 Perl_croak(aTHX_ PL_no_modify);
3378 }
3379#endif
3380
dd69841b
BB
3381 if(GvGP((GV*)sstr)) {
3382 /* If source has method cache entry, clear it */
3383 if(GvCVGEN(sstr)) {
3384 SvREFCNT_dec(GvCV(sstr));
3385 GvCV(sstr) = NULL;
3386 GvCVGEN(sstr) = 0;
3387 }
3388 /* If source has a real method, then a method is
3389 going to change */
3390 else if(GvCV((GV*)sstr)) {
70cd14a1 3391 mro_changes = 1;
dd69841b
BB
3392 }
3393 }
3394
3395 /* If dest already had a real method, that's a change as well */
70cd14a1
CB
3396 if(!mro_changes && GvGP((GV*)dstr) && GvCVu((GV*)dstr)) {
3397 mro_changes = 1;
dd69841b
BB
3398 }
3399
70cd14a1
CB
3400 if(strEQ(GvNAME((GV*)dstr),"ISA"))
3401 mro_changes = 2;
3402
f7877b28 3403 gp_free((GV*)dstr);
2e5b91de 3404 isGV_with_GP_off(dstr);
5d0301b7 3405 (void)SvOK_off(dstr);
2e5b91de 3406 isGV_with_GP_on(dstr);
dedf8e73 3407 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3408 GvGP(dstr) = gp_ref(GvGP(sstr));
3409 if (SvTAINTED(sstr))
3410 SvTAINT(dstr);
3411 if (GvIMPORTED(dstr) != GVf_IMPORTED
3412 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3413 {
3414 GvIMPORTED_on(dstr);
3415 }
3416 GvMULTI_on(dstr);
70cd14a1
CB
3417 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3418 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
5d0301b7
NC
3419 return;
3420}
3421
b8473700 3422static void
7bc54cea 3423S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
7918f24d 3424{
b8473700
NC
3425 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3426 SV *dref = NULL;
3427 const int intro = GvINTRO(dstr);
2440974c 3428 SV **location;
3386d083 3429 U8 import_flag = 0;
27242d61
NC
3430 const U32 stype = SvTYPE(sref);
3431
7918f24d 3432 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
b8473700
NC
3433
3434#ifdef GV_UNIQUE_CHECK
3435 if (GvUNIQUE((GV*)dstr)) {
3436 Perl_croak(aTHX_ PL_no_modify);
3437 }
3438#endif
3439
3440 if (intro) {
3441 GvINTRO_off(dstr); /* one-shot flag */
3442 GvLINE(dstr) = CopLINE(PL_curcop);
3443 GvEGV(dstr) = (GV*)dstr;
3444 }
3445 GvMULTI_on(dstr);
27242d61 3446 switch (stype) {
b8473700 3447 case SVt_PVCV:
27242d61
NC
3448 location = (SV **) &GvCV(dstr);
3449 import_flag = GVf_IMPORTED_CV;
3450 goto common;
3451 case SVt_PVHV:
3452 location = (SV **) &GvHV(dstr);
3453 import_flag = GVf_IMPORTED_HV;
3454 goto common;
3455 case SVt_PVAV:
3456 location = (SV **) &GvAV(dstr);
3457 import_flag = GVf_IMPORTED_AV;
3458 goto common;
3459 case SVt_PVIO:
3460 location = (SV **) &GvIOp(dstr);
3461 goto common;
3462 case SVt_PVFM:
3463 location = (SV **) &GvFORM(dstr);
3464 default:
3465 location = &GvSV(dstr);
3466 import_flag = GVf_IMPORTED_SV;
3467 common:
b8473700 3468 if (intro) {
27242d61 3469 if (stype == SVt_PVCV) {
5f2fca8a
BB
3470 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (CV*)sref || GvCVGEN(dstr))) {*/
3471 if (GvCVGEN(dstr)) {
27242d61
NC
3472 SvREFCNT_dec(GvCV(dstr));
3473 GvCV(dstr) = NULL;
3474 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
27242d61 3475 }
b8473700 3476 }
27242d61 3477 SAVEGENERICSV(*location);
b8473700
NC
3478 }
3479 else
27242d61 3480 dref = *location;
5f2fca8a 3481 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
27242d61 3482 CV* const cv = (CV*)*location;
b8473700
NC
3483 if (cv) {
3484 if (!GvCVGEN((GV*)dstr) &&
3485 (CvROOT(cv) || CvXSUB(cv)))
3486 {
3487 /* Redefining a sub - warning is mandatory if
3488 it was a const and its value changed. */
3489 if (CvCONST(cv) && CvCONST((CV*)sref)
3490 && cv_const_sv(cv) == cv_const_sv((CV*)sref)) {
6f207bd3 3491 NOOP;
b8473700
NC
3492 /* They are 2 constant subroutines generated from
3493 the same constant. This probably means that
3494 they are really the "same" proxy subroutine
3495 instantiated in 2 places. Most likely this is
3496 when a constant is exported twice. Don't warn.
3497 */
3498 }
3499 else if (ckWARN(WARN_REDEFINE)
3500 || (CvCONST(cv)
3501 && (!CvCONST((CV*)sref)
3502 || sv_cmp(cv_const_sv(cv),
3503 cv_const_sv((CV*)sref))))) {
3504 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3505 (const char *)
3506 (CvCONST(cv)
3507 ? "Constant subroutine %s::%s redefined"
3508 : "Subroutine %s::%s redefined"),
b8473700
NC
3509 HvNAME_get(GvSTASH((GV*)dstr)),
3510 GvENAME((GV*)dstr));
3511 }
3512 }
3513 if (!intro)
cbf82dd0
NC
3514 cv_ckproto_len(cv, (GV*)dstr,
3515 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3516 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3517 }
b8473700
NC
3518 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3519 GvASSUMECV_on(dstr);
dd69841b 3520 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
b8473700 3521 }
2440974c 3522 *location = sref;
3386d083
NC
3523 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3524 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3525 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3526 }
3527 break;
3528 }
b37c2d43 3529 SvREFCNT_dec(dref);
b8473700
NC
3530 if (SvTAINTED(sstr))
3531 SvTAINT(dstr);
3532 return;
3533}
3534
8d6d96c1 3535void
7bc54cea 3536Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
8d6d96c1 3537{
97aff369 3538 dVAR;
8990e307
LW
3539 register U32 sflags;
3540 register int dtype;
42d0e0b7 3541 register svtype stype;
463ee0b2 3542
7918f24d
NC
3543 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3544
79072805
LW
3545 if (sstr == dstr)
3546 return;
29f4f0ab
NC
3547
3548 if (SvIS_FREED(dstr)) {
3549 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
be2597df 3550 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
29f4f0ab 3551 }
765f542d 3552 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3553 if (!sstr)
3280af22 3554 sstr = &PL_sv_undef;
29f4f0ab 3555 if (SvIS_FREED(sstr)) {
6c9570dc
MHM
3556 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3557 (void*)sstr, (void*)dstr);
29f4f0ab 3558 }
8990e307
LW
3559 stype = SvTYPE(sstr);
3560 dtype = SvTYPE(dstr);
79072805 3561
52944de8 3562 (void)SvAMAGIC_off(dstr);
7a5fa8a2 3563 if ( SvVOK(dstr) )
ece467f9
JP
3564 {
3565 /* need to nuke the magic */
3566 mg_free(dstr);
ece467f9 3567 }
9e7bc3e8 3568
463ee0b2 3569 /* There's a lot of redundancy below but we're going for speed here */
79072805 3570
8990e307 3571 switch (stype) {
79072805 3572 case SVt_NULL:
aece5585 3573 undef_sstr:
20408e3c
GS
3574 if (dtype != SVt_PVGV) {
3575 (void)SvOK_off(dstr);
3576 return;
3577 }
3578 break;
463ee0b2 3579 case SVt_IV:
aece5585
GA
3580 if (SvIOK(sstr)) {
3581 switch (dtype) {
3582 case SVt_NULL:
8990e307 3583 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3584 break;
3585 case SVt_NV:
aece5585 3586 case SVt_PV:
a0d0e21e 3587 sv_upgrade(dstr, SVt_PVIV);
aece5585 3588 break;
010be86b
NC
3589 case SVt_PVGV:
3590 goto end_of_first_switch;
aece5585
GA
3591 }
3592 (void)SvIOK_only(dstr);
45977657 3593 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3594 if (SvIsUV(sstr))
3595 SvIsUV_on(dstr);
37c25af0
NC
3596 /* SvTAINTED can only be true if the SV has taint magic, which in
3597 turn means that the SV type is PVMG (or greater). This is the
3598 case statement for SVt_IV, so this cannot be true (whatever gcov
3599 may say). */
3600 assert(!SvTAINTED(sstr));
aece5585 3601 return;
8990e307 3602 }
4df7f6af
NC
3603 if (!SvROK(sstr))
3604 goto undef_sstr;
3605 if (dtype < SVt_PV && dtype != SVt_IV)
3606 sv_upgrade(dstr, SVt_IV);
3607 break;
aece5585 3608
463ee0b2 3609 case SVt_NV:
aece5585
GA
3610 if (SvNOK(sstr)) {
3611 switch (dtype) {
3612 case SVt_NULL:
3613 case SVt_IV:
8990e307 3614 sv_upgrade(dstr, SVt_NV);
aece5585 3615 break;
aece5585
GA
3616 case SVt_PV:
3617 case SVt_PVIV:
a0d0e21e 3618 sv_upgrade(dstr, SVt_PVNV);
aece5585 3619 break;
010be86b
NC
3620 case SVt_PVGV:
3621 goto end_of_first_switch;
aece5585 3622 }
9d6ce603 3623 SvNV_set(dstr, SvNVX(sstr));
aece5585 3624 (void)SvNOK_only(dstr);
37c25af0
NC
3625 /* SvTAINTED can only be true if the SV has taint magic, which in
3626 turn means that the SV type is PVMG (or greater). This is the
3627 case statement for SVt_NV, so this cannot be true (whatever gcov
3628 may say). */
3629 assert(!SvTAINTED(sstr));
aece5585 3630 return;
8990e307 3631 }
aece5585
GA
3632 goto undef_sstr;
3633
fc36a67e 3634 case SVt_PVFM:
f8c7b90f 3635#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3636 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3637 if (dtype < SVt_PVIV)
3638 sv_upgrade(dstr, SVt_PVIV);
3639 break;
3640 }
3641 /* Fall through */
3642#endif
fd44068c 3643 case SVt_REGEXP: