3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
71 /* ============================================================================
73 =head1 Allocation and deallocation of SVs.
75 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
76 sv, av, hv...) contains type and reference count information, and for
77 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
78 contains fields specific to each type. Some types store all they need
79 in the head, so don't have a body.
81 In all but the most memory-paranoid configurations (ex: PURIFY), heads
82 and bodies are allocated out of arenas, which by default are
83 approximately 4K chunks of memory parcelled up into N heads or bodies.
84 Sv-bodies are allocated by their sv-type, guaranteeing size
85 consistency needed to allocate safely from arrays.
87 For SV-heads, the first slot in each arena is reserved, and holds a
88 link to the next arena, some flags, and a note of the number of slots.
89 Snaked through each arena chain is a linked list of free items; when
90 this becomes empty, an extra arena is allocated and divided up into N
91 items which are threaded into the free list.
93 SV-bodies are similar, but they use arena-sets by default, which
94 separate the link and info from the arena itself, and reclaim the 1st
95 slot in the arena. SV-bodies are further described later.
97 The following global variables are associated with arenas:
99 PL_sv_arenaroot pointer to list of SV arenas
100 PL_sv_root pointer to list of free SV structures
102 PL_body_arenas head of linked-list of body arenas
103 PL_body_roots[] array of pointers to list of free bodies of svtype
104 arrays are indexed by the svtype needed
106 A few special SV heads are not allocated from an arena, but are
107 instead directly created in the interpreter structure, eg PL_sv_undef.
108 The size of arenas can be changed from the default by setting
109 PERL_ARENA_SIZE appropriately at compile time.
111 The SV arena serves the secondary purpose of allowing still-live SVs
112 to be located and destroyed during final cleanup.
114 At the lowest level, the macros new_SV() and del_SV() grab and free
115 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
116 to return the SV to the free list with error checking.) new_SV() calls
117 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
118 SVs in the free list have their SvTYPE field set to all ones.
120 At the time of very final cleanup, sv_free_arenas() is called from
121 perl_destruct() to physically free all the arenas allocated since the
122 start of the interpreter.
124 The function visit() scans the SV arenas list, and calls a specified
125 function for each SV it finds which is still live - ie which has an SvTYPE
126 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
127 following functions (specified as [function that calls visit()] / [function
128 called by visit() for each SV]):
130 sv_report_used() / do_report_used()
131 dump all remaining SVs (debugging aid)
133 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
134 do_clean_named_io_objs(),do_curse()
135 Attempt to free all objects pointed to by RVs,
136 try to do the same for all objects indir-
137 ectly referenced by typeglobs too, and
138 then do a final sweep, cursing any
139 objects that remain. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) STMT_START { \
186 if ((sv)->sv_debug_file) PerlMemShared_free((sv)->sv_debug_file); \
188 # define DEBUG_SV_SERIAL(sv) \
189 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
190 PTR2UV(sv), (long)(sv)->sv_debug_serial))
192 # define FREE_SV_DEBUG_FILE(sv)
193 # define DEBUG_SV_SERIAL(sv) NOOP
197 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
198 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
199 /* Whilst I'd love to do this, it seems that things like to check on
201 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
203 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
204 PoisonNew(&SvREFCNT(sv), 1, U32)
206 # define SvARENA_CHAIN(sv) SvANY(sv)
207 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
208 # define POSION_SV_HEAD(sv)
211 /* Mark an SV head as unused, and add to free list.
213 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
214 * its refcount artificially decremented during global destruction, so
215 * there may be dangling pointers to it. The last thing we want in that
216 * case is for it to be reused. */
218 #define plant_SV(p) \
220 const U32 old_flags = SvFLAGS(p); \
221 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
222 DEBUG_SV_SERIAL(p); \
223 FREE_SV_DEBUG_FILE(p); \
225 SvFLAGS(p) = SVTYPEMASK; \
226 if (!(old_flags & SVf_BREAK)) { \
227 SvARENA_CHAIN_SET(p, PL_sv_root); \
233 #define uproot_SV(p) \
236 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
241 /* make some more SVs by adding another arena */
248 char *chunk; /* must use New here to match call to */
249 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
250 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
255 /* new_SV(): return a new, empty SV head */
257 #ifdef DEBUG_LEAKING_SCALARS
258 /* provide a real function for a debugger to play with */
260 S_new_SV(pTHX_ const char *file, int line, const char *func)
267 sv = S_more_sv(aTHX);
271 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
272 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
278 sv->sv_debug_inpad = 0;
279 sv->sv_debug_parent = NULL;
280 sv->sv_debug_file = PL_curcop ? savesharedpv(CopFILE(PL_curcop)): NULL;
282 sv->sv_debug_serial = PL_sv_serial++;
284 MEM_LOG_NEW_SV(sv, file, line, func);
285 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
286 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
290 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
298 (p) = S_more_sv(aTHX); \
302 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
307 /* del_SV(): return an empty SV head to the free list */
320 S_del_sv(pTHX_ SV *p)
324 PERL_ARGS_ASSERT_DEL_SV;
329 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
330 const SV * const sv = sva + 1;
331 const SV * const svend = &sva[SvREFCNT(sva)];
332 if (p >= sv && p < svend) {
338 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
339 "Attempt to free non-arena SV: 0x%"UVxf
340 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
347 #else /* ! DEBUGGING */
349 #define del_SV(p) plant_SV(p)
351 #endif /* DEBUGGING */
355 =head1 SV Manipulation Functions
357 =for apidoc sv_add_arena
359 Given a chunk of memory, link it to the head of the list of arenas,
360 and split it into a list of free SVs.
366 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
369 SV *const sva = MUTABLE_SV(ptr);
373 PERL_ARGS_ASSERT_SV_ADD_ARENA;
375 /* The first SV in an arena isn't an SV. */
376 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
377 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
378 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
380 PL_sv_arenaroot = sva;
381 PL_sv_root = sva + 1;
383 svend = &sva[SvREFCNT(sva) - 1];
386 SvARENA_CHAIN_SET(sv, (sv + 1));
390 /* Must always set typemask because it's always checked in on cleanup
391 when the arenas are walked looking for objects. */
392 SvFLAGS(sv) = SVTYPEMASK;
395 SvARENA_CHAIN_SET(sv, 0);
399 SvFLAGS(sv) = SVTYPEMASK;
402 /* visit(): call the named function for each non-free SV in the arenas
403 * whose flags field matches the flags/mask args. */
406 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
412 PERL_ARGS_ASSERT_VISIT;
414 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
415 const SV * const svend = &sva[SvREFCNT(sva)];
417 for (sv = sva + 1; sv < svend; ++sv) {
418 if (SvTYPE(sv) != (svtype)SVTYPEMASK
419 && (sv->sv_flags & mask) == flags
432 /* called by sv_report_used() for each live SV */
435 do_report_used(pTHX_ SV *const sv)
437 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
438 PerlIO_printf(Perl_debug_log, "****\n");
445 =for apidoc sv_report_used
447 Dump the contents of all SVs not yet freed (debugging aid).
453 Perl_sv_report_used(pTHX)
456 visit(do_report_used, 0, 0);
462 /* called by sv_clean_objs() for each live SV */
465 do_clean_objs(pTHX_ SV *const ref)
470 SV * const target = SvRV(ref);
471 if (SvOBJECT(target)) {
472 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
473 if (SvWEAKREF(ref)) {
474 sv_del_backref(target, ref);
480 SvREFCNT_dec_NN(target);
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
508 SvREFCNT_dec_NN(obj);
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
514 SvREFCNT_dec_NN(obj);
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
520 SvREFCNT_dec_NN(obj);
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
526 SvREFCNT_dec_NN(obj);
528 SvREFCNT_dec_NN(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
549 SvREFCNT_dec_NN(obj);
551 SvREFCNT_dec_NN(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* IVs are in the head, so the allocation size is 0. */
886 sizeof(IV), /* This is used to copy out the IV body. */
887 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
888 NOARENA /* IVS don't need an arena */, 0
891 { sizeof(NV), sizeof(NV),
892 STRUCT_OFFSET(XPVNV, xnv_u),
893 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
895 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
896 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
897 + STRUCT_OFFSET(XPV, xpv_cur),
898 SVt_PV, FALSE, NONV, HASARENA,
899 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
901 { sizeof(XINVLIST) - STRUCT_OFFSET(XPV, xpv_cur),
902 copy_length(XINVLIST, is_offset) - STRUCT_OFFSET(XPV, xpv_cur),
903 + STRUCT_OFFSET(XPV, xpv_cur),
904 SVt_INVLIST, TRUE, NONV, HASARENA,
905 FIT_ARENA(0, sizeof(XINVLIST) - STRUCT_OFFSET(XPV, xpv_cur)) },
907 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
908 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
909 + STRUCT_OFFSET(XPV, xpv_cur),
910 SVt_PVIV, FALSE, NONV, HASARENA,
911 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
913 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
914 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
915 + STRUCT_OFFSET(XPV, xpv_cur),
916 SVt_PVNV, FALSE, HADNV, HASARENA,
917 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
919 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
920 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
925 SVt_REGEXP, TRUE, NONV, HASARENA,
926 FIT_ARENA(0, sizeof(regexp))
929 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
930 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
932 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
933 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
936 copy_length(XPVAV, xav_alloc),
938 SVt_PVAV, TRUE, NONV, HASARENA,
939 FIT_ARENA(0, sizeof(XPVAV)) },
942 copy_length(XPVHV, xhv_max),
944 SVt_PVHV, TRUE, NONV, HASARENA,
945 FIT_ARENA(0, sizeof(XPVHV)) },
950 SVt_PVCV, TRUE, NONV, HASARENA,
951 FIT_ARENA(0, sizeof(XPVCV)) },
956 SVt_PVFM, TRUE, NONV, NOARENA,
957 FIT_ARENA(20, sizeof(XPVFM)) },
962 SVt_PVIO, TRUE, NONV, HASARENA,
963 FIT_ARENA(24, sizeof(XPVIO)) },
966 #define new_body_allocated(sv_type) \
967 (void *)((char *)S_new_body(aTHX_ sv_type) \
968 - bodies_by_type[sv_type].offset)
970 /* return a thing to the free list */
972 #define del_body(thing, root) \
974 void ** const thing_copy = (void **)thing; \
975 *thing_copy = *root; \
976 *root = (void*)thing_copy; \
981 #define new_XNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVNV() safemalloc(sizeof(XPVNV))
983 #define new_XPVMG() safemalloc(sizeof(XPVMG))
985 #define del_XPVGV(p) safefree(p)
989 #define new_XNV() new_body_allocated(SVt_NV)
990 #define new_XPVNV() new_body_allocated(SVt_PVNV)
991 #define new_XPVMG() new_body_allocated(SVt_PVMG)
993 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
994 &PL_body_roots[SVt_PVGV])
998 /* no arena for you! */
1000 #define new_NOARENA(details) \
1001 safemalloc((details)->body_size + (details)->offset)
1002 #define new_NOARENAZ(details) \
1003 safecalloc((details)->body_size + (details)->offset, 1)
1006 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1007 const size_t arena_size)
1010 void ** const root = &PL_body_roots[sv_type];
1011 struct arena_desc *adesc;
1012 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1016 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1017 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1018 static bool done_sanity_check;
1020 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1021 * variables like done_sanity_check. */
1022 if (!done_sanity_check) {
1023 unsigned int i = SVt_LAST;
1025 done_sanity_check = TRUE;
1028 assert (bodies_by_type[i].type == i);
1034 /* may need new arena-set to hold new arena */
1035 if (!aroot || aroot->curr >= aroot->set_size) {
1036 struct arena_set *newroot;
1037 Newxz(newroot, 1, struct arena_set);
1038 newroot->set_size = ARENAS_PER_SET;
1039 newroot->next = aroot;
1041 PL_body_arenas = (void *) newroot;
1042 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1045 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1046 curr = aroot->curr++;
1047 adesc = &(aroot->set[curr]);
1048 assert(!adesc->arena);
1050 Newx(adesc->arena, good_arena_size, char);
1051 adesc->size = good_arena_size;
1052 adesc->utype = sv_type;
1053 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1054 curr, (void*)adesc->arena, (UV)good_arena_size));
1056 start = (char *) adesc->arena;
1058 /* Get the address of the byte after the end of the last body we can fit.
1059 Remember, this is integer division: */
1060 end = start + good_arena_size / body_size * body_size;
1062 /* computed count doesn't reflect the 1st slot reservation */
1063 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1064 DEBUG_m(PerlIO_printf(Perl_debug_log,
1065 "arena %p end %p arena-size %d (from %d) type %d "
1067 (void*)start, (void*)end, (int)good_arena_size,
1068 (int)arena_size, sv_type, (int)body_size,
1069 (int)good_arena_size / (int)body_size));
1071 DEBUG_m(PerlIO_printf(Perl_debug_log,
1072 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1073 (void*)start, (void*)end,
1074 (int)arena_size, sv_type, (int)body_size,
1075 (int)good_arena_size / (int)body_size));
1077 *root = (void *)start;
1080 /* Where the next body would start: */
1081 char * const next = start + body_size;
1084 /* This is the last body: */
1085 assert(next == end);
1087 *(void **)start = 0;
1091 *(void**) start = (void *)next;
1096 /* grab a new thing from the free list, allocating more if necessary.
1097 The inline version is used for speed in hot routines, and the
1098 function using it serves the rest (unless PURIFY).
1100 #define new_body_inline(xpv, sv_type) \
1102 void ** const r3wt = &PL_body_roots[sv_type]; \
1103 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1104 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1105 bodies_by_type[sv_type].body_size,\
1106 bodies_by_type[sv_type].arena_size)); \
1107 *(r3wt) = *(void**)(xpv); \
1113 S_new_body(pTHX_ const svtype sv_type)
1117 new_body_inline(xpv, sv_type);
1123 static const struct body_details fake_rv =
1124 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1127 =for apidoc sv_upgrade
1129 Upgrade an SV to a more complex form. Generally adds a new body type to the
1130 SV, then copies across as much information as possible from the old body.
1131 It croaks if the SV is already in a more complex form than requested. You
1132 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1133 before calling C<sv_upgrade>, and hence does not croak. See also
1140 Perl_sv_upgrade(pTHX_ SV *const sv, svtype new_type)
1145 const svtype old_type = SvTYPE(sv);
1146 const struct body_details *new_type_details;
1147 const struct body_details *old_type_details
1148 = bodies_by_type + old_type;
1149 SV *referant = NULL;
1151 PERL_ARGS_ASSERT_SV_UPGRADE;
1153 if (old_type == new_type)
1156 /* This clause was purposefully added ahead of the early return above to
1157 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1158 inference by Nick I-S that it would fix other troublesome cases. See
1159 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1161 Given that shared hash key scalars are no longer PVIV, but PV, there is
1162 no longer need to unshare so as to free up the IVX slot for its proper
1163 purpose. So it's safe to move the early return earlier. */
1165 if (new_type > SVt_PVMG && SvIsCOW(sv)) {
1166 sv_force_normal_flags(sv, 0);
1169 old_body = SvANY(sv);
1171 /* Copying structures onto other structures that have been neatly zeroed
1172 has a subtle gotcha. Consider XPVMG
1174 +------+------+------+------+------+-------+-------+
1175 | NV | CUR | LEN | IV | MAGIC | STASH |
1176 +------+------+------+------+------+-------+-------+
1177 0 4 8 12 16 20 24 28
1179 where NVs are aligned to 8 bytes, so that sizeof that structure is
1180 actually 32 bytes long, with 4 bytes of padding at the end:
1182 +------+------+------+------+------+-------+-------+------+
1183 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1184 +------+------+------+------+------+-------+-------+------+
1185 0 4 8 12 16 20 24 28 32
1187 so what happens if you allocate memory for this structure:
1189 +------+------+------+------+------+-------+-------+------+------+...
1190 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1191 +------+------+------+------+------+-------+-------+------+------+...
1192 0 4 8 12 16 20 24 28 32 36
1194 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1195 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1196 started out as zero once, but it's quite possible that it isn't. So now,
1197 rather than a nicely zeroed GP, you have it pointing somewhere random.
1200 (In fact, GP ends up pointing at a previous GP structure, because the
1201 principle cause of the padding in XPVMG getting garbage is a copy of
1202 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1203 this happens to be moot because XPVGV has been re-ordered, with GP
1204 no longer after STASH)
1206 So we are careful and work out the size of used parts of all the
1214 referant = SvRV(sv);
1215 old_type_details = &fake_rv;
1216 if (new_type == SVt_NV)
1217 new_type = SVt_PVNV;
1219 if (new_type < SVt_PVIV) {
1220 new_type = (new_type == SVt_NV)
1221 ? SVt_PVNV : SVt_PVIV;
1226 if (new_type < SVt_PVNV) {
1227 new_type = SVt_PVNV;
1231 assert(new_type > SVt_PV);
1232 assert(SVt_IV < SVt_PV);
1233 assert(SVt_NV < SVt_PV);
1240 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1241 there's no way that it can be safely upgraded, because perl.c
1242 expects to Safefree(SvANY(PL_mess_sv)) */
1243 assert(sv != PL_mess_sv);
1244 /* This flag bit is used to mean other things in other scalar types.
1245 Given that it only has meaning inside the pad, it shouldn't be set
1246 on anything that can get upgraded. */
1247 assert(!SvPAD_TYPED(sv));
1250 if (UNLIKELY(old_type_details->cant_upgrade))
1251 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1252 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1255 if (UNLIKELY(old_type > new_type))
1256 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1257 (int)old_type, (int)new_type);
1259 new_type_details = bodies_by_type + new_type;
1261 SvFLAGS(sv) &= ~SVTYPEMASK;
1262 SvFLAGS(sv) |= new_type;
1264 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1265 the return statements above will have triggered. */
1266 assert (new_type != SVt_NULL);
1269 assert(old_type == SVt_NULL);
1270 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1274 assert(old_type == SVt_NULL);
1275 SvANY(sv) = new_XNV();
1280 assert(new_type_details->body_size);
1283 assert(new_type_details->arena);
1284 assert(new_type_details->arena_size);
1285 /* This points to the start of the allocated area. */
1286 new_body_inline(new_body, new_type);
1287 Zero(new_body, new_type_details->body_size, char);
1288 new_body = ((char *)new_body) - new_type_details->offset;
1290 /* We always allocated the full length item with PURIFY. To do this
1291 we fake things so that arena is false for all 16 types.. */
1292 new_body = new_NOARENAZ(new_type_details);
1294 SvANY(sv) = new_body;
1295 if (new_type == SVt_PVAV) {
1299 if (old_type_details->body_size) {
1302 /* It will have been zeroed when the new body was allocated.
1303 Lets not write to it, in case it confuses a write-back
1309 #ifndef NODEFAULT_SHAREKEYS
1310 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1312 /* start with PERL_HASH_DEFAULT_HvMAX+1 buckets: */
1313 HvMAX(sv) = PERL_HASH_DEFAULT_HvMAX;
1316 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1317 The target created by newSVrv also is, and it can have magic.
1318 However, it never has SvPVX set.
1320 if (old_type == SVt_IV) {
1322 } else if (old_type >= SVt_PV) {
1323 assert(SvPVX_const(sv) == 0);
1326 if (old_type >= SVt_PVMG) {
1327 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1328 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1330 sv->sv_u.svu_array = NULL; /* or svu_hash */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1350 assert(new_type_details->body_size);
1351 /* We always allocated the full length item with PURIFY. To do this
1352 we fake things so that arena is false for all 16 types.. */
1353 if(new_type_details->arena) {
1354 /* This points to the start of the allocated area. */
1355 new_body_inline(new_body, new_type);
1356 Zero(new_body, new_type_details->body_size, char);
1357 new_body = ((char *)new_body) - new_type_details->offset;
1359 new_body = new_NOARENAZ(new_type_details);
1361 SvANY(sv) = new_body;
1363 if (old_type_details->copy) {
1364 /* There is now the potential for an upgrade from something without
1365 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1366 int offset = old_type_details->offset;
1367 int length = old_type_details->copy;
1369 if (new_type_details->offset > old_type_details->offset) {
1370 const int difference
1371 = new_type_details->offset - old_type_details->offset;
1372 offset += difference;
1373 length -= difference;
1375 assert (length >= 0);
1377 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1381 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1382 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1383 * correct 0.0 for us. Otherwise, if the old body didn't have an
1384 * NV slot, but the new one does, then we need to initialise the
1385 * freshly created NV slot with whatever the correct bit pattern is
1387 if (old_type_details->zero_nv && !new_type_details->zero_nv
1388 && !isGV_with_GP(sv))
1392 if (UNLIKELY(new_type == SVt_PVIO)) {
1393 IO * const io = MUTABLE_IO(sv);
1394 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1397 /* Clear the stashcache because a new IO could overrule a package
1399 DEBUG_o(Perl_deb(aTHX_ "sv_upgrade clearing PL_stashcache\n"));
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (UNLIKELY(new_type == SVt_REGEXP))
1406 sv->sv_u.svu_rx = (regexp *)new_body;
1407 else if (old_type < SVt_PV) {
1408 /* referant will be NULL unless the old type was SVt_IV emulating
1410 sv->sv_u.svu_rv = referant;
1414 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1415 (unsigned long)new_type);
1418 if (old_type > SVt_IV) {
1422 /* Note that there is an assumption that all bodies of types that
1423 can be upgraded came from arenas. Only the more complex non-
1424 upgradable types are allowed to be directly malloc()ed. */
1425 assert(old_type_details->arena);
1426 del_body((void*)((char*)old_body + old_type_details->offset),
1427 &PL_body_roots[old_type]);
1433 =for apidoc sv_backoff
1435 Remove any string offset. You should normally use the C<SvOOK_off> macro
1442 Perl_sv_backoff(pTHX_ SV *const sv)
1445 const char * const s = SvPVX_const(sv);
1447 PERL_ARGS_ASSERT_SV_BACKOFF;
1448 PERL_UNUSED_CONTEXT;
1451 assert(SvTYPE(sv) != SVt_PVHV);
1452 assert(SvTYPE(sv) != SVt_PVAV);
1454 SvOOK_offset(sv, delta);
1456 SvLEN_set(sv, SvLEN(sv) + delta);
1457 SvPV_set(sv, SvPVX(sv) - delta);
1458 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1459 SvFLAGS(sv) &= ~SVf_OOK;
1466 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1467 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1468 Use the C<SvGROW> wrapper instead.
1473 static void S_sv_uncow(pTHX_ SV * const sv, const U32 flags);
1476 Perl_sv_grow(pTHX_ SV *const sv, STRLEN newlen)
1480 PERL_ARGS_ASSERT_SV_GROW;
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1507 if (SvIsCOW(sv)) S_sv_uncow(aTHX_ sv, 0);
1508 s = SvPVX_mutable(sv);
1511 #ifdef PERL_NEW_COPY_ON_WRITE
1512 /* the new COW scheme uses SvPVX(sv)[SvLEN(sv)-1] (if spare)
1513 * to store the COW count. So in general, allocate one more byte than
1514 * asked for, to make it likely this byte is always spare: and thus
1515 * make more strings COW-able.
1516 * If the new size is a big power of two, don't bother: we assume the
1517 * caller wanted a nice 2^N sized block and will be annoyed at getting
1523 if (newlen > SvLEN(sv)) { /* need more room? */
1524 STRLEN minlen = SvCUR(sv);
1525 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1526 if (newlen < minlen)
1528 #ifndef Perl_safesysmalloc_size
1529 newlen = PERL_STRLEN_ROUNDUP(newlen);
1531 if (SvLEN(sv) && s) {
1532 s = (char*)saferealloc(s, newlen);
1535 s = (char*)safemalloc(newlen);
1536 if (SvPVX_const(sv) && SvCUR(sv)) {
1537 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1541 #ifdef Perl_safesysmalloc_size
1542 /* Do this here, do it once, do it right, and then we will never get
1543 called back into sv_grow() unless there really is some growing
1545 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1547 SvLEN_set(sv, newlen);
1554 =for apidoc sv_setiv
1556 Copies an integer into the given SV, upgrading first if necessary.
1557 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1563 Perl_sv_setiv(pTHX_ SV *const sv, const IV i)
1567 PERL_ARGS_ASSERT_SV_SETIV;
1569 SV_CHECK_THINKFIRST_COW_DROP(sv);
1570 switch (SvTYPE(sv)) {
1573 sv_upgrade(sv, SVt_IV);
1576 sv_upgrade(sv, SVt_PVIV);
1580 if (!isGV_with_GP(sv))
1587 /* diag_listed_as: Can't coerce %s to %s in %s */
1588 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1592 (void)SvIOK_only(sv); /* validate number */
1598 =for apidoc sv_setiv_mg
1600 Like C<sv_setiv>, but also handles 'set' magic.
1606 Perl_sv_setiv_mg(pTHX_ SV *const sv, const IV i)
1608 PERL_ARGS_ASSERT_SV_SETIV_MG;
1615 =for apidoc sv_setuv
1617 Copies an unsigned integer into the given SV, upgrading first if necessary.
1618 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1624 Perl_sv_setuv(pTHX_ SV *const sv, const UV u)
1626 PERL_ARGS_ASSERT_SV_SETUV;
1628 /* With the if statement to ensure that integers are stored as IVs whenever
1630 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1633 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1635 If you wish to remove the following if statement, so that this routine
1636 (and its callers) always return UVs, please benchmark to see what the
1637 effect is. Modern CPUs may be different. Or may not :-)
1639 if (u <= (UV)IV_MAX) {
1640 sv_setiv(sv, (IV)u);
1649 =for apidoc sv_setuv_mg
1651 Like C<sv_setuv>, but also handles 'set' magic.
1657 Perl_sv_setuv_mg(pTHX_ SV *const sv, const UV u)
1659 PERL_ARGS_ASSERT_SV_SETUV_MG;
1666 =for apidoc sv_setnv
1668 Copies a double into the given SV, upgrading first if necessary.
1669 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1675 Perl_sv_setnv(pTHX_ SV *const sv, const NV num)
1679 PERL_ARGS_ASSERT_SV_SETNV;
1681 SV_CHECK_THINKFIRST_COW_DROP(sv);
1682 switch (SvTYPE(sv)) {
1685 sv_upgrade(sv, SVt_NV);
1689 sv_upgrade(sv, SVt_PVNV);
1693 if (!isGV_with_GP(sv))
1700 /* diag_listed_as: Can't coerce %s to %s in %s */
1701 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1706 (void)SvNOK_only(sv); /* validate number */
1711 =for apidoc sv_setnv_mg
1713 Like C<sv_setnv>, but also handles 'set' magic.
1719 Perl_sv_setnv_mg(pTHX_ SV *const sv, const NV num)
1721 PERL_ARGS_ASSERT_SV_SETNV_MG;
1727 /* Return a cleaned-up, printable version of sv, for non-numeric, or
1728 * not incrementable warning display.
1729 * Originally part of S_not_a_number().
1730 * The return value may be != tmpbuf.
1734 S_sv_display(pTHX_ SV *const sv, char *tmpbuf, STRLEN tmpbuf_size) {
1737 PERL_ARGS_ASSERT_SV_DISPLAY;
1740 SV *dsv = newSVpvs_flags("", SVs_TEMP);
1741 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1744 const char * const limit = tmpbuf + tmpbuf_size - 8;
1745 /* each *s can expand to 4 chars + "...\0",
1746 i.e. need room for 8 chars */
1748 const char *s = SvPVX_const(sv);
1749 const char * const end = s + SvCUR(sv);
1750 for ( ; s < end && d < limit; s++ ) {
1752 if (ch & 128 && !isPRINT_LC(ch)) {
1761 else if (ch == '\r') {
1765 else if (ch == '\f') {
1769 else if (ch == '\\') {
1773 else if (ch == '\0') {
1777 else if (isPRINT_LC(ch))
1796 /* Print an "isn't numeric" warning, using a cleaned-up,
1797 * printable version of the offending string
1801 S_not_a_number(pTHX_ SV *const sv)
1807 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1809 pv = sv_display(sv, tmpbuf, sizeof(tmpbuf));
1812 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1813 /* diag_listed_as: Argument "%s" isn't numeric%s */
1814 "Argument \"%s\" isn't numeric in %s", pv,
1817 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1818 /* diag_listed_as: Argument "%s" isn't numeric%s */
1819 "Argument \"%s\" isn't numeric", pv);
1823 S_not_incrementable(pTHX_ SV *const sv) {
1828 PERL_ARGS_ASSERT_NOT_INCREMENTABLE;
1830 pv = sv_display(sv, tmpbuf, sizeof(tmpbuf));
1832 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1833 "Argument \"%s\" treated as 0 in increment (++)", pv);
1837 =for apidoc looks_like_number
1839 Test if the content of an SV looks like a number (or is a number).
1840 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1841 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1848 Perl_looks_like_number(pTHX_ SV *const sv)
1853 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1855 if (SvPOK(sv) || SvPOKp(sv)) {
1856 sbegin = SvPV_nomg_const(sv, len);
1859 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1860 return grok_number(sbegin, len, NULL);
1864 S_glob_2number(pTHX_ GV * const gv)
1866 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1868 /* We know that all GVs stringify to something that is not-a-number,
1869 so no need to test that. */
1870 if (ckWARN(WARN_NUMERIC))
1872 SV *const buffer = sv_newmortal();
1873 gv_efullname3(buffer, gv, "*");
1874 not_a_number(buffer);
1876 /* We just want something true to return, so that S_sv_2iuv_common
1877 can tail call us and return true. */
1881 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1882 until proven guilty, assume that things are not that bad... */
1887 As 64 bit platforms often have an NV that doesn't preserve all bits of
1888 an IV (an assumption perl has been based on to date) it becomes necessary
1889 to remove the assumption that the NV always carries enough precision to
1890 recreate the IV whenever needed, and that the NV is the canonical form.
1891 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1892 precision as a side effect of conversion (which would lead to insanity
1893 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1894 1) to distinguish between IV/UV/NV slots that have cached a valid
1895 conversion where precision was lost and IV/UV/NV slots that have a
1896 valid conversion which has lost no precision
1897 2) to ensure that if a numeric conversion to one form is requested that
1898 would lose precision, the precise conversion (or differently
1899 imprecise conversion) is also performed and cached, to prevent
1900 requests for different numeric formats on the same SV causing
1901 lossy conversion chains. (lossless conversion chains are perfectly
1906 SvIOKp is true if the IV slot contains a valid value
1907 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1908 SvNOKp is true if the NV slot contains a valid value
1909 SvNOK is true only if the NV value is accurate
1912 while converting from PV to NV, check to see if converting that NV to an
1913 IV(or UV) would lose accuracy over a direct conversion from PV to
1914 IV(or UV). If it would, cache both conversions, return NV, but mark
1915 SV as IOK NOKp (ie not NOK).
1917 While converting from PV to IV, check to see if converting that IV to an
1918 NV would lose accuracy over a direct conversion from PV to NV. If it
1919 would, cache both conversions, flag similarly.
1921 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1922 correctly because if IV & NV were set NV *always* overruled.
1923 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1924 changes - now IV and NV together means that the two are interchangeable:
1925 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1927 The benefit of this is that operations such as pp_add know that if
1928 SvIOK is true for both left and right operands, then integer addition
1929 can be used instead of floating point (for cases where the result won't
1930 overflow). Before, floating point was always used, which could lead to
1931 loss of precision compared with integer addition.
1933 * making IV and NV equal status should make maths accurate on 64 bit
1935 * may speed up maths somewhat if pp_add and friends start to use
1936 integers when possible instead of fp. (Hopefully the overhead in
1937 looking for SvIOK and checking for overflow will not outweigh the
1938 fp to integer speedup)
1939 * will slow down integer operations (callers of SvIV) on "inaccurate"
1940 values, as the change from SvIOK to SvIOKp will cause a call into
1941 sv_2iv each time rather than a macro access direct to the IV slot
1942 * should speed up number->string conversion on integers as IV is
1943 favoured when IV and NV are equally accurate
1945 ####################################################################
1946 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1947 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1948 On the other hand, SvUOK is true iff UV.
1949 ####################################################################
1951 Your mileage will vary depending your CPU's relative fp to integer
1955 #ifndef NV_PRESERVES_UV
1956 # define IS_NUMBER_UNDERFLOW_IV 1
1957 # define IS_NUMBER_UNDERFLOW_UV 2
1958 # define IS_NUMBER_IV_AND_UV 2
1959 # define IS_NUMBER_OVERFLOW_IV 4
1960 # define IS_NUMBER_OVERFLOW_UV 5
1962 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1964 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1966 S_sv_2iuv_non_preserve(pTHX_ SV *const sv
1974 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1976 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));
1977 if (SvNVX(sv) < (NV)IV_MIN) {
1978 (void)SvIOKp_on(sv);
1980 SvIV_set(sv, IV_MIN);
1981 return IS_NUMBER_UNDERFLOW_IV;
1983 if (SvNVX(sv) > (NV)UV_MAX) {
1984 (void)SvIOKp_on(sv);
1987 SvUV_set(sv, UV_MAX);
1988 return IS_NUMBER_OVERFLOW_UV;
1990 (void)SvIOKp_on(sv);
1992 /* Can't use strtol etc to convert this string. (See truth table in
1994 if (SvNVX(sv) <= (UV)IV_MAX) {
1995 SvIV_set(sv, I_V(SvNVX(sv)));
1996 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1997 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1999 /* Integer is imprecise. NOK, IOKp */
2001 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
2004 SvUV_set(sv, U_V(SvNVX(sv)));
2005 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2006 if (SvUVX(sv) == UV_MAX) {
2007 /* As we know that NVs don't preserve UVs, UV_MAX cannot
2008 possibly be preserved by NV. Hence, it must be overflow.
2010 return IS_NUMBER_OVERFLOW_UV;
2012 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2014 /* Integer is imprecise. NOK, IOKp */
2016 return IS_NUMBER_OVERFLOW_IV;
2018 #endif /* !NV_PRESERVES_UV*/
2021 S_sv_2iuv_common(pTHX_ SV *const sv)
2025 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2028 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2029 * without also getting a cached IV/UV from it at the same time
2030 * (ie PV->NV conversion should detect loss of accuracy and cache
2031 * IV or UV at same time to avoid this. */
2032 /* IV-over-UV optimisation - choose to cache IV if possible */
2034 if (SvTYPE(sv) == SVt_NV)
2035 sv_upgrade(sv, SVt_PVNV);
2037 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2038 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2039 certainly cast into the IV range at IV_MAX, whereas the correct
2040 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2042 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2043 if (Perl_isnan(SvNVX(sv))) {
2049 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2050 SvIV_set(sv, I_V(SvNVX(sv)));
2051 if (SvNVX(sv) == (NV) SvIVX(sv)
2052 #ifndef NV_PRESERVES_UV
2053 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2054 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2055 /* Don't flag it as "accurately an integer" if the number
2056 came from a (by definition imprecise) NV operation, and
2057 we're outside the range of NV integer precision */
2061 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2063 /* scalar has trailing garbage, eg "42a" */
2065 DEBUG_c(PerlIO_printf(Perl_debug_log,
2066 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2072 /* IV not precise. No need to convert from PV, as NV
2073 conversion would already have cached IV if it detected
2074 that PV->IV would be better than PV->NV->IV
2075 flags already correct - don't set public IOK. */
2076 DEBUG_c(PerlIO_printf(Perl_debug_log,
2077 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2082 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2083 but the cast (NV)IV_MIN rounds to a the value less (more
2084 negative) than IV_MIN which happens to be equal to SvNVX ??
2085 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2086 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2087 (NV)UVX == NVX are both true, but the values differ. :-(
2088 Hopefully for 2s complement IV_MIN is something like
2089 0x8000000000000000 which will be exact. NWC */
2092 SvUV_set(sv, U_V(SvNVX(sv)));
2094 (SvNVX(sv) == (NV) SvUVX(sv))
2095 #ifndef NV_PRESERVES_UV
2096 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2097 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2098 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2099 /* Don't flag it as "accurately an integer" if the number
2100 came from a (by definition imprecise) NV operation, and
2101 we're outside the range of NV integer precision */
2107 DEBUG_c(PerlIO_printf(Perl_debug_log,
2108 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2114 else if (SvPOKp(sv)) {
2116 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2117 /* We want to avoid a possible problem when we cache an IV/ a UV which
2118 may be later translated to an NV, and the resulting NV is not
2119 the same as the direct translation of the initial string
2120 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2121 be careful to ensure that the value with the .456 is around if the
2122 NV value is requested in the future).
2124 This means that if we cache such an IV/a UV, we need to cache the
2125 NV as well. Moreover, we trade speed for space, and do not
2126 cache the NV if we are sure it's not needed.
2129 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2130 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2131 == IS_NUMBER_IN_UV) {
2132 /* It's definitely an integer, only upgrade to PVIV */
2133 if (SvTYPE(sv) < SVt_PVIV)
2134 sv_upgrade(sv, SVt_PVIV);
2136 } else if (SvTYPE(sv) < SVt_PVNV)
2137 sv_upgrade(sv, SVt_PVNV);
2139 /* If NVs preserve UVs then we only use the UV value if we know that
2140 we aren't going to call atof() below. If NVs don't preserve UVs
2141 then the value returned may have more precision than atof() will
2142 return, even though value isn't perfectly accurate. */
2143 if ((numtype & (IS_NUMBER_IN_UV
2144 #ifdef NV_PRESERVES_UV
2147 )) == IS_NUMBER_IN_UV) {
2148 /* This won't turn off the public IOK flag if it was set above */
2149 (void)SvIOKp_on(sv);
2151 if (!(numtype & IS_NUMBER_NEG)) {
2153 if (value <= (UV)IV_MAX) {
2154 SvIV_set(sv, (IV)value);
2156 /* it didn't overflow, and it was positive. */
2157 SvUV_set(sv, value);
2161 /* 2s complement assumption */
2162 if (value <= (UV)IV_MIN) {
2163 SvIV_set(sv, -(IV)value);
2165 /* Too negative for an IV. This is a double upgrade, but
2166 I'm assuming it will be rare. */
2167 if (SvTYPE(sv) < SVt_PVNV)
2168 sv_upgrade(sv, SVt_PVNV);
2172 SvNV_set(sv, -(NV)value);
2173 SvIV_set(sv, IV_MIN);
2177 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2178 will be in the previous block to set the IV slot, and the next
2179 block to set the NV slot. So no else here. */
2181 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2182 != IS_NUMBER_IN_UV) {
2183 /* It wasn't an (integer that doesn't overflow the UV). */
2184 SvNV_set(sv, Atof(SvPVX_const(sv)));
2186 if (! numtype && ckWARN(WARN_NUMERIC))
2189 #if defined(USE_LONG_DOUBLE)
2190 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2191 PTR2UV(sv), SvNVX(sv)));
2193 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2194 PTR2UV(sv), SvNVX(sv)));
2197 #ifdef NV_PRESERVES_UV
2198 (void)SvIOKp_on(sv);
2200 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2201 SvIV_set(sv, I_V(SvNVX(sv)));
2202 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2205 NOOP; /* Integer is imprecise. NOK, IOKp */
2207 /* UV will not work better than IV */
2209 if (SvNVX(sv) > (NV)UV_MAX) {
2211 /* Integer is inaccurate. NOK, IOKp, is UV */
2212 SvUV_set(sv, UV_MAX);
2214 SvUV_set(sv, U_V(SvNVX(sv)));
2215 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2216 NV preservse UV so can do correct comparison. */
2217 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2220 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2225 #else /* NV_PRESERVES_UV */
2226 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2227 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2228 /* The IV/UV slot will have been set from value returned by
2229 grok_number above. The NV slot has just been set using
2232 assert (SvIOKp(sv));
2234 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2235 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2236 /* Small enough to preserve all bits. */
2237 (void)SvIOKp_on(sv);
2239 SvIV_set(sv, I_V(SvNVX(sv)));
2240 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2242 /* Assumption: first non-preserved integer is < IV_MAX,
2243 this NV is in the preserved range, therefore: */
2244 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2246 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);
2250 0 0 already failed to read UV.
2251 0 1 already failed to read UV.
2252 1 0 you won't get here in this case. IV/UV
2253 slot set, public IOK, Atof() unneeded.
2254 1 1 already read UV.
2255 so there's no point in sv_2iuv_non_preserve() attempting
2256 to use atol, strtol, strtoul etc. */
2258 sv_2iuv_non_preserve (sv, numtype);
2260 sv_2iuv_non_preserve (sv);
2264 #endif /* NV_PRESERVES_UV */
2265 /* It might be more code efficient to go through the entire logic above
2266 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2267 gets complex and potentially buggy, so more programmer efficient
2268 to do it this way, by turning off the public flags: */
2270 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2274 if (isGV_with_GP(sv))
2275 return glob_2number(MUTABLE_GV(sv));
2277 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2279 if (SvTYPE(sv) < SVt_IV)
2280 /* Typically the caller expects that sv_any is not NULL now. */
2281 sv_upgrade(sv, SVt_IV);
2282 /* Return 0 from the caller. */
2289 =for apidoc sv_2iv_flags
2291 Return the integer value of an SV, doing any necessary string
2292 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2293 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2299 Perl_sv_2iv_flags(pTHX_ SV *const sv, const I32 flags)
2306 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2307 && SvTYPE(sv) != SVt_PVFM);
2309 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2315 if (flags & SV_SKIP_OVERLOAD)
2317 tmpstr = AMG_CALLunary(sv, numer_amg);
2318 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2319 return SvIV(tmpstr);
2322 return PTR2IV(SvRV(sv));
2325 if (SvVALID(sv) || isREGEXP(sv)) {
2326 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2327 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2328 In practice they are extremely unlikely to actually get anywhere
2329 accessible by user Perl code - the only way that I'm aware of is when
2330 a constant subroutine which is used as the second argument to index.
2332 Regexps have no SvIVX and SvNVX fields.
2334 assert(isREGEXP(sv) || SvPOKp(sv));
2337 const char * const ptr =
2338 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2340 = grok_number(ptr, SvCUR(sv), &value);
2342 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2343 == IS_NUMBER_IN_UV) {
2344 /* It's definitely an integer */
2345 if (numtype & IS_NUMBER_NEG) {
2346 if (value < (UV)IV_MIN)
2349 if (value < (UV)IV_MAX)
2354 if (ckWARN(WARN_NUMERIC))
2357 return I_V(Atof(ptr));
2361 if (SvTHINKFIRST(sv)) {
2362 #ifdef PERL_OLD_COPY_ON_WRITE
2364 sv_force_normal_flags(sv, 0);
2367 if (SvREADONLY(sv) && !SvOK(sv)) {
2368 if (ckWARN(WARN_UNINITIALIZED))
2375 if (S_sv_2iuv_common(aTHX_ sv))
2379 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2380 PTR2UV(sv),SvIVX(sv)));
2381 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2385 =for apidoc sv_2uv_flags
2387 Return the unsigned integer value of an SV, doing any necessary string
2388 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2389 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2395 Perl_sv_2uv_flags(pTHX_ SV *const sv, const I32 flags)
2402 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2408 if (flags & SV_SKIP_OVERLOAD)
2410 tmpstr = AMG_CALLunary(sv, numer_amg);
2411 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2412 return SvUV(tmpstr);
2415 return PTR2UV(SvRV(sv));
2418 if (SvVALID(sv) || isREGEXP(sv)) {
2419 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2420 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2421 Regexps have no SvIVX and SvNVX fields. */
2422 assert(isREGEXP(sv) || SvPOKp(sv));
2425 const char * const ptr =
2426 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2428 = grok_number(ptr, SvCUR(sv), &value);
2430 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2431 == IS_NUMBER_IN_UV) {
2432 /* It's definitely an integer */
2433 if (!(numtype & IS_NUMBER_NEG))
2437 if (ckWARN(WARN_NUMERIC))
2440 return U_V(Atof(ptr));
2444 if (SvTHINKFIRST(sv)) {
2445 #ifdef PERL_OLD_COPY_ON_WRITE
2447 sv_force_normal_flags(sv, 0);
2450 if (SvREADONLY(sv) && !SvOK(sv)) {
2451 if (ckWARN(WARN_UNINITIALIZED))
2458 if (S_sv_2iuv_common(aTHX_ sv))
2462 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2463 PTR2UV(sv),SvUVX(sv)));
2464 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2468 =for apidoc sv_2nv_flags
2470 Return the num value of an SV, doing any necessary string or integer
2471 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2472 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2478 Perl_sv_2nv_flags(pTHX_ SV *const sv, const I32 flags)
2483 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2484 && SvTYPE(sv) != SVt_PVFM);
2485 if (SvGMAGICAL(sv) || SvVALID(sv) || isREGEXP(sv)) {
2486 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2487 the same flag bit as SVf_IVisUV, so must not let them cache NVs.
2488 Regexps have no SvIVX and SvNVX fields. */
2490 if (flags & SV_GMAGIC)
2494 if (SvPOKp(sv) && !SvIOKp(sv)) {
2495 ptr = SvPVX_const(sv);
2497 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2498 !grok_number(ptr, SvCUR(sv), NULL))
2504 return (NV)SvUVX(sv);
2506 return (NV)SvIVX(sv);
2512 ptr = RX_WRAPPED((REGEXP *)sv);
2515 assert(SvTYPE(sv) >= SVt_PVMG);
2516 /* This falls through to the report_uninit near the end of the
2518 } else if (SvTHINKFIRST(sv)) {
2523 if (flags & SV_SKIP_OVERLOAD)
2525 tmpstr = AMG_CALLunary(sv, numer_amg);
2526 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2527 return SvNV(tmpstr);
2530 return PTR2NV(SvRV(sv));
2532 #ifdef PERL_OLD_COPY_ON_WRITE
2534 sv_force_normal_flags(sv, 0);
2537 if (SvREADONLY(sv) && !SvOK(sv)) {
2538 if (ckWARN(WARN_UNINITIALIZED))
2543 if (SvTYPE(sv) < SVt_NV) {
2544 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2545 sv_upgrade(sv, SVt_NV);
2546 #ifdef USE_LONG_DOUBLE
2548 STORE_NUMERIC_LOCAL_SET_STANDARD();
2549 PerlIO_printf(Perl_debug_log,
2550 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2551 PTR2UV(sv), SvNVX(sv));
2552 RESTORE_NUMERIC_LOCAL();
2556 STORE_NUMERIC_LOCAL_SET_STANDARD();
2557 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2558 PTR2UV(sv), SvNVX(sv));
2559 RESTORE_NUMERIC_LOCAL();
2563 else if (SvTYPE(sv) < SVt_PVNV)
2564 sv_upgrade(sv, SVt_PVNV);
2569 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2570 #ifdef NV_PRESERVES_UV
2576 /* Only set the public NV OK flag if this NV preserves the IV */
2577 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2579 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2580 : (SvIVX(sv) == I_V(SvNVX(sv))))
2586 else if (SvPOKp(sv)) {
2588 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2589 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2591 #ifdef NV_PRESERVES_UV
2592 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2593 == IS_NUMBER_IN_UV) {
2594 /* It's definitely an integer */
2595 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2597 SvNV_set(sv, Atof(SvPVX_const(sv)));
2603 SvNV_set(sv, Atof(SvPVX_const(sv)));
2604 /* Only set the public NV OK flag if this NV preserves the value in
2605 the PV at least as well as an IV/UV would.
2606 Not sure how to do this 100% reliably. */
2607 /* if that shift count is out of range then Configure's test is
2608 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2610 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2611 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2612 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2613 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2614 /* Can't use strtol etc to convert this string, so don't try.
2615 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2618 /* value has been set. It may not be precise. */
2619 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2620 /* 2s complement assumption for (UV)IV_MIN */
2621 SvNOK_on(sv); /* Integer is too negative. */
2626 if (numtype & IS_NUMBER_NEG) {
2627 SvIV_set(sv, -(IV)value);
2628 } else if (value <= (UV)IV_MAX) {
2629 SvIV_set(sv, (IV)value);
2631 SvUV_set(sv, value);
2635 if (numtype & IS_NUMBER_NOT_INT) {
2636 /* I believe that even if the original PV had decimals,
2637 they are lost beyond the limit of the FP precision.
2638 However, neither is canonical, so both only get p
2639 flags. NWC, 2000/11/25 */
2640 /* Both already have p flags, so do nothing */
2642 const NV nv = SvNVX(sv);
2643 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2644 if (SvIVX(sv) == I_V(nv)) {
2647 /* It had no "." so it must be integer. */
2651 /* between IV_MAX and NV(UV_MAX).
2652 Could be slightly > UV_MAX */
2654 if (numtype & IS_NUMBER_NOT_INT) {
2655 /* UV and NV both imprecise. */
2657 const UV nv_as_uv = U_V(nv);
2659 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2668 /* It might be more code efficient to go through the entire logic above
2669 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2670 gets complex and potentially buggy, so more programmer efficient
2671 to do it this way, by turning off the public flags: */
2673 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2674 #endif /* NV_PRESERVES_UV */
2677 if (isGV_with_GP(sv)) {
2678 glob_2number(MUTABLE_GV(sv));
2682 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2684 assert (SvTYPE(sv) >= SVt_NV);
2685 /* Typically the caller expects that sv_any is not NULL now. */
2686 /* XXX Ilya implies that this is a bug in callers that assume this
2687 and ideally should be fixed. */
2690 #if defined(USE_LONG_DOUBLE)
2692 STORE_NUMERIC_LOCAL_SET_STANDARD();
2693 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2694 PTR2UV(sv), SvNVX(sv));
2695 RESTORE_NUMERIC_LOCAL();
2699 STORE_NUMERIC_LOCAL_SET_STANDARD();
2700 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2701 PTR2UV(sv), SvNVX(sv));
2702 RESTORE_NUMERIC_LOCAL();
2711 Return an SV with the numeric value of the source SV, doing any necessary
2712 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2713 access this function.
2719 Perl_sv_2num(pTHX_ SV *const sv)
2721 PERL_ARGS_ASSERT_SV_2NUM;
2726 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2727 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2728 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2729 return sv_2num(tmpsv);
2731 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2734 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2735 * UV as a string towards the end of buf, and return pointers to start and
2738 * We assume that buf is at least TYPE_CHARS(UV) long.
2742 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2744 char *ptr = buf + TYPE_CHARS(UV);
2745 char * const ebuf = ptr;
2748 PERL_ARGS_ASSERT_UIV_2BUF;
2760 *--ptr = '0' + (char)(uv % 10);
2769 =for apidoc sv_2pv_flags
2771 Returns a pointer to the string value of an SV, and sets *lp to its length.
2772 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2773 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2774 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2780 Perl_sv_2pv_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
2790 assert (SvTYPE(sv) != SVt_PVAV && SvTYPE(sv) != SVt_PVHV
2791 && SvTYPE(sv) != SVt_PVFM);
2792 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2797 if (flags & SV_SKIP_OVERLOAD)
2799 tmpstr = AMG_CALLunary(sv, string_amg);
2800 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2801 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2803 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2807 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2808 if (flags & SV_CONST_RETURN) {
2809 pv = (char *) SvPVX_const(tmpstr);
2811 pv = (flags & SV_MUTABLE_RETURN)
2812 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2815 *lp = SvCUR(tmpstr);
2817 pv = sv_2pv_flags(tmpstr, lp, flags);
2830 SV *const referent = SvRV(sv);
2834 retval = buffer = savepvn("NULLREF", len);
2835 } else if (SvTYPE(referent) == SVt_REGEXP &&
2836 (!(PL_curcop->cop_hints & HINT_NO_AMAGIC) ||
2837 amagic_is_enabled(string_amg))) {
2838 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2842 /* If the regex is UTF-8 we want the containing scalar to
2843 have an UTF-8 flag too */
2850 *lp = RX_WRAPLEN(re);
2852 return RX_WRAPPED(re);
2854 const char *const typestr = sv_reftype(referent, 0);
2855 const STRLEN typelen = strlen(typestr);
2856 UV addr = PTR2UV(referent);
2857 const char *stashname = NULL;
2858 STRLEN stashnamelen = 0; /* hush, gcc */
2859 const char *buffer_end;
2861 if (SvOBJECT(referent)) {
2862 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2865 stashname = HEK_KEY(name);
2866 stashnamelen = HEK_LEN(name);
2868 if (HEK_UTF8(name)) {
2874 stashname = "__ANON__";
2877 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2878 + 2 * sizeof(UV) + 2 /* )\0 */;
2880 len = typelen + 3 /* (0x */
2881 + 2 * sizeof(UV) + 2 /* )\0 */;
2884 Newx(buffer, len, char);
2885 buffer_end = retval = buffer + len;
2887 /* Working backwards */
2891 *--retval = PL_hexdigit[addr & 15];
2892 } while (addr >>= 4);
2898 memcpy(retval, typestr, typelen);
2902 retval -= stashnamelen;
2903 memcpy(retval, stashname, stashnamelen);
2905 /* retval may not necessarily have reached the start of the
2907 assert (retval >= buffer);
2909 len = buffer_end - retval - 1; /* -1 for that \0 */
2921 if (flags & SV_MUTABLE_RETURN)
2922 return SvPVX_mutable(sv);
2923 if (flags & SV_CONST_RETURN)
2924 return (char *)SvPVX_const(sv);
2929 /* I'm assuming that if both IV and NV are equally valid then
2930 converting the IV is going to be more efficient */
2931 const U32 isUIOK = SvIsUV(sv);
2932 char buf[TYPE_CHARS(UV)];
2936 if (SvTYPE(sv) < SVt_PVIV)
2937 sv_upgrade(sv, SVt_PVIV);
2938 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2940 /* inlined from sv_setpvn */
2941 s = SvGROW_mutable(sv, len + 1);
2942 Move(ptr, s, len, char);
2947 else if (SvNOK(sv)) {
2948 if (SvTYPE(sv) < SVt_PVNV)
2949 sv_upgrade(sv, SVt_PVNV);
2950 if (SvNVX(sv) == 0.0) {
2951 s = SvGROW_mutable(sv, 2);
2956 /* The +20 is pure guesswork. Configure test needed. --jhi */
2957 s = SvGROW_mutable(sv, NV_DIG + 20);
2958 /* some Xenix systems wipe out errno here */
2960 #ifndef USE_LOCALE_NUMERIC
2961 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2964 /* Gconvert always uses the current locale. That's the right thing
2965 * to do if we're supposed to be using locales. But otherwise, we
2966 * want the result to be based on the C locale, so we need to
2967 * change to the C locale during the Gconvert and then change back.
2968 * But if we're already in the C locale (PL_numeric_standard is
2969 * TRUE in that case), no need to do any changing */
2970 if (PL_numeric_standard || IN_SOME_LOCALE_FORM_RUNTIME) {
2971 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2973 /* If the radix character is UTF-8, and actually is in the
2974 * output, turn on the UTF-8 flag for the scalar */
2975 if (! PL_numeric_standard
2976 && PL_numeric_radix_sv && SvUTF8(PL_numeric_radix_sv)
2977 && instr(s, SvPVX_const(PL_numeric_radix_sv)))
2983 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2984 setlocale(LC_NUMERIC, "C");
2985 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2986 setlocale(LC_NUMERIC, loc);
2991 /* We don't call SvPOK_on(), because it may come to pass that the
2992 * locale changes so that the stringification we just did is no
2993 * longer correct. We will have to re-stringify every time it is
3004 else if (isGV_with_GP(sv)) {
3005 GV *const gv = MUTABLE_GV(sv);
3006 SV *const buffer = sv_newmortal();
3008 gv_efullname3(buffer, gv, "*");
3010 assert(SvPOK(buffer));
3014 *lp = SvCUR(buffer);
3015 return SvPVX(buffer);
3017 else if (isREGEXP(sv)) {
3018 if (lp) *lp = RX_WRAPLEN((REGEXP *)sv);
3019 return RX_WRAPPED((REGEXP *)sv);
3024 if (flags & SV_UNDEF_RETURNS_NULL)
3026 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
3028 /* Typically the caller expects that sv_any is not NULL now. */
3029 if (!SvREADONLY(sv) && SvTYPE(sv) < SVt_PV)
3030 sv_upgrade(sv, SVt_PV);
3035 const STRLEN len = s - SvPVX_const(sv);
3040 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3041 PTR2UV(sv),SvPVX_const(sv)));
3042 if (flags & SV_CONST_RETURN)
3043 return (char *)SvPVX_const(sv);
3044 if (flags & SV_MUTABLE_RETURN)
3045 return SvPVX_mutable(sv);
3050 =for apidoc sv_copypv
3052 Copies a stringified representation of the source SV into the
3053 destination SV. Automatically performs any necessary mg_get and
3054 coercion of numeric values into strings. Guaranteed to preserve
3055 UTF8 flag even from overloaded objects. Similar in nature to
3056 sv_2pv[_flags] but operates directly on an SV instead of just the
3057 string. Mostly uses sv_2pv_flags to do its work, except when that
3058 would lose the UTF-8'ness of the PV.
3060 =for apidoc sv_copypv_nomg
3062 Like sv_copypv, but doesn't invoke get magic first.
3064 =for apidoc sv_copypv_flags
3066 Implementation of sv_copypv and sv_copypv_nomg. Calls get magic iff flags
3073 Perl_sv_copypv(pTHX_ SV *const dsv, SV *const ssv)
3075 PERL_ARGS_ASSERT_SV_COPYPV;
3077 sv_copypv_flags(dsv, ssv, 0);
3081 Perl_sv_copypv_flags(pTHX_ SV *const dsv, SV *const ssv, const I32 flags)
3086 PERL_ARGS_ASSERT_SV_COPYPV_FLAGS;
3088 if ((flags & SV_GMAGIC) && SvGMAGICAL(ssv))
3090 s = SvPV_nomg_const(ssv,len);
3091 sv_setpvn(dsv,s,len);
3099 =for apidoc sv_2pvbyte
3101 Return a pointer to the byte-encoded representation of the SV, and set *lp
3102 to its length. May cause the SV to be downgraded from UTF-8 as a
3105 Usually accessed via the C<SvPVbyte> macro.
3111 Perl_sv_2pvbyte(pTHX_ SV *sv, STRLEN *const lp)
3113 PERL_ARGS_ASSERT_SV_2PVBYTE;
3116 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3117 || isGV_with_GP(sv) || SvROK(sv)) {
3118 SV *sv2 = sv_newmortal();
3119 sv_copypv_nomg(sv2,sv);
3122 sv_utf8_downgrade(sv,0);
3123 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3127 =for apidoc sv_2pvutf8
3129 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3130 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3132 Usually accessed via the C<SvPVutf8> macro.
3138 Perl_sv_2pvutf8(pTHX_ SV *sv, STRLEN *const lp)
3140 PERL_ARGS_ASSERT_SV_2PVUTF8;
3142 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3143 || isGV_with_GP(sv) || SvROK(sv))
3144 sv = sv_mortalcopy(sv);
3147 sv_utf8_upgrade_nomg(sv);
3148 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3153 =for apidoc sv_2bool
3155 This macro is only used by sv_true() or its macro equivalent, and only if
3156 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3157 It calls sv_2bool_flags with the SV_GMAGIC flag.
3159 =for apidoc sv_2bool_flags
3161 This function is only used by sv_true() and friends, and only if
3162 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3163 contain SV_GMAGIC, then it does an mg_get() first.
3170 Perl_sv_2bool_flags(pTHX_ SV *const sv, const I32 flags)
3174 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3176 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3182 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3183 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3184 return cBOOL(SvTRUE(tmpsv));
3186 return SvRV(sv) != 0;
3190 RX_WRAPLEN(sv) > 1 || (RX_WRAPLEN(sv) && *RX_WRAPPED(sv) != '0');
3191 return SvTRUE_common(sv, isGV_with_GP(sv) ? 1 : 0);
3195 =for apidoc sv_utf8_upgrade
3197 Converts the PV of an SV to its UTF-8-encoded form.
3198 Forces the SV to string form if it is not already.
3199 Will C<mg_get> on C<sv> if appropriate.
3200 Always sets the SvUTF8 flag to avoid future validity checks even
3201 if the whole string is the same in UTF-8 as not.
3202 Returns the number of bytes in the converted string
3204 This is not a general purpose byte encoding to Unicode interface:
3205 use the Encode extension for that.
3207 =for apidoc sv_utf8_upgrade_nomg
3209 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3211 =for apidoc sv_utf8_upgrade_flags
3213 Converts the PV of an SV to its UTF-8-encoded form.
3214 Forces the SV to string form if it is not already.
3215 Always sets the SvUTF8 flag to avoid future validity checks even
3216 if all the bytes are invariant in UTF-8.
3217 If C<flags> has C<SV_GMAGIC> bit set,
3218 will C<mg_get> on C<sv> if appropriate, else not.
3219 Returns the number of bytes in the converted string
3220 C<sv_utf8_upgrade> and
3221 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3223 This is not a general purpose byte encoding to Unicode interface:
3224 use the Encode extension for that.
3228 The grow version is currently not externally documented. It adds a parameter,
3229 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3230 have free after it upon return. This allows the caller to reserve extra space
3231 that it intends to fill, to avoid extra grows.
3233 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3234 which can be used to tell this function to not first check to see if there are
3235 any characters that are different in UTF-8 (variant characters) which would
3236 force it to allocate a new string to sv, but to assume there are. Typically
3237 this flag is used by a routine that has already parsed the string to find that
3238 there are such characters, and passes this information on so that the work
3239 doesn't have to be repeated.
3241 (One might think that the calling routine could pass in the position of the
3242 first such variant, so it wouldn't have to be found again. But that is not the
3243 case, because typically when the caller is likely to use this flag, it won't be
3244 calling this routine unless it finds something that won't fit into a byte.
3245 Otherwise it tries to not upgrade and just use bytes. But some things that
3246 do fit into a byte are variants in utf8, and the caller may not have been
3247 keeping track of these.)
3249 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3250 isn't guaranteed due to having other routines do the work in some input cases,
3251 or if the input is already flagged as being in utf8.
3253 The speed of this could perhaps be improved for many cases if someone wanted to
3254 write a fast function that counts the number of variant characters in a string,
3255 especially if it could return the position of the first one.
3260 Perl_sv_utf8_upgrade_flags_grow(pTHX_ SV *const sv, const I32 flags, STRLEN extra)
3264 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3266 if (sv == &PL_sv_undef)
3268 if (!SvPOK_nog(sv)) {
3270 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3271 (void) sv_2pv_flags(sv,&len, flags);
3273 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3277 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3282 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3287 S_sv_uncow(aTHX_ sv, 0);
3290 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3291 sv_recode_to_utf8(sv, PL_encoding);
3292 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3296 if (SvCUR(sv) == 0) {
3297 if (extra) SvGROW(sv, extra);
3298 } else { /* Assume Latin-1/EBCDIC */
3299 /* This function could be much more efficient if we
3300 * had a FLAG in SVs to signal if there are any variant
3301 * chars in the PV. Given that there isn't such a flag
3302 * make the loop as fast as possible (although there are certainly ways
3303 * to speed this up, eg. through vectorization) */
3304 U8 * s = (U8 *) SvPVX_const(sv);
3305 U8 * e = (U8 *) SvEND(sv);
3307 STRLEN two_byte_count = 0;
3309 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3311 /* See if really will need to convert to utf8. We mustn't rely on our
3312 * incoming SV being well formed and having a trailing '\0', as certain
3313 * code in pp_formline can send us partially built SVs. */
3317 if (NATIVE_IS_INVARIANT(ch)) continue;
3319 t--; /* t already incremented; re-point to first variant */
3324 /* utf8 conversion not needed because all are invariants. Mark as
3325 * UTF-8 even if no variant - saves scanning loop */
3327 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3332 /* Here, the string should be converted to utf8, either because of an
3333 * input flag (two_byte_count = 0), or because a character that
3334 * requires 2 bytes was found (two_byte_count = 1). t points either to
3335 * the beginning of the string (if we didn't examine anything), or to
3336 * the first variant. In either case, everything from s to t - 1 will
3337 * occupy only 1 byte each on output.
3339 * There are two main ways to convert. One is to create a new string
3340 * and go through the input starting from the beginning, appending each
3341 * converted value onto the new string as we go along. It's probably
3342 * best to allocate enough space in the string for the worst possible
3343 * case rather than possibly running out of space and having to
3344 * reallocate and then copy what we've done so far. Since everything
3345 * from s to t - 1 is invariant, the destination can be initialized
3346 * with these using a fast memory copy
3348 * The other way is to figure out exactly how big the string should be
3349 * by parsing the entire input. Then you don't have to make it big
3350 * enough to handle the worst possible case, and more importantly, if
3351 * the string you already have is large enough, you don't have to
3352 * allocate a new string, you can copy the last character in the input
3353 * string to the final position(s) that will be occupied by the
3354 * converted string and go backwards, stopping at t, since everything
3355 * before that is invariant.
3357 * There are advantages and disadvantages to each method.
3359 * In the first method, we can allocate a new string, do the memory
3360 * copy from the s to t - 1, and then proceed through the rest of the
3361 * string byte-by-byte.
3363 * In the second method, we proceed through the rest of the input
3364 * string just calculating how big the converted string will be. Then
3365 * there are two cases:
3366 * 1) if the string has enough extra space to handle the converted
3367 * value. We go backwards through the string, converting until we
3368 * get to the position we are at now, and then stop. If this
3369 * position is far enough along in the string, this method is
3370 * faster than the other method. If the memory copy were the same
3371 * speed as the byte-by-byte loop, that position would be about
3372 * half-way, as at the half-way mark, parsing to the end and back
3373 * is one complete string's parse, the same amount as starting
3374 * over and going all the way through. Actually, it would be
3375 * somewhat less than half-way, as it's faster to just count bytes
3376 * than to also copy, and we don't have the overhead of allocating
3377 * a new string, changing the scalar to use it, and freeing the
3378 * existing one. But if the memory copy is fast, the break-even
3379 * point is somewhere after half way. The counting loop could be
3380 * sped up by vectorization, etc, to move the break-even point
3381 * further towards the beginning.
3382 * 2) if the string doesn't have enough space to handle the converted
3383 * value. A new string will have to be allocated, and one might
3384 * as well, given that, start from the beginning doing the first
3385 * method. We've spent extra time parsing the string and in
3386 * exchange all we've gotten is that we know precisely how big to
3387 * make the new one. Perl is more optimized for time than space,
3388 * so this case is a loser.
3389 * So what I've decided to do is not use the 2nd method unless it is
3390 * guaranteed that a new string won't have to be allocated, assuming
3391 * the worst case. I also decided not to put any more conditions on it
3392 * than this, for now. It seems likely that, since the worst case is
3393 * twice as big as the unknown portion of the string (plus 1), we won't
3394 * be guaranteed enough space, causing us to go to the first method,
3395 * unless the string is short, or the first variant character is near
3396 * the end of it. In either of these cases, it seems best to use the
3397 * 2nd method. The only circumstance I can think of where this would
3398 * be really slower is if the string had once had much more data in it
3399 * than it does now, but there is still a substantial amount in it */
3402 STRLEN invariant_head = t - s;
3403 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3404 if (SvLEN(sv) < size) {
3406 /* Here, have decided to allocate a new string */
3411 Newx(dst, size, U8);
3413 /* If no known invariants at the beginning of the input string,
3414 * set so starts from there. Otherwise, can use memory copy to
3415 * get up to where we are now, and then start from here */
3417 if (invariant_head <= 0) {
3420 Copy(s, dst, invariant_head, char);
3421 d = dst + invariant_head;
3425 append_utf8_from_native_byte(*t, &d);
3429 SvPV_free(sv); /* No longer using pre-existing string */
3430 SvPV_set(sv, (char*)dst);
3431 SvCUR_set(sv, d - dst);
3432 SvLEN_set(sv, size);
3435 /* Here, have decided to get the exact size of the string.
3436 * Currently this happens only when we know that there is
3437 * guaranteed enough space to fit the converted string, so
3438 * don't have to worry about growing. If two_byte_count is 0,
3439 * then t points to the first byte of the string which hasn't
3440 * been examined yet. Otherwise two_byte_count is 1, and t
3441 * points to the first byte in the string that will expand to
3442 * two. Depending on this, start examining at t or 1 after t.
3445 U8 *d = t + two_byte_count;
3448 /* Count up the remaining bytes that expand to two */
3451 const U8 chr = *d++;
3452 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3455 /* The string will expand by just the number of bytes that
3456 * occupy two positions. But we are one afterwards because of
3457 * the increment just above. This is the place to put the
3458 * trailing NUL, and to set the length before we decrement */
3460 d += two_byte_count;
3461 SvCUR_set(sv, d - s);
3465 /* Having decremented d, it points to the position to put the
3466 * very last byte of the expanded string. Go backwards through
3467 * the string, copying and expanding as we go, stopping when we
3468 * get to the part that is invariant the rest of the way down */
3472 if (NATIVE_IS_INVARIANT(*e)) {
3475 *d-- = UTF8_EIGHT_BIT_LO(*e);
3476 *d-- = UTF8_EIGHT_BIT_HI(*e);
3482 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3483 /* Update pos. We do it at the end rather than during
3484 * the upgrade, to avoid slowing down the common case
3485 * (upgrade without pos).
3486 * pos can be stored as either bytes or characters. Since
3487 * this was previously a byte string we can just turn off
3488 * the bytes flag. */
3489 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3491 mg->mg_flags &= ~MGf_BYTES;
3493 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3494 magic_setutf8(sv,mg); /* clear UTF8 cache */
3499 /* Mark as UTF-8 even if no variant - saves scanning loop */
3505 =for apidoc sv_utf8_downgrade
3507 Attempts to convert the PV of an SV from characters to bytes.
3508 If the PV contains a character that cannot fit
3509 in a byte, this conversion will fail;
3510 in this case, either returns false or, if C<fail_ok> is not
3513 This is not a general purpose Unicode to byte encoding interface:
3514 use the Encode extension for that.
3520 Perl_sv_utf8_downgrade(pTHX_ SV *const sv, const bool fail_ok)
3524 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3526 if (SvPOKp(sv) && SvUTF8(sv)) {
3530 int mg_flags = SV_GMAGIC;
3533 S_sv_uncow(aTHX_ sv, 0);
3535 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3537 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3538 if (mg && mg->mg_len > 0 && mg->mg_flags & MGf_BYTES) {
3539 mg->mg_len = sv_pos_b2u_flags(sv, mg->mg_len,
3540 SV_GMAGIC|SV_CONST_RETURN);
3541 mg_flags = 0; /* sv_pos_b2u does get magic */
3543 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3544 magic_setutf8(sv,mg); /* clear UTF8 cache */
3547 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3549 if (!utf8_to_bytes(s, &len)) {
3554 Perl_croak(aTHX_ "Wide character in %s",
3557 Perl_croak(aTHX_ "Wide character");
3568 =for apidoc sv_utf8_encode
3570 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3571 flag off so that it looks like octets again.
3577 Perl_sv_utf8_encode(pTHX_ SV *const sv)
3579 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3581 if (SvREADONLY(sv)) {
3582 sv_force_normal_flags(sv, 0);
3584 (void) sv_utf8_upgrade(sv);
3589 =for apidoc sv_utf8_decode
3591 If the PV of the SV is an octet sequence in UTF-8
3592 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3593 so that it looks like a character. If the PV contains only single-byte
3594 characters, the C<SvUTF8> flag stays off.
3595 Scans PV for validity and returns false if the PV is invalid UTF-8.
3601 Perl_sv_utf8_decode(pTHX_ SV *const sv)
3603 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3606 const U8 *start, *c;
3609 /* The octets may have got themselves encoded - get them back as
3612 if (!sv_utf8_downgrade(sv, TRUE))
3615 /* it is actually just a matter of turning the utf8 flag on, but
3616 * we want to make sure everything inside is valid utf8 first.
3618 c = start = (const U8 *) SvPVX_const(sv);
3619 if (!is_utf8_string(c, SvCUR(sv)))
3621 e = (const U8 *) SvEND(sv);
3624 if (!UTF8_IS_INVARIANT(ch)) {
3629 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3630 /* XXX Is this dead code? XS_utf8_decode calls SvSETMAGIC
3631 after this, clearing pos. Does anything on CPAN
3633 /* adjust pos to the start of a UTF8 char sequence */
3634 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3636 I32 pos = mg->mg_len;
3638 for (c = start + pos; c > start; c--) {
3639 if (UTF8_IS_START(*c))
3642 mg->mg_len = c - start;
3645 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3646 magic_setutf8(sv,mg); /* clear UTF8 cache */
3653 =for apidoc sv_setsv
3655 Copies the contents of the source SV C<ssv> into the destination SV
3656 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3657 function if the source SV needs to be reused. Does not handle 'set' magic.
3658 Loosely speaking, it performs a copy-by-value, obliterating any previous
3659 content of the destination.
3661 You probably want to use one of the assortment of wrappers, such as
3662 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3663 C<SvSetMagicSV_nosteal>.
3665 =for apidoc sv_setsv_flags
3667 Copies the contents of the source SV C<ssv> into the destination SV
3668 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3669 function if the source SV needs to be reused. Does not handle 'set' magic.
3670 Loosely speaking, it performs a copy-by-value, obliterating any previous
3671 content of the destination.
3672 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3673 C<ssv> if appropriate, else not. If the C<flags>
3674 parameter has the C<NOSTEAL> bit set then the
3675 buffers of temps will not be stolen. <sv_setsv>
3676 and C<sv_setsv_nomg> are implemented in terms of this function.
3678 You probably want to use one of the assortment of wrappers, such as
3679 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3680 C<SvSetMagicSV_nosteal>.
3682 This is the primary function for copying scalars, and most other
3683 copy-ish functions and macros use this underneath.
3689 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3691 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3692 HV *old_stash = NULL;
3694 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3696 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3697 const char * const name = GvNAME(sstr);
3698 const STRLEN len = GvNAMELEN(sstr);
3700 if (dtype >= SVt_PV) {
3706 SvUPGRADE(dstr, SVt_PVGV);
3707 (void)SvOK_off(dstr);
3708 /* We have to turn this on here, even though we turn it off
3709 below, as GvSTASH will fail an assertion otherwise. */
3710 isGV_with_GP_on(dstr);
3712 GvSTASH(dstr) = GvSTASH(sstr);
3714 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3715 gv_name_set(MUTABLE_GV(dstr), name, len,
3716 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3717 SvFAKE_on(dstr); /* can coerce to non-glob */
3720 if(GvGP(MUTABLE_GV(sstr))) {
3721 /* If source has method cache entry, clear it */
3723 SvREFCNT_dec(GvCV(sstr));
3724 GvCV_set(sstr, NULL);
3727 /* If source has a real method, then a method is
3730 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3736 /* If dest already had a real method, that's a change as well */
3738 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3739 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3744 /* We don't need to check the name of the destination if it was not a
3745 glob to begin with. */
3746 if(dtype == SVt_PVGV) {
3747 const char * const name = GvNAME((const GV *)dstr);
3750 /* The stash may have been detached from the symbol table, so
3752 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3756 const STRLEN len = GvNAMELEN(dstr);
3757 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3758 || (len == 1 && name[0] == ':')) {
3761 /* Set aside the old stash, so we can reset isa caches on
3763 if((old_stash = GvHV(dstr)))
3764 /* Make sure we do not lose it early. */
3765 SvREFCNT_inc_simple_void_NN(
3766 sv_2mortal((SV *)old_stash)
3772 gp_free(MUTABLE_GV(dstr));
3773 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3774 (void)SvOK_off(dstr);
3775 isGV_with_GP_on(dstr);
3776 GvINTRO_off(dstr); /* one-shot flag */
3777 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3778 if (SvTAINTED(sstr))
3780 if (GvIMPORTED(dstr) != GVf_IMPORTED
3781 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3783 GvIMPORTED_on(dstr);
3786 if(mro_changes == 2) {
3787 if (GvAV((const GV *)sstr)) {
3789 SV * const sref = (SV *)GvAV((const GV *)dstr);
3790 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3791 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3792 AV * const ary = newAV();
3793 av_push(ary, mg->mg_obj); /* takes the refcount */
3794 mg->mg_obj = (SV *)ary;
3796 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3798 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3800 mro_isa_changed_in(GvSTASH(dstr));
3802 else if(mro_changes == 3) {
3803 HV * const stash = GvHV(dstr);
3804 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3810 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3811 if (GvIO(dstr) && dtype == SVt_PVGV) {
3812 DEBUG_o(Perl_deb(aTHX_
3813 "glob_assign_glob clearing PL_stashcache\n"));
3814 /* It's a cache. It will rebuild itself quite happily.
3815 It's a lot of effort to work out exactly which key (or keys)
3816 might be invalidated by the creation of the this file handle.
3818 hv_clear(PL_stashcache);
3824 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3826 SV * const sref = SvRV(sstr);
3828 const int intro = GvINTRO(dstr);
3831 const U32 stype = SvTYPE(sref);
3833 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3836 GvINTRO_off(dstr); /* one-shot flag */
3837 GvLINE(dstr) = CopLINE(PL_curcop);
3838 GvEGV(dstr) = MUTABLE_GV(dstr);
3843 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3844 import_flag = GVf_IMPORTED_CV;
3847 location = (SV **) &GvHV(dstr);
3848 import_flag = GVf_IMPORTED_HV;
3851 location = (SV **) &GvAV(dstr);
3852 import_flag = GVf_IMPORTED_AV;
3855 location = (SV **) &GvIOp(dstr);
3858 location = (SV **) &GvFORM(dstr);
3861 location = &GvSV(dstr);
3862 import_flag = GVf_IMPORTED_SV;
3865 if (stype == SVt_PVCV) {
3866 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3867 if (GvCVGEN(dstr)) {
3868 SvREFCNT_dec(GvCV(dstr));
3869 GvCV_set(dstr, NULL);
3870 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3873 /* SAVEt_GVSLOT takes more room on the savestack and has more
3874 overhead in leave_scope than SAVEt_GENERIC_SV. But for CVs
3875 leave_scope needs access to the GV so it can reset method
3876 caches. We must use SAVEt_GVSLOT whenever the type is
3877 SVt_PVCV, even if the stash is anonymous, as the stash may
3878 gain a name somehow before leave_scope. */
3879 if (stype == SVt_PVCV) {
3880 /* There is no save_pushptrptrptr. Creating it for this
3881 one call site would be overkill. So inline the ss add
3885 SS_ADD_PTR(location);
3886 SS_ADD_PTR(SvREFCNT_inc(*location));
3887 SS_ADD_UV(SAVEt_GVSLOT);
3890 else SAVEGENERICSV(*location);
3893 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3894 CV* const cv = MUTABLE_CV(*location);
3896 if (!GvCVGEN((const GV *)dstr) &&
3897 (CvROOT(cv) || CvXSUB(cv)) &&
3898 /* redundant check that avoids creating the extra SV
3899 most of the time: */
3900 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3902 SV * const new_const_sv =
3903 CvCONST((const CV *)sref)
3904 ? cv_const_sv((const CV *)sref)
3906 report_redefined_cv(
3907 sv_2mortal(Perl_newSVpvf(aTHX_
3910 HvNAME_HEK(GvSTASH((const GV *)dstr))
3912 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3915 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3919 cv_ckproto_len_flags(cv, (const GV *)dstr,
3920 SvPOK(sref) ? CvPROTO(sref) : NULL,
3921 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3922 SvPOK(sref) ? SvUTF8(sref) : 0);
3924 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3925 GvASSUMECV_on(dstr);
3926 if(GvSTASH(dstr)) gv_method_changed(dstr); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3928 *location = SvREFCNT_inc_simple_NN(sref);
3929 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3930 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3931 GvFLAGS(dstr) |= import_flag;
3933 if (stype == SVt_PVHV) {
3934 const char * const name = GvNAME((GV*)dstr);
3935 const STRLEN len = GvNAMELEN(dstr);
3938 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3939 || (len == 1 && name[0] == ':')
3941 && (!dref || HvENAME_get(dref))
3944 (HV *)sref, (HV *)dref,
3950 stype == SVt_PVAV && sref != dref
3951 && strEQ(GvNAME((GV*)dstr), "ISA")
3952 /* The stash may have been detached from the symbol table, so
3953 check its name before doing anything. */
3954 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3957 MAGIC * const omg = dref && SvSMAGICAL(dref)
3958 ? mg_find(dref, PERL_MAGIC_isa)
3960 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3961 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3962 AV * const ary = newAV();
3963 av_push(ary, mg->mg_obj); /* takes the refcount */
3964 mg->mg_obj = (SV *)ary;
3967 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3968 SV **svp = AvARRAY((AV *)omg->mg_obj);
3969 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3973 SvREFCNT_inc_simple_NN(*svp++)
3979 SvREFCNT_inc_simple_NN(omg->mg_obj)
3983 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3988 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3990 mg = mg_find(sref, PERL_MAGIC_isa);
3992 /* Since the *ISA assignment could have affected more than
3993 one stash, don't call mro_isa_changed_in directly, but let
3994 magic_clearisa do it for us, as it already has the logic for
3995 dealing with globs vs arrays of globs. */
3997 Perl_magic_clearisa(aTHX_ NULL, mg);
3999 else if (stype == SVt_PVIO) {
4000 DEBUG_o(Perl_deb(aTHX_ "glob_assign_ref clearing PL_stashcache\n"));
4001 /* It's a cache. It will rebuild itself quite happily.
4002 It's a lot of effort to work out exactly which key (or keys)
4003 might be invalidated by the creation of the this file handle.
4005 hv_clear(PL_stashcache);
4009 if (!intro) SvREFCNT_dec(dref);
4010 if (SvTAINTED(sstr))
4015 /* Work around compiler warnings about unsigned >= THRESHOLD when thres-
4017 #if SV_COW_THRESHOLD
4018 # define GE_COW_THRESHOLD(len) ((len) >= SV_COW_THRESHOLD)
4020 # define GE_COW_THRESHOLD(len) 1
4022 #if SV_COWBUF_THRESHOLD
4023 # define GE_COWBUF_THRESHOLD(len) ((len) >= SV_COWBUF_THRESHOLD)
4025 # define GE_COWBUF_THRESHOLD(len) 1
4029 Perl_sv_setsv_flags(pTHX_ SV *dstr, SV* sstr, const I32 flags)
4036 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
4041 if (SvIS_FREED(dstr)) {
4042 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
4043 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
4045 SV_CHECK_THINKFIRST_COW_DROP(dstr);
4047 sstr = &PL_sv_undef;
4048 if (SvIS_FREED(sstr)) {
4049 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
4050 (void*)sstr, (void*)dstr);
4052 stype = SvTYPE(sstr);
4053 dtype = SvTYPE(dstr);
4055 /* There's a lot of redundancy below but we're going for speed here */
4060 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
4061 (void)SvOK_off(dstr);
4069 sv_upgrade(dstr, SVt_IV);
4073 sv_upgrade(dstr, SVt_PVIV);
4077 goto end_of_first_switch;
4079 (void)SvIOK_only(dstr);
4080 SvIV_set(dstr, SvIVX(sstr));
4083 /* SvTAINTED can only be true if the SV has taint magic, which in
4084 turn means that the SV type is PVMG (or greater). This is the
4085 case statement for SVt_IV, so this cannot be true (whatever gcov
4087 assert(!SvTAINTED(sstr));
4092 if (dtype < SVt_PV && dtype != SVt_IV)
4093 sv_upgrade(dstr, SVt_IV);
4101 sv_upgrade(dstr, SVt_NV);
4105 sv_upgrade(dstr, SVt_PVNV);
4109 goto end_of_first_switch;
4111 SvNV_set(dstr, SvNVX(sstr));
4112 (void)SvNOK_only(dstr);
4113 /* SvTAINTED can only be true if the SV has taint magic, which in
4114 turn means that the SV type is PVMG (or greater). This is the
4115 case statement for SVt_NV, so this cannot be true (whatever gcov
4117 assert(!SvTAINTED(sstr));
4124 sv_upgrade(dstr, SVt_PV);
4127 if (dtype < SVt_PVIV)
4128 sv_upgrade(dstr, SVt_PVIV);
4131 if (dtype < SVt_PVNV)
4132 sv_upgrade(dstr, SVt_PVNV);
4136 const char * const type = sv_reftype(sstr,0);
4138 /* diag_listed_as: Bizarre copy of %s */
4139 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4141 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4147 if (dtype < SVt_REGEXP)
4149 if (dtype >= SVt_PV) {
4155 sv_upgrade(dstr, SVt_REGEXP);
4163 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4165 if (SvTYPE(sstr) != stype)
4166 stype = SvTYPE(sstr);
4168 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4169 glob_assign_glob(dstr, sstr, dtype);
4172 if (stype == SVt_PVLV)
4174 if (isREGEXP(sstr)) goto upgregexp;
4175 SvUPGRADE(dstr, SVt_PVNV);
4178 SvUPGRADE(dstr, (svtype)stype);
4180 end_of_first_switch:
4182 /* dstr may have been upgraded. */
4183 dtype = SvTYPE(dstr);
4184 sflags = SvFLAGS(sstr);
4186 if (dtype == SVt_PVCV) {
4187 /* Assigning to a subroutine sets the prototype. */
4190 const char *const ptr = SvPV_const(sstr, len);
4192 SvGROW(dstr, len + 1);
4193 Copy(ptr, SvPVX(dstr), len + 1, char);
4194 SvCUR_set(dstr, len);
4196 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4197 CvAUTOLOAD_off(dstr);
4202 else if (dtype == SVt_PVAV || dtype == SVt_PVHV || dtype == SVt_PVFM) {
4203 const char * const type = sv_reftype(dstr,0);
4205 /* diag_listed_as: Cannot copy to %s */
4206 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4208 Perl_croak(aTHX_ "Cannot copy to %s", type);
4209 } else if (sflags & SVf_ROK) {
4210 if (isGV_with_GP(dstr)
4211 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4214 if (GvIMPORTED(dstr) != GVf_IMPORTED
4215 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4217 GvIMPORTED_on(dstr);
4222 glob_assign_glob(dstr, sstr, dtype);
4226 if (dtype >= SVt_PV) {
4227 if (isGV_with_GP(dstr)) {
4228 glob_assign_ref(dstr, sstr);
4231 if (SvPVX_const(dstr)) {
4237 (void)SvOK_off(dstr);
4238 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4239 SvFLAGS(dstr) |= sflags & SVf_ROK;
4240 assert(!(sflags & SVp_NOK));
4241 assert(!(sflags & SVp_IOK));
4242 assert(!(sflags & SVf_NOK));
4243 assert(!(sflags & SVf_IOK));
4245 else if (isGV_with_GP(dstr)) {
4246 if (!(sflags & SVf_OK)) {
4247 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4248 "Undefined value assigned to typeglob");
4251 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4252 if (dstr != (const SV *)gv) {
4253 const char * const name = GvNAME((const GV *)dstr);
4254 const STRLEN len = GvNAMELEN(dstr);
4255 HV *old_stash = NULL;
4256 bool reset_isa = FALSE;
4257 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4258 || (len == 1 && name[0] == ':')) {
4259 /* Set aside the old stash, so we can reset isa caches
4260 on its subclasses. */
4261 if((old_stash = GvHV(dstr))) {
4262 /* Make sure we do not lose it early. */
4263 SvREFCNT_inc_simple_void_NN(
4264 sv_2mortal((SV *)old_stash)
4271 gp_free(MUTABLE_GV(dstr));
4272 GvGP_set(dstr, gp_ref(GvGP(gv)));
4275 HV * const stash = GvHV(dstr);
4277 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4287 else if ((dtype == SVt_REGEXP || dtype == SVt_PVLV)
4288 && (stype == SVt_REGEXP || isREGEXP(sstr))) {
4289 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4291 else if (sflags & SVp_POK) {
4293 const STRLEN cur = SvCUR(sstr);
4294 const STRLEN len = SvLEN(sstr);
4297 * Check to see if we can just swipe the string. If so, it's a
4298 * possible small lose on short strings, but a big win on long ones.
4299 * It might even be a win on short strings if SvPVX_const(dstr)
4300 * has to be allocated and SvPVX_const(sstr) has to be freed.
4301 * Likewise if we can set up COW rather than doing an actual copy, we
4302 * drop to the else clause, as the swipe code and the COW setup code
4303 * have much in common.
4306 /* Whichever path we take through the next code, we want this true,
4307 and doing it now facilitates the COW check. */
4308 (void)SvPOK_only(dstr);
4311 /* If we're already COW then this clause is not true, and if COW
4312 is allowed then we drop down to the else and make dest COW
4313 with us. If caller hasn't said that we're allowed to COW
4314 shared hash keys then we don't do the COW setup, even if the
4315 source scalar is a shared hash key scalar. */
4316 (((flags & SV_COW_SHARED_HASH_KEYS)
4317 ? !(sflags & SVf_IsCOW)
4318 #ifdef PERL_NEW_COPY_ON_WRITE
4320 ((!GE_COWBUF_THRESHOLD(cur) && SvLEN(dstr) > cur)
4321 /* If this is a regular (non-hek) COW, only so many COW
4322 "copies" are possible. */
4323 || CowREFCNT(sstr) == SV_COW_REFCNT_MAX))
4325 : 1 /* If making a COW copy is forbidden then the behaviour we
4326 desire is as if the source SV isn't actually already
4327 COW, even if it is. So we act as if the source flags
4328 are not COW, rather than actually testing them. */
4330 #ifndef PERL_ANY_COW
4331 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4332 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4333 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4334 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4335 but in turn, it's somewhat dead code, never expected to go
4336 live, but more kept as a placeholder on how to do it better
4337 in a newer implementation. */
4338 /* If we are COW and dstr is a suitable target then we drop down
4339 into the else and make dest a COW of us. */
4340 || (SvFLAGS(dstr) & SVf_BREAK)
4345 #ifdef PERL_NEW_COPY_ON_WRITE
4346 /* slated for free anyway (and not COW)? */
4347 (sflags & (SVs_TEMP|SVf_IsCOW)) == SVs_TEMP &&
4349 (sflags & SVs_TEMP) && /* slated for free anyway? */
4351 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4352 (!(flags & SV_NOSTEAL)) &&
4353 /* and we're allowed to steal temps */
4354 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4355 len) /* and really is a string */
4357 && ((flags & SV_COW_SHARED_HASH_KEYS)
4358 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4359 # ifdef PERL_OLD_COPY_ON_WRITE
4360 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4361 && SvTYPE(sstr) >= SVt_PVIV && len
4363 && !(SvFLAGS(dstr) & SVf_BREAK)
4364 && !(sflags & SVf_IsCOW)
4365 && GE_COW_THRESHOLD(cur) && cur+1 < len
4366 && (GE_COWBUF_THRESHOLD(cur) || SvLEN(dstr) < cur+1)
4372 /* Failed the swipe test, and it's not a shared hash key either.
4373 Have to copy the string. */
4374 SvGROW(dstr, cur + 1); /* inlined from sv_setpvn */
4375 Move(SvPVX_const(sstr),SvPVX(dstr),cur,char);
4376 SvCUR_set(dstr, cur);
4377 *SvEND(dstr) = '\0';
4379 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4381 /* Either it's a shared hash key, or it's suitable for
4382 copy-on-write or we can swipe the string. */
4384 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4390 if (!(sflags & SVf_IsCOW)) {
4392 # ifdef PERL_OLD_COPY_ON_WRITE
4393 /* Make the source SV into a loop of 1.
4394 (about to become 2) */
4395 SV_COW_NEXT_SV_SET(sstr, sstr);
4397 CowREFCNT(sstr) = 0;
4402 /* Initial code is common. */
4403 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4408 /* making another shared SV. */
4411 # ifdef PERL_OLD_COPY_ON_WRITE
4412 assert (SvTYPE(dstr) >= SVt_PVIV);
4413 /* SvIsCOW_normal */
4414 /* splice us in between source and next-after-source. */
4415 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4416 SV_COW_NEXT_SV_SET(sstr, dstr);
4420 SvPV_set(dstr, SvPVX_mutable(sstr));
4424 /* SvIsCOW_shared_hash */
4425 DEBUG_C(PerlIO_printf(Perl_debug_log,
4426 "Copy on write: Sharing hash\n"));
4428 assert (SvTYPE(dstr) >= SVt_PV);
4430 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4432 SvLEN_set(dstr, len);
4433 SvCUR_set(dstr, cur);
4437 { /* Passes the swipe test. */
4438 SvPV_set(dstr, SvPVX_mutable(sstr));
4439 SvLEN_set(dstr, SvLEN(sstr));
4440 SvCUR_set(dstr, SvCUR(sstr));
4443 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4444 SvPV_set(sstr, NULL);
4450 if (sflags & SVp_NOK) {
4451 SvNV_set(dstr, SvNVX(sstr));
4453 if (sflags & SVp_IOK) {
4454 SvIV_set(dstr, SvIVX(sstr));
4455 /* Must do this otherwise some other overloaded use of 0x80000000
4456 gets confused. I guess SVpbm_VALID */
4457 if (sflags & SVf_IVisUV)
4460 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4462 const MAGIC * const smg = SvVSTRING_mg(sstr);
4464 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4465 smg->mg_ptr, smg->mg_len);
4466 SvRMAGICAL_on(dstr);
4470 else if (sflags & (SVp_IOK|SVp_NOK)) {
4471 (void)SvOK_off(dstr);
4472 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4473 if (sflags & SVp_IOK) {
4474 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4475 SvIV_set(dstr, SvIVX(sstr));
4477 if (sflags & SVp_NOK) {
4478 SvNV_set(dstr, SvNVX(sstr));
4482 if (isGV_with_GP(sstr)) {
4483 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4486 (void)SvOK_off(dstr);
4488 if (SvTAINTED(sstr))
4493 =for apidoc sv_setsv_mg
4495 Like C<sv_setsv>, but also handles 'set' magic.
4501 Perl_sv_setsv_mg(pTHX_ SV *const dstr, SV *const sstr)
4503 PERL_ARGS_ASSERT_SV_SETSV_MG;
4505 sv_setsv(dstr,sstr);
4510 # ifdef PERL_OLD_COPY_ON_WRITE
4511 # define SVt_COW SVt_PVIV
4513 # define SVt_COW SVt_PV
4516 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4518 STRLEN cur = SvCUR(sstr);
4519 STRLEN len = SvLEN(sstr);
4522 PERL_ARGS_ASSERT_SV_SETSV_COW;
4525 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4526 (void*)sstr, (void*)dstr);
4533 if (SvTHINKFIRST(dstr))
4534 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4535 else if (SvPVX_const(dstr))
4536 Safefree(SvPVX_mutable(dstr));
4540 SvUPGRADE(dstr, SVt_COW);
4542 assert (SvPOK(sstr));
4543 assert (SvPOKp(sstr));
4544 # ifdef PERL_OLD_COPY_ON_WRITE
4545 assert (!SvIOK(sstr));
4546 assert (!SvIOKp(sstr));
4547 assert (!SvNOK(sstr));
4548 assert (!SvNOKp(sstr));
4551 if (SvIsCOW(sstr)) {
4553 if (SvLEN(sstr) == 0) {
4554 /* source is a COW shared hash key. */
4555 DEBUG_C(PerlIO_printf(Perl_debug_log,
4556 "Fast copy on write: Sharing hash\n"));
4557 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4560 # ifdef PERL_OLD_COPY_ON_WRITE
4561 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4563 assert(SvCUR(sstr)+1 < SvLEN(sstr));
4564 assert(CowREFCNT(sstr) < SV_COW_REFCNT_MAX);
4567 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4568 SvUPGRADE(sstr, SVt_COW);
4570 DEBUG_C(PerlIO_printf(Perl_debug_log,
4571 "Fast copy on write: Converting sstr to COW\n"));
4572 # ifdef PERL_OLD_COPY_ON_WRITE
4573 SV_COW_NEXT_SV_SET(dstr, sstr);
4575 CowREFCNT(sstr) = 0;
4578 # ifdef PERL_OLD_COPY_ON_WRITE
4579 SV_COW_NEXT_SV_SET(sstr, dstr);
4583 new_pv = SvPVX_mutable(sstr);
4586 SvPV_set(dstr, new_pv);
4587 SvFLAGS(dstr) = (SVt_COW|SVf_POK|SVp_POK|SVf_IsCOW);
4590 SvLEN_set(dstr, len);
4591 SvCUR_set(dstr, cur);
4600 =for apidoc sv_setpvn
4602 Copies a string into an SV. The C<len> parameter indicates the number of
4603 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4604 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4610 Perl_sv_setpvn(pTHX_ SV *const sv, const char *const ptr, const STRLEN len)
4615 PERL_ARGS_ASSERT_SV_SETPVN;
4617 SV_CHECK_THINKFIRST_COW_DROP(sv);
4623 /* len is STRLEN which is unsigned, need to copy to signed */
4626 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4629 SvUPGRADE(sv, SVt_PV);
4631 dptr = SvGROW(sv, len + 1);
4632 Move(ptr,dptr,len,char);
4635 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4637 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4641 =for apidoc sv_setpvn_mg
4643 Like C<sv_setpvn>, but also handles 'set' magic.
4649 Perl_sv_setpvn_mg(pTHX_ SV *const sv, const char *const ptr, const STRLEN len)
4651 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4653 sv_setpvn(sv,ptr,len);
4658 =for apidoc sv_setpv
4660 Copies a string into an SV. The string must be null-terminated. Does not
4661 handle 'set' magic. See C<sv_setpv_mg>.
4667 Perl_sv_setpv(pTHX_ SV *const sv, const char *const ptr)
4672 PERL_ARGS_ASSERT_SV_SETPV;
4674 SV_CHECK_THINKFIRST_COW_DROP(sv);
4680 SvUPGRADE(sv, SVt_PV);
4682 SvGROW(sv, len + 1);
4683 Move(ptr,SvPVX(sv),len+1,char);
4685 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4687 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4691 =for apidoc sv_setpv_mg
4693 Like C<sv_setpv>, but also handles 'set' magic.
4699 Perl_sv_setpv_mg(pTHX_ SV *const sv, const char *const ptr)
4701 PERL_ARGS_ASSERT_SV_SETPV_MG;
4708 Perl_sv_sethek(pTHX_ SV *const sv, const HEK *const hek)
4712 PERL_ARGS_ASSERT_SV_SETHEK;
4718 if (HEK_LEN(hek) == HEf_SVKEY) {
4719 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4722 const int flags = HEK_FLAGS(hek);
4723 if (flags & HVhek_WASUTF8) {
4724 STRLEN utf8_len = HEK_LEN(hek);
4725 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4726 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4729 } else if (flags & HVhek_UNSHARED) {
4730 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4733 else SvUTF8_off(sv);
4737 SV_CHECK_THINKFIRST_COW_DROP(sv);
4738 SvUPGRADE(sv, SVt_PV);
4740 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4741 SvCUR_set(sv, HEK_LEN(hek));
4747 else SvUTF8_off(sv);
4755 =for apidoc sv_usepvn_flags
4757 Tells an SV to use C<ptr> to find its string value. Normally the
4758 string is stored inside the SV but sv_usepvn allows the SV to use an
4759 outside string. The C<ptr> should point to memory that was allocated
4760 by C<malloc>. It must be the start of a mallocked block
4761 of memory, and not a pointer to the middle of it. The
4762 string length, C<len>, must be supplied. By default
4763 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4764 so that pointer should not be freed or used by the programmer after
4765 giving it to sv_usepvn, and neither should any pointers from "behind"
4766 that pointer (e.g. ptr + 1) be used.
4768 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4769 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4770 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4771 C<len>, and already meets the requirements for storing in C<SvPVX>).
4777 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4782 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4784 SV_CHECK_THINKFIRST_COW_DROP(sv);
4785 SvUPGRADE(sv, SVt_PV);
4788 if (flags & SV_SMAGIC)
4792 if (SvPVX_const(sv))
4796 if (flags & SV_HAS_TRAILING_NUL)
4797 assert(ptr[len] == '\0');
4800 allocate = (flags & SV_HAS_TRAILING_NUL)
4802 #ifdef Perl_safesysmalloc_size
4805 PERL_STRLEN_ROUNDUP(len + 1);
4807 if (flags & SV_HAS_TRAILING_NUL) {
4808 /* It's long enough - do nothing.
4809 Specifically Perl_newCONSTSUB is relying on this. */
4812 /* Force a move to shake out bugs in callers. */
4813 char *new_ptr = (char*)safemalloc(allocate);
4814 Copy(ptr, new_ptr, len, char);
4815 PoisonFree(ptr,len,char);
4819 ptr = (char*) saferealloc (ptr, allocate);
4822 #ifdef Perl_safesysmalloc_size
4823 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4825 SvLEN_set(sv, allocate);
4829 if (!(flags & SV_HAS_TRAILING_NUL)) {
4832 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4834 if (flags & SV_SMAGIC)
4838 #ifdef PERL_OLD_COPY_ON_WRITE
4839 /* Need to do this *after* making the SV normal, as we need the buffer
4840 pointer to remain valid until after we've copied it. If we let go too early,
4841 another thread could invalidate it by unsharing last of the same hash key
4842 (which it can do by means other than releasing copy-on-write Svs)
4843 or by changing the other copy-on-write SVs in the loop. */
4845 S_sv_release_COW(pTHX_ SV *sv, const char *pvx, SV *after)
4847 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4849 { /* this SV was SvIsCOW_normal(sv) */
4850 /* we need to find the SV pointing to us. */
4851 SV *current = SV_COW_NEXT_SV(after);
4853 if (current == sv) {
4854 /* The SV we point to points back to us (there were only two of us
4856 Hence other SV is no longer copy on write either. */
4859 /* We need to follow the pointers around the loop. */
4861 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4864 /* don't loop forever if the structure is bust, and we have
4865 a pointer into a closed loop. */
4866 assert (current != after);
4867 assert (SvPVX_const(current) == pvx);
4869 /* Make the SV before us point to the SV after us. */
4870 SV_COW_NEXT_SV_SET(current, after);
4876 =for apidoc sv_force_normal_flags
4878 Undo various types of fakery on an SV, where fakery means
4879 "more than" a string: if the PV is a shared string, make
4880 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4881 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4882 we do the copy, and is also used locally; if this is a
4883 vstring, drop the vstring magic. If C<SV_COW_DROP_PV> is set
4884 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4885 SvPOK_off rather than making a copy. (Used where this
4886 scalar is about to be set to some other value.) In addition,
4887 the C<flags> parameter gets passed to C<sv_unref_flags()>
4888 when unreffing. C<sv_force_normal> calls this function
4889 with flags set to 0.
4895 S_sv_uncow(pTHX_ SV * const sv, const U32 flags)
4899 assert(SvIsCOW(sv));
4902 const char * const pvx = SvPVX_const(sv);
4903 const STRLEN len = SvLEN(sv);
4904 const STRLEN cur = SvCUR(sv);
4905 # ifdef PERL_OLD_COPY_ON_WRITE
4906 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4907 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4908 we'll fail an assertion. */
4909 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4913 PerlIO_printf(Perl_debug_log,
4914 "Copy on write: Force normal %ld\n",
4919 # ifdef PERL_NEW_COPY_ON_WRITE
4920 if (len && CowREFCNT(sv) == 0)
4921 /* We own the buffer ourselves. */
4927 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4928 # ifdef PERL_NEW_COPY_ON_WRITE
4929 /* Must do this first, since the macro uses SvPVX. */
4930 if (len) CowREFCNT(sv)--;
4934 if (flags & SV_COW_DROP_PV) {
4935 /* OK, so we don't need to copy our buffer. */
4938 SvGROW(sv, cur + 1);
4939 Move(pvx,SvPVX(sv),cur,char);
4944 # ifdef PERL_OLD_COPY_ON_WRITE
4945 sv_release_COW(sv, pvx, next);
4948 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4955 const char * const pvx = SvPVX_const(sv);
4956 const STRLEN len = SvCUR(sv);
4960 if (flags & SV_COW_DROP_PV) {
4961 /* OK, so we don't need to copy our buffer. */
4964 SvGROW(sv, len + 1);
4965 Move(pvx,SvPVX(sv),len,char);
4968 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4974 Perl_sv_force_normal_flags(pTHX_ SV *const sv, const U32 flags)
4976 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4979 Perl_croak_no_modify();
4980 else if (SvIsCOW(sv))
4981 S_sv_uncow(aTHX_ sv, flags);
4983 sv_unref_flags(sv, flags);
4984 else if (SvFAKE(sv) && isGV_with_GP(sv))
4985 sv_unglob(sv, flags);
4986 else if (SvFAKE(sv) && isREGEXP(sv)) {
4987 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4988 to sv_unglob. We only need it here, so inline it. */
4989 const bool islv = SvTYPE(sv) == SVt_PVLV;
4990 const svtype new_type =
4991 islv ? SVt_NULL : SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4992 SV *const temp = newSV_type(new_type);
4993 regexp *const temp_p = ReANY((REGEXP *)sv);
4995 if (new_type == SVt_PVMG) {
4996 SvMAGIC_set(temp, SvMAGIC(sv));
4997 SvMAGIC_set(sv, NULL);
4998 SvSTASH_set(temp, SvSTASH(sv));
4999 SvSTASH_set(sv, NULL);
5001 if (!islv) SvCUR_set(temp, SvCUR(sv));
5002 /* Remember that SvPVX is in the head, not the body. But
5003 RX_WRAPPED is in the body. */
5004 assert(ReANY((REGEXP *)sv)->mother_re);
5005 /* Their buffer is already owned by someone else. */
5006 if (flags & SV_COW_DROP_PV) {
5007 /* SvLEN is already 0. For SVt_REGEXP, we have a brand new
5008 zeroed body. For SVt_PVLV, it should have been set to 0
5009 before turning into a regexp. */
5010 assert(!SvLEN(islv ? sv : temp));
5011 sv->sv_u.svu_pv = 0;
5014 sv->sv_u.svu_pv = savepvn(RX_WRAPPED((REGEXP *)sv), SvCUR(sv));
5015 SvLEN_set(islv ? sv : temp, SvCUR(sv)+1);
5019 /* Now swap the rest of the bodies. */
5023 SvFLAGS(sv) &= ~SVTYPEMASK;
5024 SvFLAGS(sv) |= new_type;
5025 SvANY(sv) = SvANY(temp);
5028 SvFLAGS(temp) &= ~(SVTYPEMASK);
5029 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
5030 SvANY(temp) = temp_p;
5031 temp->sv_u.svu_rx = (regexp *)temp_p;
5033 SvREFCNT_dec_NN(temp);
5035 else if (SvVOK(sv)) sv_unmagic(sv, PERL_MAGIC_vstring);
5041 Efficient removal of characters from the beginning of the string buffer.
5042 SvPOK(sv), or at least SvPOKp(sv), must be true and the C<ptr> must be a
5043 pointer to somewhere inside the string buffer. The C<ptr> becomes the first
5044 character of the adjusted string. Uses the "OOK hack". On return, only
5045 SvPOK(sv) and SvPOKp(sv) among the OK flags will be true.
5047 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
5048 refer to the same chunk of data.
5050 The unfortunate similarity of this function's name to that of Perl's C<chop>
5051 operator is strictly coincidental. This function works from the left;
5052 C<chop> works from the right.
5058 Perl_sv_chop(pTHX_ SV *const sv, const char *const ptr)
5069 PERL_ARGS_ASSERT_SV_CHOP;
5071 if (!ptr || !SvPOKp(sv))
5073 delta = ptr - SvPVX_const(sv);
5075 /* Nothing to do. */
5078 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
5079 if (delta > max_delta)
5080 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
5081 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
5082 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
5083 SV_CHECK_THINKFIRST(sv);
5084 SvPOK_only_UTF8(sv);
5087 if (!SvLEN(sv)) { /* make copy of shared string */
5088 const char *pvx = SvPVX_const(sv);
5089 const STRLEN len = SvCUR(sv);
5090 SvGROW(sv, len + 1);
5091 Move(pvx,SvPVX(sv),len,char);
5097 SvOOK_offset(sv, old_delta);
5099 SvLEN_set(sv, SvLEN(sv) - delta);
5100 SvCUR_set(sv, SvCUR(sv) - delta);
5101 SvPV_set(sv, SvPVX(sv) + delta);
5103 p = (U8 *)SvPVX_const(sv);
5106 /* how many bytes were evacuated? we will fill them with sentinel
5107 bytes, except for the part holding the new offset of course. */
5110 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
5112 assert(evacn <= delta + old_delta);
5116 /* This sets 'delta' to the accumulated value of all deltas so far */
5120 /* If 'delta' fits in a byte, store it just prior to the new beginning of
5121 * the string; otherwise store a 0 byte there and store 'delta' just prior
5122 * to that, using as many bytes as a STRLEN occupies. Thus it overwrites a
5123 * portion of the chopped part of the string */
5124 if (delta < 0x100) {
5128 p -= sizeof(STRLEN);
5129 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
5133 /* Fill the preceding buffer with sentinals to verify that no-one is
5143 =for apidoc sv_catpvn
5145 Concatenates the string onto the end of the string which is in the SV. The
5146 C<len> indicates number of bytes to copy. If the SV has the UTF-8
5147 status set, then the bytes appended should be valid UTF-8.
5148 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
5150 =for apidoc sv_catpvn_flags
5152 Concatenates the string onto the end of the string which is in the SV. The
5153 C<len> indicates number of bytes to copy. If the SV has the UTF-8
5154 status set, then the bytes appended should be valid UTF-8.
5155 If C<flags> has the C<SV_SMAGIC> bit set, will
5156 C<mg_set> on C<dsv> afterwards if appropriate.
5157 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
5158 in terms of this function.
5164 Perl_sv_catpvn_flags(pTHX_ SV *const dsv, const char *sstr, const STRLEN slen, const I32 flags)
5168 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
5170 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
5171 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5173 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5174 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5175 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5178 else SvGROW(dsv, dlen + slen + 1);
5180 sstr = SvPVX_const(dsv);
5181 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5182 SvCUR_set(dsv, SvCUR(dsv) + slen);
5185 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5186 const char * const send = sstr + slen;
5189 /* Something this code does not account for, which I think is
5190 impossible; it would require the same pv to be treated as
5191 bytes *and* utf8, which would indicate a bug elsewhere. */
5192 assert(sstr != dstr);
5194 SvGROW(dsv, dlen + slen * 2 + 1);
5195 d = (U8 *)SvPVX(dsv) + dlen;
5197 while (sstr < send) {
5198 append_utf8_from_native_byte(*sstr, &d);
5201 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5204 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5206 if (flags & SV_SMAGIC)
5211 =for apidoc sv_catsv
5213 Concatenates the string from SV C<ssv> onto the end of the string in SV
5214 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5215 Handles 'get' magic on both SVs, but no 'set' magic. See C<sv_catsv_mg> and
5218 =for apidoc sv_catsv_flags
5220 Concatenates the string from SV C<ssv> onto the end of the string in SV
5221 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5222 If C<flags> include C<SV_GMAGIC> bit set, will call C<mg_get> on both SVs if
5223 appropriate. If C<flags> include C<SV_SMAGIC>, C<mg_set> will be called on
5224 the modified SV afterward, if appropriate. C<sv_catsv>, C<sv_catsv_nomg>,
5225 and C<sv_catsv_mg> are implemented in terms of this function.
5230 Perl_sv_catsv_flags(pTHX_ SV *const dsv, SV *const ssv, const I32 flags)
5234 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5238 const char *spv = SvPV_flags_const(ssv, slen, flags);
5240 if (flags & SV_GMAGIC)
5242 sv_catpvn_flags(dsv, spv, slen,
5243 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5244 if (flags & SV_SMAGIC)
5251 =for apidoc sv_catpv
5253 Concatenates the string onto the end of the string which is in the SV.
5254 If the SV has the UTF-8 status set, then the bytes appended should be
5255 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5260 Perl_sv_catpv(pTHX_ SV *const sv, const char *ptr)
5267 PERL_ARGS_ASSERT_SV_CATPV;
5271 junk = SvPV_force(sv, tlen);
5273 SvGROW(sv, tlen + len + 1);
5275 ptr = SvPVX_const(sv);
5276 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5277 SvCUR_set(sv, SvCUR(sv) + len);
5278 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5283 =for apidoc sv_catpv_flags
5285 Concatenates the string onto the end of the string which is in the SV.
5286 If the SV has the UTF-8 status set, then the bytes appended should
5287 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5288 on the modified SV if appropriate.
5294 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5296 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5297 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5301 =for apidoc sv_catpv_mg
5303 Like C<sv_catpv>, but also handles 'set' magic.
5309 Perl_sv_catpv_mg(pTHX_ SV *const sv, const char *const ptr)
5311 PERL_ARGS_ASSERT_SV_CATPV_MG;
5320 Creates a new SV. A non-zero C<len> parameter indicates the number of
5321 bytes of preallocated string space the SV should have. An extra byte for a
5322 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5323 space is allocated.) The reference count for the new SV is set to 1.
5325 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5326 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5327 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5328 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5329 modules supporting older perls.
5335 Perl_newSV(pTHX_ const STRLEN len)
5342 sv_upgrade(sv, SVt_PV);
5343 SvGROW(sv, len + 1);
5348 =for apidoc sv_magicext
5350 Adds magic to an SV, upgrading it if necessary. Applies the
5351 supplied vtable and returns a pointer to the magic added.
5353 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5354 In particular, you can add magic to SvREADONLY SVs, and add more than
5355 one instance of the same 'how'.
5357 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5358 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5359 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5360 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5362 (This is now used as a subroutine by C<sv_magic>.)
5367 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5368 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5373 PERL_ARGS_ASSERT_SV_MAGICEXT;
5375 if (SvTYPE(sv)==SVt_PVAV) { assert (!AvPAD_NAMELIST(sv)); }
5377 SvUPGRADE(sv, SVt_PVMG);
5378 Newxz(mg, 1, MAGIC);
5379 mg->mg_moremagic = SvMAGIC(sv);
5380 SvMAGIC_set(sv, mg);
5382 /* Sometimes a magic contains a reference loop, where the sv and
5383 object refer to each other. To prevent a reference loop that
5384 would prevent such objects being freed, we look for such loops
5385 and if we find one we avoid incrementing the object refcount.
5387 Note we cannot do this to avoid self-tie loops as intervening RV must
5388 have its REFCNT incremented to keep it in existence.
5391 if (!obj || obj == sv ||
5392 how == PERL_MAGIC_arylen ||
5393 how == PERL_MAGIC_symtab ||
5394 (SvTYPE(obj) == SVt_PVGV &&
5395 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5396 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5397 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5402 mg->mg_obj = SvREFCNT_inc_simple(obj);
5403 mg->mg_flags |= MGf_REFCOUNTED;
5406 /* Normal self-ties simply pass a null object, and instead of
5407 using mg_obj directly, use the SvTIED_obj macro to produce a
5408 new RV as needed. For glob "self-ties", we are tieing the PVIO
5409 with an RV obj pointing to the glob containing the PVIO. In
5410 this case, to avoid a reference loop, we need to weaken the
5414 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5415 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5421 mg->mg_len = namlen;
5424 mg->mg_ptr = savepvn(name, namlen);
5425 else if (namlen == HEf_SVKEY) {
5426 /* Yes, this is casting away const. This is only for the case of
5427 HEf_SVKEY. I think we need to document this aberation of the
5428 constness of the API, rather than making name non-const, as
5429 that change propagating outwards a long way. */
5430 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5432 mg->mg_ptr = (char *) name;
5434 mg->mg_virtual = (MGVTBL *) vtable;
5441 Perl_sv_magicext_mglob(pTHX_ SV *sv)
5443 PERL_ARGS_ASSERT_SV_MAGICEXT_MGLOB;
5444 if (SvTYPE(sv) == SVt_PVLV && LvTYPE(sv) == 'y') {
5445 /* This sv is only a delegate. //g magic must be attached to
5450 #ifdef PERL_OLD_COPY_ON_WRITE
5452 sv_force_normal_flags(sv, 0);
5454 return sv_magicext(sv, NULL, PERL_MAGIC_regex_global,
5455 &PL_vtbl_mglob, 0, 0);
5459 =for apidoc sv_magic
5461 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5462 necessary, then adds a new magic item of type C<how> to the head of the
5465 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5466 handling of the C<name> and C<namlen> arguments.
5468 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5469 to add more than one instance of the same 'how'.
5475 Perl_sv_magic(pTHX_ SV *const sv, SV *const obj, const int how,
5476 const char *const name, const I32 namlen)
5479 const MGVTBL *vtable;
5482 unsigned int vtable_index;
5484 PERL_ARGS_ASSERT_SV_MAGIC;
5486 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5487 || ((flags = PL_magic_data[how]),
5488 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5489 > magic_vtable_max))
5490 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5492 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5493 Useful for attaching extension internal data to perl vars.
5494 Note that multiple extensions may clash if magical scalars
5495 etc holding private data from one are passed to another. */
5497 vtable = (vtable_index == magic_vtable_max)
5498 ? NULL : PL_magic_vtables + vtable_index;
5500 #ifdef PERL_OLD_COPY_ON_WRITE
5502 sv_force_normal_flags(sv, 0);
5504 if (SvREADONLY(sv)) {
5506 !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5509 Perl_croak_no_modify();
5512 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5513 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5514 /* sv_magic() refuses to add a magic of the same 'how' as an
5517 if (how == PERL_MAGIC_taint)
5523 /* Force pos to be stored as characters, not bytes. */
5524 if (SvMAGICAL(sv) && DO_UTF8(sv)
5525 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
5527 && mg->mg_flags & MGf_BYTES) {
5528 mg->mg_len = (SSize_t)sv_pos_b2u_flags(sv, (STRLEN)mg->mg_len,
5530 mg->mg_flags &= ~MGf_BYTES;
5533 /* Rest of work is done else where */
5534 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5537 case PERL_MAGIC_taint:
5540 case PERL_MAGIC_ext:
5541 case PERL_MAGIC_dbfile:
5548 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5555 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5557 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5558 for (mg = *mgp; mg; mg = *mgp) {
5559 const MGVTBL* const virt = mg->mg_virtual;
5560 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5561 *mgp = mg->mg_moremagic;
5562 if (virt && virt->svt_free)
5563 virt->svt_free(aTHX_ sv, mg);
5564 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5566 Safefree(mg->mg_ptr);
5567 else if (mg->mg_len == HEf_SVKEY)
5568 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5569 else if (mg->mg_type == PERL_MAGIC_utf8)
5570 Safefree(mg->mg_ptr);
5572 if (mg->mg_flags & MGf_REFCOUNTED)
5573 SvREFCNT_dec(mg->mg_obj);
5577 mgp = &mg->mg_moremagic;
5580 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5581 mg_magical(sv); /* else fix the flags now */
5585 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5591 =for apidoc sv_unmagic
5593 Removes all magic of type C<type> from an SV.
5599 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5601 PERL_ARGS_ASSERT_SV_UNMAGIC;
5602 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5606 =for apidoc sv_unmagicext
5608 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5614 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5616 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5617 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5621 =for apidoc sv_rvweaken
5623 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5624 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5625 push a back-reference to this RV onto the array of backreferences
5626 associated with that magic. If the RV is magical, set magic will be
5627 called after the RV is cleared.
5633 Perl_sv_rvweaken(pTHX_ SV *const sv)
5637 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5639 if (!SvOK(sv)) /* let undefs pass */
5642 Perl_croak(aTHX_ "Can't weaken a nonreference");
5643 else if (SvWEAKREF(sv)) {
5644 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5647 else if (SvREADONLY(sv)) croak_no_modify();
5649 Perl_sv_add_backref(aTHX_ tsv, sv);
5651 SvREFCNT_dec_NN(tsv);
5655 /* Give tsv backref magic if it hasn't already got it, then push a
5656 * back-reference to sv onto the array associated with the backref magic.
5658 * As an optimisation, if there's only one backref and it's not an AV,
5659 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5660 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5664 /* A discussion about the backreferences array and its refcount:
5666 * The AV holding the backreferences is pointed to either as the mg_obj of
5667 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5668 * xhv_backreferences field. The array is created with a refcount
5669 * of 2. This means that if during global destruction the array gets
5670 * picked on before its parent to have its refcount decremented by the
5671 * random zapper, it won't actually be freed, meaning it's still there for
5672 * when its parent gets freed.
5674 * When the parent SV is freed, the extra ref is killed by
5675 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5676 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5678 * When a single backref SV is stored directly, it is not reference
5683 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5690 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5692 /* find slot to store array or singleton backref */
5694 if (SvTYPE(tsv) == SVt_PVHV) {
5695 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5698 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5700 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5701 mg = mg_find(tsv, PERL_MAGIC_backref);
5703 svp = &(mg->mg_obj);
5706 /* create or retrieve the array */
5708 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5709 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5714 SvREFCNT_inc_simple_void(av);
5715 /* av now has a refcnt of 2; see discussion above */
5717 /* move single existing backref to the array */
5719 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5723 mg->mg_flags |= MGf_REFCOUNTED;
5726 av = MUTABLE_AV(*svp);
5729 /* optimisation: store single backref directly in HvAUX or mg_obj */
5733 /* push new backref */
5734 assert(SvTYPE(av) == SVt_PVAV);
5735 if (AvFILLp(av) >= AvMAX(av)) {
5736 av_extend(av, AvFILLp(av)+1);
5738 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5741 /* delete a back-reference to ourselves from the backref magic associated
5742 * with the SV we point to.
5746 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5751 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5753 if (SvTYPE(tsv) == SVt_PVHV) {
5755 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5757 else if (SvIS_FREED(tsv) && PL_phase == PERL_PHASE_DESTRUCT) {
5758 /* It's possible for the the last (strong) reference to tsv to have
5759 become freed *before* the last thing holding a weak reference.
5760 If both survive longer than the backreferences array, then when
5761 the referent's reference count drops to 0 and it is freed, it's
5762 not able to chase the backreferences, so they aren't NULLed.
5764 For example, a CV holds a weak reference to its stash. If both the
5765 CV and the stash survive longer than the backreferences array,
5766 and the CV gets picked for the SvBREAK() treatment first,
5767 *and* it turns out that the stash is only being kept alive because
5768 of an our variable in the pad of the CV, then midway during CV
5769 destruction the stash gets freed, but CvSTASH() isn't set to NULL.
5770 It ends up pointing to the freed HV. Hence it's chased in here, and
5771 if this block wasn't here, it would hit the !svp panic just below.
5773 I don't believe that "better" destruction ordering is going to help
5774 here - during global destruction there's always going to be the
5775 chance that something goes out of order. We've tried to make it
5776 foolproof before, and it only resulted in evolutionary pressure on
5777 fools. Which made us look foolish for our hubris. :-(
5783 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5784 svp = mg ? &(mg->mg_obj) : NULL;
5788 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5790 /* It's possible that sv is being freed recursively part way through the
5791 freeing of tsv. If this happens, the backreferences array of tsv has
5792 already been freed, and so svp will be NULL. If this is the case,
5793 we should not panic. Instead, nothing needs doing, so return. */
5794 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5796 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5797 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5800 if (SvTYPE(*svp) == SVt_PVAV) {
5804 AV * const av = (AV*)*svp;
5806 assert(!SvIS_FREED(av));
5810 /* for an SV with N weak references to it, if all those
5811 * weak refs are deleted, then sv_del_backref will be called
5812 * N times and O(N^2) compares will be done within the backref
5813 * array. To ameliorate this potential slowness, we:
5814 * 1) make sure this code is as tight as possible;
5815 * 2) when looking for SV, look for it at both the head and tail of the
5816 * array first before searching the rest, since some create/destroy
5817 * patterns will cause the backrefs to be freed in order.
5824 SV **p = &svp[fill];
5825 SV *const topsv = *p;
5832 /* We weren't the last entry.
5833 An unordered list has this property that you
5834 can take the last element off the end to fill
5835 the hole, and it's still an unordered list :-)
5841 break; /* should only be one */
5848 AvFILLp(av) = fill-1;
5850 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5851 /* freed AV; skip */
5854 /* optimisation: only a single backref, stored directly */
5856 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5863 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5869 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5874 /* after multiple passes through Perl_sv_clean_all() for a thingy
5875 * that has badly leaked, the backref array may have gotten freed,
5876 * since we only protect it against 1 round of cleanup */
5877 if (SvIS_FREED(av)) {
5878 if (PL_in_clean_all) /* All is fair */
5881 "panic: magic_killbackrefs (freed backref AV/SV)");
5885 is_array = (SvTYPE(av) == SVt_PVAV);
5887 assert(!SvIS_FREED(av));
5890 last = svp + AvFILLp(av);
5893 /* optimisation: only a single backref, stored directly */
5899 while (svp <= last) {
5901 SV *const referrer = *svp;
5902 if (SvWEAKREF(referrer)) {
5903 /* XXX Should we check that it hasn't changed? */
5904 assert(SvROK(referrer));
5905 SvRV_set(referrer, 0);
5907 SvWEAKREF_off(referrer);
5908 SvSETMAGIC(referrer);
5909 } else if (SvTYPE(referrer) == SVt_PVGV ||
5910 SvTYPE(referrer) == SVt_PVLV) {
5911 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5912 /* You lookin' at me? */
5913 assert(GvSTASH(referrer));
5914 assert(GvSTASH(referrer) == (const HV *)sv);
5915 GvSTASH(referrer) = 0;
5916 } else if (SvTYPE(referrer) == SVt_PVCV ||
5917 SvTYPE(referrer) == SVt_PVFM) {
5918 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5919 /* You lookin' at me? */
5920 assert(CvSTASH(referrer));
5921 assert(CvSTASH(referrer) == (const HV *)sv);
5922 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5925 assert(SvTYPE(sv) == SVt_PVGV);
5926 /* You lookin' at me? */
5927 assert(CvGV(referrer));
5928 assert(CvGV(referrer) == (const GV *)sv);
5929 anonymise_cv_maybe(MUTABLE_GV(sv),
5930 MUTABLE_CV(referrer));
5935 "panic: magic_killbackrefs (flags=%"UVxf")",
5936 (UV)SvFLAGS(referrer));
5947 SvREFCNT_dec_NN(av); /* remove extra count added by sv_add_backref() */
5953 =for apidoc sv_insert
5955 Inserts a string at the specified offset/length within the SV. Similar to
5956 the Perl substr() function. Handles get magic.
5958 =for apidoc sv_insert_flags
5960 Same as C<sv_insert>, but the extra C<flags> are passed to the
5961 C<SvPV_force_flags> that applies to C<bigstr>.
5967 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5974 SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5977 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5980 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5981 SvPV_force_flags(bigstr, curlen, flags);
5982 (void)SvPOK_only_UTF8(bigstr);
5983 if (offset + len > curlen) {
5984 SvGROW(bigstr, offset+len+1);
5985 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5986 SvCUR_set(bigstr, offset+len);
5990 i = littlelen - len;
5991 if (i > 0) { /* string might grow */
5992 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5993 mid = big + offset + len;
5994 midend = bigend = big + SvCUR(bigstr);
5997 while (midend > mid) /* shove everything down */
5998 *--bigend = *--midend;
5999 Move(little,big+offset,littlelen,char);
6000 SvCUR_set(bigstr, SvCUR(bigstr) + i);
6005 Move(little,SvPVX(bigstr)+offset,len,char);
6010 big = SvPVX(bigstr);
6013 bigend = big + SvCUR(bigstr);
6015 if (midend > bigend)
6016 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
6019 if (mid - big > bigend - midend) { /* faster to shorten from end */
6021 Move(little, mid, littlelen,char);
6024 i = bigend - midend;
6026 Move(midend, mid, i,char);
6030 SvCUR_set(bigstr, mid - big);
6032 else if ((i = mid - big)) { /* faster from front */
6033 midend -= littlelen;
6035 Move(big, midend - i, i, char);
6036 sv_chop(bigstr,midend-i);
6038 Move(little, mid, littlelen,char);
6040 else if (littlelen) {
6041 midend -= littlelen;
6042 sv_chop(bigstr,midend);
6043 Move(little,midend,littlelen,char);
6046 sv_chop(bigstr,midend);
6052 =for apidoc sv_replace
6054 Make the first argument a copy of the second, then delete the original.
6055 The target SV physically takes over ownership of the body of the source SV
6056 and inherits its flags; however, the target keeps any magic it owns,
6057 and any magic in the source is discarded.
6058 Note that this is a rather specialist SV copying operation; most of the
6059 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
6065 Perl_sv_replace(pTHX_ SV *const sv, SV *const nsv)
6068 const U32 refcnt = SvREFCNT(sv);
6070 PERL_ARGS_ASSERT_SV_REPLACE;
6072 SV_CHECK_THINKFIRST_COW_DROP(sv);
6073 if (SvREFCNT(nsv) != 1) {
6074 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
6075 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
6077 if (SvMAGICAL(sv)) {
6081 sv_upgrade(nsv, SVt_PVMG);
6082 SvMAGIC_set(nsv, SvMAGIC(sv));
6083 SvFLAGS(nsv) |= SvMAGICAL(sv);
6085 SvMAGIC_set(sv, NULL);
6089 assert(!SvREFCNT(sv));
6090 #ifdef DEBUG_LEAKING_SCALARS
6091 sv->sv_flags = nsv->sv_flags;
6092 sv->sv_any = nsv->sv_any;
6093 sv->sv_refcnt = nsv->sv_refcnt;
6094 sv->sv_u = nsv->sv_u;
6096 StructCopy(nsv,sv,SV);
6098 if(SvTYPE(sv) == SVt_IV) {
6100 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
6104 #ifdef PERL_OLD_COPY_ON_WRITE
6105 if (SvIsCOW_normal(nsv)) {
6106 /* We need to follow the pointers around the loop to make the
6107 previous SV point to sv, rather than nsv. */
6110 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
6113 assert(SvPVX_const(current) == SvPVX_const(nsv));
6115 /* Make the SV before us point to the SV after us. */
6117 PerlIO_printf(Perl_debug_log, "previous is\n");
6119 PerlIO_printf(Perl_debug_log,
6120 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
6121 (UV) SV_COW_NEXT_SV(current), (UV) sv);
6123 SV_COW_NEXT_SV_SET(current, sv);
6126 SvREFCNT(sv) = refcnt;
6127 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
6132 /* We're about to free a GV which has a CV that refers back to us.
6133 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
6137 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
6142 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
6145 assert(SvREFCNT(gv) == 0);
6146 assert(isGV(gv) && isGV_with_GP(gv));
6148 assert(!CvANON(cv));
6149 assert(CvGV(cv) == gv);
6150 assert(!CvNAMED(cv));
6152 /* will the CV shortly be freed by gp_free() ? */
6153 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
6154 SvANY(cv)->xcv_gv_u.xcv_gv = NULL;
6158 /* if not, anonymise: */
6159 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
6160 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
6161 : newSVpvn_flags( "__ANON__", 8, 0 );
6162 sv_catpvs(gvname, "::__ANON__");
6163 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
6164 SvREFCNT_dec_NN(gvname);
6168 SvANY(cv)->xcv_gv_u.xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
6173 =for apidoc sv_clear
6175 Clear an SV: call any destructors, free up any memory used by the body,
6176 and free the body itself. The SV's head is I<not> freed, although
6177 its type is set to all 1's so that it won't inadvertently be assumed
6178 to be live during global destruction etc.
6179 This function should only be called when REFCNT is zero. Most of the time
6180 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
6187 Perl_sv_clear(pTHX_ SV *const orig_sv)
6192 const struct body_details *sv_type_details;
6198 PERL_ARGS_ASSERT_SV_CLEAR;
6200 /* within this loop, sv is the SV currently being freed, and
6201 * iter_sv is the most recent AV or whatever that's being iterated
6202 * over to provide more SVs */
6208 assert(SvREFCNT(sv) == 0);
6209 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
6211 if (type <= SVt_IV) {
6212 /* See the comment in sv.h about the collusion between this
6213 * early return and the overloading of the NULL slots in the
6217 SvFLAGS(sv) &= SVf_BREAK;
6218 SvFLAGS(sv) |= SVTYPEMASK;
6222 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6224 if (type >= SVt_PVMG) {
6226 if (!curse(sv, 1)) goto get_next_sv;
6227 type = SvTYPE(sv); /* destructor may have changed it */
6229 /* Free back-references before magic, in case the magic calls
6230 * Perl code that has weak references to sv. */
6231 if (type == SVt_PVHV) {
6232 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6236 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6237 SvREFCNT_dec(SvOURSTASH(sv));
6239 else if (type == SVt_PVAV && AvPAD_NAMELIST(sv)) {
6240 assert(!SvMAGICAL(sv));
6241 } else if (SvMAGIC(sv)) {
6242 /* Free back-references before other types of magic. */
6243 sv_unmagic(sv, PERL_MAGIC_backref);
6247 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6248 SvREFCNT_dec(SvSTASH(sv));
6251 /* case SVt_INVLIST: */
6254 IoIFP(sv) != PerlIO_stdin() &&
6255 IoIFP(sv) != PerlIO_stdout() &&
6256 IoIFP(sv) != PerlIO_stderr() &&
6257 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6259 io_close(MUTABLE_IO(sv), FALSE);
6261 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6262 PerlDir_close(IoDIRP(sv));
6263 IoDIRP(sv) = (DIR*)NULL;
6264 Safefree(IoTOP_NAME(sv));
6265 Safefree(IoFMT_NAME(sv));
6266 Safefree(IoBOTTOM_NAME(sv));
6267 if ((const GV *)sv == PL_statgv)
6271 /* FIXME for plugins */
6273 pregfree2((REGEXP*) sv);
6277 cv_undef(MUTABLE_CV(sv));
6278 /* If we're in a stash, we don't own a reference to it.
6279 * However it does have a back reference to us, which needs to
6281 if ((stash = CvSTASH(sv)))
6282 sv_del_backref(MUTABLE_SV(stash), sv);
6285 if (PL_last_swash_hv == (const HV *)sv) {
6286 PL_last_swash_hv = NULL;
6288 if (HvTOTALKEYS((HV*)sv) > 0) {
6290 /* this statement should match the one at the beginning of
6291 * hv_undef_flags() */
6292 if ( PL_phase != PERL_PHASE_DESTRUCT
6293 && (name = HvNAME((HV*)sv)))
6295 if (PL_stashcache) {
6296 DEBUG_o(Perl_deb(aTHX_ "sv_clear clearing PL_stashcache for '%"SVf"'\n",
6298 (void)hv_delete(PL_stashcache, name,
6299 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6301 hv_name_set((HV*)sv, NULL, 0, 0);
6304 /* save old iter_sv in unused SvSTASH field */
6305 assert(!SvOBJECT(sv));
6306 SvSTASH(sv) = (HV*)iter_sv;
6309 /* save old hash_index in unused SvMAGIC field */
6310 assert(!SvMAGICAL(sv));
6311 assert(!SvMAGIC(sv));
6312 ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index = hash_index;
6315 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6316 goto get_next_sv; /* process this new sv */
6318 /* free empty hash */
6319 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6320 assert(!HvARRAY((HV*)sv));
6324 AV* av = MUTABLE_AV(sv);
6325 if (PL_comppad == av) {
6329 if (AvREAL(av) && AvFILLp(av) > -1) {
6330 next_sv = AvARRAY(av)[AvFILLp(av)--];
6331 /* save old iter_sv in top-most slot of AV,
6332 * and pray that it doesn't get wiped in the meantime */
6333 AvARRAY(av)[AvMAX(av)] = iter_sv;
6335 goto get_next_sv; /* process this new sv */
6337 Safefree(AvALLOC(av));
6342 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6343 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6344 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6345 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6347 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6348 SvREFCNT_dec(LvTARG(sv));
6349 if (isREGEXP(sv)) goto freeregexp;
6351 if (isGV_with_GP(sv)) {
6352 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6353 && HvENAME_get(stash))
6354 mro_method_changed_in(stash);
6355 gp_free(MUTABLE_GV(sv));
6357 unshare_hek(GvNAME_HEK(sv));
6358 /* If we're in a stash, we don't own a reference to it.
6359 * However it does have a back reference to us, which
6360 * needs to be cleared. */
6361 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6362 sv_del_backref(MUTABLE_SV(stash), sv);
6364 /* FIXME. There are probably more unreferenced pointers to SVs
6365 * in the interpreter struct that we should check and tidy in
6366 * a similar fashion to this: */
6367 /* See also S_sv_unglob, which does the same thing. */
6368 if ((const GV *)sv == PL_last_in_gv)
6369 PL_last_in_gv = NULL;
6370 else if ((const GV *)sv == PL_statgv)
6372 else if ((const GV *)sv == PL_stderrgv)
6380 /* Don't bother with SvOOK_off(sv); as we're only going to
6384 SvOOK_offset(sv, offset);
6385 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6386 /* Don't even bother with turning off the OOK flag. */
6391 SV * const target = SvRV(sv);
6393 sv_del_backref(target, sv);
6399 else if (SvPVX_const(sv)
6400 && !(SvTYPE(sv) == SVt_PVIO
6401 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6405 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6409 # ifdef PERL_OLD_COPY_ON_WRITE
6410 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6412 if (CowREFCNT(sv)) {
6418 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6422 # ifdef PERL_OLD_COPY_ON_WRITE
6426 Safefree(SvPVX_mutable(sv));
6430 else if (SvPVX_const(sv) && SvLEN(sv)
6431 && !(SvTYPE(sv) == SVt_PVIO
6432 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6433 Safefree(SvPVX_mutable(sv));
6434 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6435 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6445 SvFLAGS(sv) &= SVf_BREAK;
6446 SvFLAGS(sv) |= SVTYPEMASK;
6448 sv_type_details = bodies_by_type + type;
6449 if (sv_type_details->arena) {
6450 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6451 &PL_body_roots[type]);
6453 else if (sv_type_details->body_size) {
6454 safefree(SvANY(sv));
6458 /* caller is responsible for freeing the head of the original sv */
6459 if (sv != orig_sv && !SvREFCNT(sv))
6462 /* grab and free next sv, if any */
6470 else if (!iter_sv) {
6472 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6473 AV *const av = (AV*)iter_sv;
6474 if (AvFILLp(av) > -1) {
6475 sv = AvARRAY(av)[AvFILLp(av)--];
6477 else { /* no more elements of current AV to free */
6480 /* restore previous value, squirrelled away */
6481 iter_sv = AvARRAY(av)[AvMAX(av)];
6482 Safefree(AvALLOC(av));
6485 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6486 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6487 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6488 /* no more elements of current HV to free */
6491 /* Restore previous values of iter_sv and hash_index,
6492 * squirrelled away */
6493 assert(!SvOBJECT(sv));
6494 iter_sv = (SV*)SvSTASH(sv);
6495 assert(!SvMAGICAL(sv));
6496 hash_index = ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index;
6498 /* perl -DA does not like rubbish in SvMAGIC. */
6502 /* free any remaining detritus from the hash struct */
6503 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6504 assert(!HvARRAY((HV*)sv));
6509 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6513 if (!SvREFCNT(sv)) {
6517 if (--(SvREFCNT(sv)))
6521 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6522 "Attempt to free temp prematurely: SV 0x%"UVxf
6523 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6527 if (SvIMMORTAL(sv)) {
6528 /* make sure SvREFCNT(sv)==0 happens very seldom */
6529 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6538 /* This routine curses the sv itself, not the object referenced by sv. So
6539 sv does not have to be ROK. */
6542 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6545 PERL_ARGS_ASSERT_CURSE;
6546 assert(SvOBJECT(sv));
6548 if (PL_defstash && /* Still have a symbol table? */
6554 stash = SvSTASH(sv);
6555 assert(SvTYPE(stash) == SVt_PVHV);
6556 if (HvNAME(stash)) {
6557 CV* destructor = NULL;
6558 assert (SvOOK(stash));
6559 if (!SvOBJECT(stash)) destructor = (CV *)SvSTASH(stash);
6560 if (!destructor || HvMROMETA(stash)->destroy_gen
6561 != PL_sub_generation)
6564 gv_fetchmeth_autoload(stash, "DESTROY", 7, 0);
6565 if (gv) destructor = GvCV(gv);
6566 if (!SvOBJECT(stash))
6569 destructor ? (HV *)destructor : ((HV *)0)+1;
6570 HvAUX(stash)->xhv_mro_meta->destroy_gen =
6574 assert(!destructor || destructor == ((CV *)0)+1
6575 || SvTYPE(destructor) == SVt_PVCV);
6576 if (destructor && destructor != ((CV *)0)+1
6577 /* A constant subroutine can have no side effects, so
6578 don't bother calling it. */
6579 && !CvCONST(destructor)
6580 /* Don't bother calling an empty destructor or one that
6581 returns immediately. */
6582 && (CvISXSUB(destructor)
6583 || (CvSTART(destructor)
6584 && (CvSTART(destructor)->op_next->op_type
6586 && (CvSTART(destructor)->op_next->op_type
6588 || CvSTART(destructor)->op_next->op_next->op_type
6594 SV* const tmpref = newRV(sv);
6595 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6597 PUSHSTACKi(PERLSI_DESTROY);
6602 call_sv(MUTABLE_SV(destructor),
6603 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6607 if(SvREFCNT(tmpref) < 2) {
6608 /* tmpref is not kept alive! */
6610 SvRV_set(tmpref, NULL);
6613 SvREFCNT_dec_NN(tmpref);
6616 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6619 if (check_refcnt && SvREFCNT(sv)) {
6620 if (PL_in_clean_objs)
6622 "DESTROY created new reference to dead object '%"HEKf"'",
6623 HEKfARG(HvNAME_HEK(stash)));
6624 /* DESTROY gave object new lease on life */
6630 HV * const stash = SvSTASH(sv);
6631 /* Curse before freeing the stash, as freeing the stash could cause
6632 a recursive call into S_curse. */
6633 SvOBJECT_off(sv); /* Curse the object. */
6634 SvSTASH_set(sv,0); /* SvREFCNT_dec may try to read this */
6635 SvREFCNT_dec(stash); /* possibly of changed persuasion */
6641 =for apidoc sv_newref
6643 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6650 Perl_sv_newref(pTHX_ SV *const sv)
6652 PERL_UNUSED_CONTEXT;
6661 Decrement an SV's reference count, and if it drops to zero, call
6662 C<sv_clear> to invoke destructors and free up any memory used by
6663 the body; finally, deallocate the SV's head itself.
6664 Normally called via a wrapper macro C<SvREFCNT_dec>.
6670 Perl_sv_free(pTHX_ SV *const sv)
6676 /* Private helper function for SvREFCNT_dec().
6677 * Called with rc set to original SvREFCNT(sv), where rc == 0 or 1 */
6680 Perl_sv_free2(pTHX_ SV *const sv, const U32 rc)
6684 PERL_ARGS_ASSERT_SV_FREE2;
6686 if (LIKELY( rc == 1 )) {
6692 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6693 "Attempt to free temp prematurely: SV 0x%"UVxf
6694 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6698 if (SvIMMORTAL(sv)) {
6699 /* make sure SvREFCNT(sv)==0 happens very seldom */
6700 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6704 if (! SvREFCNT(sv)) /* may have have been resurrected */
6709 /* handle exceptional cases */
6713 if (SvFLAGS(sv) & SVf_BREAK)
6714 /* this SV's refcnt has been artificially decremented to
6715 * trigger cleanup */
6717 if (PL_in_clean_all) /* All is fair */
6719 if (SvIMMORTAL(sv)) {
6720 /* make sure SvREFCNT(sv)==0 happens very seldom */
6721 SvREFCNT(sv) = SvREFCNT_IMMORTAL;
6724 if (ckWARN_d(WARN_INTERNAL)) {
6725 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6726 Perl_dump_sv_child(aTHX_ sv);
6728 #ifdef DEBUG_LEAKING_SCALARS
6731 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6732 if (PL_warnhook == PERL_WARNHOOK_FATAL
6733 || ckDEAD(packWARN(WARN_INTERNAL))) {
6734 /* Don't let Perl_warner cause us to escape our fate: */
6738 /* This may not return: */
6739 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6740 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6741 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6744 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6754 Returns the length of the string in the SV. Handles magic and type
6755 coercion and sets the UTF8 flag appropriately. See also C<SvCUR>, which
6756 gives raw access to the xpv_cur slot.
6762 Perl_sv_len(pTHX_ SV *const sv)
6769 (void)SvPV_const(sv, len);
6774 =for apidoc sv_len_utf8
6776 Returns the number of characters in the string in an SV, counting wide
6777 UTF-8 bytes as a single character. Handles magic and type coercion.
6783 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6784 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6785 * (Note that the mg_len is not the length of the mg_ptr field.
6786 * This allows the cache to store the character length of the string without
6787 * needing to malloc() extra storage to attach to the mg_ptr.)
6792 Perl_sv_len_utf8(pTHX_ SV *const sv)
6798 return sv_len_utf8_nomg(sv);
6802 Perl_sv_len_utf8_nomg(pTHX_ SV * const sv)
6806 const U8 *s = (U8*)SvPV_nomg_const(sv, len);
6808 PERL_ARGS_ASSERT_SV_LEN_UTF8_NOMG;
6810 if (PL_utf8cache && SvUTF8(sv)) {
6812 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6814 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6815 if (mg->mg_len != -1)
6818 /* We can use the offset cache for a headstart.
6819 The longer value is stored in the first pair. */
6820 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6822 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6826 if (PL_utf8cache < 0) {
6827 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6828 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6832 ulen = Perl_utf8_length(aTHX_ s, s + len);
6833 utf8_mg_len_cache_update(sv, &mg, ulen);
6837 return SvUTF8(sv) ? Perl_utf8_length(aTHX_ s, s + len) : len;
6840 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6843 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6844 STRLEN *const uoffset_p, bool *const at_end)
6846 const U8 *s = start;
6847 STRLEN uoffset = *uoffset_p;
6849 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6851 while (s < send && uoffset) {
6858 else if (s > send) {
6860 /* This is the existing behaviour. Possibly it should be a croak, as
6861 it's actually a bounds error */
6864 *uoffset_p -= uoffset;
6868 /* Given the length of the string in both bytes and UTF-8 characters, decide
6869 whether to walk forwards or backwards to find the byte corresponding to
6870 the passed in UTF-8 offset. */
6872 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6873 STRLEN uoffset, const STRLEN uend)
6875 STRLEN backw = uend - uoffset;
6877 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6879 if (uoffset < 2 * backw) {
6880 /* The assumption is that going forwards is twice the speed of going
6881 forward (that's where the 2 * backw comes from).
6882 (The real figure of course depends on the UTF-8 data.) */
6883 const U8 *s = start;
6885 while (s < send && uoffset--)
6895 while (UTF8_IS_CONTINUATION(*send))
6898 return send - start;
6901 /* For the string representation of the given scalar, find the byte
6902 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6903 give another position in the string, *before* the sought offset, which
6904 (which is always true, as 0, 0 is a valid pair of positions), which should
6905 help reduce the amount of linear searching.
6906 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6907 will be used to reduce the amount of linear searching. The cache will be
6908 created if necessary, and the found value offered to it for update. */
6910 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6911 const U8 *const send, STRLEN uoffset,
6912 STRLEN uoffset0, STRLEN boffset0)
6914 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6916 bool at_end = FALSE;
6918 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6920 assert (uoffset >= uoffset0);
6925 if (!SvREADONLY(sv) && !SvGMAGICAL(sv) && SvPOK(sv)
6927 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6928 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6929 if ((*mgp)->mg_ptr) {
6930 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6931 if (cache[0] == uoffset) {
6932 /* An exact match. */
6935 if (cache[2] == uoffset) {
6936 /* An exact match. */
6940 if (cache[0] < uoffset) {
6941 /* The cache already knows part of the way. */
6942 if (cache[0] > uoffset0) {
6943 /* The cache knows more than the passed in pair */
6944 uoffset0 = cache[0];
6945 boffset0 = cache[1];
6947 if ((*mgp)->mg_len != -1) {
6948 /* And we know the end too. */
6950 + sv_pos_u2b_midway(start + boffset0, send,
6952 (*mgp)->mg_len - uoffset0);
6954 uoffset -= uoffset0;
6956 + sv_pos_u2b_forwards(start + boffset0,
6957 send, &uoffset, &at_end);
6958 uoffset += uoffset0;
6961 else if (cache[2] < uoffset) {
6962 /* We're between the two cache entries. */
6963 if (cache[2] > uoffset0) {
6964 /* and the cache knows more than the passed in pair */
6965 uoffset0 = cache[2];
6966 boffset0 = cache[3];
6970 + sv_pos_u2b_midway(start + boffset0,
6973 cache[0] - uoffset0);
6976 + sv_pos_u2b_midway(start + boffset0,
6979 cache[2] - uoffset0);
6983 else if ((*mgp)->mg_len != -1) {
6984 /* If we can take advantage of a passed in offset, do so. */
6985 /* In fact, offset0 is either 0, or less than offset, so don't
6986 need to worry about the other possibility. */
6988 + sv_pos_u2b_midway(start + boffset0, send,
6990 (*mgp)->mg_len - uoffset0);
6995 if (!found || PL_utf8cache < 0) {
6996 STRLEN real_boffset;
6997 uoffset -= uoffset0;
6998 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6999 send, &uoffset, &at_end);
7000 uoffset += uoffset0;
7002 if (found && PL_utf8cache < 0)
7003 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
7005 boffset = real_boffset;
7008 if (PL_utf8cache && !SvGMAGICAL(sv) && SvPOK(sv)) {
7010 utf8_mg_len_cache_update(sv, mgp, uoffset);
7012 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
7019 =for apidoc sv_pos_u2b_flags
7021 Converts the offset from a count of UTF-8 chars from
7022 the start of the string, to a count of the equivalent number of bytes; if
7023 lenp is non-zero, it does the same to lenp, but this time starting from
7024 the offset, rather than from the start
7025 of the string. Handles type coercion.
7026 I<flags> is passed to C<SvPV_flags>, and usually should be
7027 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
7033 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
7034 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7035 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
7040 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
7047 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
7049 start = (U8*)SvPV_flags(sv, len, flags);
7051 const U8 * const send = start + len;
7053 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
7056 && *lenp /* don't bother doing work for 0, as its bytes equivalent
7057 is 0, and *lenp is already set to that. */) {
7058 /* Convert the relative offset to absolute. */
7059 const STRLEN uoffset2 = uoffset + *lenp;
7060 const STRLEN boffset2
7061 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
7062 uoffset, boffset) - boffset;
7076 =for apidoc sv_pos_u2b
7078 Converts the value pointed to by offsetp from a count of UTF-8 chars from
7079 the start of the string, to a count of the equivalent number of bytes; if
7080 lenp is non-zero, it does the same to lenp, but this time starting from
7081 the offset, rather than from the start of the string. Handles magic and
7084 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
7091 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
7092 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7093 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
7097 /* This function is subject to size and sign problems */
7100 Perl_sv_pos_u2b(pTHX_ SV *const sv, I32 *const offsetp, I32 *const lenp)
7102 PERL_ARGS_ASSERT_SV_POS_U2B;
7105 STRLEN ulen = (STRLEN)*lenp;
7106 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
7107 SV_GMAGIC|SV_CONST_RETURN);
7110 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
7111 SV_GMAGIC|SV_CONST_RETURN);
7116 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
7119 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
7120 if (SvREADONLY(sv) || SvGMAGICAL(sv) || !SvPOK(sv))
7123 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
7124 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
7125 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
7129 (*mgp)->mg_len = ulen;
7132 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
7133 byte length pairing. The (byte) length of the total SV is passed in too,
7134 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
7135 may not have updated SvCUR, so we can't rely on reading it directly.
7137 The proffered utf8/byte length pairing isn't used if the cache already has
7138 two pairs, and swapping either for the proffered pair would increase the
7139 RMS of the intervals between known byte offsets.
7141 The cache itself consists of 4 STRLEN values
7142 0: larger UTF-8 offset
7143 1: corresponding byte offset
7144 2: smaller UTF-8 offset
7145 3: corresponding byte offset
7147 Unused cache pairs have the value 0, 0.
7148 Keeping the cache "backwards" means that the invariant of
7149 cache[0] >= cache[2] is maintained even with empty slots, which means that
7150 the code that uses it doesn't need to worry if only 1 entry has actually
7151 been set to non-zero. It also makes the "position beyond the end of the
7152 cache" logic much simpler, as the first slot is always the one to start
7156 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
7157 const STRLEN utf8, const STRLEN blen)
7161 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
7166 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
7167 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
7168 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
7170 (*mgp)->mg_len = -1;
7174 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
7175 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
7176 (*mgp)->mg_ptr = (char *) cache;
7180 if (PL_utf8cache < 0 && SvPOKp(sv)) {
7181 /* SvPOKp() because it's possible that sv has string overloading, and
7182 therefore is a reference, hence SvPVX() is actually a pointer.
7183 This cures the (very real) symptoms of RT 69422, but I'm not actually
7184 sure whether we should even be caching the results of UTF-8
7185 operations on overloading, given that nothing stops overloading
7186 returning a different value every time it's called. */
7187 const U8 *start = (const U8 *) SvPVX_const(sv);
7188 const STRLEN realutf8 = utf8_length(start, start + byte);
7190 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
7194 /* Cache is held with the later position first, to simplify the code
7195 that deals with unbounded ends. */
7197 ASSERT_UTF8_CACHE(cache);
7198 if (cache[1] == 0) {
7199 /* Cache is totally empty */
7202 } else if (cache[3] == 0) {
7203 if (byte > cache[1]) {
7204 /* New one is larger, so goes first. */
7205 cache[2] = cache[0];
7206 cache[3] = cache[1];
7214 #define THREEWAY_SQUARE(a,b,c,d) \
7215 ((float)((d) - (c))) * ((float)((d) - (c))) \
7216 + ((float)((c) - (b))) * ((float)((c) - (b))) \
7217 + ((float)((b) - (a))) * ((float)((b) - (a)))
7219 /* Cache has 2 slots in use, and we know three potential pairs.
7220 Keep the two that give the lowest RMS distance. Do the
7221 calculation in bytes simply because we always know the byte
7222 length. squareroot has the same ordering as the positive value,
7223 so don't bother with the actual square root. */
7224 if (byte > cache[1]) {
7225 /* New position is after the existing pair of pairs. */
7226 const float keep_earlier
7227 = THREEWAY_SQUARE(0, cache[3], byte, blen);
7228 const float keep_later
7229 = THREEWAY_SQUARE(0, cache[1], byte, blen);
7231 if (keep_later < keep_earlier) {
7232 cache[2] = cache[0];
7233 cache[3] = cache[1];
7242 else if (byte > cache[3]) {
7243 /* New position is between the existing pair of pairs. */
7244 const float keep_earlier
7245 = THREEWAY_SQUARE(0, cache[3], byte, blen);
7246 const float keep_later
7247 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7249 if (keep_later < keep_earlier) {
7259 /* New position is before the existing pair of pairs. */
7260 const float keep_earlier
7261 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7262 const float keep_later
7263 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7265 if (keep_later < keep_earlier) {
7270 cache[0] = cache[2];
7271 cache[1] = cache[3];
7277 ASSERT_UTF8_CACHE(cache);
7280 /* We already know all of the way, now we may be able to walk back. The same
7281 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7282 backward is half the speed of walking forward. */
7284 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7285 const U8 *end, STRLEN endu)
7287 const STRLEN forw = target - s;
7288 STRLEN backw = end - target;
7290 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7292 if (forw < 2 * backw) {
7293 return utf8_length(s, target);
7296 while (end > target) {
7298 while (UTF8_IS_CONTINUATION(*end)) {
7307 =for apidoc sv_pos_b2u_flags
7309 Converts the offset from a count of bytes from the start of the string, to
7310 a count of the equivalent number of UTF-8 chars. Handles type coercion.
7311 I<flags> is passed to C<SvPV_flags>, and usually should be
7312 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
7318 * sv_pos_b2u_flags() uses, like sv_pos_u2b_flags(), the mg_ptr of the
7319 * potential PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8
7324 Perl_sv_pos_b2u_flags(pTHX_ SV *const sv, STRLEN const offset, U32 flags)
7327 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7333 PERL_ARGS_ASSERT_SV_POS_B2U_FLAGS;
7335 s = (const U8*)SvPV_flags(sv, blen, flags);
7338 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7339 ", byte=%"UVuf, (UV)blen, (UV)offset);
7345 && SvTYPE(sv) >= SVt_PVMG
7346 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7349 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7350 if (cache[1] == offset) {
7351 /* An exact match. */
7354 if (cache[3] == offset) {
7355 /* An exact match. */
7359 if (cache[1] < offset) {
7360 /* We already know part of the way. */
7361 if (mg->mg_len != -1) {
7362 /* Actually, we know the end too. */
7364 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7365 s + blen, mg->mg_len - cache[0]);
7367 len = cache[0] + utf8_length(s + cache[1], send);
7370 else if (cache[3] < offset) {
7371 /* We're between the two cached pairs, so we do the calculation
7372 offset by the byte/utf-8 positions for the earlier pair,
7373 then add the utf-8 characters from the string start to
7375 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7376 s + cache[1], cache[0] - cache[2])
7380 else { /* cache[3] > offset */
7381 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7385 ASSERT_UTF8_CACHE(cache);
7387 } else if (mg->mg_len != -1) {
7388 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7392 if (!found || PL_utf8cache < 0) {
7393 const STRLEN real_len = utf8_length(s, send);
7395 if (found && PL_utf8cache < 0)
7396 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7402 utf8_mg_len_cache_update(sv, &mg, len);
7404 utf8_mg_pos_cache_update(sv, &mg, offset, len, blen);
7411 =for apidoc sv_pos_b2u
7413 Converts the value pointed to by offsetp from a count of bytes from the
7414 start of the string, to a count of the equivalent number of UTF-8 chars.
7415 Handles magic and type coercion.
7417 Use C<sv_pos_b2u_flags> in preference, which correctly handles strings
7424 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7425 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7430 Perl_sv_pos_b2u(pTHX_ SV *const sv, I32 *const offsetp)
7432 PERL_ARGS_ASSERT_SV_POS_B2U;
7437 *offsetp = (I32)sv_pos_b2u_flags(sv, (STRLEN)*offsetp,
7438 SV_GMAGIC|SV_CONST_RETURN);
7442 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7443 STRLEN real, SV *const sv)
7445 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7447 /* As this is debugging only code, save space by keeping this test here,
7448 rather than inlining it in all the callers. */
7449 if (from_cache == real)
7452 /* Need to turn the assertions off otherwise we may recurse infinitely
7453 while printing error messages. */
7454 SAVEI8(PL_utf8cache);
7456 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7457 func, (UV) from_cache, (UV) real, SVfARG(sv));
7463 Returns a boolean indicating whether the strings in the two SVs are
7464 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7465 coerce its args to strings if necessary.
7467 =for apidoc sv_eq_flags
7469 Returns a boolean indicating whether the strings in the two SVs are
7470 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7471 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7477 Perl_sv_eq_flags(pTHX_ SV *sv1, SV *sv2, const U32 flags)
7485 SV* svrecode = NULL;
7492 /* if pv1 and pv2 are the same, second SvPV_const call may
7493 * invalidate pv1 (if we are handling magic), so we may need to
7495 if (sv1 == sv2 && flags & SV_GMAGIC
7496 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7497 pv1 = SvPV_const(sv1, cur1);
7498 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7500 pv1 = SvPV_flags_const(sv1, cur1, flags);
7508 pv2 = SvPV_flags_const(sv2, cur2, flags);
7510 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7511 /* Differing utf8ness.
7512 * Do not UTF8size the comparands as a side-effect. */
7515 svrecode = newSVpvn(pv2, cur2);
7516 sv_recode_to_utf8(svrecode, PL_encoding);
7517 pv2 = SvPV_const(svrecode, cur2);
7520 svrecode = newSVpvn(pv1, cur1);
7521 sv_recode_to_utf8(svrecode, PL_encoding);
7522 pv1 = SvPV_const(svrecode, cur1);
7524 /* Now both are in UTF-8. */
7526 SvREFCNT_dec_NN(svrecode);
7532 /* sv1 is the UTF-8 one */
7533 return bytes_cmp_utf8((const U8*)pv2, cur2,
7534 (const U8*)pv1, cur1) == 0;
7537 /* sv2 is the UTF-8 one */
7538 return bytes_cmp_utf8((const U8*)pv1, cur1,
7539 (const U8*)pv2, cur2) == 0;
7545 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7547 SvREFCNT_dec(svrecode);
7555 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7556 string in C<sv1> is less than, equal to, or greater than the string in
7557 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7558 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7560 =for apidoc sv_cmp_flags
7562 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7563 string in C<sv1> is less than, equal to, or greater than the string in
7564 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7565 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7566 also C<sv_cmp_locale_flags>.
7572 Perl_sv_cmp(pTHX_ SV *const sv1, SV *const sv2)
7574 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7578 Perl_sv_cmp_flags(pTHX_ SV *const sv1, SV *const sv2,
7583 const char *pv1, *pv2;
7585 SV *svrecode = NULL;
7592 pv1 = SvPV_flags_const(sv1, cur1, flags);
7599 pv2 = SvPV_flags_const(sv2, cur2, flags);
7601 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7602 /* Differing utf8ness.
7603 * Do not UTF8size the comparands as a side-effect. */
7606 svrecode = newSVpvn(pv2, cur2);
7607 sv_recode_to_utf8(svrecode, PL_encoding);
7608 pv2 = SvPV_const(svrecode, cur2);
7611 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7612 (const U8*)pv1, cur1);
7613 return retval ? retval < 0 ? -1 : +1 : 0;
7618 svrecode = newSVpvn(pv1, cur1);
7619 sv_recode_to_utf8(svrecode, PL_encoding);
7620 pv1 = SvPV_const(svrecode, cur1);
7623 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7624 (const U8*)pv2, cur2);
7625 return retval ? retval < 0 ? -1 : +1 : 0;
7631 cmp = cur2 ? -1 : 0;
7635 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7638 cmp = retval < 0 ? -1 : 1;
7639 } else if (cur1 == cur2) {
7642 cmp = cur1 < cur2 ? -1 : 1;
7646 SvREFCNT_dec(svrecode);
7652 =for apidoc sv_cmp_locale
7654 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7655 'use bytes' aware, handles get magic, and will coerce its args to strings
7656 if necessary. See also C<sv_cmp>.
7658 =for apidoc sv_cmp_locale_flags
7660 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7661 'use bytes' aware and will coerce its args to strings if necessary. If the
7662 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7668 Perl_sv_cmp_locale(pTHX_ SV *const sv1, SV *const sv2)
7670 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7674 Perl_sv_cmp_locale_flags(pTHX_ SV *const sv1, SV *const sv2,
7678 #ifdef USE_LOCALE_COLLATE
7684 if (PL_collation_standard)
7688 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7690 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7692 if (!pv1 || !len1) {
7703 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7706 return retval < 0 ? -1 : 1;
7709 * When the result of collation is equality, that doesn't mean
7710 * that there are no differences -- some locales exclude some
7711 * characters from consideration. So to avoid false equalities,
7712 * we use the raw string as a tiebreaker.
7718 #endif /* USE_LOCALE_COLLATE */
7720 return sv_cmp(sv1, sv2);
7724 #ifdef USE_LOCALE_COLLATE
7727 =for apidoc sv_collxfrm
7729 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7730 C<sv_collxfrm_flags>.
7732 =for apidoc sv_collxfrm_flags
7734 Add Collate Transform magic to an SV if it doesn't already have it. If the
7735 flags contain SV_GMAGIC, it handles get-magic.
7737 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7738 scalar data of the variable, but transformed to such a format that a normal
7739 memory comparison can be used to compare the data according to the locale
7746 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7751 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7753 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7754 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7760 Safefree(mg->mg_ptr);
7761 s = SvPV_flags_const(sv, len, flags);
7762 if ((xf = mem_collxfrm(s, len, &xlen))) {
7764 #ifdef PERL_OLD_COPY_ON_WRITE
7766 sv_force_normal_flags(sv, 0);
7768 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7782 if (mg && mg->mg_ptr) {
7784 return mg->mg_ptr + sizeof(PL_collation_ix);
7792 #endif /* USE_LOCALE_COLLATE */
7795 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7797 SV * const tsv = newSV(0);
7800 sv_gets(tsv, fp, 0);
7801 sv_utf8_upgrade_nomg(tsv);
7802 SvCUR_set(sv,append);
7805 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7809 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7812 const STRLEN recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7813 /* Grab the size of the record we're getting */
7814 char *buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7822 /* With a true, record-oriented file on VMS, we need to use read directly
7823 * to ensure that we respect RMS record boundaries. The user is responsible
7824 * for providing a PL_rs value that corresponds to the FAB$W_MRS (maximum
7825 * record size) field. N.B. This is likely to produce invalid results on
7826 * varying-width character data when a record ends mid-character.
7828 fd = PerlIO_fileno(fp);
7830 && PerlLIO_fstat(fd, &st) == 0
7831 && (st.st_fab_rfm == FAB$C_VAR
7832 || st.st_fab_rfm == FAB$C_VFC
7833 || st.st_fab_rfm == FAB$C_FIX)) {
7835 bytesread = PerlLIO_read(fd, buffer, recsize);
7837 else /* in-memory file from PerlIO::Scalar
7838 * or not a record-oriented file
7842 bytesread = PerlIO_read(fp, buffer, recsize);
7844 /* At this point, the logic in sv_get() means that sv will
7845 be treated as utf-8 if the handle is utf8.
7847 if (PerlIO_isutf8(fp) && bytesread > 0) {
7848 char *bend = buffer + bytesread;
7849 char *bufp = buffer;
7850 size_t charcount = 0;
7851 bool charstart = TRUE;
7854 while (charcount < recsize) {
7855 /* count accumulated characters */
7856 while (bufp < bend) {
7858 skip = UTF8SKIP(bufp);
7860 if (bufp + skip > bend) {
7861 /* partial at the end */
7872 if (charcount < recsize) {
7874 STRLEN bufp_offset = bufp - buffer;
7875 SSize_t morebytesread;
7877 /* originally I read enough to fill any incomplete
7878 character and the first byte of the next
7879 character if needed, but if there's many
7880 multi-byte encoded characters we're going to be
7881 making a read call for every character beyond
7882 the original read size.
7884 So instead, read the rest of the character if
7885 any, and enough bytes to match at least the
7886 start bytes for each character we're going to
7890 readsize = recsize - charcount;
7892 readsize = skip - (bend - bufp) + recsize - charcount - 1;
7893 buffer = SvGROW(sv, append + bytesread + readsize + 1) + append;
7894 bend = buffer + bytesread;
7895 morebytesread = PerlIO_read(fp, bend, readsize);
7896 if (morebytesread <= 0) {
7897 /* we're done, if we still have incomplete
7898 characters the check code in sv_gets() will
7901 I'd originally considered doing
7902 PerlIO_ungetc() on all but the lead
7903 character of the incomplete character, but
7904 read() doesn't do that, so I don't.
7909 /* prepare to scan some more */
7910 bytesread += morebytesread;
7911 bend = buffer + bytesread;
7912 bufp = buffer + bufp_offset;
7920 SvCUR_set(sv, bytesread + append);
7921 buffer[bytesread] = '\0';
7922 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7928 Get a line from the filehandle and store it into the SV, optionally
7929 appending to the currently-stored string. If C<append> is not 0, the
7930 line is appended to the SV instead of overwriting it. C<append> should
7931 be set to the byte offset that the appended string should start at
7932 in the SV (typically, C<SvCUR(sv)> is a suitable choice).
7938 Perl_sv_gets(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7949 PERL_ARGS_ASSERT_SV_GETS;
7951 if (SvTHINKFIRST(sv))
7952 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7953 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7955 However, perlbench says it's slower, because the existing swipe code
7956 is faster than copy on write.
7957 Swings and roundabouts. */
7958 SvUPGRADE(sv, SVt_PV);
7961 if (PerlIO_isutf8(fp)) {
7963 sv_utf8_upgrade_nomg(sv);
7964 sv_pos_u2b(sv,&append,0);
7966 } else if (SvUTF8(sv)) {
7967 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7975 if (PerlIO_isutf8(fp))
7978 if (IN_PERL_COMPILETIME) {
7979 /* we always read code in line mode */
7983 else if (RsSNARF(PL_rs)) {
7984 /* If it is a regular disk file use size from stat() as estimate
7985 of amount we are going to read -- may result in mallocing
7986 more memory than we really need if the layers below reduce
7987 the size we read (e.g. CRLF or a gzip layer).
7990 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7991 const Off_t offset = PerlIO_tell(fp);
7992 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7993 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7999 else if (RsRECORD(PL_rs)) {
8000 return S_sv_gets_read_record(aTHX_ sv, fp, append);
8002 else if (RsPARA(PL_rs)) {
8008 /* Get $/ i.e. PL_rs into same encoding as stream wants */
8009 if (PerlIO_isutf8(fp)) {
8010 rsptr = SvPVutf8(PL_rs, rslen);
8013 if (SvUTF8(PL_rs)) {
8014 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
8015 Perl_croak(aTHX_ "Wide character in $/");
8018 rsptr = SvPV_const(PL_rs, rslen);
8022 rslast = rslen ? rsptr[rslen - 1] : '\0';
8024 if (rspara) { /* have to do this both before and after */
8025 do { /* to make sure file boundaries work right */
8028 i = PerlIO_getc(fp);
8032 PerlIO_ungetc(fp,i);
8038 /* See if we know enough about I/O mechanism to cheat it ! */
8040 /* This used to be #ifdef test - it is made run-time test for ease
8041 of abstracting out stdio interface. One call should be cheap
8042 enough here - and may even be a macro allowing compile
8046 if (PerlIO_fast_gets(fp)) {
8049 * We're going to steal some values from the stdio struct
8050 * and put EVERYTHING in the innermost loop into registers.
8056 #if defined(VMS) && defined(PERLIO_IS_STDIO)
8057 /* An ungetc()d char is handled separately from the regular
8058 * buffer, so we getc() it back out and stuff it in the buffer.
8060 i = PerlIO_getc(fp);
8061 if (i == EOF) return 0;
8062 *(--((*fp)->_ptr)) = (unsigned char) i;
8066 /* Here is some breathtakingly efficient cheating */
8068 cnt = PerlIO_get_cnt(fp); /* get count into register */
8069 /* make sure we have the room */
8070 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
8071 /* Not room for all of it
8072 if we are looking for a separator and room for some
8074 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
8075 /* just process what we have room for */
8076 shortbuffered = cnt - SvLEN(sv) + append + 1;
8077 cnt -= shortbuffered;
8081 /* remember that cnt can be negative */
8082 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
8087 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
8088 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
8089 DEBUG_P(PerlIO_printf(Perl_debug_log,
8090 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
8091 DEBUG_P(PerlIO_printf(Perl_debug_log,
8092 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
8093 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
8094 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
8099 while (cnt > 0) { /* this | eat */
8101 if ((*bp++ = *ptr++) == rslast) /* really | dust */
8102 goto thats_all_folks; /* screams | sed :-) */
8106 Copy(ptr, bp, cnt, char); /* this | eat */
8107 bp += cnt; /* screams | dust */
8108 ptr += cnt; /* louder | sed :-) */
8110 assert (!shortbuffered);
8111 goto cannot_be_shortbuffered;
8115 if (shortbuffered) { /* oh well, must extend */
8116 cnt = shortbuffered;
8118 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
8120 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
8121 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
8125 cannot_be_shortbuffered:
8126 DEBUG_P(PerlIO_printf(Perl_debug_log,
8127 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
8128 PTR2UV(ptr),(long)cnt));
8129 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
8131 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
8132 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
8133 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
8134 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8136 /* This used to call 'filbuf' in stdio form, but as that behaves like
8137 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
8138 another abstraction. */
8139 i = PerlIO_getc(fp); /* get more characters */
8141 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
8142 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
8143 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
8144 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8146 cnt = PerlIO_get_cnt(fp);
8147 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
8148 DEBUG_P(PerlIO_printf(Perl_debug_log,
8149 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
8151 if (i == EOF) /* all done for ever? */
8152 goto thats_really_all_folks;
8154 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
8156 SvGROW(sv, bpx + cnt + 2);
8157 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
8159 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
8161 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
8162 goto thats_all_folks;
8166 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
8167 memNE((char*)bp - rslen, rsptr, rslen))
8168 goto screamer; /* go back to the fray */
8169 thats_really_all_folks:
8171 cnt += shortbuffered;
8172 DEBUG_P(PerlIO_printf(Perl_debug_log,
8173 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
8174 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
8175 DEBUG_P(PerlIO_printf(Perl_debug_log,
8176 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
8177 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
8178 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
8180 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
8181 DEBUG_P(PerlIO_printf(Perl_debug_log,
8182 "Screamer: done, len=%ld, string=|%.*s|\n",
8183 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
8187 /*The big, slow, and stupid way. */
8188 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
8189 STDCHAR *buf = NULL;
8190 Newx(buf, 8192, STDCHAR);
8198 const STDCHAR * const bpe = buf + sizeof(buf);
8200 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
8201 ; /* keep reading */
8205 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
8206 /* Accommodate broken VAXC compiler, which applies U8 cast to
8207 * both args of ?: operator, causing EOF to change into 255
8210 i = (U8)buf[cnt - 1];
8216 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
8218 sv_catpvn_nomg(sv, (char *) buf, cnt);
8220 sv_setpvn(sv, (char *) buf, cnt); /* "nomg" is implied */
8222 if (i != EOF && /* joy */
8224 SvCUR(sv) < rslen ||
8225 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
8229 * If we're reading from a TTY and we get a short read,
8230 * indicating that the user hit his EOF character, we need
8231 * to notice it now, because if we try to read from the TTY
8232 * again, the EOF condition will disappear.
8234 * The comparison of cnt to sizeof(buf) is an optimization
8235 * that prevents unnecessary calls to feof().
8239 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
8243 #ifdef USE_HEAP_INSTEAD_OF_STACK
8248 if (rspara) { /* have to do this both before and after */
8249 while (i != EOF) { /* to make sure file boundaries work right */
8250 i = PerlIO_getc(fp);
8252 PerlIO_ungetc(fp,i);
8258 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
8264 Auto-increment of the value in the SV, doing string to numeric conversion
8265 if necessary. Handles 'get' magic and operator overloading.
8271 Perl_sv_inc(pTHX_ SV *const sv)
8280 =for apidoc sv_inc_nomg
8282 Auto-increment of the value in the SV, doing string to numeric conversion
8283 if necessary. Handles operator overloading. Skips handling 'get' magic.
8289 Perl_sv_inc_nomg(pTHX_ SV *const sv)
8297 if (SvTHINKFIRST(sv)) {
8298 if (SvREADONLY(sv)) {
8299 Perl_croak_no_modify();
8303 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
8305 i = PTR2IV(SvRV(sv));
8309 else sv_force_normal_flags(sv, 0);
8311 flags = SvFLAGS(sv);
8312 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
8313 /* It's (privately or publicly) a float, but not tested as an
8314 integer, so test it to see. */
8316 flags = SvFLAGS(sv);
8318 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8319 /* It's publicly an integer, or privately an integer-not-float */
8320 #ifdef PERL_PRESERVE_IVUV
8324 if (SvUVX(sv) == UV_MAX)
8325 sv_setnv(sv, UV_MAX_P1);
8327 (void)SvIOK_only_UV(sv);
8328 SvUV_set(sv, SvUVX(sv) + 1);
8330 if (SvIVX(sv) == IV_MAX)
8331 sv_setuv(sv, (UV)IV_MAX + 1);
8333 (void)SvIOK_only(sv);
8334 SvIV_set(sv, SvIVX(sv) + 1);
8339 if (flags & SVp_NOK) {
8340 const NV was = SvNVX(sv);
8341 if (NV_OVERFLOWS_INTEGERS_AT &&
8342 was >= NV_OVERFLOWS_INTEGERS_AT) {
8343 /* diag_listed_as: Lost precision when %s %f by 1 */
8344 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8345 "Lost precision when incrementing %" NVff " by 1",
8348 (void)SvNOK_only(sv);
8349 SvNV_set(sv, was + 1.0);
8353 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
8354 if ((flags & SVTYPEMASK) < SVt_PVIV)
8355 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
8356 (void)SvIOK_only(sv);
8361 while (isALPHA(*d)) d++;
8362 while (isDIGIT(*d)) d++;
8363 if (d < SvEND(sv)) {
8364 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8365 #ifdef PERL_PRESERVE_IVUV
8366 /* Got to punt this as an integer if needs be, but we don't issue
8367 warnings. Probably ought to make the sv_iv_please() that does
8368 the conversion if possible, and silently. */
8369 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8370 /* Need to try really hard to see if it's an integer.
8371 9.22337203685478e+18 is an integer.
8372 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8373 so $a="9.22337203685478e+18"; $a+0; $a++
8374 needs to be the same as $a="9.22337203685478e+18"; $a++
8381 /* sv_2iv *should* have made this an NV */
8382 if (flags & SVp_NOK) {
8383 (void)SvNOK_only(sv);
8384 SvNV_set(sv, SvNVX(sv) + 1.0);
8387 /* I don't think we can get here. Maybe I should assert this
8388 And if we do get here I suspect that sv_setnv will croak. NWC
8390 #if defined(USE_LONG_DOUBLE)
8391 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
8392 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8394 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8395 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8398 #endif /* PERL_PRESERVE_IVUV */
8399 if (!numtype && ckWARN(WARN_NUMERIC))
8400 not_incrementable(sv);
8401 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8405 while (d >= SvPVX_const(sv)) {
8413 /* MKS: The original code here died if letters weren't consecutive.
8414 * at least it didn't have to worry about non-C locales. The
8415 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8416 * arranged in order (although not consecutively) and that only
8417 * [A-Za-z] are accepted by isALPHA in the C locale.
8419 if (*d != 'z' && *d != 'Z') {
8420 do { ++*d; } while (!isALPHA(*d));
8423 *(d--) -= 'z' - 'a';
8428 *(d--) -= 'z' - 'a' + 1;
8432 /* oh,oh, the number grew */
8433 SvGROW(sv, SvCUR(sv) + 2);
8434 SvCUR_set(sv, SvCUR(sv) + 1);
8435 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8446 Auto-decrement of the value in the SV, doing string to numeric conversion
8447 if necessary. Handles 'get' magic and operator overloading.
8453 Perl_sv_dec(pTHX_ SV *const sv)
8463 =for apidoc sv_dec_nomg
8465 Auto-decrement of the value in the SV, doing string to numeric conversion
8466 if necessary. Handles operator overloading. Skips handling 'get' magic.
8472 Perl_sv_dec_nomg(pTHX_ SV *const sv)
8479 if (SvTHINKFIRST(sv)) {
8480 if (SvREADONLY(sv)) {
8481 Perl_croak_no_modify();
8485 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8487 i = PTR2IV(SvRV(sv));
8491 else sv_force_normal_flags(sv, 0);
8493 /* Unlike sv_inc we don't have to worry about string-never-numbers
8494 and keeping them magic. But we mustn't warn on punting */
8495 flags = SvFLAGS(sv);
8496 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8497 /* It's publicly an integer, or privately an integer-not-float */
8498 #ifdef PERL_PRESERVE_IVUV
8502 if (SvUVX(sv) == 0) {
8503 (void)SvIOK_only(sv);
8507 (void)SvIOK_only_UV(sv);
8508 SvUV_set(sv, SvUVX(sv) - 1);
8511 if (SvIVX(sv) == IV_MIN) {
8512 sv_setnv(sv, (NV)IV_MIN);
8516 (void)SvIOK_only(sv);
8517 SvIV_set(sv, SvIVX(sv) - 1);
8522 if (flags & SVp_NOK) {
8525 const NV was = SvNVX(sv);
8526 if (NV_OVERFLOWS_INTEGERS_AT &&
8527 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8528 /* diag_listed_as: Lost precision when %s %f by 1 */
8529 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8530 "Lost precision when decrementing %" NVff " by 1",
8533 (void)SvNOK_only(sv);
8534 SvNV_set(sv, was - 1.0);
8538 if (!(flags & SVp_POK)) {
8539 if ((flags & SVTYPEMASK) < SVt_PVIV)
8540 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8542 (void)SvIOK_only(sv);
8545 #ifdef PERL_PRESERVE_IVUV
8547 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8548 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8549 /* Need to try really hard to see if it's an integer.
8550 9.22337203685478e+18 is an integer.
8551 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8552 so $a="9.22337203685478e+18"; $a+0; $a--
8553 needs to be the same as $a="9.22337203685478e+18"; $a--
8560 /* sv_2iv *should* have made this an NV */
8561 if (flags & SVp_NOK) {
8562 (void)SvNOK_only(sv);
8563 SvNV_set(sv, SvNVX(sv) - 1.0);
8566 /* I don't think we can get here. Maybe I should assert this
8567 And if we do get here I suspect that sv_setnv will croak. NWC
8569 #if defined(USE_LONG_DOUBLE)
8570 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
8571 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8573 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8574 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8578 #endif /* PERL_PRESERVE_IVUV */
8579 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8582 /* this define is used to eliminate a chunk of duplicated but shared logic
8583 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8584 * used anywhere but here - yves
8586 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8589 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8593 =for apidoc sv_mortalcopy
8595 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8596 The new SV is marked as mortal. It will be destroyed "soon", either by an
8597 explicit call to FREETMPS, or by an implicit call at places such as
8598 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8603 /* Make a string that will exist for the duration of the expression
8604 * evaluation. Actually, it may have to last longer than that, but
8605 * hopefully we won't free it until it has been assigned to a
8606 * permanent location. */
8609 Perl_sv_mortalcopy_flags(pTHX_ SV *const oldstr, U32 flags)
8614 if (flags & SV_GMAGIC)
8615 SvGETMAGIC(oldstr); /* before new_SV, in case it dies */
8617 sv_setsv_flags(sv,oldstr,flags & ~SV_GMAGIC);
8618 PUSH_EXTEND_MORTAL__SV_C(sv);
8624 =for apidoc sv_newmortal
8626 Creates a new null SV which is mortal. The reference count of the SV is
8627 set to 1. It will be destroyed "soon", either by an explicit call to
8628 FREETMPS, or by an implicit call at places such as statement boundaries.
8629 See also C<sv_mortalcopy> and C<sv_2mortal>.
8635 Perl_sv_newmortal(pTHX)
8641 SvFLAGS(sv) = SVs_TEMP;
8642 PUSH_EXTEND_MORTAL__SV_C(sv);
8648 =for apidoc newSVpvn_flags
8650 Creates a new SV and copies a string into it. The reference count for the
8651 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8652 string. You are responsible for ensuring that the source string is at least
8653 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8654 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8655 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8656 returning. If C<SVf_UTF8> is set, C<s>
8657 is considered to be in UTF-8 and the
8658 C<SVf_UTF8> flag will be set on the new SV.
8659 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8661 #define newSVpvn_utf8(s, len, u) \
8662 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8668 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8673 /* All the flags we don't support must be zero.
8674 And we're new code so I'm going to assert this from the start. */
8675 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8677 sv_setpvn(sv,s,len);
8679 /* This code used to do a sv_2mortal(), however we now unroll the call to
8680 * sv_2mortal() and do what it does ourselves here. Since we have asserted
8681 * that flags can only have the SVf_UTF8 and/or SVs_TEMP flags set above we
8682 * can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8683 * in turn means we dont need to mask out the SVf_UTF8 flag below, which
8684 * means that we eliminate quite a few steps than it looks - Yves
8685 * (explaining patch by gfx) */
8687 SvFLAGS(sv) |= flags;
8689 if(flags & SVs_TEMP){
8690 PUSH_EXTEND_MORTAL__SV_C(sv);
8697 =for apidoc sv_2mortal
8699 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8700 by an explicit call to FREETMPS, or by an implicit call at places such as
8701 statement boundaries. SvTEMP() is turned on which means that the SV's
8702 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8703 and C<sv_mortalcopy>.
8709 Perl_sv_2mortal(pTHX_ SV *const sv)
8716 PUSH_EXTEND_MORTAL__SV_C(sv);
8724 Creates a new SV and copies a string into it. The reference count for the
8725 SV is set to 1. If C<len> is zero, Perl will compute the length using
8726 strlen(). For efficiency, consider using C<newSVpvn> instead.
8732 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8738 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8743 =for apidoc newSVpvn
8745 Creates a new SV and copies a buffer into it, which may contain NUL characters
8746 (C<\0>) and other binary data. The reference count for the SV is set to 1.
8747 Note that if C<len> is zero, Perl will create a zero length (Perl) string. You
8748 are responsible for ensuring that the source buffer is at least
8749 C<len> bytes long. If the C<buffer> argument is NULL the new SV will be
8756 Perl_newSVpvn(pTHX_ const char *const buffer, const STRLEN len)
8762 sv_setpvn(sv,buffer,len);
8767 =for apidoc newSVhek
8769 Creates a new SV from the hash key structure. It will generate scalars that
8770 point to the shared string table where possible. Returns a new (undefined)
8771 SV if the hek is NULL.
8777 Perl_newSVhek(pTHX_ const HEK *const hek)
8787 if (HEK_LEN(hek) == HEf_SVKEY) {
8788 return newSVsv(*(SV**)HEK_KEY(hek));
8790 const int flags = HEK_FLAGS(hek);
8791 if (flags & HVhek_WASUTF8) {
8793 Andreas would like keys he put in as utf8 to come back as utf8
8795 STRLEN utf8_len = HEK_LEN(hek);
8796 SV * const sv = newSV_type(SVt_PV);
8797 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8798 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8799 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8802 } else if (flags & HVhek_UNSHARED) {
8803 /* A hash that isn't using shared hash keys has to have
8804 the flag in every key so that we know not to try to call
8805 share_hek_hek on it. */
8807 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8812 /* This will be overwhelminly the most common case. */
8814 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8815 more efficient than sharepvn(). */
8819 sv_upgrade(sv, SVt_PV);
8820 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8821 SvCUR_set(sv, HEK_LEN(hek));
8833 =for apidoc newSVpvn_share
8835 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8836 table. If the string does not already exist in the table, it is
8837 created first. Turns on the SvIsCOW flag (or READONLY
8838 and FAKE in 5.16 and earlier). If the C<hash> parameter
8839 is non-zero, that value is used; otherwise the hash is computed.
8840 The string's hash can later be retrieved from the SV
8841 with the C<SvSHARED_HASH()> macro. The idea here is
8842 that as the string table is used for shared hash keys these strings will have
8843 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8849 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8853 bool is_utf8 = FALSE;
8854 const char *const orig_src = src;
8857 STRLEN tmplen = -len;
8859 /* See the note in hv.c:hv_fetch() --jhi */
8860 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8864 PERL_HASH(hash, src, len);
8866 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8867 changes here, update it there too. */
8868 sv_upgrade(sv, SVt_PV);
8869 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8876 if (src != orig_src)
8882 =for apidoc newSVpv_share
8884 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8891 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8893 return newSVpvn_share(src, strlen(src), hash);
8896 #if defined(PERL_IMPLICIT_CONTEXT)
8898 /* pTHX_ magic can't cope with varargs, so this is a no-context
8899 * version of the main function, (which may itself be aliased to us).
8900 * Don't access this version directly.
8904 Perl_newSVpvf_nocontext(const char *const pat, ...)
8910 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8912 va_start(args, pat);
8913 sv = vnewSVpvf(pat, &args);
8920 =for apidoc newSVpvf
8922 Creates a new SV and initializes it with the string formatted like
8929 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8934 PERL_ARGS_ASSERT_NEWSVPVF;
8936 va_start(args, pat);
8937 sv = vnewSVpvf(pat, &args);
8942 /* backend for newSVpvf() and newSVpvf_nocontext() */
8945 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8950 PERL_ARGS_ASSERT_VNEWSVPVF;
8953 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8960 Creates a new SV and copies a floating point value into it.
8961 The reference count for the SV is set to 1.
8967 Perl_newSVnv(pTHX_ const NV n)
8980 Creates a new SV and copies an integer into it. The reference count for the
8987 Perl_newSViv(pTHX_ const IV i)
9000 Creates a new SV and copies an unsigned integer into it.
9001 The reference count for the SV is set to 1.
9007 Perl_newSVuv(pTHX_ const UV u)
9018 =for apidoc newSV_type
9020 Creates a new SV, of the type specified. The reference count for the new SV
9027 Perl_newSV_type(pTHX_ const svtype type)
9032 sv_upgrade(sv, type);
9037 =for apidoc newRV_noinc
9039 Creates an RV wrapper for an SV. The reference count for the original
9040 SV is B<not> incremented.
9046 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
9049 SV *sv = newSV_type(SVt_IV);
9051 PERL_ARGS_ASSERT_NEWRV_NOINC;
9054 SvRV_set(sv, tmpRef);
9059 /* newRV_inc is the official function name to use now.
9060 * newRV_inc is in fact #defined to newRV in sv.h
9064 Perl_newRV(pTHX_ SV *const sv)
9068 PERL_ARGS_ASSERT_NEWRV;
9070 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
9076 Creates a new SV which is an exact duplicate of the original SV.
9083 Perl_newSVsv(pTHX_ SV *const old)
9090 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
9091 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
9094 /* Do this here, otherwise we leak the new SV if this croaks. */
9097 /* SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
9098 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
9099 sv_setsv_flags(sv, old, SV_NOSTEAL);
9104 =for apidoc sv_reset
9106 Underlying implementation for the C<reset> Perl function.
9107 Note that the perl-level function is vaguely deprecated.
9113 Perl_sv_reset(pTHX_ const char *s, HV *const stash)
9115 PERL_ARGS_ASSERT_SV_RESET;
9117 sv_resetpvn(*s ? s : NULL, strlen(s), stash);
9121 Perl_sv_resetpvn(pTHX_ const char *s, STRLEN len, HV * const stash)
9124 char todo[PERL_UCHAR_MAX+1];
9127 if (!stash || SvTYPE(stash) != SVt_PVHV)
9130 if (!s) { /* reset ?? searches */
9131 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
9133 const U32 count = mg->mg_len / sizeof(PMOP**);
9134 PMOP **pmp = (PMOP**) mg->mg_ptr;
9135 PMOP *const *const end = pmp + count;
9139 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
9141 (*pmp)->op_pmflags &= ~PMf_USED;
9149 /* reset variables */
9151 if (!HvARRAY(stash))
9154 Zero(todo, 256, char);
9158 I32 i = (unsigned char)*s;
9162 max = (unsigned char)*s++;
9163 for ( ; i <= max; i++) {
9166 for (i = 0; i <= (I32) HvMAX(stash); i++) {
9168 for (entry = HvARRAY(stash)[i];
9170 entry = HeNEXT(entry))
9175 if (!todo[(U8)*HeKEY(entry)])
9177 gv = MUTABLE_GV(HeVAL(entry));
9179 if (sv && !SvREADONLY(sv)) {
9180 SV_CHECK_THINKFIRST_COW_DROP(sv);
9181 if (!isGV(sv)) SvOK_off(sv);
9186 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
9197 Using various gambits, try to get an IO from an SV: the IO slot if its a
9198 GV; or the recursive result if we're an RV; or the IO slot of the symbol
9199 named after the PV if we're a string.
9201 'Get' magic is ignored on the sv passed in, but will be called on
9202 C<SvRV(sv)> if sv is an RV.
9208 Perl_sv_2io(pTHX_ SV *const sv)
9213 PERL_ARGS_ASSERT_SV_2IO;
9215 switch (SvTYPE(sv)) {
9217 io = MUTABLE_IO(sv);
9221 if (isGV_with_GP(sv)) {
9222 gv = MUTABLE_GV(sv);
9225 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
9226 HEKfARG(GvNAME_HEK(gv)));
9232 Perl_croak(aTHX_ PL_no_usym, "filehandle");
9234 SvGETMAGIC(SvRV(sv));
9235 return sv_2io(SvRV(sv));
9237 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
9244 if (SvGMAGICAL(sv)) {
9245 newsv = sv_newmortal();
9246 sv_setsv_nomg(newsv, sv);
9248 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
9258 Using various gambits, try to get a CV from an SV; in addition, try if
9259 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
9260 The flags in C<lref> are passed to gv_fetchsv.
9266 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
9272 PERL_ARGS_ASSERT_SV_2CV;
9279 switch (SvTYPE(sv)) {
9283 return MUTABLE_CV(sv);
9293 sv = amagic_deref_call(sv, to_cv_amg);
9296 if (SvTYPE(sv) == SVt_PVCV) {
9297 cv = MUTABLE_CV(sv);
9302 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
9303 gv = MUTABLE_GV(sv);
9305 Perl_croak(aTHX_ "Not a subroutine reference");
9307 else if (isGV_with_GP(sv)) {
9308 gv = MUTABLE_GV(sv);
9311 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
9318 /* Some flags to gv_fetchsv mean don't really create the GV */
9319 if (!isGV_with_GP(gv)) {
9324 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
9325 /* XXX this is probably not what they think they're getting.
9326 * It has the same effect as "sub name;", i.e. just a forward
9337 Returns true if the SV has a true value by Perl's rules.
9338 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9339 instead use an in-line version.
9345 Perl_sv_true(pTHX_ SV *const sv)
9350 const XPV* const tXpv = (XPV*)SvANY(sv);
9352 (tXpv->xpv_cur > 1 ||
9353 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9360 return SvIVX(sv) != 0;
9363 return SvNVX(sv) != 0.0;
9365 return sv_2bool(sv);
9371 =for apidoc sv_pvn_force
9373 Get a sensible string out of the SV somehow.
9374 A private implementation of the C<SvPV_force> macro for compilers which
9375 can't cope with complex macro expressions. Always use the macro instead.
9377 =for apidoc sv_pvn_force_flags
9379 Get a sensible string out of the SV somehow.
9380 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9381 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9382 implemented in terms of this function.
9383 You normally want to use the various wrapper macros instead: see
9384 C<SvPV_force> and C<SvPV_force_nomg>
9390 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9394 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9396 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9397 if (SvTHINKFIRST(sv) && (!SvROK(sv) || SvREADONLY(sv)))
9398 sv_force_normal_flags(sv, 0);
9408 if (SvTYPE(sv) > SVt_PVLV
9409 || isGV_with_GP(sv))
9410 /* diag_listed_as: Can't coerce %s to %s in %s */
9411 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9413 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9420 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9423 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9424 SvGROW(sv, len + 1);
9425 Move(s,SvPVX(sv),len,char);
9427 SvPVX(sv)[len] = '\0';
9430 SvPOK_on(sv); /* validate pointer */
9432 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9433 PTR2UV(sv),SvPVX_const(sv)));
9436 (void)SvPOK_only_UTF8(sv);
9437 return SvPVX_mutable(sv);
9441 =for apidoc sv_pvbyten_force
9443 The backend for the C<SvPVbytex_force> macro. Always use the macro
9450 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9452 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9454 sv_pvn_force(sv,lp);
9455 sv_utf8_downgrade(sv,0);
9461 =for apidoc sv_pvutf8n_force
9463 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9470 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9472 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9475 sv_utf8_upgrade_nomg(sv);
9481 =for apidoc sv_reftype
9483 Returns a string describing what the SV is a reference to.
9489 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9491 PERL_ARGS_ASSERT_SV_REFTYPE;
9492 if (ob && SvOBJECT(sv)) {
9493 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9496 switch (SvTYPE(sv)) {
9511 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9512 /* tied lvalues should appear to be
9513 * scalars for backwards compatibility */
9514 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9515 ? "SCALAR" : "LVALUE");
9516 case SVt_PVAV: return "ARRAY";
9517 case SVt_PVHV: return "HASH";
9518 case SVt_PVCV: return "CODE";
9519 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9520 ? "GLOB" : "SCALAR");
9521 case SVt_PVFM: return "FORMAT";
9522 case SVt_PVIO: return "IO";
9523 case SVt_INVLIST: return "INVLIST";
9524 case SVt_REGEXP: return "REGEXP";
9525 default: return "UNKNOWN";
9533 Returns a SV describing what the SV passed in is a reference to.
9539 Perl_sv_ref(pTHX_ SV *dst, const SV *const sv, const int ob)
9541 PERL_ARGS_ASSERT_SV_REF;
9544 dst = sv_newmortal();
9546 if (ob && SvOBJECT(sv)) {
9547 HvNAME_get(SvSTASH(sv))
9548 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9549 : sv_setpvn(dst, "__ANON__", 8);
9552 const char * reftype = sv_reftype(sv, 0);
9553 sv_setpv(dst, reftype);
9559 =for apidoc sv_isobject
9561 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9562 object. If the SV is not an RV, or if the object is not blessed, then this
9569 Perl_sv_isobject(pTHX_ SV *sv)
9585 Returns a boolean indicating whether the SV is blessed into the specified
9586 class. This does not check for subtypes; use C<sv_derived_from> to verify
9587 an inheritance relationship.
9593 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9597 PERL_ARGS_ASSERT_SV_ISA;
9607 hvname = HvNAME_get(SvSTASH(sv));
9611 return strEQ(hvname, name);
9617 Creates a new SV for the existing RV, C<rv>, to point to. If C<rv> is not an
9618 RV then it will be upgraded to one. If C<classname> is non-null then the new
9619 SV will be blessed in the specified package. The new SV is returned and its
9620 reference count is 1. The reference count 1 is owned by C<rv>.
9626 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9631 PERL_ARGS_ASSERT_NEWSVRV;
9635 SV_CHECK_THINKFIRST_COW_DROP(rv);
9637 if (SvTYPE(rv) >= SVt_PVMG) {
9638 const U32 refcnt = SvREFCNT(rv);
9642 SvREFCNT(rv) = refcnt;
9644 sv_upgrade(rv, SVt_IV);
9645 } else if (SvROK(rv)) {
9646 SvREFCNT_dec(SvRV(rv));
9648 prepare_SV_for_RV(rv);
9656 HV* const stash = gv_stashpv(classname, GV_ADD);
9657 (void)sv_bless(rv, stash);
9663 =for apidoc sv_setref_pv
9665 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9666 argument will be upgraded to an RV. That RV will be modified to point to
9667 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9668 into the SV. The C<classname> argument indicates the package for the
9669 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9670 will have a reference count of 1, and the RV will be returned.
9672 Do not use with other Perl types such as HV, AV, SV, CV, because those
9673 objects will become corrupted by the pointer copy process.
9675 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9681 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9685 PERL_ARGS_ASSERT_SV_SETREF_PV;
9688 sv_setsv(rv, &PL_sv_undef);
9692 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9697 =for apidoc sv_setref_iv
9699 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9700 argument will be upgraded to an RV. That RV will be modified to point to
9701 the new SV. The C<classname> argument indicates the package for the
9702 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9703 will have a reference count of 1, and the RV will be returned.
9709 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9711 PERL_ARGS_ASSERT_SV_SETREF_IV;
9713 sv_setiv(newSVrv(rv,classname), iv);
9718 =for apidoc sv_setref_uv
9720 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9721 argument will be upgraded to an RV. That RV will be modified to point to
9722 the new SV. The C<classname> argument indicates the package for the
9723 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9724 will have a reference count of 1, and the RV will be returned.
9730 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9732 PERL_ARGS_ASSERT_SV_SETREF_UV;
9734 sv_setuv(newSVrv(rv,classname), uv);
9739 =for apidoc sv_setref_nv
9741 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9742 argument will be upgraded to an RV. That RV will be modified to point to
9743 the new SV. The C<classname> argument indicates the package for the
9744 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9745 will have a reference count of 1, and the RV will be returned.
9751 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9753 PERL_ARGS_ASSERT_SV_SETREF_NV;
9755 sv_setnv(newSVrv(rv,classname), nv);
9760 =for apidoc sv_setref_pvn
9762 Copies a string into a new SV, optionally blessing the SV. The length of the
9763 string must be specified with C<n>. The C<rv> argument will be upgraded to
9764 an RV. That RV will be modified to point to the new SV. The C<classname>
9765 argument indicates the package for the blessing. Set C<classname> to
9766 C<NULL> to avoid the blessing. The new SV will have a reference count
9767 of 1, and the RV will be returned.
9769 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9775 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9776 const char *const pv, const STRLEN n)
9778 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9780 sv_setpvn(newSVrv(rv,classname), pv, n);
9785 =for apidoc sv_bless
9787 Blesses an SV into a specified package. The SV must be an RV. The package
9788 must be designated by its stash (see C<gv_stashpv()>). The reference count
9789 of the SV is unaffected.
9795 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9800 PERL_ARGS_ASSERT_SV_BLESS;
9804 Perl_croak(aTHX_ "Can't bless non-reference value");
9806 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9807 if (SvREADONLY(tmpRef))
9808 Perl_croak_no_modify();
9809 if (SvOBJECT(tmpRef)) {
9810 SvREFCNT_dec(SvSTASH(tmpRef));
9813 SvOBJECT_on(tmpRef);
9814 SvUPGRADE(tmpRef, SVt_PVMG);
9815 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9817 if(SvSMAGICAL(tmpRef))
9818 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9826 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9827 * as it is after unglobbing it.
9830 PERL_STATIC_INLINE void
9831 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9836 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9838 PERL_ARGS_ASSERT_SV_UNGLOB;
9840 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9842 if (!(flags & SV_COW_DROP_PV))
9843 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9846 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9847 && HvNAME_get(stash))
9848 mro_method_changed_in(stash);
9849 gp_free(MUTABLE_GV(sv));
9852 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9856 if (GvNAME_HEK(sv)) {
9857 unshare_hek(GvNAME_HEK(sv));
9859 isGV_with_GP_off(sv);
9861 if(SvTYPE(sv) == SVt_PVGV) {
9862 /* need to keep SvANY(sv) in the right arena */
9863 xpvmg = new_XPVMG();
9864 StructCopy(SvANY(sv), xpvmg, XPVMG);
9865 del_XPVGV(SvANY(sv));
9868 SvFLAGS(sv) &= ~SVTYPEMASK;
9869 SvFLAGS(sv) |= SVt_PVMG;
9872 /* Intentionally not calling any local SET magic, as this isn't so much a
9873 set operation as merely an internal storage change. */
9874 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9875 else sv_setsv_flags(sv, temp, 0);
9877 if ((const GV *)sv == PL_last_in_gv)
9878 PL_last_in_gv = NULL;
9879 else if ((const GV *)sv == PL_statgv)
9884 =for apidoc sv_unref_flags
9886 Unsets the RV status of the SV, and decrements the reference count of
9887 whatever was being referenced by the RV. This can almost be thought of
9888 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9889 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9890 (otherwise the decrementing is conditional on the reference count being
9891 different from one or the reference being a readonly SV).
9898 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9900 SV* const target = SvRV(ref);
9902 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9904 if (SvWEAKREF(ref)) {
9905 sv_del_backref(target, ref);
9907 SvRV_set(ref, NULL);
9910 SvRV_set(ref, NULL);
9912 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9913 assigned to as BEGIN {$a = \"Foo"} will fail. */
9914 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9915 SvREFCNT_dec_NN(target);
9916 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9917 sv_2mortal(target); /* Schedule for freeing later */
9921 =for apidoc sv_untaint
9923 Untaint an SV. Use C<SvTAINTED_off> instead.
9929 Perl_sv_untaint(pTHX_ SV *const sv)
9931 PERL_ARGS_ASSERT_SV_UNTAINT;
9933 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9934 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9941 =for apidoc sv_tainted
9943 Test an SV for taintedness. Use C<SvTAINTED> instead.
9949 Perl_sv_tainted(pTHX_ SV *const sv)
9951 PERL_ARGS_ASSERT_SV_TAINTED;
9953 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9954 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9955 if (mg && (mg->mg_len & 1) )
9962 =for apidoc sv_setpviv
9964 Copies an integer into the given SV, also updating its string value.
9965 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9971 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9973 char buf[TYPE_CHARS(UV)];
9975 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9977 PERL_ARGS_ASSERT_SV_SETPVIV;
9979 sv_setpvn(sv, ptr, ebuf - ptr);
9983 =for apidoc sv_setpviv_mg
9985 Like C<sv_setpviv>, but also handles 'set' magic.
9991 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9993 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9999 #if defined(PERL_IMPLICIT_CONTEXT)
10001 /* pTHX_ magic can't cope with varargs, so this is a no-context
10002 * version of the main function, (which may itself be aliased to us).
10003 * Don't access this version directly.
10007 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
10012 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
10014 va_start(args, pat);
10015 sv_vsetpvf(sv, pat, &args);
10019 /* pTHX_ magic can't cope with varargs, so this is a no-context
10020 * version of the main function, (which may itself be aliased to us).
10021 * Don't access this version directly.
10025 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
10030 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
10032 va_start(args, pat);
10033 sv_vsetpvf_mg(sv, pat, &args);
10039 =for apidoc sv_setpvf
10041 Works like C<sv_catpvf> but copies the text into the SV instead of
10042 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
10048 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
10052 PERL_ARGS_ASSERT_SV_SETPVF;
10054 va_start(args, pat);
10055 sv_vsetpvf(sv, pat, &args);
10060 =for apidoc sv_vsetpvf
10062 Works like C<sv_vcatpvf> but copies the text into the SV instead of
10063 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
10065 Usually used via its frontend C<sv_setpvf>.
10071 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10073 PERL_ARGS_ASSERT_SV_VSETPVF;
10075 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10079 =for apidoc sv_setpvf_mg
10081 Like C<sv_setpvf>, but also handles 'set' magic.
10087 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
10091 PERL_ARGS_ASSERT_SV_SETPVF_MG;
10093 va_start(args, pat);
10094 sv_vsetpvf_mg(sv, pat, &args);
10099 =for apidoc sv_vsetpvf_mg
10101 Like C<sv_vsetpvf>, but also handles 'set' magic.
10103 Usually used via its frontend C<sv_setpvf_mg>.
10109 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10111 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
10113 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10117 #if defined(PERL_IMPLICIT_CONTEXT)
10119 /* pTHX_ magic can't cope with varargs, so this is a no-context
10120 * version of the main function, (which may itself be aliased to us).
10121 * Don't access this version directly.
10125 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
10130 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
10132 va_start(args, pat);
10133 sv_vcatpvf(sv, pat, &args);
10137 /* pTHX_ magic can't cope with varargs, so this is a no-context
10138 * version of the main function, (which may itself be aliased to us).
10139 * Don't access this version directly.
10143 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
10148 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
10150 va_start(args, pat);
10151 sv_vcatpvf_mg(sv, pat, &args);
10157 =for apidoc sv_catpvf
10159 Processes its arguments like C<sprintf> and appends the formatted
10160 output to an SV. If the appended data contains "wide" characters
10161 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
10162 and characters >255 formatted with %c), the original SV might get
10163 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
10164 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
10165 valid UTF-8; if the original SV was bytes, the pattern should be too.
10170 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
10174 PERL_ARGS_ASSERT_SV_CATPVF;
10176 va_start(args, pat);
10177 sv_vcatpvf(sv, pat, &args);
10182 =for apidoc sv_vcatpvf
10184 Processes its arguments like C<vsprintf> and appends the formatted output
10185 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
10187 Usually used via its frontend C<sv_catpvf>.
10193 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10195 PERL_ARGS_ASSERT_SV_VCATPVF;
10197 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10201 =for apidoc sv_catpvf_mg
10203 Like C<sv_catpvf>, but also handles 'set' magic.
10209 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
10213 PERL_ARGS_ASSERT_SV_CATPVF_MG;
10215 va_start(args, pat);
10216 sv_vcatpvf_mg(sv, pat, &args);
10221 =for apidoc sv_vcatpvf_mg
10223 Like C<sv_vcatpvf>, but also handles 'set' magic.
10225 Usually used via its frontend C<sv_catpvf_mg>.
10231 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
10233 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
10235 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
10240 =for apidoc sv_vsetpvfn
10242 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
10245 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
10251 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10252 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10254 PERL_ARGS_ASSERT_SV_VSETPVFN;
10257 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, 0);
10262 * Warn of missing argument to sprintf, and then return a defined value
10263 * to avoid inappropriate "use of uninit" warnings [perl #71000].
10265 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
10267 S_vcatpvfn_missing_argument(pTHX) {
10268 if (ckWARN(WARN_MISSING)) {
10269 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
10270 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
10277 S_expect_number(pTHX_ char **const pattern)
10282 PERL_ARGS_ASSERT_EXPECT_NUMBER;
10284 switch (**pattern) {
10285 case '1': case '2': case '3':
10286 case '4': case '5': case '6':
10287 case '7': case '8': case '9':
10288 var = *(*pattern)++ - '0';
10289 while (isDIGIT(**pattern)) {
10290 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
10292 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
10300 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
10302 const int neg = nv < 0;
10305 PERL_ARGS_ASSERT_F0CONVERT;
10313 if (uv & 1 && uv == nv)
10314 uv--; /* Round to even */
10316 const unsigned dig = uv % 10;
10318 } while (uv /= 10);
10329 =for apidoc sv_vcatpvfn
10331 =for apidoc sv_vcatpvfn_flags
10333 Processes its arguments like C<vsprintf> and appends the formatted output
10334 to an SV. Uses an array of SVs if the C style variable argument list is
10335 missing (NULL). When running with taint checks enabled, indicates via
10336 C<maybe_tainted> if results are untrustworthy (often due to the use of
10339 If called as C<sv_vcatpvfn> or flags include C<SV_GMAGIC>, calls get magic.
10341 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10346 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10347 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10348 vec_utf8 = DO_UTF8(vecsv);
10350 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10353 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10354 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10356 PERL_ARGS_ASSERT_SV_VCATPVFN;
10358 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, SV_GMAGIC|SV_SMAGIC);
10362 Perl_sv_vcatpvfn_flags(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10363 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted,
10369 const char *patend;
10372 static const char nullstr[] = "(null)";
10374 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10375 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10377 /* Times 4: a decimal digit takes more than 3 binary digits.
10378 * NV_DIG: mantissa takes than many decimal digits.
10379 * Plus 32: Playing safe. */
10380 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10381 /* large enough for "%#.#f" --chip */
10382 /* what about long double NVs? --jhi */
10384 PERL_ARGS_ASSERT_SV_VCATPVFN_FLAGS;
10385 PERL_UNUSED_ARG(maybe_tainted);
10387 if (flags & SV_GMAGIC)
10390 /* no matter what, this is a string now */
10391 (void)SvPV_force_nomg(sv, origlen);
10393 /* special-case "", "%s", and "%-p" (SVf - see below) */
10396 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10398 const char * const s = va_arg(*args, char*);
10399 sv_catpv_nomg(sv, s ? s : nullstr);
10401 else if (svix < svmax) {
10402 /* we want get magic on the source but not the target. sv_catsv can't do that, though */
10403 SvGETMAGIC(*svargs);
10404 sv_catsv_nomg(sv, *svargs);
10407 S_vcatpvfn_missing_argument(aTHX);
10410 if (args && patlen == 3 && pat[0] == '%' &&
10411 pat[1] == '-' && pat[2] == 'p') {
10412 argsv = MUTABLE_SV(va_arg(*args, void*));
10413 sv_catsv_nomg(sv, argsv);
10417 #ifndef USE_LONG_DOUBLE
10418 /* special-case "%.<number>[gf]" */
10419 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10420 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10421 unsigned digits = 0;
10425 while (*pp >= '0' && *pp <= '9')
10426 digits = 10 * digits + (*pp++ - '0');
10427 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10428 const NV nv = SvNV(*svargs);
10430 /* Add check for digits != 0 because it seems that some
10431 gconverts are buggy in this case, and we don't yet have
10432 a Configure test for this. */
10433 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10434 /* 0, point, slack */
10435 Gconvert(nv, (int)digits, 0, ebuf);
10436 sv_catpv_nomg(sv, ebuf);
10437 if (*ebuf) /* May return an empty string for digits==0 */
10440 } else if (!digits) {
10443 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10444 sv_catpvn_nomg(sv, p, l);
10450 #endif /* !USE_LONG_DOUBLE */
10452 if (!args && svix < svmax && DO_UTF8(*svargs))
10455 patend = (char*)pat + patlen;
10456 for (p = (char*)pat; p < patend; p = q) {
10459 bool vectorize = FALSE;
10460 bool vectorarg = FALSE;
10461 bool vec_utf8 = FALSE;
10467 bool has_precis = FALSE;
10469 const I32 osvix = svix;
10470 bool is_utf8 = FALSE; /* is this item utf8? */
10471 #ifdef HAS_LDBL_SPRINTF_BUG
10472 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10473 with sfio - Allen <allens@cpan.org> */
10474 bool fix_ldbl_sprintf_bug = FALSE;
10478 U8 utf8buf[UTF8_MAXBYTES+1];
10479 STRLEN esignlen = 0;
10481 const char *eptr = NULL;
10482 const char *fmtstart;
10485 const U8 *vecstr = NULL;
10492 /* we need a long double target in case HAS_LONG_DOUBLE but
10493 not USE_LONG_DOUBLE
10495 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10503 const char *dotstr = ".";
10504 STRLEN dotstrlen = 1;
10505 I32 efix = 0; /* explicit format parameter index */
10506 I32 ewix = 0; /* explicit width index */
10507 I32 epix = 0; /* explicit precision index */
10508 I32 evix = 0; /* explicit vector index */
10509 bool asterisk = FALSE;
10511 /* echo everything up to the next format specification */
10512 for (q = p; q < patend && *q != '%'; ++q) ;
10514 if (has_utf8 && !pat_utf8)
10515 sv_catpvn_nomg_utf8_upgrade(sv, p, q - p, nsv);
10517 sv_catpvn_nomg(sv, p, q - p);
10526 We allow format specification elements in this order:
10527 \d+\$ explicit format parameter index
10529 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10530 0 flag (as above): repeated to allow "v02"
10531 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10532 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10534 [%bcdefginopsuxDFOUX] format (mandatory)
10539 As of perl5.9.3, printf format checking is on by default.
10540 Internally, perl uses %p formats to provide an escape to
10541 some extended formatting. This block deals with those
10542 extensions: if it does not match, (char*)q is reset and
10543 the normal format processing code is used.
10545 Currently defined extensions are:
10546 %p include pointer address (standard)
10547 %-p (SVf) include an SV (previously %_)
10548 %-<num>p include an SV with precision <num>
10550 %3p include a HEK with precision of 256
10551 %4p char* preceded by utf8 flag and length
10552 %<num>p (where num is 1 or > 4) reserved for future
10555 Robin Barker 2005-07-14 (but modified since)
10557 %1p (VDf) removed. RMB 2007-10-19
10564 else if (strnEQ(q, UTF8f, sizeof(UTF8f)-1)) { /* UTF8f */
10565 /* The argument has already gone through cBOOL, so the cast
10567 is_utf8 = (bool)va_arg(*args, int);
10568 elen = va_arg(*args, UV);
10569 eptr = va_arg(*args, char *);
10570 q += sizeof(UTF8f)-1;
10573 n = expect_number(&q);
10575 if (sv) { /* SVf */
10580 argsv = MUTABLE_SV(va_arg(*args, void*));
10581 eptr = SvPV_const(argsv, elen);
10582 if (DO_UTF8(argsv))
10586 else if (n==2 || n==3) { /* HEKf */
10587 HEK * const hek = va_arg(*args, HEK *);
10588 eptr = HEK_KEY(hek);
10589 elen = HEK_LEN(hek);
10590 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10591 if (n==3) precis = 256, has_precis = TRUE;
10595 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10596 "internal %%<num>p might conflict with future printf extensions");
10602 if ( (width = expect_number(&q)) ) {
10617 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10646 if ( (ewix = expect_number(&q)) )
10655 if ((vectorarg = asterisk)) {
10668 width = expect_number(&q);
10671 if (vectorize && vectorarg) {
10672 /* vectorizing, but not with the default "." */
10674 vecsv = va_arg(*args, SV*);
10676 vecsv = (evix > 0 && evix <= svmax)
10677 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10679 vecsv = svix < svmax
10680 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10682 dotstr = SvPV_const(vecsv, dotstrlen);
10683 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10684 bad with tied or overloaded values that return UTF8. */
10685 if (DO_UTF8(vecsv))
10687 else if (has_utf8) {
10688 vecsv = sv_mortalcopy(vecsv);
10689 sv_utf8_upgrade(vecsv);
10690 dotstr = SvPV_const(vecsv, dotstrlen);
10697 i = va_arg(*args, int);
10699 i = (ewix ? ewix <= svmax : svix < svmax) ?
10700 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10702 width = (i < 0) ? -i : i;
10712 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10714 /* XXX: todo, support specified precision parameter */
10718 i = va_arg(*args, int);
10720 i = (ewix ? ewix <= svmax : svix < svmax)
10721 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10723 has_precis = !(i < 0);
10727 while (isDIGIT(*q))
10728 precis = precis * 10 + (*q++ - '0');
10737 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10738 vecsv = svargs[efix ? efix-1 : svix++];
10739 vecstr = (U8*)SvPV_const(vecsv,veclen);
10740 vec_utf8 = DO_UTF8(vecsv);
10742 /* if this is a version object, we need to convert
10743 * back into v-string notation and then let the
10744 * vectorize happen normally
10746 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10747 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10748 Perl_ck_warner_d(aTHX_ packWARN(WARN_PRINTF),
10749 "vector argument not supported with alpha versions");
10752 vecsv = sv_newmortal();
10753 scan_vstring((char *)vecstr, (char *)vecstr + veclen,
10755 vecstr = (U8*)SvPV_const(vecsv, veclen);
10756 vec_utf8 = DO_UTF8(vecsv);
10770 case 'I': /* Ix, I32x, and I64x */
10771 # ifdef USE_64_BIT_INT
10772 if (q[1] == '6' && q[2] == '4') {
10778 if (q[1] == '3' && q[2] == '2') {
10782 # ifdef USE_64_BIT_INT
10788 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10800 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10801 if (*q == 'l') { /* lld, llf */
10810 if (*++q == 'h') { /* hhd, hhu */
10839 if (!vectorize && !args) {
10841 const I32 i = efix-1;
10842 argsv = (i >= 0 && i < svmax)
10843 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10845 argsv = (svix >= 0 && svix < svmax)
10846 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10850 switch (c = *q++) {
10857 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10859 (!NATIVE_IS_INVARIANT(uv) && SvUTF8(sv)))
10861 eptr = (char*)utf8buf;
10862 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10876 eptr = va_arg(*args, char*);
10878 elen = strlen(eptr);
10880 eptr = (char *)nullstr;
10881 elen = sizeof nullstr - 1;
10885 eptr = SvPV_const(argsv, elen);
10886 if (DO_UTF8(argsv)) {
10887 STRLEN old_precis = precis;
10888 if (has_precis && precis < elen) {
10889 STRLEN ulen = sv_or_pv_len_utf8(argsv, eptr, elen);
10890 STRLEN p = precis > ulen ? ulen : precis;
10891 precis = sv_or_pv_pos_u2b(argsv, eptr, p, 0);
10892 /* sticks at end */
10894 if (width) { /* fudge width (can't fudge elen) */
10895 if (has_precis && precis < elen)
10896 width += precis - old_precis;
10899 elen - sv_or_pv_len_utf8(argsv,eptr,elen);
10906 if (has_precis && precis < elen)
10913 if (alt || vectorize)
10915 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10936 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10945 esignbuf[esignlen++] = plus;
10949 case 'c': iv = (char)va_arg(*args, int); break;
10950 case 'h': iv = (short)va_arg(*args, int); break;
10951 case 'l': iv = va_arg(*args, long); break;
10952 case 'V': iv = va_arg(*args, IV); break;
10953 case 'z': iv = va_arg(*args, SSize_t); break;
10954 case 't': iv = va_arg(*args, ptrdiff_t); break;
10955 default: iv = va_arg(*args, int); break;
10957 case 'j': iv = va_arg(*args, intmax_t); break;
10961 iv = va_arg(*args, Quad_t); break;
10968 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10970 case 'c': iv = (char)tiv; break;
10971 case 'h': iv = (short)tiv; break;
10972 case 'l': iv = (long)tiv; break;
10974 default: iv = tiv; break;
10977 iv = (Quad_t)tiv; break;
10983 if ( !vectorize ) /* we already set uv above */
10988 esignbuf[esignlen++] = plus;
10992 esignbuf[esignlen++] = '-';
11036 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
11047 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
11048 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
11049 case 'l': uv = va_arg(*args, unsigned long); break;
11050 case 'V': uv = va_arg(*args, UV); break;
11051 case 'z': uv = va_arg(*args, Size_t); break;
11052 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
11054 case 'j': uv = va_arg(*args, uintmax_t); break;
11056 default: uv = va_arg(*args, unsigned); break;
11059 uv = va_arg(*args, Uquad_t); break;
11066 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
11068 case 'c': uv = (unsigned char)tuv; break;
11069 case 'h': uv = (unsigned short)tuv; break;
11070 case 'l': uv = (unsigned long)tuv; break;
11072 default: uv = tuv; break;
11075 uv = (Uquad_t)tuv; break;
11084 char *ptr = ebuf + sizeof ebuf;
11085 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
11091 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
11095 } while (uv >>= 4);
11097 esignbuf[esignlen++] = '0';
11098 esignbuf[esignlen++] = c; /* 'x' or 'X' */
11104 *--ptr = '0' + dig;
11105 } while (uv >>= 3);
11106 if (alt && *ptr != '0')
11112 *--ptr = '0' + dig;
11113 } while (uv >>= 1);
11115 esignbuf[esignlen++] = '0';
11116 esignbuf[esignlen++] = c;
11119 default: /* it had better be ten or less */
11122 *--ptr = '0' + dig;
11123 } while (uv /= base);
11126 elen = (ebuf + sizeof ebuf) - ptr;
11130 zeros = precis - elen;
11131 else if (precis == 0 && elen == 1 && *eptr == '0'
11132 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
11135 /* a precision nullifies the 0 flag. */
11142 /* FLOATING POINT */
11145 c = 'f'; /* maybe %F isn't supported here */
11147 case 'e': case 'E':
11149 case 'g': case 'G':
11153 /* This is evil, but floating point is even more evil */
11155 /* for SV-style calling, we can only get NV
11156 for C-style calling, we assume %f is double;
11157 for simplicity we allow any of %Lf, %llf, %qf for long double
11161 #if defined(USE_LONG_DOUBLE)
11165 /* [perl #20339] - we should accept and ignore %lf rather than die */
11169 #if defined(USE_LONG_DOUBLE)
11170 intsize = args ? 0 : 'q';
11174 #if defined(HAS_LONG_DOUBLE)
11187 /* now we need (long double) if intsize == 'q', else (double) */
11189 #if LONG_DOUBLESIZE > DOUBLESIZE
11191 va_arg(*args, long double) :
11192 va_arg(*args, double)
11194 va_arg(*args, double)
11199 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
11200 else. frexp() has some unspecified behaviour for those three */
11201 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
11203 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
11204 will cast our (long double) to (double) */
11205 (void)Perl_frexp(nv, &i);
11206 if (i == PERL_INT_MIN)
11207 Perl_die(aTHX_ "panic: frexp");
11209 need = BIT_DIGITS(i);
11211 need += has_precis ? precis : 6; /* known default */
11216 #ifdef HAS_LDBL_SPRINTF_BUG
11217 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
11218 with sfio - Allen <allens@cpan.org> */
11221 # define MY_DBL_MAX DBL_MAX
11222 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
11223 # if DOUBLESIZE >= 8
11224 # define MY_DBL_MAX 1.7976931348623157E+308L
11226 # define MY_DBL_MAX 3.40282347E+38L
11230 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
11231 # define MY_DBL_MAX_BUG 1L
11233 # define MY_DBL_MAX_BUG MY_DBL_MAX
11237 # define MY_DBL_MIN DBL_MIN
11238 # else /* XXX guessing! -Allen */
11239 # if DOUBLESIZE >= 8
11240 # define MY_DBL_MIN 2.2250738585072014E-308L
11242 # define MY_DBL_MIN 1.17549435E-38L
11246 if ((intsize == 'q') && (c == 'f') &&
11247 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
11248 (need < DBL_DIG)) {
11249 /* it's going to be short enough that
11250 * long double precision is not needed */
11252 if ((nv <= 0L) && (nv >= -0L))
11253 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
11255 /* would use Perl_fp_class as a double-check but not
11256 * functional on IRIX - see perl.h comments */
11258 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
11259 /* It's within the range that a double can represent */
11260 #if defined(DBL_MAX) && !defined(DBL_MIN)
11261 if ((nv >= ((long double)1/DBL_MAX)) ||
11262 (nv <= (-(long double)1/DBL_MAX)))
11264 fix_ldbl_sprintf_bug = TRUE;
11267 if (fix_ldbl_sprintf_bug == TRUE) {
11277 # undef MY_DBL_MAX_BUG
11280 #endif /* HAS_LDBL_SPRINTF_BUG */
11282 need += 20; /* fudge factor */
11283 if (PL_efloatsize < need) {
11284 Safefree(PL_efloatbuf);
11285 PL_efloatsize = need + 20; /* more fudge */
11286 Newx(PL_efloatbuf, PL_efloatsize, char);
11287 PL_efloatbuf[0] = '\0';
11290 if ( !(width || left || plus || alt) && fill != '0'
11291 && has_precis && intsize != 'q' ) { /* Shortcuts */
11292 /* See earlier comment about buggy Gconvert when digits,
11294 if ( c == 'g' && precis) {
11295 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
11296 /* May return an empty string for digits==0 */
11297 if (*PL_efloatbuf) {
11298 elen = strlen(PL_efloatbuf);
11299 goto float_converted;
11301 } else if ( c == 'f' && !precis) {
11302 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
11307 char *ptr = ebuf + sizeof ebuf;
11310 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
11311 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
11312 if (intsize == 'q') {
11313 /* Copy the one or more characters in a long double
11314 * format before the 'base' ([efgEFG]) character to
11315 * the format string. */
11316 static char const prifldbl[] = PERL_PRIfldbl;
11317 char const *p = prifldbl + sizeof(prifldbl) - 3;
11318 while (p >= prifldbl) { *--ptr = *p--; }
11323 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11328 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11340 /* No taint. Otherwise we are in the strange situation
11341 * where printf() taints but print($float) doesn't.
11343 #if defined(HAS_LONG_DOUBLE)
11344 elen = ((intsize == 'q')
11345 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
11346 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
11348 elen = my_sprintf(PL_efloatbuf, ptr, nv);
11352 eptr = PL_efloatbuf;
11354 #ifdef USE_LOCALE_NUMERIC
11355 if (PL_numeric_radix_sv && SvUTF8(PL_numeric_radix_sv)
11356 && instr(eptr, SvPVX_const(PL_numeric_radix_sv)))
11369 i = SvCUR(sv) - origlen;
11372 case 'c': *(va_arg(*args, char*)) = i; break;
11373 case 'h': *(va_arg(*args, short*)) = i; break;
11374 default: *(va_arg(*args, int*)) = i; break;
11375 case 'l': *(va_arg(*args, long*)) = i; break;
11376 case 'V': *(va_arg(*args, IV*)) = i; break;
11377 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11378 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11380 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11384 *(va_arg(*args, Quad_t*)) = i; break;
11391 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11392 continue; /* not "break" */
11399 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11400 && ckWARN(WARN_PRINTF))
11402 SV * const msg = sv_newmortal();
11403 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11404 (PL_op->op_type == OP_PRTF) ? "" : "s");
11405 if (fmtstart < patend) {
11406 const char * const fmtend = q < patend ? q : patend;
11408 sv_catpvs(msg, "\"%");
11409 for (f = fmtstart; f < fmtend; f++) {
11411 sv_catpvn_nomg(msg, f, 1);
11413 Perl_sv_catpvf(aTHX_ msg,
11414 "\\%03"UVof, (UV)*f & 0xFF);
11417 sv_catpvs(msg, "\"");
11419 sv_catpvs(msg, "end of string");
11421 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11424 /* output mangled stuff ... */
11430 /* ... right here, because formatting flags should not apply */
11431 SvGROW(sv, SvCUR(sv) + elen + 1);
11433 Copy(eptr, p, elen, char);
11436 SvCUR_set(sv, p - SvPVX_const(sv));
11438 continue; /* not "break" */
11441 if (is_utf8 != has_utf8) {
11444 sv_utf8_upgrade(sv);
11447 const STRLEN old_elen = elen;
11448 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11449 sv_utf8_upgrade(nsv);
11450 eptr = SvPVX_const(nsv);
11453 if (width) { /* fudge width (can't fudge elen) */
11454 width += elen - old_elen;
11460 have = esignlen + zeros + elen;
11462 croak_memory_wrap();
11464 need = (have > width ? have : width);
11467 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11468 croak_memory_wrap();
11469 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11471 if (esignlen && fill == '0') {
11473 for (i = 0; i < (int)esignlen; i++)
11474 *p++ = esignbuf[i];
11476 if (gap && !left) {
11477 memset(p, fill, gap);
11480 if (esignlen && fill != '0') {
11482 for (i = 0; i < (int)esignlen; i++)
11483 *p++ = esignbuf[i];
11487 for (i = zeros; i; i--)
11491 Copy(eptr, p, elen, char);
11495 memset(p, ' ', gap);
11500 Copy(dotstr, p, dotstrlen, char);
11504 vectorize = FALSE; /* done iterating over vecstr */
11511 SvCUR_set(sv, p - SvPVX_const(sv));
11520 /* =========================================================================
11522 =head1 Cloning an interpreter
11524 All the macros and functions in this section are for the private use of
11525 the main function, perl_clone().
11527 The foo_dup() functions make an exact copy of an existing foo thingy.
11528 During the course of a cloning, a hash table is used to map old addresses
11529 to new addresses. The table is created and manipulated with the
11530 ptr_table_* functions.
11534 * =========================================================================*/
11537 #if defined(USE_ITHREADS)
11539 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11540 #ifndef GpREFCNT_inc
11541 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11545 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11546 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11547 If this changes, please unmerge ss_dup.
11548 Likewise, sv_dup_inc_multiple() relies on this fact. */
11549 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11550 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11551 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11552 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11553 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11554 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11555 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11556 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11557 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11558 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11559 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11560 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11561 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11563 /* clone a parser */
11566 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11570 PERL_ARGS_ASSERT_PARSER_DUP;
11575 /* look for it in the table first */
11576 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11580 /* create anew and remember what it is */
11581 Newxz(parser, 1, yy_parser);
11582 ptr_table_store(PL_ptr_table, proto, parser);
11584 /* XXX these not yet duped */
11585 parser->old_parser = NULL;
11586 parser->stack = NULL;
11588 parser->stack_size = 0;
11589 /* XXX parser->stack->state = 0; */
11591 /* XXX eventually, just Copy() most of the parser struct ? */
11593 parser->lex_brackets = proto->lex_brackets;
11594 parser->lex_casemods = proto->lex_casemods;
11595 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11596 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11597 parser->lex_casestack = savepvn(proto->lex_casestack,
11598 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11599 parser->lex_defer = proto->lex_defer;
11600 parser->lex_dojoin = proto->lex_dojoin;
11601 parser->lex_expect = proto->lex_expect;
11602 parser->lex_formbrack = proto->lex_formbrack;
11603 parser->lex_inpat = proto->lex_inpat;
11604 parser->lex_inwhat = proto->lex_inwhat;
11605 parser->lex_op = proto->lex_op;
11606 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11607 parser->lex_starts = proto->lex_starts;
11608 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11609 parser->multi_close = proto->multi_close;
11610 parser->multi_open = proto->multi_open;
11611 parser->multi_start = proto->multi_start;
11612 parser->multi_end = proto->multi_end;
11613 parser->preambled = proto->preambled;
11614 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11615 parser->linestr = sv_dup_inc(proto->linestr, param);
11616 parser->expect = proto->expect;
11617 parser->copline = proto->copline;
11618 parser->last_lop_op = proto->last_lop_op;
11619 parser->lex_state = proto->lex_state;
11620 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11621 /* rsfp_filters entries have fake IoDIRP() */
11622 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11623 parser->in_my = proto->in_my;
11624 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11625 parser->error_count = proto->error_count;
11628 parser->linestr = sv_dup_inc(proto->linestr, param);
11631 char * const ols = SvPVX(proto->linestr);
11632 char * const ls = SvPVX(parser->linestr);
11634 parser->bufptr = ls + (proto->bufptr >= ols ?
11635 proto->bufptr - ols : 0);
11636 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11637 proto->oldbufptr - ols : 0);
11638 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11639 proto->oldoldbufptr - ols : 0);
11640 parser->linestart = ls + (proto->linestart >= ols ?
11641 proto->linestart - ols : 0);
11642 parser->last_uni = ls + (proto->last_uni >= ols ?
11643 proto->last_uni - ols : 0);
11644 parser->last_lop = ls + (proto->last_lop >= ols ?
11645 proto->last_lop - ols : 0);
11647 parser->bufend = ls + SvCUR(parser->linestr);
11650 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11654 parser->endwhite = proto->endwhite;
11655 parser->faketokens = proto->faketokens;
11656 parser->lasttoke = proto->lasttoke;
11657 parser->nextwhite = proto->nextwhite;
11658 parser->realtokenstart = proto->realtokenstart;
11659 parser->skipwhite = proto->skipwhite;
11660 parser->thisclose = proto->thisclose;
11661 parser->thismad = proto->thismad;
11662 parser->thisopen = proto->thisopen;
11663 parser->thisstuff = proto->thisstuff;
11664 parser->thistoken = proto->thistoken;
11665 parser->thiswhite = proto->thiswhite;
11667 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11668 parser->curforce = proto->curforce;
11670 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11671 Copy(proto->nexttype, parser->nexttype, 5, I32);
11672 parser->nexttoke = proto->nexttoke;
11675 /* XXX should clone saved_curcop here, but we aren't passed
11676 * proto_perl; so do it in perl_clone_using instead */
11682 /* duplicate a file handle */
11685 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11689 PERL_ARGS_ASSERT_FP_DUP;
11690 PERL_UNUSED_ARG(type);
11693 return (PerlIO*)NULL;
11695 /* look for it in the table first */
11696 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11700 /* create anew and remember what it is */
11701 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11702 ptr_table_store(PL_ptr_table, fp, ret);
11706 /* duplicate a directory handle */
11709 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11715 const Direntry_t *dirent;
11716 char smallbuf[256];
11722 PERL_UNUSED_CONTEXT;
11723 PERL_ARGS_ASSERT_DIRP_DUP;
11728 /* look for it in the table first */
11729 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11735 PERL_UNUSED_ARG(param);
11739 /* open the current directory (so we can switch back) */
11740 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11742 /* chdir to our dir handle and open the present working directory */
11743 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11744 PerlDir_close(pwd);
11745 return (DIR *)NULL;
11747 /* Now we should have two dir handles pointing to the same dir. */
11749 /* Be nice to the calling code and chdir back to where we were. */
11750 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11752 /* We have no need of the pwd handle any more. */
11753 PerlDir_close(pwd);
11756 # define d_namlen(d) (d)->d_namlen
11758 # define d_namlen(d) strlen((d)->d_name)
11760 /* Iterate once through dp, to get the file name at the current posi-
11761 tion. Then step back. */
11762 pos = PerlDir_tell(dp);
11763 if ((dirent = PerlDir_read(dp))) {
11764 len = d_namlen(dirent);
11765 if (len <= sizeof smallbuf) name = smallbuf;
11766 else Newx(name, len, char);
11767 Move(dirent->d_name, name, len, char);
11769 PerlDir_seek(dp, pos);
11771 /* Iterate through the new dir handle, till we find a file with the
11773 if (!dirent) /* just before the end */
11775 pos = PerlDir_tell(ret);
11776 if (PerlDir_read(ret)) continue; /* not there yet */
11777 PerlDir_seek(ret, pos); /* step back */
11781 const long pos0 = PerlDir_tell(ret);
11783 pos = PerlDir_tell(ret);
11784 if ((dirent = PerlDir_read(ret))) {
11785 if (len == d_namlen(dirent)
11786 && memEQ(name, dirent->d_name, len)) {
11788 PerlDir_seek(ret, pos); /* step back */
11791 /* else we are not there yet; keep iterating */
11793 else { /* This is not meant to happen. The best we can do is
11794 reset the iterator to the beginning. */
11795 PerlDir_seek(ret, pos0);
11802 if (name && name != smallbuf)
11807 ret = win32_dirp_dup(dp, param);
11810 /* pop it in the pointer table */
11812 ptr_table_store(PL_ptr_table, dp, ret);
11817 /* duplicate a typeglob */
11820 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11824 PERL_ARGS_ASSERT_GP_DUP;
11828 /* look for it in the table first */
11829 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11833 /* create anew and remember what it is */
11835 ptr_table_store(PL_ptr_table, gp, ret);
11838 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11839 on Newxz() to do this for us. */
11840 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11841 ret->gp_io = io_dup_inc(gp->gp_io, param);
11842 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11843 ret->gp_av = av_dup_inc(gp->gp_av, param);
11844 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11845 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11846 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11847 ret->gp_cvgen = gp->gp_cvgen;
11848 ret->gp_line = gp->gp_line;
11849 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11853 /* duplicate a chain of magic */
11856 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11858 MAGIC *mgret = NULL;
11859 MAGIC **mgprev_p = &mgret;
11861 PERL_ARGS_ASSERT_MG_DUP;
11863 for (; mg; mg = mg->mg_moremagic) {
11866 if ((param->flags & CLONEf_JOIN_IN)
11867 && mg->mg_type == PERL_MAGIC_backref)
11868 /* when joining, we let the individual SVs add themselves to
11869 * backref as needed. */
11872 Newx(nmg, 1, MAGIC);
11874 mgprev_p = &(nmg->mg_moremagic);
11876 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11877 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11878 from the original commit adding Perl_mg_dup() - revision 4538.
11879 Similarly there is the annotation "XXX random ptr?" next to the
11880 assignment to nmg->mg_ptr. */
11883 /* FIXME for plugins
11884 if (nmg->mg_type == PERL_MAGIC_qr) {
11885 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11889 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11890 ? nmg->mg_type == PERL_MAGIC_backref
11891 /* The backref AV has its reference
11892 * count deliberately bumped by 1 */
11893 ? SvREFCNT_inc(av_dup_inc((const AV *)
11894 nmg->mg_obj, param))
11895 : sv_dup_inc(nmg->mg_obj, param)
11896 : sv_dup(nmg->mg_obj, param);
11898 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11899 if (nmg->mg_len > 0) {
11900 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11901 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11902 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11904 AMT * const namtp = (AMT*)nmg->mg_ptr;
11905 sv_dup_inc_multiple((SV**)(namtp->table),
11906 (SV**)(namtp->table), NofAMmeth, param);
11909 else if (nmg->mg_len == HEf_SVKEY)
11910 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11912 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11913 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11919 #endif /* USE_ITHREADS */
11921 struct ptr_tbl_arena {
11922 struct ptr_tbl_arena *next;
11923 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11926 /* create a new pointer-mapping table */
11929 Perl_ptr_table_new(pTHX)
11932 PERL_UNUSED_CONTEXT;
11934 Newx(tbl, 1, PTR_TBL_t);
11935 tbl->tbl_max = 511;
11936 tbl->tbl_items = 0;
11937 tbl->tbl_arena = NULL;
11938 tbl->tbl_arena_next = NULL;
11939 tbl->tbl_arena_end = NULL;
11940 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11944 #define PTR_TABLE_HASH(ptr) \
11945 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11947 /* map an existing pointer using a table */
11949 STATIC PTR_TBL_ENT_t *
11950 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11952 PTR_TBL_ENT_t *tblent;
11953 const UV hash = PTR_TABLE_HASH(sv);
11955 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11957 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11958 for (; tblent; tblent = tblent->next) {
11959 if (tblent->oldval == sv)
11966 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11968 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11970 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11971 PERL_UNUSED_CONTEXT;
11973 return tblent ? tblent->newval : NULL;
11976 /* add a new entry to a pointer-mapping table */
11979 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11981 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11983 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11984 PERL_UNUSED_CONTEXT;
11987 tblent->newval = newsv;
11989 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11991 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11992 struct ptr_tbl_arena *new_arena;
11994 Newx(new_arena, 1, struct ptr_tbl_arena);
11995 new_arena->next = tbl->tbl_arena;
11996 tbl->tbl_arena = new_arena;
11997 tbl->tbl_arena_next = new_arena->array;
11998 tbl->tbl_arena_end = new_arena->array
11999 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
12002 tblent = tbl->tbl_arena_next++;
12004 tblent->oldval = oldsv;
12005 tblent->newval = newsv;
12006 tblent->next = tbl->tbl_ary[entry];
12007 tbl->tbl_ary[entry] = tblent;
12009 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
12010 ptr_table_split(tbl);
12014 /* double the hash bucket size of an existing ptr table */
12017 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
12019 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
12020 const UV oldsize = tbl->tbl_max + 1;
12021 UV newsize = oldsize * 2;
12024 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
12025 PERL_UNUSED_CONTEXT;
12027 Renew(ary, newsize, PTR_TBL_ENT_t*);
12028 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
12029 tbl->tbl_max = --newsize;
12030 tbl->tbl_ary = ary;
12031 for (i=0; i < oldsize; i++, ary++) {
12032 PTR_TBL_ENT_t **entp = ary;
12033 PTR_TBL_ENT_t *ent = *ary;
12034 PTR_TBL_ENT_t **curentp;
12037 curentp = ary + oldsize;
12039 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
12041 ent->next = *curentp;
12051 /* remove all the entries from a ptr table */
12052 /* Deprecated - will be removed post 5.14 */
12055 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
12057 if (tbl && tbl->tbl_items) {
12058 struct ptr_tbl_arena *arena = tbl->tbl_arena;
12060 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
12063 struct ptr_tbl_arena *next = arena->next;
12069 tbl->tbl_items = 0;
12070 tbl->tbl_arena = NULL;
12071 tbl->tbl_arena_next = NULL;
12072 tbl->tbl_arena_end = NULL;
12076 /* clear and free a ptr table */
12079 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
12081 struct ptr_tbl_arena *arena;
12087 arena = tbl->tbl_arena;
12090 struct ptr_tbl_arena *next = arena->next;
12096 Safefree(tbl->tbl_ary);
12100 #if defined(USE_ITHREADS)
12103 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
12105 PERL_ARGS_ASSERT_RVPV_DUP;
12107 assert(!isREGEXP(sstr));
12109 if (SvWEAKREF(sstr)) {
12110 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
12111 if (param->flags & CLONEf_JOIN_IN) {
12112 /* if joining, we add any back references individually rather
12113 * than copying the whole backref array */
12114 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
12118 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
12120 else if (SvPVX_const(sstr)) {
12121 /* Has something there */
12123 /* Normal PV - clone whole allocated space */
12124 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
12125 /* sstr may not be that normal, but actually copy on write.
12126 But we are a true, independent SV, so: */
12130 /* Special case - not normally malloced for some reason */
12131 if (isGV_with_GP(sstr)) {
12132 /* Don't need to do anything here. */
12134 else if ((SvIsCOW(sstr))) {
12135 /* A "shared" PV - clone it as "shared" PV */
12137 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
12141 /* Some other special case - random pointer */
12142 SvPV_set(dstr, (char *) SvPVX_const(sstr));
12147 /* Copy the NULL */
12148 SvPV_set(dstr, NULL);
12152 /* duplicate a list of SVs. source and dest may point to the same memory. */
12154 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
12155 SSize_t items, CLONE_PARAMS *const param)
12157 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
12159 while (items-- > 0) {
12160 *dest++ = sv_dup_inc(*source++, param);
12166 /* duplicate an SV of any type (including AV, HV etc) */
12169 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12174 PERL_ARGS_ASSERT_SV_DUP_COMMON;
12176 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
12177 #ifdef DEBUG_LEAKING_SCALARS_ABORT
12182 /* look for it in the table first */
12183 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
12187 if(param->flags & CLONEf_JOIN_IN) {
12188 /** We are joining here so we don't want do clone
12189 something that is bad **/
12190 if (SvTYPE(sstr) == SVt_PVHV) {
12191 const HEK * const hvname = HvNAME_HEK(sstr);
12193 /** don't clone stashes if they already exist **/
12194 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
12195 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
12196 ptr_table_store(PL_ptr_table, sstr, dstr);
12200 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
12201 HV *stash = GvSTASH(sstr);
12202 const HEK * hvname;
12203 if (stash && (hvname = HvNAME_HEK(stash))) {
12204 /** don't clone GVs if they already exist **/
12206 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
12207 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
12209 stash, GvNAME(sstr),
12215 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
12216 ptr_table_store(PL_ptr_table, sstr, *svp);
12223 /* create anew and remember what it is */
12226 #ifdef DEBUG_LEAKING_SCALARS
12227 dstr->sv_debug_optype = sstr->sv_debug_optype;
12228 dstr->sv_debug_line = sstr->sv_debug_line;
12229 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
12230 dstr->sv_debug_parent = (SV*)sstr;
12231 FREE_SV_DEBUG_FILE(dstr);
12232 dstr->sv_debug_file = savesharedpv(sstr->sv_debug_file);
12235 ptr_table_store(PL_ptr_table, sstr, dstr);
12238 SvFLAGS(dstr) = SvFLAGS(sstr);
12239 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
12240 SvREFCNT(dstr) = 0; /* must be before any other dups! */
12243 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
12244 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
12245 (void*)PL_watch_pvx, SvPVX_const(sstr));
12248 /* don't clone objects whose class has asked us not to */
12249 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
12254 switch (SvTYPE(sstr)) {
12256 SvANY(dstr) = NULL;
12259 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
12261 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
12263 SvIV_set(dstr, SvIVX(sstr));
12267 SvANY(dstr) = new_XNV();
12268 SvNV_set(dstr, SvNVX(sstr));
12272 /* These are all the types that need complex bodies allocating. */
12274 const svtype sv_type = SvTYPE(sstr);
12275 const struct body_details *const sv_type_details
12276 = bodies_by_type + sv_type;
12280 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
12296 assert(sv_type_details->body_size);
12297 if (sv_type_details->arena) {
12298 new_body_inline(new_body, sv_type);
12300 = (void*)((char*)new_body - sv_type_details->offset);
12302 new_body = new_NOARENA(sv_type_details);
12306 SvANY(dstr) = new_body;
12309 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
12310 ((char*)SvANY(dstr)) + sv_type_details->offset,
12311 sv_type_details->copy, char);
12313 Copy(((char*)SvANY(sstr)),
12314 ((char*)SvANY(dstr)),
12315 sv_type_details->body_size + sv_type_details->offset, char);
12318 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
12319 && !isGV_with_GP(dstr)
12321 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
12322 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
12324 /* The Copy above means that all the source (unduplicated) pointers
12325 are now in the destination. We can check the flags and the
12326 pointers in either, but it's possible that there's less cache
12327 missing by always going for the destination.
12328 FIXME - instrument and check that assumption */
12329 if (sv_type >= SVt_PVMG) {
12330 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
12331 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
12332 } else if (sv_type == SVt_PVAV && AvPAD_NAMELIST(dstr)) {
12334 } else if (SvMAGIC(dstr))
12335 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
12336 if (SvOBJECT(dstr) && SvSTASH(dstr))
12337 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
12338 else SvSTASH_set(dstr, 0); /* don't copy DESTROY cache */
12341 /* The cast silences a GCC warning about unhandled types. */
12342 switch ((int)sv_type) {
12353 /* FIXME for plugins */
12354 dstr->sv_u.svu_rx = ((REGEXP *)dstr)->sv_any;
12355 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
12358 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
12359 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
12360 LvTARG(dstr) = dstr;
12361 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
12362 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
12364 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
12365 if (isREGEXP(sstr)) goto duprex;
12367 /* non-GP case already handled above */
12368 if(isGV_with_GP(sstr)) {
12369 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
12370 /* Don't call sv_add_backref here as it's going to be
12371 created as part of the magic cloning of the symbol
12372 table--unless this is during a join and the stash
12373 is not actually being cloned. */
12374 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12375 at the point of this comment. */
12376 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12377 if (param->flags & CLONEf_JOIN_IN)
12378 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12379 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12380 (void)GpREFCNT_inc(GvGP(dstr));
12384 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12385 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12386 /* I have no idea why fake dirp (rsfps)
12387 should be treated differently but otherwise
12388 we end up with leaks -- sky*/
12389 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12390 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12391 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12393 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12394 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12395 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12396 if (IoDIRP(dstr)) {
12397 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12400 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12402 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12404 if (IoOFP(dstr) == IoIFP(sstr))
12405 IoOFP(dstr) = IoIFP(dstr);
12407 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12408 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12409 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12410 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12413 /* avoid cloning an empty array */
12414 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12415 SV **dst_ary, **src_ary;
12416 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12418 src_ary = AvARRAY((const AV *)sstr);
12419 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12420 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12421 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12422 AvALLOC((const AV *)dstr) = dst_ary;
12423 if (AvREAL((const AV *)sstr)) {
12424 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12428 while (items-- > 0)
12429 *dst_ary++ = sv_dup(*src_ary++, param);
12431 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12432 while (items-- > 0) {
12433 *dst_ary++ = &PL_sv_undef;
12437 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12438 AvALLOC((const AV *)dstr) = (SV**)NULL;
12439 AvMAX( (const AV *)dstr) = -1;
12440 AvFILLp((const AV *)dstr) = -1;
12444 if (HvARRAY((const HV *)sstr)) {
12446 const bool sharekeys = !!HvSHAREKEYS(sstr);
12447 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12448 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12450 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12451 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12453 HvARRAY(dstr) = (HE**)darray;
12454 while (i <= sxhv->xhv_max) {
12455 const HE * const source = HvARRAY(sstr)[i];
12456 HvARRAY(dstr)[i] = source
12457 ? he_dup(source, sharekeys, param) : 0;
12461 const struct xpvhv_aux * const saux = HvAUX(sstr);
12462 struct xpvhv_aux * const daux = HvAUX(dstr);
12463 /* This flag isn't copied. */
12466 if (saux->xhv_name_count) {
12467 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12469 = saux->xhv_name_count < 0
12470 ? -saux->xhv_name_count
12471 : saux->xhv_name_count;
12472 HEK **shekp = sname + count;
12474 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12475 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12476 while (shekp-- > sname) {
12478 *dhekp = hek_dup(*shekp, param);
12482 daux->xhv_name_u.xhvnameu_name
12483 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12486 daux->xhv_name_count = saux->xhv_name_count;
12488 daux->xhv_fill_lazy = saux->xhv_fill_lazy;
12489 daux->xhv_riter = saux->xhv_riter;
12490 daux->xhv_eiter = saux->xhv_eiter
12491 ? he_dup(saux->xhv_eiter,
12492 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12493 /* backref array needs refcnt=2; see sv_add_backref */
12494 daux->xhv_backreferences =
12495 (param->flags & CLONEf_JOIN_IN)
12496 /* when joining, we let the individual GVs and
12497 * CVs add themselves to backref as
12498 * needed. This avoids pulling in stuff
12499 * that isn't required, and simplifies the
12500 * case where stashes aren't cloned back
12501 * if they already exist in the parent
12504 : saux->xhv_backreferences
12505 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12506 ? MUTABLE_AV(SvREFCNT_inc(
12507 sv_dup_inc((const SV *)
12508 saux->xhv_backreferences, param)))
12509 : MUTABLE_AV(sv_dup((const SV *)
12510 saux->xhv_backreferences, param))
12513 daux->xhv_mro_meta = saux->xhv_mro_meta
12514 ? mro_meta_dup(saux->xhv_mro_meta, param)
12517 /* Record stashes for possible cloning in Perl_clone(). */
12519 av_push(param->stashes, dstr);
12523 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12526 if (!(param->flags & CLONEf_COPY_STACKS)) {
12531 /* NOTE: not refcounted */
12532 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12533 hv_dup(CvSTASH(dstr), param);
12534 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12535 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12536 if (!CvISXSUB(dstr)) {
12538 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12540 CvSLABBED_off(dstr);
12541 } else if (CvCONST(dstr)) {
12542 CvXSUBANY(dstr).any_ptr =
12543 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12545 assert(!CvSLABBED(dstr));
12546 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12548 SvANY((CV *)dstr)->xcv_gv_u.xcv_hek =
12549 share_hek_hek(CvNAME_HEK((CV *)sstr));
12550 /* don't dup if copying back - CvGV isn't refcounted, so the
12551 * duped GV may never be freed. A bit of a hack! DAPM */
12553 SvANY(MUTABLE_CV(dstr))->xcv_gv_u.xcv_gv =
12555 ? gv_dup_inc(CvGV(sstr), param)
12556 : (param->flags & CLONEf_JOIN_IN)
12558 : gv_dup(CvGV(sstr), param);
12560 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12562 CvWEAKOUTSIDE(sstr)
12563 ? cv_dup( CvOUTSIDE(dstr), param)
12564 : cv_dup_inc(CvOUTSIDE(dstr), param);
12574 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12576 PERL_ARGS_ASSERT_SV_DUP_INC;
12577 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12581 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12583 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12584 PERL_ARGS_ASSERT_SV_DUP;
12586 /* Track every SV that (at least initially) had a reference count of 0.
12587 We need to do this by holding an actual reference to it in this array.
12588 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12589 (akin to the stashes hash, and the perl stack), we come unstuck if
12590 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12591 thread) is manipulated in a CLONE method, because CLONE runs before the
12592 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12593 (and fix things up by giving each a reference via the temps stack).
12594 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12595 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12596 before the walk of unreferenced happens and a reference to that is SV
12597 added to the temps stack. At which point we have the same SV considered
12598 to be in use, and free to be re-used. Not good.
12600 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12601 assert(param->unreferenced);
12602 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12608 /* duplicate a context */
12611 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12613 PERL_CONTEXT *ncxs;
12615 PERL_ARGS_ASSERT_CX_DUP;
12618 return (PERL_CONTEXT*)NULL;
12620 /* look for it in the table first */
12621 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12625 /* create anew and remember what it is */
12626 Newx(ncxs, max + 1, PERL_CONTEXT);
12627 ptr_table_store(PL_ptr_table, cxs, ncxs);
12628 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12631 PERL_CONTEXT * const ncx = &ncxs[ix];
12632 if (CxTYPE(ncx) == CXt_SUBST) {
12633 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12636 ncx->blk_oldcop = (COP*)any_dup(ncx->blk_oldcop, param->proto_perl);
12637 switch (CxTYPE(ncx)) {
12639 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12640 ? cv_dup_inc(ncx->blk_sub.cv, param)
12641 : cv_dup(ncx->blk_sub.cv,param));
12642 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12643 ? av_dup_inc(ncx->blk_sub.argarray,
12646 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12648 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12649 ncx->blk_sub.oldcomppad);
12652 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12654 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12655 ncx->blk_eval.cv = cv_dup(ncx->blk_eval.cv, param);
12657 case CXt_LOOP_LAZYSV:
12658 ncx->blk_loop.state_u.lazysv.end
12659 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12660 /* We are taking advantage of av_dup_inc and sv_dup_inc
12661 actually being the same function, and order equivalence of
12663 We can assert the later [but only at run time :-(] */
12664 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12665 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12667 ncx->blk_loop.state_u.ary.ary
12668 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12669 case CXt_LOOP_LAZYIV:
12670 case CXt_LOOP_PLAIN:
12671 if (CxPADLOOP(ncx)) {
12672 ncx->blk_loop.itervar_u.oldcomppad
12673 = (PAD*)ptr_table_fetch(PL_ptr_table,
12674 ncx->blk_loop.itervar_u.oldcomppad);
12676 ncx->blk_loop.itervar_u.gv
12677 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12682 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12683 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12684 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12699 /* duplicate a stack info structure */
12702 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12706 PERL_ARGS_ASSERT_SI_DUP;
12709 return (PERL_SI*)NULL;
12711 /* look for it in the table first */
12712 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12716 /* create anew and remember what it is */
12717 Newxz(nsi, 1, PERL_SI);
12718 ptr_table_store(PL_ptr_table, si, nsi);
12720 nsi->si_stack = av_dup_inc(si->si_stack, param);
12721 nsi->si_cxix = si->si_cxix;
12722 nsi->si_cxmax = si->si_cxmax;
12723 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12724 nsi->si_type = si->si_type;
12725 nsi->si_prev = si_dup(si->si_prev, param);
12726 nsi->si_next = si_dup(si->si_next, param);
12727 nsi->si_markoff = si->si_markoff;
12732 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12733 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12734 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12735 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12736 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12737 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12738 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12739 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12740 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12741 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12742 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12743 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12744 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12745 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12746 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12747 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12750 #define pv_dup_inc(p) SAVEPV(p)
12751 #define pv_dup(p) SAVEPV(p)
12752 #define svp_dup_inc(p,pp) any_dup(p,pp)
12754 /* map any object to the new equivent - either something in the
12755 * ptr table, or something in the interpreter structure
12759 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12763 PERL_ARGS_ASSERT_ANY_DUP;
12766 return (void*)NULL;
12768 /* look for it in the table first */
12769 ret = ptr_table_fetch(PL_ptr_table, v);
12773 /* see if it is part of the interpreter structure */
12774 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12775 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12783 /* duplicate the save stack */
12786 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12789 ANY * const ss = proto_perl->Isavestack;
12790 const I32 max = proto_perl->Isavestack_max;
12791 I32 ix = proto_perl->Isavestack_ix;
12804 void (*dptr) (void*);
12805 void (*dxptr) (pTHX_ void*);
12807 PERL_ARGS_ASSERT_SS_DUP;
12809 Newxz(nss, max, ANY);
12812 const UV uv = POPUV(ss,ix);
12813 const U8 type = (U8)uv & SAVE_MASK;
12815 TOPUV(nss,ix) = uv;
12817 case SAVEt_CLEARSV:
12818 case SAVEt_CLEARPADRANGE:
12820 case SAVEt_HELEM: /* hash element */
12821 sv = (const SV *)POPPTR(ss,ix);
12822 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12824 case SAVEt_ITEM: /* normal string */
12825 case SAVEt_GVSV: /* scalar slot in GV */
12826 case SAVEt_SV: /* scalar reference */
12827 sv = (const SV *)POPPTR(ss,ix);
12828 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12831 case SAVEt_MORTALIZESV:
12832 case SAVEt_READONLY_OFF:
12833 sv = (const SV *)POPPTR(ss,ix);
12834 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12836 case SAVEt_SHARED_PVREF: /* char* in shared space */
12837 c = (char*)POPPTR(ss,ix);
12838 TOPPTR(nss,ix) = savesharedpv(c);
12839 ptr = POPPTR(ss,ix);
12840 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12842 case SAVEt_GENERIC_SVREF: /* generic sv */
12843 case SAVEt_SVREF: /* scalar reference */
12844 sv = (const SV *)POPPTR(ss,ix);
12845 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12846 ptr = POPPTR(ss,ix);
12847 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12849 case SAVEt_GVSLOT: /* any slot in GV */
12850 sv = (const SV *)POPPTR(ss,ix);
12851 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12852 ptr = POPPTR(ss,ix);
12853 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12854 sv = (const SV *)POPPTR(ss,ix);
12855 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12857 case SAVEt_HV: /* hash reference */
12858 case SAVEt_AV: /* array reference */
12859 sv = (const SV *) POPPTR(ss,ix);
12860 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12862 case SAVEt_COMPPAD:
12864 sv = (const SV *) POPPTR(ss,ix);
12865 TOPPTR(nss,ix) = sv_dup(sv, param);
12867 case SAVEt_INT: /* int reference */
12868 ptr = POPPTR(ss,ix);
12869 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12870 intval = (int)POPINT(ss,ix);
12871 TOPINT(nss,ix) = intval;
12873 case SAVEt_LONG: /* long reference */
12874 ptr = POPPTR(ss,ix);
12875 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12876 longval = (long)POPLONG(ss,ix);
12877 TOPLONG(nss,ix) = longval;
12879 case SAVEt_I32: /* I32 reference */
12880 ptr = POPPTR(ss,ix);
12881 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12883 TOPINT(nss,ix) = i;
12885 case SAVEt_IV: /* IV reference */
12886 case SAVEt_STRLEN: /* STRLEN/size_t ref */
12887 ptr = POPPTR(ss,ix);
12888 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12890 TOPIV(nss,ix) = iv;
12892 case SAVEt_HPTR: /* HV* reference */
12893 case SAVEt_APTR: /* AV* reference */
12894 case SAVEt_SPTR: /* SV* reference */
12895 ptr = POPPTR(ss,ix);
12896 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12897 sv = (const SV *)POPPTR(ss,ix);
12898 TOPPTR(nss,ix) = sv_dup(sv, param);
12900 case SAVEt_VPTR: /* random* reference */
12901 ptr = POPPTR(ss,ix);
12902 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12904 case SAVEt_INT_SMALL:
12905 case SAVEt_I32_SMALL:
12906 case SAVEt_I16: /* I16 reference */
12907 case SAVEt_I8: /* I8 reference */
12909 ptr = POPPTR(ss,ix);
12910 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12912 case SAVEt_GENERIC_PVREF: /* generic char* */
12913 case SAVEt_PPTR: /* char* reference */
12914 ptr = POPPTR(ss,ix);
12915 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12916 c = (char*)POPPTR(ss,ix);
12917 TOPPTR(nss,ix) = pv_dup(c);
12919 case SAVEt_GP: /* scalar reference */
12920 gp = (GP*)POPPTR(ss,ix);
12921 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12922 (void)GpREFCNT_inc(gp);
12923 gv = (const GV *)POPPTR(ss,ix);
12924 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12927 ptr = POPPTR(ss,ix);
12928 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12929 /* these are assumed to be refcounted properly */
12931 switch (((OP*)ptr)->op_type) {
12933 case OP_LEAVESUBLV:
12937 case OP_LEAVEWRITE:
12938 TOPPTR(nss,ix) = ptr;
12941 (void) OpREFCNT_inc(o);
12945 TOPPTR(nss,ix) = NULL;
12950 TOPPTR(nss,ix) = NULL;
12952 case SAVEt_FREECOPHH:
12953 ptr = POPPTR(ss,ix);
12954 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12956 case SAVEt_ADELETE:
12957 av = (const AV *)POPPTR(ss,ix);
12958 TOPPTR(nss,ix) = av_dup_inc(av, param);
12960 TOPINT(nss,ix) = i;
12963 hv = (const HV *)POPPTR(ss,ix);
12964 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12966 TOPINT(nss,ix) = i;
12969 c = (char*)POPPTR(ss,ix);
12970 TOPPTR(nss,ix) = pv_dup_inc(c);
12972 case SAVEt_STACK_POS: /* Position on Perl stack */
12974 TOPINT(nss,ix) = i;
12976 case SAVEt_DESTRUCTOR:
12977 ptr = POPPTR(ss,ix);
12978 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12979 dptr = POPDPTR(ss,ix);
12980 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12981 any_dup(FPTR2DPTR(void *, dptr),
12984 case SAVEt_DESTRUCTOR_X:
12985 ptr = POPPTR(ss,ix);
12986 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12987 dxptr = POPDXPTR(ss,ix);
12988 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12989 any_dup(FPTR2DPTR(void *, dxptr),
12992 case SAVEt_REGCONTEXT:
12994 ix -= uv >> SAVE_TIGHT_SHIFT;
12996 case SAVEt_AELEM: /* array element */
12997 sv = (const SV *)POPPTR(ss,ix);
12998 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
13000 TOPINT(nss,ix) = i;
13001 av = (const AV *)POPPTR(ss,ix);
13002 TOPPTR(nss,ix) = av_dup_inc(av, param);
13005 ptr = POPPTR(ss,ix);
13006 TOPPTR(nss,ix) = ptr;
13009 ptr = POPPTR(ss,ix);
13010 ptr = cophh_copy((COPHH*)ptr);
13011 TOPPTR(nss,ix) = ptr;
13013 TOPINT(nss,ix) = i;
13014 if (i & HINT_LOCALIZE_HH) {
13015 hv = (const HV *)POPPTR(ss,ix);
13016 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
13019 case SAVEt_PADSV_AND_MORTALIZE:
13020 longval = (long)POPLONG(ss,ix);
13021 TOPLONG(nss,ix) = longval;
13022 ptr = POPPTR(ss,ix);
13023 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
13024 sv = (const SV *)POPPTR(ss,ix);
13025 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
13027 case SAVEt_SET_SVFLAGS:
13029 TOPINT(nss,ix) = i;
13031 TOPINT(nss,ix) = i;
13032 sv = (const SV *)POPPTR(ss,ix);
13033 TOPPTR(nss,ix) = sv_dup(sv, param);
13035 case SAVEt_COMPILE_WARNINGS:
13036 ptr = POPPTR(ss,ix);
13037 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
13040 ptr = POPPTR(ss,ix);
13041 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
13045 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
13053 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
13054 * flag to the result. This is done for each stash before cloning starts,
13055 * so we know which stashes want their objects cloned */
13058 do_mark_cloneable_stash(pTHX_ SV *const sv)
13060 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
13062 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
13063 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
13064 if (cloner && GvCV(cloner)) {
13071 mXPUSHs(newSVhek(hvname));
13073 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
13080 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
13088 =for apidoc perl_clone
13090 Create and return a new interpreter by cloning the current one.
13092 perl_clone takes these flags as parameters:
13094 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
13095 without it we only clone the data and zero the stacks,
13096 with it we copy the stacks and the new perl interpreter is
13097 ready to run at the exact same point as the previous one.
13098 The pseudo-fork code uses COPY_STACKS while the
13099 threads->create doesn't.
13101 CLONEf_KEEP_PTR_TABLE -
13102 perl_clone keeps a ptr_table with the pointer of the old
13103 variable as a key and the new variable as a value,
13104 this allows it to check if something has been cloned and not
13105 clone it again but rather just use the value and increase the
13106 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
13107 the ptr_table using the function
13108 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
13109 reason to keep it around is if you want to dup some of your own
13110 variable who are outside the graph perl scans, example of this
13111 code is in threads.xs create.
13113 CLONEf_CLONE_HOST -
13114 This is a win32 thing, it is ignored on unix, it tells perls
13115 win32host code (which is c++) to clone itself, this is needed on
13116 win32 if you want to run two threads at the same time,
13117 if you just want to do some stuff in a separate perl interpreter
13118 and then throw it away and return to the original one,
13119 you don't need to do anything.
13124 /* XXX the above needs expanding by someone who actually understands it ! */
13125 EXTERN_C PerlInterpreter *
13126 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
13129 perl_clone(PerlInterpreter *proto_perl, UV flags)
13132 #ifdef PERL_IMPLICIT_SYS
13134 PERL_ARGS_ASSERT_PERL_CLONE;
13136 /* perlhost.h so we need to call into it
13137 to clone the host, CPerlHost should have a c interface, sky */
13139 if (flags & CLONEf_CLONE_HOST) {
13140 return perl_clone_host(proto_perl,flags);
13142 return perl_clone_using(proto_perl, flags,
13144 proto_perl->IMemShared,
13145 proto_perl->IMemParse,
13147 proto_perl->IStdIO,
13151 proto_perl->IProc);
13155 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
13156 struct IPerlMem* ipM, struct IPerlMem* ipMS,
13157 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
13158 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
13159 struct IPerlDir* ipD, struct IPerlSock* ipS,
13160 struct IPerlProc* ipP)
13162 /* XXX many of the string copies here can be optimized if they're
13163 * constants; they need to be allocated as common memory and just
13164 * their pointers copied. */
13167 CLONE_PARAMS clone_params;
13168 CLONE_PARAMS* const param = &clone_params;
13170 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
13172 PERL_ARGS_ASSERT_PERL_CLONE_USING;
13173 #else /* !PERL_IMPLICIT_SYS */
13175 CLONE_PARAMS clone_params;
13176 CLONE_PARAMS* param = &clone_params;
13177 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
13179 PERL_ARGS_ASSERT_PERL_CLONE;
13180 #endif /* PERL_IMPLICIT_SYS */
13182 /* for each stash, determine whether its objects should be cloned */
13183 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
13184 PERL_SET_THX(my_perl);
13187 PoisonNew(my_perl, 1, PerlInterpreter);
13190 PL_defstash = NULL; /* may be used by perl malloc() */
13193 PL_scopestack_name = 0;
13195 PL_savestack_ix = 0;
13196 PL_savestack_max = -1;
13197 PL_sig_pending = 0;
13199 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
13200 # ifdef DEBUG_LEAKING_SCALARS
13201 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
13203 #else /* !DEBUGGING */
13204 Zero(my_perl, 1, PerlInterpreter);
13205 #endif /* DEBUGGING */
13207 #ifdef PERL_IMPLICIT_SYS
13208 /* host pointers */
13210 PL_MemShared = ipMS;
13211 PL_MemParse = ipMP;
13218 #endif /* PERL_IMPLICIT_SYS */
13221 param->flags = flags;
13222 /* Nothing in the core code uses this, but we make it available to
13223 extensions (using mg_dup). */
13224 param->proto_perl = proto_perl;
13225 /* Likely nothing will use this, but it is initialised to be consistent
13226 with Perl_clone_params_new(). */
13227 param->new_perl = my_perl;
13228 param->unreferenced = NULL;
13231 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
13233 PL_body_arenas = NULL;
13234 Zero(&PL_body_roots, 1, PL_body_roots);
13238 PL_sv_arenaroot = NULL;
13240 PL_debug = proto_perl->Idebug;
13242 /* dbargs array probably holds garbage */
13245 PL_compiling = proto_perl->Icompiling;
13247 /* pseudo environmental stuff */
13248 PL_origargc = proto_perl->Iorigargc;
13249 PL_origargv = proto_perl->Iorigargv;
13251 #if !NO_TAINT_SUPPORT
13252 /* Set tainting stuff before PerlIO_debug can possibly get called */
13253 PL_tainting = proto_perl->Itainting;
13254 PL_taint_warn = proto_perl->Itaint_warn;
13256 PL_tainting = FALSE;
13257 PL_taint_warn = FALSE;
13260 PL_minus_c = proto_perl->Iminus_c;
13262 PL_localpatches = proto_perl->Ilocalpatches;
13263 PL_splitstr = proto_perl->Isplitstr;
13264 PL_minus_n = proto_perl->Iminus_n;
13265 PL_minus_p = proto_perl->Iminus_p;
13266 PL_minus_l = proto_perl->Iminus_l;
13267 PL_minus_a = proto_perl->Iminus_a;
13268 PL_minus_E = proto_perl->Iminus_E;
13269 PL_minus_F = proto_perl->Iminus_F;
13270 PL_doswitches = proto_perl->Idoswitches;
13271 PL_dowarn = proto_perl->Idowarn;
13272 #ifdef PERL_SAWAMPERSAND
13273 PL_sawampersand = proto_perl->Isawampersand;
13275 PL_unsafe = proto_perl->Iunsafe;
13276 PL_perldb = proto_perl->Iperldb;
13277 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
13278 PL_exit_flags = proto_perl->Iexit_flags;
13280 /* XXX time(&PL_basetime) when asked for? */
13281 PL_basetime = proto_perl->Ibasetime;
13283 PL_maxsysfd = proto_perl->Imaxsysfd;
13284 PL_statusvalue = proto_perl->Istatusvalue;
13286 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
13288 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
13291 /* RE engine related */
13292 PL_regmatch_slab = NULL;
13293 PL_reg_curpm = NULL;
13295 PL_sub_generation = proto_perl->Isub_generation;
13297 /* funky return mechanisms */
13298 PL_forkprocess = proto_perl->Iforkprocess;
13300 /* internal state */
13301 PL_maxo = proto_perl->Imaxo;
13303 PL_main_start = proto_perl->Imain_start;
13304 PL_eval_root = proto_perl->Ieval_root;
13305 PL_eval_start = proto_perl->Ieval_start;
13307 PL_filemode = proto_perl->Ifilemode;
13308 PL_lastfd = proto_perl->Ilastfd;
13309 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
13312 PL_gensym = proto_perl->Igensym;
13314 PL_laststatval = proto_perl->Ilaststatval;
13315 PL_laststype = proto_perl->Ilaststype;
13318 PL_profiledata = NULL;
13320 PL_generation = proto_perl->Igeneration;
13322 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13323 PL_in_clean_all = proto_perl->Iin_clean_all;
13325 PL_delaymagic_uid = proto_perl->Idelaymagic_uid;
13326 PL_delaymagic_euid = proto_perl->Idelaymagic_euid;
13327 PL_delaymagic_gid = proto_perl->Idelaymagic_gid;
13328 PL_delaymagic_egid = proto_perl->Idelaymagic_egid;
13329 PL_nomemok = proto_perl->Inomemok;
13330 PL_an = proto_perl->Ian;
13331 PL_evalseq = proto_perl->Ievalseq;
13332 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13333 PL_origalen = proto_perl->Iorigalen;
13335 PL_sighandlerp = proto_perl->Isighandlerp;
13337 PL_runops = proto_perl->Irunops;
13339 PL_subline = proto_perl->Isubline;
13342 PL_cryptseen = proto_perl->Icryptseen;
13345 #ifdef USE_LOCALE_COLLATE
13346 PL_collation_ix = proto_perl->Icollation_ix;
13347 PL_collation_standard = proto_perl->Icollation_standard;
13348 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13349 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13350 #endif /* USE_LOCALE_COLLATE */
13352 #ifdef USE_LOCALE_NUMERIC
13353 PL_numeric_standard = proto_perl->Inumeric_standard;
13354 PL_numeric_local = proto_perl->Inumeric_local;
13355 #endif /* !USE_LOCALE_NUMERIC */
13357 /* Did the locale setup indicate UTF-8? */
13358 PL_utf8locale = proto_perl->Iutf8locale;
13359 /* Unicode features (see perlrun/-C) */
13360 PL_unicode = proto_perl->Iunicode;
13362 /* Pre-5.8 signals control */
13363 PL_signals = proto_perl->Isignals;
13365 /* times() ticks per second */
13366 PL_clocktick = proto_perl->Iclocktick;
13368 /* Recursion stopper for PerlIO_find_layer */
13369 PL_in_load_module = proto_perl->Iin_load_module;
13371 /* sort() routine */
13372 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13374 /* Not really needed/useful since the reenrant_retint is "volatile",
13375 * but do it for consistency's sake. */
13376 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13378 /* Hooks to shared SVs and locks. */
13379 PL_sharehook = proto_perl->Isharehook;
13380 PL_lockhook = proto_perl->Ilockhook;
13381 PL_unlockhook = proto_perl->Iunlockhook;
13382 PL_threadhook = proto_perl->Ithreadhook;
13383 PL_destroyhook = proto_perl->Idestroyhook;
13384 PL_signalhook = proto_perl->Isignalhook;
13386 PL_globhook = proto_perl->Iglobhook;
13389 PL_last_swash_hv = NULL; /* reinits on demand */
13390 PL_last_swash_klen = 0;
13391 PL_last_swash_key[0]= '\0';
13392 PL_last_swash_tmps = (U8*)NULL;
13393 PL_last_swash_slen = 0;
13395 PL_srand_called = proto_perl->Isrand_called;
13397 if (flags & CLONEf_COPY_STACKS) {
13398 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13399 PL_tmps_ix = proto_perl->Itmps_ix;
13400 PL_tmps_max = proto_perl->Itmps_max;
13401 PL_tmps_floor = proto_perl->Itmps_floor;
13403 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13404 * NOTE: unlike the others! */
13405 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13406 PL_scopestack_max = proto_perl->Iscopestack_max;
13408 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13409 * NOTE: unlike the others! */
13410 PL_savestack_ix = proto_perl->Isavestack_ix;
13411 PL_savestack_max = proto_perl->Isavestack_max;
13414 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13415 PL_top_env = &PL_start_env;
13417 PL_op = proto_perl->Iop;
13420 PL_Xpv = (XPV*)NULL;
13421 my_perl->Ina = proto_perl->Ina;
13423 PL_statbuf = proto_perl->Istatbuf;
13424 PL_statcache = proto_perl->Istatcache;
13427 PL_timesbuf = proto_perl->Itimesbuf;
13430 #if !NO_TAINT_SUPPORT
13431 PL_tainted = proto_perl->Itainted;
13433 PL_tainted = FALSE;
13435 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13437 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13439 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13440 PL_restartop = proto_perl->Irestartop;
13441 PL_in_eval = proto_perl->Iin_eval;
13442 PL_delaymagic = proto_perl->Idelaymagic;
13443 PL_phase = proto_perl->Iphase;
13444 PL_localizing = proto_perl->Ilocalizing;
13446 PL_hv_fetch_ent_mh = NULL;
13447 PL_modcount = proto_perl->Imodcount;
13448 PL_lastgotoprobe = NULL;
13449 PL_dumpindent = proto_perl->Idumpindent;
13451 PL_efloatbuf = NULL; /* reinits on demand */
13452 PL_efloatsize = 0; /* reinits on demand */
13456 PL_colorset = 0; /* reinits PL_colors[] */
13457 /*PL_colors[6] = {0,0,0,0,0,0};*/
13459 /* Pluggable optimizer */
13460 PL_peepp = proto_perl->Ipeepp;
13461 PL_rpeepp = proto_perl->Irpeepp;
13462 /* op_free() hook */
13463 PL_opfreehook = proto_perl->Iopfreehook;
13465 #ifdef USE_REENTRANT_API
13466 /* XXX: things like -Dm will segfault here in perlio, but doing
13467 * PERL_SET_CONTEXT(proto_perl);
13468 * breaks too many other things
13470 Perl_reentrant_init(aTHX);
13473 /* create SV map for pointer relocation */
13474 PL_ptr_table = ptr_table_new();
13476 /* initialize these special pointers as early as possible */
13478 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13479 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13480 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13482 /* create (a non-shared!) shared string table */
13483 PL_strtab = newHV();
13484 HvSHAREKEYS_off(PL_strtab);
13485 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13486 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13488 Zero(PL_sv_consts, SV_CONSTS_COUNT, SV*);
13490 /* This PV will be free'd special way so must set it same way op.c does */
13491 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13492 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13494 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13495 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13496 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13497 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13499 param->stashes = newAV(); /* Setup array of objects to call clone on */
13500 /* This makes no difference to the implementation, as it always pushes
13501 and shifts pointers to other SVs without changing their reference
13502 count, with the array becoming empty before it is freed. However, it
13503 makes it conceptually clear what is going on, and will avoid some
13504 work inside av.c, filling slots between AvFILL() and AvMAX() with
13505 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13506 AvREAL_off(param->stashes);
13508 if (!(flags & CLONEf_COPY_STACKS)) {
13509 param->unreferenced = newAV();
13512 #ifdef PERLIO_LAYERS
13513 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13514 PerlIO_clone(aTHX_ proto_perl, param);
13517 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13518 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13519 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13520 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13521 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13522 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13525 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13526 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13527 PL_inplace = SAVEPV(proto_perl->Iinplace);
13528 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13530 /* magical thingies */
13532 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13534 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13535 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13536 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13539 /* Clone the regex array */
13540 /* ORANGE FIXME for plugins, probably in the SV dup code.
13541 newSViv(PTR2IV(CALLREGDUPE(
13542 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13544 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13545 PL_regex_pad = AvARRAY(PL_regex_padav);
13547 PL_stashpadmax = proto_perl->Istashpadmax;
13548 PL_stashpadix = proto_perl->Istashpadix ;
13549 Newx(PL_stashpad, PL_stashpadmax, HV *);
13552 for (; o < PL_stashpadmax; ++o)
13553 PL_stashpad[o] = hv_dup(proto_perl->Istashpad[o], param);
13556 /* shortcuts to various I/O objects */
13557 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13558 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13559 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13560 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13561 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13562 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13563 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13565 /* shortcuts to regexp stuff */
13566 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13568 /* shortcuts to misc objects */
13569 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13571 /* shortcuts to debugging objects */
13572 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13573 PL_DBline = gv_dup(proto_perl->IDBline, param);
13574 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13575 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13576 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13577 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13579 /* symbol tables */
13580 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13581 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13582 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13583 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13584 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13586 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13587 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13588 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13589 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13590 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13591 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13592 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13593 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13595 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13597 /* subprocess state */
13598 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13600 if (proto_perl->Iop_mask)
13601 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13604 /* PL_asserting = proto_perl->Iasserting; */
13606 /* current interpreter roots */
13607 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13609 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13612 /* runtime control stuff */
13613 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13615 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13617 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13619 /* interpreter atexit processing */
13620 PL_exitlistlen = proto_perl->Iexitlistlen;
13621 if (PL_exitlistlen) {
13622 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13623 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13626 PL_exitlist = (PerlExitListEntry*)NULL;
13628 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13629 if (PL_my_cxt_size) {
13630 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13631 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13632 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13633 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13634 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13638 PL_my_cxt_list = (void**)NULL;
13639 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13640 PL_my_cxt_keys = (const char**)NULL;
13643 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13644 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13645 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13646 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13648 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13650 PAD_CLONE_VARS(proto_perl, param);
13652 #ifdef HAVE_INTERP_INTERN
13653 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13656 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13658 #ifdef PERL_USES_PL_PIDSTATUS
13659 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13661 PL_osname = SAVEPV(proto_perl->Iosname);
13662 PL_parser = parser_dup(proto_perl->Iparser, param);
13664 /* XXX this only works if the saved cop has already been cloned */
13665 if (proto_perl->Iparser) {
13666 PL_parser->saved_curcop = (COP*)any_dup(
13667 proto_perl->Iparser->saved_curcop,
13671 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13673 #ifdef USE_LOCALE_COLLATE
13674 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13675 #endif /* USE_LOCALE_COLLATE */
13677 #ifdef USE_LOCALE_NUMERIC
13678 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13679 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13680 #endif /* !USE_LOCALE_NUMERIC */
13682 /* Unicode inversion lists */
13683 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13684 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13685 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13687 PL_NonL1NonFinalFold = sv_dup_inc(proto_perl->INonL1NonFinalFold, param);
13688 PL_HasMultiCharFold= sv_dup_inc(proto_perl->IHasMultiCharFold, param);
13690 /* utf8 character class swashes */
13691 for (i = 0; i < POSIX_SWASH_COUNT; i++) {
13692 PL_utf8_swash_ptrs[i] = sv_dup_inc(proto_perl->Iutf8_swash_ptrs[i], param);
13694 for (i = 0; i < POSIX_CC_COUNT; i++) {
13695 PL_Posix_ptrs[i] = sv_dup_inc(proto_perl->IPosix_ptrs[i], param);
13696 PL_L1Posix_ptrs[i] = sv_dup_inc(proto_perl->IL1Posix_ptrs[i], param);
13697 PL_XPosix_ptrs[i] = sv_dup_inc(proto_perl->IXPosix_ptrs[i], param);
13699 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13700 PL_utf8_X_regular_begin = sv_dup_inc(proto_perl->Iutf8_X_regular_begin, param);
13701 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13702 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13703 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13704 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13705 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13706 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13707 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13708 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13709 PL_utf8_perl_idcont = sv_dup_inc(proto_perl->Iutf8_perl_idcont, param);
13710 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13711 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13712 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13713 PL_utf8_charname_begin = sv_dup_inc(proto_perl->Iutf8_charname_begin, param);
13714 PL_utf8_charname_continue = sv_dup_inc(proto_perl->Iutf8_charname_continue, param);
13716 if (proto_perl->Ipsig_pend) {
13717 Newxz(PL_psig_pend, SIG_SIZE, int);
13720 PL_psig_pend = (int*)NULL;
13723 if (proto_perl->Ipsig_name) {
13724 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13725 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13727 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13730 PL_psig_ptr = (SV**)NULL;
13731 PL_psig_name = (SV**)NULL;
13734 if (flags & CLONEf_COPY_STACKS) {
13735 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13736 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13737 PL_tmps_ix+1, param);
13739 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13740 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13741 Newxz(PL_markstack, i, I32);
13742 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13743 - proto_perl->Imarkstack);
13744 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13745 - proto_perl->Imarkstack);
13746 Copy(proto_perl->Imarkstack, PL_markstack,
13747 PL_markstack_ptr - PL_markstack + 1, I32);
13749 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13750 * NOTE: unlike the others! */
13751 Newxz(PL_scopestack, PL_scopestack_max, I32);
13752 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13755 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13756 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13758 /* reset stack AV to correct length before its duped via
13759 * PL_curstackinfo */
13760 AvFILLp(proto_perl->Icurstack) =
13761 proto_perl->Istack_sp - proto_perl->Istack_base;
13763 /* NOTE: si_dup() looks at PL_markstack */
13764 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13766 /* PL_curstack = PL_curstackinfo->si_stack; */
13767 PL_curstack = av_dup(proto_perl->Icurstack, param);
13768 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13770 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13771 PL_stack_base = AvARRAY(PL_curstack);
13772 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13773 - proto_perl->Istack_base);
13774 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13776 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13777 PL_savestack = ss_dup(proto_perl, param);
13781 ENTER; /* perl_destruct() wants to LEAVE; */
13784 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13785 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13787 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13788 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13789 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13790 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13791 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13792 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13794 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13796 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13797 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13798 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13800 PL_stashcache = newHV();
13802 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13803 proto_perl->Iwatchaddr);
13804 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13805 if (PL_debug && PL_watchaddr) {
13806 PerlIO_printf(Perl_debug_log,
13807 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13808 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13809 PTR2UV(PL_watchok));
13812 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13813 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13814 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13816 /* Call the ->CLONE method, if it exists, for each of the stashes
13817 identified by sv_dup() above.
13819 while(av_len(param->stashes) != -1) {
13820 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13821 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13822 if (cloner && GvCV(cloner)) {
13827 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13829 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13835 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13836 ptr_table_free(PL_ptr_table);
13837 PL_ptr_table = NULL;
13840 if (!(flags & CLONEf_COPY_STACKS)) {
13841 unreferenced_to_tmp_stack(param->unreferenced);
13844 SvREFCNT_dec(param->stashes);
13846 /* orphaned? eg threads->new inside BEGIN or use */
13847 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13848 SvREFCNT_inc_simple_void(PL_compcv);
13849 SAVEFREESV(PL_compcv);
13856 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13858 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13860 if (AvFILLp(unreferenced) > -1) {
13861 SV **svp = AvARRAY(unreferenced);
13862 SV **const last = svp + AvFILLp(unreferenced);
13866 if (SvREFCNT(*svp) == 1)
13868 } while (++svp <= last);
13870 EXTEND_MORTAL(count);
13871 svp = AvARRAY(unreferenced);
13874 if (SvREFCNT(*svp) == 1) {
13875 /* Our reference is the only one to this SV. This means that
13876 in this thread, the scalar effectively has a 0 reference.
13877 That doesn't work (cleanup never happens), so donate our
13878 reference to it onto the save stack. */
13879 PL_tmps_stack[++PL_tmps_ix] = *svp;
13881 /* As an optimisation, because we are already walking the
13882 entire array, instead of above doing either
13883 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13884 release our reference to the scalar, so that at the end of
13885 the array owns zero references to the scalars it happens to
13886 point to. We are effectively converting the array from
13887 AvREAL() on to AvREAL() off. This saves the av_clear()
13888 (triggered by the SvREFCNT_dec(unreferenced) below) from
13889 walking the array a second time. */
13890 SvREFCNT_dec(*svp);
13893 } while (++svp <= last);
13894 AvREAL_off(unreferenced);
13896 SvREFCNT_dec_NN(unreferenced);
13900 Perl_clone_params_del(CLONE_PARAMS *param)
13902 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13904 PerlInterpreter *const to = param->new_perl;
13906 PerlInterpreter *const was = PERL_GET_THX;
13908 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13914 SvREFCNT_dec(param->stashes);
13915 if (param->unreferenced)
13916 unreferenced_to_tmp_stack(param->unreferenced);
13926 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13929 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13930 does a dTHX; to get the context from thread local storage.
13931 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13932 a version that passes in my_perl. */
13933 PerlInterpreter *const was = PERL_GET_THX;
13934 CLONE_PARAMS *param;
13936 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13942 /* Given that we've set the context, we can do this unshared. */
13943 Newx(param, 1, CLONE_PARAMS);
13946 param->proto_perl = from;
13947 param->new_perl = to;
13948 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13949 AvREAL_off(param->stashes);
13950 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13958 #endif /* USE_ITHREADS */
13961 Perl_init_constants(pTHX)
13963 SvREFCNT(&PL_sv_undef) = SvREFCNT_IMMORTAL;
13964 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
13965 SvANY(&PL_sv_undef) = NULL;
13967 SvANY(&PL_sv_no) = new_XPVNV();
13968 SvREFCNT(&PL_sv_no) = SvREFCNT_IMMORTAL;
13969 SvFLAGS(&PL_sv_no) = SVt_PVNV|SVf_READONLY
13970 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13973 SvANY(&PL_sv_yes) = new_XPVNV();
13974 SvREFCNT(&PL_sv_yes) = SvREFCNT_IMMORTAL;
13975 SvFLAGS(&PL_sv_yes) = SVt_PVNV|SVf_READONLY
13976 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13979 SvPV_set(&PL_sv_no, (char*)PL_No);
13980 SvCUR_set(&PL_sv_no, 0);
13981 SvLEN_set(&PL_sv_no, 0);
13982 SvIV_set(&PL_sv_no, 0);
13983 SvNV_set(&PL_sv_no, 0);
13985 SvPV_set(&PL_sv_yes, (char*)PL_Yes);
13986 SvCUR_set(&PL_sv_yes, 1);
13987 SvLEN_set(&PL_sv_yes, 0);
13988 SvIV_set(&PL_sv_yes, 1);
13989 SvNV_set(&PL_sv_yes, 1);
13993 =head1 Unicode Support
13995 =for apidoc sv_recode_to_utf8
13997 The encoding is assumed to be an Encode object, on entry the PV
13998 of the sv is assumed to be octets in that encoding, and the sv
13999 will be converted into Unicode (and UTF-8).
14001 If the sv already is UTF-8 (or if it is not POK), or if the encoding
14002 is not a reference, nothing is done to the sv. If the encoding is not
14003 an C<Encode::XS> Encoding object, bad things will happen.
14004 (See F<lib/encoding.pm> and L<Encode>.)
14006 The PV of the sv is returned.
14011 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
14015 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
14017 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
14031 Passing sv_yes is wrong - it needs to be or'ed set of constants
14032 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
14033 remove converted chars from source.
14035 Both will default the value - let them.
14037 XPUSHs(&PL_sv_yes);
14040 call_method("decode", G_SCALAR);
14044 s = SvPV_const(uni, len);
14045 if (s != SvPVX_const(sv)) {
14046 SvGROW(sv, len + 1);
14047 Move(s, SvPVX(sv), len + 1, char);
14048 SvCUR_set(sv, len);
14052 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
14053 /* clear pos and any utf8 cache */
14054 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
14057 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
14058 magic_setutf8(sv,mg); /* clear UTF8 cache */
14063 return SvPOKp(sv) ? SvPVX(sv) : NULL;
14067 =for apidoc sv_cat_decode
14069 The encoding is assumed to be an Encode object, the PV of the ssv is
14070 assumed to be octets in that encoding and decoding the input starts
14071 from the position which (PV + *offset) pointed to. The dsv will be
14072 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
14073 when the string tstr appears in decoding output or the input ends on
14074 the PV of the ssv. The value which the offset points will be modified
14075 to the last input position on the ssv.
14077 Returns TRUE if the terminator was found, else returns FALSE.
14082 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
14083 SV *ssv, int *offset, char *tstr, int tlen)
14088 PERL_ARGS_ASSERT_SV_CAT_DECODE;
14090 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
14101 offsv = newSViv(*offset);
14103 mPUSHp(tstr, tlen);
14105 call_method("cat_decode", G_SCALAR);
14107 ret = SvTRUE(TOPs);
14108 *offset = SvIV(offsv);
14114 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
14119 /* ---------------------------------------------------------------------
14121 * support functions for report_uninit()
14124 /* the maxiumum size of array or hash where we will scan looking
14125 * for the undefined element that triggered the warning */
14127 #define FUV_MAX_SEARCH_SIZE 1000
14129 /* Look for an entry in the hash whose value has the same SV as val;
14130 * If so, return a mortal copy of the key. */
14133 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
14139 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
14141 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
14142 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
14145 array = HvARRAY(hv);
14147 for (i=HvMAX(hv); i>=0; i--) {
14149 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
14150 if (HeVAL(entry) != val)
14152 if ( HeVAL(entry) == &PL_sv_undef ||
14153 HeVAL(entry) == &PL_sv_placeholder)
14157 if (HeKLEN(entry) == HEf_SVKEY)
14158 return sv_mortalcopy(HeKEY_sv(entry));
14159 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
14165 /* Look for an entry in the array whose value has the same SV as val;
14166 * If so, return the index, otherwise return -1. */
14169 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
14173 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
14175 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
14176 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
14179 if (val != &PL_sv_undef) {
14180 SV ** const svp = AvARRAY(av);
14183 for (i=AvFILLp(av); i>=0; i--)
14190 /* varname(): return the name of a variable, optionally with a subscript.
14191 * If gv is non-zero, use the name of that global, along with gvtype (one
14192 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
14193 * targ. Depending on the value of the subscript_type flag, return:
14196 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
14197 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
14198 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
14199 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
14202 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
14203 const SV *const keyname, I32 aindex, int subscript_type)
14206 SV * const name = sv_newmortal();
14207 if (gv && isGV(gv)) {
14209 buffer[0] = gvtype;
14212 /* as gv_fullname4(), but add literal '^' for $^FOO names */
14214 gv_fullname4(name, gv, buffer, 0);
14216 if ((unsigned int)SvPVX(name)[1] <= 26) {
14218 buffer[1] = SvPVX(name)[1] + 'A' - 1;
14220 /* Swap the 1 unprintable control character for the 2 byte pretty
14221 version - ie substr($name, 1, 1) = $buffer; */
14222 sv_insert(name, 1, 1, buffer, 2);
14226 CV * const cv = gv ? ((CV *)gv) : find_runcv(NULL);
14230 assert(!cv || SvTYPE(cv) == SVt_PVCV || SvTYPE(cv) == SVt_PVFM);
14232 if (!cv || !CvPADLIST(cv))
14234 av = *PadlistARRAY(CvPADLIST(cv));
14235 sv = *av_fetch(av, targ, FALSE);
14236 sv_setsv_flags(name, sv, 0);
14239 if (subscript_type == FUV_SUBSCRIPT_HASH) {
14240 SV * const sv = newSV(0);
14241 *SvPVX(name) = '$';
14242 Perl_sv_catpvf(aTHX_ name, "{%s}",
14243 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
14244 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
14245 SvREFCNT_dec_NN(sv);
14247 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
14248 *SvPVX(name) = '$';
14249 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
14251 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
14252 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
14253 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
14261 =for apidoc find_uninit_var
14263 Find the name of the undefined variable (if any) that caused the operator
14264 to issue a "Use of uninitialized value" warning.
14265 If match is true, only return a name if its value matches uninit_sv.
14266 So roughly speaking, if a unary operator (such as OP_COS) generates a
14267 warning, then following the direct child of the op may yield an
14268 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
14269 other hand, with OP_ADD there are two branches to follow, so we only print
14270 the variable name if we get an exact match.
14272 The name is returned as a mortal SV.
14274 Assumes that PL_op is the op that originally triggered the error, and that
14275 PL_comppad/PL_curpad points to the currently executing pad.
14281 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
14287 const OP *o, *o2, *kid;
14289 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
14290 uninit_sv == &PL_sv_placeholder)))
14293 switch (obase->op_type) {
14300 const bool pad = ( obase->op_type == OP_PADAV
14301 || obase->op_type == OP_PADHV
14302 || obase->op_type == OP_PADRANGE
14305 const bool hash = ( obase->op_type == OP_PADHV
14306 || obase->op_type == OP_RV2HV
14307 || (obase->op_type == OP_PADRANGE
14308 && SvTYPE(PAD_SVl(obase->op_targ)) == SVt_PVHV)
14312 int subscript_type = FUV_SUBSCRIPT_WITHIN;
14314 if (pad) { /* @lex, %lex */
14315 sv = PAD_SVl(obase->op_targ);
14319 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14320 /* @global, %global */
14321 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14324 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14326 else if (obase == PL_op) /* @{expr}, %{expr} */
14327 return find_uninit_var(cUNOPx(obase)->op_first,
14329 else /* @{expr}, %{expr} as a sub-expression */
14333 /* attempt to find a match within the aggregate */
14335 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14337 subscript_type = FUV_SUBSCRIPT_HASH;
14340 index = find_array_subscript((const AV *)sv, uninit_sv);
14342 subscript_type = FUV_SUBSCRIPT_ARRAY;
14345 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14348 return varname(gv, hash ? '%' : '@', obase->op_targ,
14349 keysv, index, subscript_type);
14353 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14355 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14356 if (!gv || !GvSTASH(gv))
14358 if (match && (GvSV(gv) != uninit_sv))
14360 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14363 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14366 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14368 return varname(NULL, '$', obase->op_targ,
14369 NULL, 0, FUV_SUBSCRIPT_NONE);
14372 gv = cGVOPx_gv(obase);
14373 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14375 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14377 case OP_AELEMFAST_LEX:
14380 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14381 if (!av || SvRMAGICAL(av))
14383 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14384 if (!svp || *svp != uninit_sv)
14387 return varname(NULL, '$', obase->op_targ,
14388 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14391 gv = cGVOPx_gv(obase);
14396 AV *const av = GvAV(gv);
14397 if (!av || SvRMAGICAL(av))
14399 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14400 if (!svp || *svp != uninit_sv)
14403 return varname(gv, '$', 0,
14404 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14409 o = cUNOPx(obase)->op_first;
14410 if (!o || o->op_type != OP_NULL ||
14411 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14413 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14418 bool negate = FALSE;
14420 if (PL_op == obase)
14421 /* $a[uninit_expr] or $h{uninit_expr} */
14422 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14425 o = cBINOPx(obase)->op_first;
14426 kid = cBINOPx(obase)->op_last;
14428 /* get the av or hv, and optionally the gv */
14430 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14431 sv = PAD_SV(o->op_targ);
14433 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14434 && cUNOPo->op_first->op_type == OP_GV)
14436 gv = cGVOPx_gv(cUNOPo->op_first);
14440 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14445 if (kid && kid->op_type == OP_NEGATE) {
14447 kid = cUNOPx(kid)->op_first;
14450 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14451 /* index is constant */
14454 kidsv = sv_2mortal(newSVpvs("-"));
14455 sv_catsv(kidsv, cSVOPx_sv(kid));
14458 kidsv = cSVOPx_sv(kid);
14462 if (obase->op_type == OP_HELEM) {
14463 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14464 if (!he || HeVAL(he) != uninit_sv)
14468 SV * const opsv = cSVOPx_sv(kid);
14469 const IV opsviv = SvIV(opsv);
14470 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14471 negate ? - opsviv : opsviv,
14473 if (!svp || *svp != uninit_sv)
14477 if (obase->op_type == OP_HELEM)
14478 return varname(gv, '%', o->op_targ,
14479 kidsv, 0, FUV_SUBSCRIPT_HASH);
14481 return varname(gv, '@', o->op_targ, NULL,
14482 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14483 FUV_SUBSCRIPT_ARRAY);
14486 /* index is an expression;
14487 * attempt to find a match within the aggregate */
14488 if (obase->op_type == OP_HELEM) {
14489 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14491 return varname(gv, '%', o->op_targ,
14492 keysv, 0, FUV_SUBSCRIPT_HASH);
14496 = find_array_subscript((const AV *)sv, uninit_sv);
14498 return varname(gv, '@', o->op_targ,
14499 NULL, index, FUV_SUBSCRIPT_ARRAY);
14504 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14506 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14512 /* only examine RHS */
14513 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14516 o = cUNOPx(obase)->op_first;
14517 if ( o->op_type == OP_PUSHMARK
14518 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)
14522 if (!o->op_sibling) {
14523 /* one-arg version of open is highly magical */
14525 if (o->op_type == OP_GV) { /* open FOO; */
14527 if (match && GvSV(gv) != uninit_sv)
14529 return varname(gv, '$', 0,
14530 NULL, 0, FUV_SUBSCRIPT_NONE);
14532 /* other possibilities not handled are:
14533 * open $x; or open my $x; should return '${*$x}'
14534 * open expr; should return '$'.expr ideally
14540 /* ops where $_ may be an implicit arg */
14545 if ( !(obase->op_flags & OPf_STACKED)) {
14546 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14547 ? PAD_SVl(obase->op_targ)
14550 sv = sv_newmortal();
14551 sv_setpvs(sv, "$_");
14560 match = 1; /* print etc can return undef on defined args */
14561 /* skip filehandle as it can't produce 'undef' warning */
14562 o = cUNOPx(obase)->op_first;
14563 if ((obase->op_flags & OPf_STACKED)
14565 ( o->op_type == OP_PUSHMARK
14566 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)))
14567 o = o->op_sibling->op_sibling;
14571 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14572 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14574 /* the following ops are capable of returning PL_sv_undef even for
14575 * defined arg(s) */
14594 case OP_GETPEERNAME:
14642 case OP_SMARTMATCH:
14651 /* XXX tmp hack: these two may call an XS sub, and currently
14652 XS subs don't have a SUB entry on the context stack, so CV and
14653 pad determination goes wrong, and BAD things happen. So, just
14654 don't try to determine the value under those circumstances.
14655 Need a better fix at dome point. DAPM 11/2007 */
14661 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14662 if (gv && GvSV(gv) == uninit_sv)
14663 return newSVpvs_flags("$.", SVs_TEMP);
14668 /* def-ness of rval pos() is independent of the def-ness of its arg */
14669 if ( !(obase->op_flags & OPf_MOD))
14674 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14675 return newSVpvs_flags("${$/}", SVs_TEMP);
14680 if (!(obase->op_flags & OPf_KIDS))
14682 o = cUNOPx(obase)->op_first;
14688 /* This loop checks all the kid ops, skipping any that cannot pos-
14689 * sibly be responsible for the uninitialized value; i.e., defined
14690 * constants and ops that return nothing. If there is only one op
14691 * left that is not skipped, then we *know* it is responsible for
14692 * the uninitialized value. If there is more than one op left, we
14693 * have to look for an exact match in the while() loop below.
14694 * Note that we skip padrange, because the individual pad ops that
14695 * it replaced are still in the tree, so we work on them instead.
14698 for (kid=o; kid; kid = kid->op_sibling) {
14700 const OPCODE type = kid->op_type;
14701 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14702 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14703 || (type == OP_PUSHMARK)
14704 || (type == OP_PADRANGE)
14708 if (o2) { /* more than one found */
14715 return find_uninit_var(o2, uninit_sv, match);
14717 /* scan all args */
14719 sv = find_uninit_var(o, uninit_sv, 1);
14731 =for apidoc report_uninit
14733 Print appropriate "Use of uninitialized variable" warning.
14739 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14743 SV* varname = NULL;
14744 if (uninit_sv && PL_curpad) {
14745 varname = find_uninit_var(PL_op, uninit_sv,0);
14747 sv_insert(varname, 0, 0, " ", 1);
14749 /* diag_listed_as: Use of uninitialized value%s */
14750 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14751 SVfARG(varname ? varname : &PL_sv_no),
14752 " in ", OP_DESC(PL_op));
14755 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14761 * c-indentation-style: bsd
14762 * c-basic-offset: 4
14763 * indent-tabs-mode: nil
14766 * ex: set ts=8 sts=4 sw=4 et: