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))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. 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(target);
485 /* XXX Might want to check arrays, etc. */
489 /* clear any slots in a GV which hold objects - except IO;
490 * called by sv_clean_objs() for each live GV */
493 do_clean_named_objs(pTHX_ SV *const sv)
497 assert(SvTYPE(sv) == SVt_PVGV);
498 assert(isGV_with_GP(sv));
502 /* freeing GP entries may indirectly free the current GV;
503 * hold onto it while we mess with the GP slots */
506 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
507 DEBUG_D((PerlIO_printf(Perl_debug_log,
508 "Cleaning named glob SV object:\n "), sv_dump(obj)));
512 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
513 DEBUG_D((PerlIO_printf(Perl_debug_log,
514 "Cleaning named glob AV object:\n "), sv_dump(obj)));
518 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
519 DEBUG_D((PerlIO_printf(Perl_debug_log,
520 "Cleaning named glob HV object:\n "), sv_dump(obj)));
524 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
525 DEBUG_D((PerlIO_printf(Perl_debug_log,
526 "Cleaning named glob CV object:\n "), sv_dump(obj)));
530 SvREFCNT_dec(sv); /* undo the inc above */
533 /* clear any IO slots in a GV which hold objects (except stderr, defout);
534 * called by sv_clean_objs() for each live GV */
537 do_clean_named_io_objs(pTHX_ SV *const sv)
541 assert(SvTYPE(sv) == SVt_PVGV);
542 assert(isGV_with_GP(sv));
543 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
547 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
548 DEBUG_D((PerlIO_printf(Perl_debug_log,
549 "Cleaning named glob IO object:\n "), sv_dump(obj)));
553 SvREFCNT_dec(sv); /* undo the inc above */
556 /* Void wrapper to pass to visit() */
558 do_curse(pTHX_ SV * const sv) {
559 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
560 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
566 =for apidoc sv_clean_objs
568 Attempt to destroy all objects not yet freed.
574 Perl_sv_clean_objs(pTHX)
578 PL_in_clean_objs = TRUE;
579 visit(do_clean_objs, SVf_ROK, SVf_ROK);
580 /* Some barnacles may yet remain, clinging to typeglobs.
581 * Run the non-IO destructors first: they may want to output
582 * error messages, close files etc */
583 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
584 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
585 /* And if there are some very tenacious barnacles clinging to arrays,
586 closures, or what have you.... */
587 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
588 olddef = PL_defoutgv;
589 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
590 if (olddef && isGV_with_GP(olddef))
591 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
592 olderr = PL_stderrgv;
593 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
594 if (olderr && isGV_with_GP(olderr))
595 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
596 SvREFCNT_dec(olddef);
597 PL_in_clean_objs = FALSE;
600 /* called by sv_clean_all() for each live SV */
603 do_clean_all(pTHX_ SV *const sv)
606 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
607 /* don't clean pid table and strtab */
610 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
611 SvFLAGS(sv) |= SVf_BREAK;
616 =for apidoc sv_clean_all
618 Decrement the refcnt of each remaining SV, possibly triggering a
619 cleanup. This function may have to be called multiple times to free
620 SVs which are in complex self-referential hierarchies.
626 Perl_sv_clean_all(pTHX)
630 PL_in_clean_all = TRUE;
631 cleaned = visit(do_clean_all, 0,0);
636 ARENASETS: a meta-arena implementation which separates arena-info
637 into struct arena_set, which contains an array of struct
638 arena_descs, each holding info for a single arena. By separating
639 the meta-info from the arena, we recover the 1st slot, formerly
640 borrowed for list management. The arena_set is about the size of an
641 arena, avoiding the needless malloc overhead of a naive linked-list.
643 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
644 memory in the last arena-set (1/2 on average). In trade, we get
645 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
646 smaller types). The recovery of the wasted space allows use of
647 small arenas for large, rare body types, by changing array* fields
648 in body_details_by_type[] below.
651 char *arena; /* the raw storage, allocated aligned */
652 size_t size; /* its size ~4k typ */
653 svtype utype; /* bodytype stored in arena */
658 /* Get the maximum number of elements in set[] such that struct arena_set
659 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
660 therefore likely to be 1 aligned memory page. */
662 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
663 - 2 * sizeof(int)) / sizeof (struct arena_desc))
666 struct arena_set* next;
667 unsigned int set_size; /* ie ARENAS_PER_SET */
668 unsigned int curr; /* index of next available arena-desc */
669 struct arena_desc set[ARENAS_PER_SET];
673 =for apidoc sv_free_arenas
675 Deallocate the memory used by all arenas. Note that all the individual SV
676 heads and bodies within the arenas must already have been freed.
681 Perl_sv_free_arenas(pTHX)
688 /* Free arenas here, but be careful about fake ones. (We assume
689 contiguity of the fake ones with the corresponding real ones.) */
691 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
692 svanext = MUTABLE_SV(SvANY(sva));
693 while (svanext && SvFAKE(svanext))
694 svanext = MUTABLE_SV(SvANY(svanext));
701 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
704 struct arena_set *current = aroot;
707 assert(aroot->set[i].arena);
708 Safefree(aroot->set[i].arena);
716 i = PERL_ARENA_ROOTS_SIZE;
718 PL_body_roots[i] = 0;
725 Here are mid-level routines that manage the allocation of bodies out
726 of the various arenas. There are 5 kinds of arenas:
728 1. SV-head arenas, which are discussed and handled above
729 2. regular body arenas
730 3. arenas for reduced-size bodies
733 Arena types 2 & 3 are chained by body-type off an array of
734 arena-root pointers, which is indexed by svtype. Some of the
735 larger/less used body types are malloced singly, since a large
736 unused block of them is wasteful. Also, several svtypes dont have
737 bodies; the data fits into the sv-head itself. The arena-root
738 pointer thus has a few unused root-pointers (which may be hijacked
739 later for arena types 4,5)
741 3 differs from 2 as an optimization; some body types have several
742 unused fields in the front of the structure (which are kept in-place
743 for consistency). These bodies can be allocated in smaller chunks,
744 because the leading fields arent accessed. Pointers to such bodies
745 are decremented to point at the unused 'ghost' memory, knowing that
746 the pointers are used with offsets to the real memory.
749 =head1 SV-Body Allocation
751 Allocation of SV-bodies is similar to SV-heads, differing as follows;
752 the allocation mechanism is used for many body types, so is somewhat
753 more complicated, it uses arena-sets, and has no need for still-live
756 At the outermost level, (new|del)_X*V macros return bodies of the
757 appropriate type. These macros call either (new|del)_body_type or
758 (new|del)_body_allocated macro pairs, depending on specifics of the
759 type. Most body types use the former pair, the latter pair is used to
760 allocate body types with "ghost fields".
762 "ghost fields" are fields that are unused in certain types, and
763 consequently don't need to actually exist. They are declared because
764 they're part of a "base type", which allows use of functions as
765 methods. The simplest examples are AVs and HVs, 2 aggregate types
766 which don't use the fields which support SCALAR semantics.
768 For these types, the arenas are carved up into appropriately sized
769 chunks, we thus avoid wasted memory for those unaccessed members.
770 When bodies are allocated, we adjust the pointer back in memory by the
771 size of the part not allocated, so it's as if we allocated the full
772 structure. (But things will all go boom if you write to the part that
773 is "not there", because you'll be overwriting the last members of the
774 preceding structure in memory.)
776 We calculate the correction using the STRUCT_OFFSET macro on the first
777 member present. If the allocated structure is smaller (no initial NV
778 actually allocated) then the net effect is to subtract the size of the NV
779 from the pointer, to return a new pointer as if an initial NV were actually
780 allocated. (We were using structures named *_allocated for this, but
781 this turned out to be a subtle bug, because a structure without an NV
782 could have a lower alignment constraint, but the compiler is allowed to
783 optimised accesses based on the alignment constraint of the actual pointer
784 to the full structure, for example, using a single 64 bit load instruction
785 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
787 This is the same trick as was used for NV and IV bodies. Ironically it
788 doesn't need to be used for NV bodies any more, because NV is now at
789 the start of the structure. IV bodies don't need it either, because
790 they are no longer allocated.
792 In turn, the new_body_* allocators call S_new_body(), which invokes
793 new_body_inline macro, which takes a lock, and takes a body off the
794 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
795 necessary to refresh an empty list. Then the lock is released, and
796 the body is returned.
798 Perl_more_bodies allocates a new arena, and carves it up into an array of N
799 bodies, which it strings into a linked list. It looks up arena-size
800 and body-size from the body_details table described below, thus
801 supporting the multiple body-types.
803 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
804 the (new|del)_X*V macros are mapped directly to malloc/free.
806 For each sv-type, struct body_details bodies_by_type[] carries
807 parameters which control these aspects of SV handling:
809 Arena_size determines whether arenas are used for this body type, and if
810 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
811 zero, forcing individual mallocs and frees.
813 Body_size determines how big a body is, and therefore how many fit into
814 each arena. Offset carries the body-pointer adjustment needed for
815 "ghost fields", and is used in *_allocated macros.
817 But its main purpose is to parameterize info needed in
818 Perl_sv_upgrade(). The info here dramatically simplifies the function
819 vs the implementation in 5.8.8, making it table-driven. All fields
820 are used for this, except for arena_size.
822 For the sv-types that have no bodies, arenas are not used, so those
823 PL_body_roots[sv_type] are unused, and can be overloaded. In
824 something of a special case, SVt_NULL is borrowed for HE arenas;
825 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
826 bodies_by_type[SVt_NULL] slot is not used, as the table is not
831 struct body_details {
832 U8 body_size; /* Size to allocate */
833 U8 copy; /* Size of structure to copy (may be shorter) */
835 unsigned int type : 4; /* We have space for a sanity check. */
836 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
837 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
838 unsigned int arena : 1; /* Allocated from an arena */
839 size_t arena_size; /* Size of arena to allocate */
847 /* With -DPURFIY we allocate everything directly, and don't use arenas.
848 This seems a rather elegant way to simplify some of the code below. */
849 #define HASARENA FALSE
851 #define HASARENA TRUE
853 #define NOARENA FALSE
855 /* Size the arenas to exactly fit a given number of bodies. A count
856 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
857 simplifying the default. If count > 0, the arena is sized to fit
858 only that many bodies, allowing arenas to be used for large, rare
859 bodies (XPVFM, XPVIO) without undue waste. The arena size is
860 limited by PERL_ARENA_SIZE, so we can safely oversize the
863 #define FIT_ARENA0(body_size) \
864 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
865 #define FIT_ARENAn(count,body_size) \
866 ( count * body_size <= PERL_ARENA_SIZE) \
867 ? count * body_size \
868 : FIT_ARENA0 (body_size)
869 #define FIT_ARENA(count,body_size) \
871 ? FIT_ARENAn (count, body_size) \
872 : FIT_ARENA0 (body_size)
874 /* Calculate the length to copy. Specifically work out the length less any
875 final padding the compiler needed to add. See the comment in sv_upgrade
876 for why copying the padding proved to be a bug. */
878 #define copy_length(type, last_member) \
879 STRUCT_OFFSET(type, last_member) \
880 + sizeof (((type*)SvANY((const SV *)0))->last_member)
882 static const struct body_details bodies_by_type[] = {
883 /* HEs use this offset for their arena. */
884 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
886 /* The bind placeholder pretends to be an RV for now.
887 Also it's marked as "can't upgrade" to stop anyone using it before it's
889 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
891 /* IVs are in the head, so the allocation size is 0. */
893 sizeof(IV), /* This is used to copy out the IV body. */
894 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
895 NOARENA /* IVS don't need an arena */, 0
898 { sizeof(NV), sizeof(NV),
899 STRUCT_OFFSET(XPVNV, xnv_u),
900 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
902 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
903 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
904 + STRUCT_OFFSET(XPV, xpv_cur),
905 SVt_PV, FALSE, NONV, HASARENA,
906 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
908 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
909 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
910 + STRUCT_OFFSET(XPV, xpv_cur),
911 SVt_PVIV, FALSE, NONV, HASARENA,
912 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
914 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
915 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
916 + STRUCT_OFFSET(XPV, xpv_cur),
917 SVt_PVNV, FALSE, HADNV, HASARENA,
918 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
920 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
921 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
926 SVt_REGEXP, FALSE, NONV, HASARENA,
927 FIT_ARENA(0, sizeof(regexp))
930 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
931 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
933 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
934 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
937 copy_length(XPVAV, xav_alloc),
939 SVt_PVAV, TRUE, NONV, HASARENA,
940 FIT_ARENA(0, sizeof(XPVAV)) },
943 copy_length(XPVHV, xhv_max),
945 SVt_PVHV, TRUE, NONV, HASARENA,
946 FIT_ARENA(0, sizeof(XPVHV)) },
951 SVt_PVCV, TRUE, NONV, HASARENA,
952 FIT_ARENA(0, sizeof(XPVCV)) },
957 SVt_PVFM, TRUE, NONV, NOARENA,
958 FIT_ARENA(20, sizeof(XPVFM)) },
963 SVt_PVIO, TRUE, NONV, HASARENA,
964 FIT_ARENA(24, sizeof(XPVIO)) },
967 #define new_body_allocated(sv_type) \
968 (void *)((char *)S_new_body(aTHX_ sv_type) \
969 - bodies_by_type[sv_type].offset)
971 /* return a thing to the free list */
973 #define del_body(thing, root) \
975 void ** const thing_copy = (void **)thing; \
976 *thing_copy = *root; \
977 *root = (void*)thing_copy; \
982 #define new_XNV() safemalloc(sizeof(XPVNV))
983 #define new_XPVNV() safemalloc(sizeof(XPVNV))
984 #define new_XPVMG() safemalloc(sizeof(XPVMG))
986 #define del_XPVGV(p) safefree(p)
990 #define new_XNV() new_body_allocated(SVt_NV)
991 #define new_XPVNV() new_body_allocated(SVt_PVNV)
992 #define new_XPVMG() new_body_allocated(SVt_PVMG)
994 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
995 &PL_body_roots[SVt_PVGV])
999 /* no arena for you! */
1001 #define new_NOARENA(details) \
1002 safemalloc((details)->body_size + (details)->offset)
1003 #define new_NOARENAZ(details) \
1004 safecalloc((details)->body_size + (details)->offset, 1)
1007 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1008 const size_t arena_size)
1011 void ** const root = &PL_body_roots[sv_type];
1012 struct arena_desc *adesc;
1013 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1017 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1018 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1019 static bool done_sanity_check;
1021 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1022 * variables like done_sanity_check. */
1023 if (!done_sanity_check) {
1024 unsigned int i = SVt_LAST;
1026 done_sanity_check = TRUE;
1029 assert (bodies_by_type[i].type == i);
1035 /* may need new arena-set to hold new arena */
1036 if (!aroot || aroot->curr >= aroot->set_size) {
1037 struct arena_set *newroot;
1038 Newxz(newroot, 1, struct arena_set);
1039 newroot->set_size = ARENAS_PER_SET;
1040 newroot->next = aroot;
1042 PL_body_arenas = (void *) newroot;
1043 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1046 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1047 curr = aroot->curr++;
1048 adesc = &(aroot->set[curr]);
1049 assert(!adesc->arena);
1051 Newx(adesc->arena, good_arena_size, char);
1052 adesc->size = good_arena_size;
1053 adesc->utype = sv_type;
1054 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1055 curr, (void*)adesc->arena, (UV)good_arena_size));
1057 start = (char *) adesc->arena;
1059 /* Get the address of the byte after the end of the last body we can fit.
1060 Remember, this is integer division: */
1061 end = start + good_arena_size / body_size * body_size;
1063 /* computed count doesn't reflect the 1st slot reservation */
1064 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1065 DEBUG_m(PerlIO_printf(Perl_debug_log,
1066 "arena %p end %p arena-size %d (from %d) type %d "
1068 (void*)start, (void*)end, (int)good_arena_size,
1069 (int)arena_size, sv_type, (int)body_size,
1070 (int)good_arena_size / (int)body_size));
1072 DEBUG_m(PerlIO_printf(Perl_debug_log,
1073 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1074 (void*)start, (void*)end,
1075 (int)arena_size, sv_type, (int)body_size,
1076 (int)good_arena_size / (int)body_size));
1078 *root = (void *)start;
1081 /* Where the next body would start: */
1082 char * const next = start + body_size;
1085 /* This is the last body: */
1086 assert(next == end);
1088 *(void **)start = 0;
1092 *(void**) start = (void *)next;
1097 /* grab a new thing from the free list, allocating more if necessary.
1098 The inline version is used for speed in hot routines, and the
1099 function using it serves the rest (unless PURIFY).
1101 #define new_body_inline(xpv, sv_type) \
1103 void ** const r3wt = &PL_body_roots[sv_type]; \
1104 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1105 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1106 bodies_by_type[sv_type].body_size,\
1107 bodies_by_type[sv_type].arena_size)); \
1108 *(r3wt) = *(void**)(xpv); \
1114 S_new_body(pTHX_ const svtype sv_type)
1118 new_body_inline(xpv, sv_type);
1124 static const struct body_details fake_rv =
1125 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1128 =for apidoc sv_upgrade
1130 Upgrade an SV to a more complex form. Generally adds a new body type to the
1131 SV, then copies across as much information as possible from the old body.
1132 It croaks if the SV is already in a more complex form than requested. You
1133 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1134 before calling C<sv_upgrade>, and hence does not croak. See also
1141 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1146 const svtype old_type = SvTYPE(sv);
1147 const struct body_details *new_type_details;
1148 const struct body_details *old_type_details
1149 = bodies_by_type + old_type;
1150 SV *referant = NULL;
1152 PERL_ARGS_ASSERT_SV_UPGRADE;
1154 if (old_type == new_type)
1157 /* This clause was purposefully added ahead of the early return above to
1158 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1159 inference by Nick I-S that it would fix other troublesome cases. See
1160 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1162 Given that shared hash key scalars are no longer PVIV, but PV, there is
1163 no longer need to unshare so as to free up the IVX slot for its proper
1164 purpose. So it's safe to move the early return earlier. */
1166 if (new_type > SVt_PVMG && SvIsCOW(sv)) {
1167 sv_force_normal_flags(sv, 0);
1170 old_body = SvANY(sv);
1172 /* Copying structures onto other structures that have been neatly zeroed
1173 has a subtle gotcha. Consider XPVMG
1175 +------+------+------+------+------+-------+-------+
1176 | NV | CUR | LEN | IV | MAGIC | STASH |
1177 +------+------+------+------+------+-------+-------+
1178 0 4 8 12 16 20 24 28
1180 where NVs are aligned to 8 bytes, so that sizeof that structure is
1181 actually 32 bytes long, with 4 bytes of padding at the end:
1183 +------+------+------+------+------+-------+-------+------+
1184 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1185 +------+------+------+------+------+-------+-------+------+
1186 0 4 8 12 16 20 24 28 32
1188 so what happens if you allocate memory for this structure:
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1192 +------+------+------+------+------+-------+-------+------+------+...
1193 0 4 8 12 16 20 24 28 32 36
1195 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1196 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1197 started out as zero once, but it's quite possible that it isn't. So now,
1198 rather than a nicely zeroed GP, you have it pointing somewhere random.
1201 (In fact, GP ends up pointing at a previous GP structure, because the
1202 principle cause of the padding in XPVMG getting garbage is a copy of
1203 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1204 this happens to be moot because XPVGV has been re-ordered, with GP
1205 no longer after STASH)
1207 So we are careful and work out the size of used parts of all the
1215 referant = SvRV(sv);
1216 old_type_details = &fake_rv;
1217 if (new_type == SVt_NV)
1218 new_type = SVt_PVNV;
1220 if (new_type < SVt_PVIV) {
1221 new_type = (new_type == SVt_NV)
1222 ? SVt_PVNV : SVt_PVIV;
1227 if (new_type < SVt_PVNV) {
1228 new_type = SVt_PVNV;
1232 assert(new_type > SVt_PV);
1233 assert(SVt_IV < SVt_PV);
1234 assert(SVt_NV < SVt_PV);
1241 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1242 there's no way that it can be safely upgraded, because perl.c
1243 expects to Safefree(SvANY(PL_mess_sv)) */
1244 assert(sv != PL_mess_sv);
1245 /* This flag bit is used to mean other things in other scalar types.
1246 Given that it only has meaning inside the pad, it shouldn't be set
1247 on anything that can get upgraded. */
1248 assert(!SvPAD_TYPED(sv));
1251 if (old_type_details->cant_upgrade)
1252 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1253 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1256 if (old_type > new_type)
1257 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1258 (int)old_type, (int)new_type);
1260 new_type_details = bodies_by_type + new_type;
1262 SvFLAGS(sv) &= ~SVTYPEMASK;
1263 SvFLAGS(sv) |= new_type;
1265 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1266 the return statements above will have triggered. */
1267 assert (new_type != SVt_NULL);
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1275 assert(old_type == SVt_NULL);
1276 SvANY(sv) = new_XNV();
1281 assert(new_type_details->body_size);
1284 assert(new_type_details->arena);
1285 assert(new_type_details->arena_size);
1286 /* This points to the start of the allocated area. */
1287 new_body_inline(new_body, new_type);
1288 Zero(new_body, new_type_details->body_size, char);
1289 new_body = ((char *)new_body) - new_type_details->offset;
1291 /* We always allocated the full length item with PURIFY. To do this
1292 we fake things so that arena is false for all 16 types.. */
1293 new_body = new_NOARENAZ(new_type_details);
1295 SvANY(sv) = new_body;
1296 if (new_type == SVt_PVAV) {
1300 if (old_type_details->body_size) {
1303 /* It will have been zeroed when the new body was allocated.
1304 Lets not write to it, in case it confuses a write-back
1310 #ifndef NODEFAULT_SHAREKEYS
1311 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1313 HvMAX(sv) = 7; /* (start with 8 buckets) */
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));
1349 assert(new_type_details->body_size);
1350 /* We always allocated the full length item with PURIFY. To do this
1351 we fake things so that arena is false for all 16 types.. */
1352 if(new_type_details->arena) {
1353 /* This points to the start of the allocated area. */
1354 new_body_inline(new_body, new_type);
1355 Zero(new_body, new_type_details->body_size, char);
1356 new_body = ((char *)new_body) - new_type_details->offset;
1358 new_body = new_NOARENAZ(new_type_details);
1360 SvANY(sv) = new_body;
1362 if (old_type_details->copy) {
1363 /* There is now the potential for an upgrade from something without
1364 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1365 int offset = old_type_details->offset;
1366 int length = old_type_details->copy;
1368 if (new_type_details->offset > old_type_details->offset) {
1369 const int difference
1370 = new_type_details->offset - old_type_details->offset;
1371 offset += difference;
1372 length -= difference;
1374 assert (length >= 0);
1376 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1380 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1381 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1382 * correct 0.0 for us. Otherwise, if the old body didn't have an
1383 * NV slot, but the new one does, then we need to initialise the
1384 * freshly created NV slot with whatever the correct bit pattern is
1386 if (old_type_details->zero_nv && !new_type_details->zero_nv
1387 && !isGV_with_GP(sv))
1391 if (new_type == SVt_PVIO) {
1392 IO * const io = MUTABLE_IO(sv);
1393 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1396 /* Clear the stashcache because a new IO could overrule a package
1398 DEBUG_o(Perl_deb(aTHX_ "sv_upgrade clearing PL_stashcache\n"));
1399 hv_clear(PL_stashcache);
1401 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1402 IoPAGE_LEN(sv) = 60;
1404 if (new_type == SVt_REGEXP)
1405 sv->sv_u.svu_rx = (regexp *)new_body;
1406 else if (old_type < SVt_PV) {
1407 /* referant will be NULL unless the old type was SVt_IV emulating
1409 sv->sv_u.svu_rv = referant;
1413 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1414 (unsigned long)new_type);
1417 if (old_type > SVt_IV) {
1421 /* Note that there is an assumption that all bodies of types that
1422 can be upgraded came from arenas. Only the more complex non-
1423 upgradable types are allowed to be directly malloc()ed. */
1424 assert(old_type_details->arena);
1425 del_body((void*)((char*)old_body + old_type_details->offset),
1426 &PL_body_roots[old_type]);
1432 =for apidoc sv_backoff
1434 Remove any string offset. You should normally use the C<SvOOK_off> macro
1441 Perl_sv_backoff(pTHX_ register SV *const sv)
1444 const char * const s = SvPVX_const(sv);
1446 PERL_ARGS_ASSERT_SV_BACKOFF;
1447 PERL_UNUSED_CONTEXT;
1450 assert(SvTYPE(sv) != SVt_PVHV);
1451 assert(SvTYPE(sv) != SVt_PVAV);
1453 SvOOK_offset(sv, delta);
1455 SvLEN_set(sv, SvLEN(sv) + delta);
1456 SvPV_set(sv, SvPVX(sv) - delta);
1457 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1458 SvFLAGS(sv) &= ~SVf_OOK;
1465 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1466 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1467 Use the C<SvGROW> wrapper instead.
1473 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1477 PERL_ARGS_ASSERT_SV_GROW;
1479 if (PL_madskills && newlen >= 0x100000) {
1480 PerlIO_printf(Perl_debug_log,
1481 "Allocation too large: %"UVxf"\n", (UV)newlen);
1483 #ifdef HAS_64K_LIMIT
1484 if (newlen >= 0x10000) {
1485 PerlIO_printf(Perl_debug_log,
1486 "Allocation too large: %"UVxf"\n", (UV)newlen);
1489 #endif /* HAS_64K_LIMIT */
1492 if (SvTYPE(sv) < SVt_PV) {
1493 sv_upgrade(sv, SVt_PV);
1494 s = SvPVX_mutable(sv);
1496 else if (SvOOK(sv)) { /* pv is offset? */
1498 s = SvPVX_mutable(sv);
1499 if (newlen > SvLEN(sv))
1500 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1501 #ifdef HAS_64K_LIMIT
1502 if (newlen >= 0x10000)
1507 s = SvPVX_mutable(sv);
1509 if (newlen > SvLEN(sv)) { /* need more room? */
1510 STRLEN minlen = SvCUR(sv);
1511 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1512 if (newlen < minlen)
1514 #ifndef Perl_safesysmalloc_size
1515 newlen = PERL_STRLEN_ROUNDUP(newlen);
1517 if (SvLEN(sv) && s) {
1518 s = (char*)saferealloc(s, newlen);
1521 s = (char*)safemalloc(newlen);
1522 if (SvPVX_const(sv) && SvCUR(sv)) {
1523 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1527 #ifdef Perl_safesysmalloc_size
1528 /* Do this here, do it once, do it right, and then we will never get
1529 called back into sv_grow() unless there really is some growing
1531 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1533 SvLEN_set(sv, newlen);
1540 =for apidoc sv_setiv
1542 Copies an integer into the given SV, upgrading first if necessary.
1543 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1549 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1553 PERL_ARGS_ASSERT_SV_SETIV;
1555 SV_CHECK_THINKFIRST_COW_DROP(sv);
1556 switch (SvTYPE(sv)) {
1559 sv_upgrade(sv, SVt_IV);
1562 sv_upgrade(sv, SVt_PVIV);
1566 if (!isGV_with_GP(sv))
1573 /* diag_listed_as: Can't coerce %s to %s in %s */
1574 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1578 (void)SvIOK_only(sv); /* validate number */
1584 =for apidoc sv_setiv_mg
1586 Like C<sv_setiv>, but also handles 'set' magic.
1592 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1594 PERL_ARGS_ASSERT_SV_SETIV_MG;
1601 =for apidoc sv_setuv
1603 Copies an unsigned integer into the given SV, upgrading first if necessary.
1604 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1610 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1612 PERL_ARGS_ASSERT_SV_SETUV;
1614 /* With the if statement to ensure that integers are stored as IVs whenever
1616 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1619 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1621 If you wish to remove the following if statement, so that this routine
1622 (and its callers) always return UVs, please benchmark to see what the
1623 effect is. Modern CPUs may be different. Or may not :-)
1625 if (u <= (UV)IV_MAX) {
1626 sv_setiv(sv, (IV)u);
1635 =for apidoc sv_setuv_mg
1637 Like C<sv_setuv>, but also handles 'set' magic.
1643 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1645 PERL_ARGS_ASSERT_SV_SETUV_MG;
1652 =for apidoc sv_setnv
1654 Copies a double into the given SV, upgrading first if necessary.
1655 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1661 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1665 PERL_ARGS_ASSERT_SV_SETNV;
1667 SV_CHECK_THINKFIRST_COW_DROP(sv);
1668 switch (SvTYPE(sv)) {
1671 sv_upgrade(sv, SVt_NV);
1675 sv_upgrade(sv, SVt_PVNV);
1679 if (!isGV_with_GP(sv))
1686 /* diag_listed_as: Can't coerce %s to %s in %s */
1687 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1692 (void)SvNOK_only(sv); /* validate number */
1697 =for apidoc sv_setnv_mg
1699 Like C<sv_setnv>, but also handles 'set' magic.
1705 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1707 PERL_ARGS_ASSERT_SV_SETNV_MG;
1713 /* Print an "isn't numeric" warning, using a cleaned-up,
1714 * printable version of the offending string
1718 S_not_a_number(pTHX_ SV *const sv)
1725 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1728 dsv = newSVpvs_flags("", SVs_TEMP);
1729 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1732 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1733 /* each *s can expand to 4 chars + "...\0",
1734 i.e. need room for 8 chars */
1736 const char *s = SvPVX_const(sv);
1737 const char * const end = s + SvCUR(sv);
1738 for ( ; s < end && d < limit; s++ ) {
1740 if (ch & 128 && !isPRINT_LC(ch)) {
1749 else if (ch == '\r') {
1753 else if (ch == '\f') {
1757 else if (ch == '\\') {
1761 else if (ch == '\0') {
1765 else if (isPRINT_LC(ch))
1782 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1783 /* diag_listed_as: Argument "%s" isn't numeric%s */
1784 "Argument \"%s\" isn't numeric in %s", pv,
1787 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1788 /* diag_listed_as: Argument "%s" isn't numeric%s */
1789 "Argument \"%s\" isn't numeric", pv);
1793 =for apidoc looks_like_number
1795 Test if the content of an SV looks like a number (or is a number).
1796 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1797 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1804 Perl_looks_like_number(pTHX_ SV *const sv)
1809 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1811 if (SvPOK(sv) || SvPOKp(sv)) {
1812 sbegin = SvPV_nomg_const(sv, len);
1815 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1816 return grok_number(sbegin, len, NULL);
1820 S_glob_2number(pTHX_ GV * const gv)
1822 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1824 /* We know that all GVs stringify to something that is not-a-number,
1825 so no need to test that. */
1826 if (ckWARN(WARN_NUMERIC))
1828 SV *const buffer = sv_newmortal();
1829 gv_efullname3(buffer, gv, "*");
1830 not_a_number(buffer);
1832 /* We just want something true to return, so that S_sv_2iuv_common
1833 can tail call us and return true. */
1837 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1838 until proven guilty, assume that things are not that bad... */
1843 As 64 bit platforms often have an NV that doesn't preserve all bits of
1844 an IV (an assumption perl has been based on to date) it becomes necessary
1845 to remove the assumption that the NV always carries enough precision to
1846 recreate the IV whenever needed, and that the NV is the canonical form.
1847 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1848 precision as a side effect of conversion (which would lead to insanity
1849 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1850 1) to distinguish between IV/UV/NV slots that have cached a valid
1851 conversion where precision was lost and IV/UV/NV slots that have a
1852 valid conversion which has lost no precision
1853 2) to ensure that if a numeric conversion to one form is requested that
1854 would lose precision, the precise conversion (or differently
1855 imprecise conversion) is also performed and cached, to prevent
1856 requests for different numeric formats on the same SV causing
1857 lossy conversion chains. (lossless conversion chains are perfectly
1862 SvIOKp is true if the IV slot contains a valid value
1863 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1864 SvNOKp is true if the NV slot contains a valid value
1865 SvNOK is true only if the NV value is accurate
1868 while converting from PV to NV, check to see if converting that NV to an
1869 IV(or UV) would lose accuracy over a direct conversion from PV to
1870 IV(or UV). If it would, cache both conversions, return NV, but mark
1871 SV as IOK NOKp (ie not NOK).
1873 While converting from PV to IV, check to see if converting that IV to an
1874 NV would lose accuracy over a direct conversion from PV to NV. If it
1875 would, cache both conversions, flag similarly.
1877 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1878 correctly because if IV & NV were set NV *always* overruled.
1879 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1880 changes - now IV and NV together means that the two are interchangeable:
1881 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1883 The benefit of this is that operations such as pp_add know that if
1884 SvIOK is true for both left and right operands, then integer addition
1885 can be used instead of floating point (for cases where the result won't
1886 overflow). Before, floating point was always used, which could lead to
1887 loss of precision compared with integer addition.
1889 * making IV and NV equal status should make maths accurate on 64 bit
1891 * may speed up maths somewhat if pp_add and friends start to use
1892 integers when possible instead of fp. (Hopefully the overhead in
1893 looking for SvIOK and checking for overflow will not outweigh the
1894 fp to integer speedup)
1895 * will slow down integer operations (callers of SvIV) on "inaccurate"
1896 values, as the change from SvIOK to SvIOKp will cause a call into
1897 sv_2iv each time rather than a macro access direct to the IV slot
1898 * should speed up number->string conversion on integers as IV is
1899 favoured when IV and NV are equally accurate
1901 ####################################################################
1902 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1903 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1904 On the other hand, SvUOK is true iff UV.
1905 ####################################################################
1907 Your mileage will vary depending your CPU's relative fp to integer
1911 #ifndef NV_PRESERVES_UV
1912 # define IS_NUMBER_UNDERFLOW_IV 1
1913 # define IS_NUMBER_UNDERFLOW_UV 2
1914 # define IS_NUMBER_IV_AND_UV 2
1915 # define IS_NUMBER_OVERFLOW_IV 4
1916 # define IS_NUMBER_OVERFLOW_UV 5
1918 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1920 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1922 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1930 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1932 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));
1933 if (SvNVX(sv) < (NV)IV_MIN) {
1934 (void)SvIOKp_on(sv);
1936 SvIV_set(sv, IV_MIN);
1937 return IS_NUMBER_UNDERFLOW_IV;
1939 if (SvNVX(sv) > (NV)UV_MAX) {
1940 (void)SvIOKp_on(sv);
1943 SvUV_set(sv, UV_MAX);
1944 return IS_NUMBER_OVERFLOW_UV;
1946 (void)SvIOKp_on(sv);
1948 /* Can't use strtol etc to convert this string. (See truth table in
1950 if (SvNVX(sv) <= (UV)IV_MAX) {
1951 SvIV_set(sv, I_V(SvNVX(sv)));
1952 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1953 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1955 /* Integer is imprecise. NOK, IOKp */
1957 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1960 SvUV_set(sv, U_V(SvNVX(sv)));
1961 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1962 if (SvUVX(sv) == UV_MAX) {
1963 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1964 possibly be preserved by NV. Hence, it must be overflow.
1966 return IS_NUMBER_OVERFLOW_UV;
1968 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1970 /* Integer is imprecise. NOK, IOKp */
1972 return IS_NUMBER_OVERFLOW_IV;
1974 #endif /* !NV_PRESERVES_UV*/
1977 S_sv_2iuv_common(pTHX_ SV *const sv)
1981 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1984 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1985 * without also getting a cached IV/UV from it at the same time
1986 * (ie PV->NV conversion should detect loss of accuracy and cache
1987 * IV or UV at same time to avoid this. */
1988 /* IV-over-UV optimisation - choose to cache IV if possible */
1990 if (SvTYPE(sv) == SVt_NV)
1991 sv_upgrade(sv, SVt_PVNV);
1993 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1994 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1995 certainly cast into the IV range at IV_MAX, whereas the correct
1996 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1998 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1999 if (Perl_isnan(SvNVX(sv))) {
2005 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2006 SvIV_set(sv, I_V(SvNVX(sv)));
2007 if (SvNVX(sv) == (NV) SvIVX(sv)
2008 #ifndef NV_PRESERVES_UV
2009 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2010 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2011 /* Don't flag it as "accurately an integer" if the number
2012 came from a (by definition imprecise) NV operation, and
2013 we're outside the range of NV integer precision */
2017 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2019 /* scalar has trailing garbage, eg "42a" */
2021 DEBUG_c(PerlIO_printf(Perl_debug_log,
2022 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2028 /* IV not precise. No need to convert from PV, as NV
2029 conversion would already have cached IV if it detected
2030 that PV->IV would be better than PV->NV->IV
2031 flags already correct - don't set public IOK. */
2032 DEBUG_c(PerlIO_printf(Perl_debug_log,
2033 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2038 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2039 but the cast (NV)IV_MIN rounds to a the value less (more
2040 negative) than IV_MIN which happens to be equal to SvNVX ??
2041 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2042 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2043 (NV)UVX == NVX are both true, but the values differ. :-(
2044 Hopefully for 2s complement IV_MIN is something like
2045 0x8000000000000000 which will be exact. NWC */
2048 SvUV_set(sv, U_V(SvNVX(sv)));
2050 (SvNVX(sv) == (NV) SvUVX(sv))
2051 #ifndef NV_PRESERVES_UV
2052 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2053 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2054 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(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 */
2063 DEBUG_c(PerlIO_printf(Perl_debug_log,
2064 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2070 else if (SvPOKp(sv)) {
2072 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2073 /* We want to avoid a possible problem when we cache an IV/ a UV which
2074 may be later translated to an NV, and the resulting NV is not
2075 the same as the direct translation of the initial string
2076 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2077 be careful to ensure that the value with the .456 is around if the
2078 NV value is requested in the future).
2080 This means that if we cache such an IV/a UV, we need to cache the
2081 NV as well. Moreover, we trade speed for space, and do not
2082 cache the NV if we are sure it's not needed.
2085 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2086 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2087 == IS_NUMBER_IN_UV) {
2088 /* It's definitely an integer, only upgrade to PVIV */
2089 if (SvTYPE(sv) < SVt_PVIV)
2090 sv_upgrade(sv, SVt_PVIV);
2092 } else if (SvTYPE(sv) < SVt_PVNV)
2093 sv_upgrade(sv, SVt_PVNV);
2095 /* If NVs preserve UVs then we only use the UV value if we know that
2096 we aren't going to call atof() below. If NVs don't preserve UVs
2097 then the value returned may have more precision than atof() will
2098 return, even though value isn't perfectly accurate. */
2099 if ((numtype & (IS_NUMBER_IN_UV
2100 #ifdef NV_PRESERVES_UV
2103 )) == IS_NUMBER_IN_UV) {
2104 /* This won't turn off the public IOK flag if it was set above */
2105 (void)SvIOKp_on(sv);
2107 if (!(numtype & IS_NUMBER_NEG)) {
2109 if (value <= (UV)IV_MAX) {
2110 SvIV_set(sv, (IV)value);
2112 /* it didn't overflow, and it was positive. */
2113 SvUV_set(sv, value);
2117 /* 2s complement assumption */
2118 if (value <= (UV)IV_MIN) {
2119 SvIV_set(sv, -(IV)value);
2121 /* Too negative for an IV. This is a double upgrade, but
2122 I'm assuming it will be rare. */
2123 if (SvTYPE(sv) < SVt_PVNV)
2124 sv_upgrade(sv, SVt_PVNV);
2128 SvNV_set(sv, -(NV)value);
2129 SvIV_set(sv, IV_MIN);
2133 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2134 will be in the previous block to set the IV slot, and the next
2135 block to set the NV slot. So no else here. */
2137 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2138 != IS_NUMBER_IN_UV) {
2139 /* It wasn't an (integer that doesn't overflow the UV). */
2140 SvNV_set(sv, Atof(SvPVX_const(sv)));
2142 if (! numtype && ckWARN(WARN_NUMERIC))
2145 #if defined(USE_LONG_DOUBLE)
2146 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2147 PTR2UV(sv), SvNVX(sv)));
2149 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2150 PTR2UV(sv), SvNVX(sv)));
2153 #ifdef NV_PRESERVES_UV
2154 (void)SvIOKp_on(sv);
2156 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2157 SvIV_set(sv, I_V(SvNVX(sv)));
2158 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2161 NOOP; /* Integer is imprecise. NOK, IOKp */
2163 /* UV will not work better than IV */
2165 if (SvNVX(sv) > (NV)UV_MAX) {
2167 /* Integer is inaccurate. NOK, IOKp, is UV */
2168 SvUV_set(sv, UV_MAX);
2170 SvUV_set(sv, U_V(SvNVX(sv)));
2171 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2172 NV preservse UV so can do correct comparison. */
2173 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2176 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2181 #else /* NV_PRESERVES_UV */
2182 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2183 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2184 /* The IV/UV slot will have been set from value returned by
2185 grok_number above. The NV slot has just been set using
2188 assert (SvIOKp(sv));
2190 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2191 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2192 /* Small enough to preserve all bits. */
2193 (void)SvIOKp_on(sv);
2195 SvIV_set(sv, I_V(SvNVX(sv)));
2196 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2198 /* Assumption: first non-preserved integer is < IV_MAX,
2199 this NV is in the preserved range, therefore: */
2200 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2202 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);
2206 0 0 already failed to read UV.
2207 0 1 already failed to read UV.
2208 1 0 you won't get here in this case. IV/UV
2209 slot set, public IOK, Atof() unneeded.
2210 1 1 already read UV.
2211 so there's no point in sv_2iuv_non_preserve() attempting
2212 to use atol, strtol, strtoul etc. */
2214 sv_2iuv_non_preserve (sv, numtype);
2216 sv_2iuv_non_preserve (sv);
2220 #endif /* NV_PRESERVES_UV */
2221 /* It might be more code efficient to go through the entire logic above
2222 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2223 gets complex and potentially buggy, so more programmer efficient
2224 to do it this way, by turning off the public flags: */
2226 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2230 if (isGV_with_GP(sv))
2231 return glob_2number(MUTABLE_GV(sv));
2233 if (!SvPADTMP(sv)) {
2234 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2237 if (SvTYPE(sv) < SVt_IV)
2238 /* Typically the caller expects that sv_any is not NULL now. */
2239 sv_upgrade(sv, SVt_IV);
2240 /* Return 0 from the caller. */
2247 =for apidoc sv_2iv_flags
2249 Return the integer value of an SV, doing any necessary string
2250 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2251 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2257 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2264 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2270 if (flags & SV_SKIP_OVERLOAD)
2272 tmpstr = AMG_CALLunary(sv, numer_amg);
2273 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2274 return SvIV(tmpstr);
2277 return PTR2IV(SvRV(sv));
2280 if (SvVALID(sv) || isREGEXP(sv)) {
2281 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2282 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2283 In practice they are extremely unlikely to actually get anywhere
2284 accessible by user Perl code - the only way that I'm aware of is when
2285 a constant subroutine which is used as the second argument to index.
2287 Regexps have no SvIVX and SvNVX fields.
2289 assert(isREGEXP(sv) || SvPOKp(sv));
2292 const char * const ptr =
2293 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2295 = grok_number(ptr, SvCUR(sv), &value);
2297 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2298 == IS_NUMBER_IN_UV) {
2299 /* It's definitely an integer */
2300 if (numtype & IS_NUMBER_NEG) {
2301 if (value < (UV)IV_MIN)
2304 if (value < (UV)IV_MAX)
2309 if (ckWARN(WARN_NUMERIC))
2312 return I_V(Atof(ptr));
2316 if (SvTHINKFIRST(sv)) {
2317 #ifdef PERL_OLD_COPY_ON_WRITE
2319 sv_force_normal_flags(sv, 0);
2322 if (SvREADONLY(sv) && !SvOK(sv)) {
2323 if (ckWARN(WARN_UNINITIALIZED))
2330 if (S_sv_2iuv_common(aTHX_ sv))
2334 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2335 PTR2UV(sv),SvIVX(sv)));
2336 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2340 =for apidoc sv_2uv_flags
2342 Return the unsigned integer value of an SV, doing any necessary string
2343 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2344 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2350 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2357 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2363 if (flags & SV_SKIP_OVERLOAD)
2365 tmpstr = AMG_CALLunary(sv, numer_amg);
2366 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2367 return SvUV(tmpstr);
2370 return PTR2UV(SvRV(sv));
2373 if (SvVALID(sv) || isREGEXP(sv)) {
2374 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2375 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2376 Regexps have no SvIVX and SvNVX fields. */
2377 assert(isREGEXP(sv) || SvPOKp(sv));
2380 const char * const ptr =
2381 isREGEXP(sv) ? RX_WRAPPED((REGEXP*)sv) : SvPVX_const(sv);
2383 = grok_number(ptr, SvCUR(sv), &value);
2385 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2386 == IS_NUMBER_IN_UV) {
2387 /* It's definitely an integer */
2388 if (!(numtype & IS_NUMBER_NEG))
2392 if (ckWARN(WARN_NUMERIC))
2395 return U_V(Atof(ptr));
2399 if (SvTHINKFIRST(sv)) {
2400 #ifdef PERL_OLD_COPY_ON_WRITE
2402 sv_force_normal_flags(sv, 0);
2405 if (SvREADONLY(sv) && !SvOK(sv)) {
2406 if (ckWARN(WARN_UNINITIALIZED))
2413 if (S_sv_2iuv_common(aTHX_ sv))
2417 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2418 PTR2UV(sv),SvUVX(sv)));
2419 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2423 =for apidoc sv_2nv_flags
2425 Return the num value of an SV, doing any necessary string or integer
2426 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2427 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2433 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2438 if (SvGMAGICAL(sv) || SvVALID(sv) || isREGEXP(sv)) {
2439 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2440 the same flag bit as SVf_IVisUV, so must not let them cache NVs.
2441 Regexps have no SvIVX and SvNVX fields. */
2443 if (flags & SV_GMAGIC)
2447 if (SvPOKp(sv) && !SvIOKp(sv)) {
2448 ptr = SvPVX_const(sv);
2450 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2451 !grok_number(ptr, SvCUR(sv), NULL))
2457 return (NV)SvUVX(sv);
2459 return (NV)SvIVX(sv);
2465 ptr = RX_WRAPPED((REGEXP *)sv);
2468 assert(SvTYPE(sv) >= SVt_PVMG);
2469 /* This falls through to the report_uninit near the end of the
2471 } else if (SvTHINKFIRST(sv)) {
2476 if (flags & SV_SKIP_OVERLOAD)
2478 tmpstr = AMG_CALLunary(sv, numer_amg);
2479 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2480 return SvNV(tmpstr);
2483 return PTR2NV(SvRV(sv));
2485 #ifdef PERL_OLD_COPY_ON_WRITE
2487 sv_force_normal_flags(sv, 0);
2490 if (SvREADONLY(sv) && !SvOK(sv)) {
2491 if (ckWARN(WARN_UNINITIALIZED))
2496 if (SvTYPE(sv) < SVt_NV) {
2497 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2498 sv_upgrade(sv, SVt_NV);
2499 #ifdef USE_LONG_DOUBLE
2501 STORE_NUMERIC_LOCAL_SET_STANDARD();
2502 PerlIO_printf(Perl_debug_log,
2503 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2504 PTR2UV(sv), SvNVX(sv));
2505 RESTORE_NUMERIC_LOCAL();
2509 STORE_NUMERIC_LOCAL_SET_STANDARD();
2510 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2511 PTR2UV(sv), SvNVX(sv));
2512 RESTORE_NUMERIC_LOCAL();
2516 else if (SvTYPE(sv) < SVt_PVNV)
2517 sv_upgrade(sv, SVt_PVNV);
2522 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2523 #ifdef NV_PRESERVES_UV
2529 /* Only set the public NV OK flag if this NV preserves the IV */
2530 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2532 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2533 : (SvIVX(sv) == I_V(SvNVX(sv))))
2539 else if (SvPOKp(sv)) {
2541 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2542 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2544 #ifdef NV_PRESERVES_UV
2545 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2546 == IS_NUMBER_IN_UV) {
2547 /* It's definitely an integer */
2548 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2550 SvNV_set(sv, Atof(SvPVX_const(sv)));
2556 SvNV_set(sv, Atof(SvPVX_const(sv)));
2557 /* Only set the public NV OK flag if this NV preserves the value in
2558 the PV at least as well as an IV/UV would.
2559 Not sure how to do this 100% reliably. */
2560 /* if that shift count is out of range then Configure's test is
2561 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2563 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2564 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2565 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2566 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2567 /* Can't use strtol etc to convert this string, so don't try.
2568 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2571 /* value has been set. It may not be precise. */
2572 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2573 /* 2s complement assumption for (UV)IV_MIN */
2574 SvNOK_on(sv); /* Integer is too negative. */
2579 if (numtype & IS_NUMBER_NEG) {
2580 SvIV_set(sv, -(IV)value);
2581 } else if (value <= (UV)IV_MAX) {
2582 SvIV_set(sv, (IV)value);
2584 SvUV_set(sv, value);
2588 if (numtype & IS_NUMBER_NOT_INT) {
2589 /* I believe that even if the original PV had decimals,
2590 they are lost beyond the limit of the FP precision.
2591 However, neither is canonical, so both only get p
2592 flags. NWC, 2000/11/25 */
2593 /* Both already have p flags, so do nothing */
2595 const NV nv = SvNVX(sv);
2596 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2597 if (SvIVX(sv) == I_V(nv)) {
2600 /* It had no "." so it must be integer. */
2604 /* between IV_MAX and NV(UV_MAX).
2605 Could be slightly > UV_MAX */
2607 if (numtype & IS_NUMBER_NOT_INT) {
2608 /* UV and NV both imprecise. */
2610 const UV nv_as_uv = U_V(nv);
2612 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2621 /* It might be more code efficient to go through the entire logic above
2622 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2623 gets complex and potentially buggy, so more programmer efficient
2624 to do it this way, by turning off the public flags: */
2626 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2627 #endif /* NV_PRESERVES_UV */
2630 if (isGV_with_GP(sv)) {
2631 glob_2number(MUTABLE_GV(sv));
2635 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2637 assert (SvTYPE(sv) >= SVt_NV);
2638 /* Typically the caller expects that sv_any is not NULL now. */
2639 /* XXX Ilya implies that this is a bug in callers that assume this
2640 and ideally should be fixed. */
2643 #if defined(USE_LONG_DOUBLE)
2645 STORE_NUMERIC_LOCAL_SET_STANDARD();
2646 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2647 PTR2UV(sv), SvNVX(sv));
2648 RESTORE_NUMERIC_LOCAL();
2652 STORE_NUMERIC_LOCAL_SET_STANDARD();
2653 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2654 PTR2UV(sv), SvNVX(sv));
2655 RESTORE_NUMERIC_LOCAL();
2664 Return an SV with the numeric value of the source SV, doing any necessary
2665 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2666 access this function.
2672 Perl_sv_2num(pTHX_ register SV *const sv)
2674 PERL_ARGS_ASSERT_SV_2NUM;
2679 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2680 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2681 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2682 return sv_2num(tmpsv);
2684 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2687 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2688 * UV as a string towards the end of buf, and return pointers to start and
2691 * We assume that buf is at least TYPE_CHARS(UV) long.
2695 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2697 char *ptr = buf + TYPE_CHARS(UV);
2698 char * const ebuf = ptr;
2701 PERL_ARGS_ASSERT_UIV_2BUF;
2713 *--ptr = '0' + (char)(uv % 10);
2722 =for apidoc sv_2pv_flags
2724 Returns a pointer to the string value of an SV, and sets *lp to its length.
2725 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2726 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2727 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2733 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2743 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2748 if (flags & SV_SKIP_OVERLOAD)
2750 tmpstr = AMG_CALLunary(sv, string_amg);
2751 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2752 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2754 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2758 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2759 if (flags & SV_CONST_RETURN) {
2760 pv = (char *) SvPVX_const(tmpstr);
2762 pv = (flags & SV_MUTABLE_RETURN)
2763 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2766 *lp = SvCUR(tmpstr);
2768 pv = sv_2pv_flags(tmpstr, lp, flags);
2781 SV *const referent = SvRV(sv);
2785 retval = buffer = savepvn("NULLREF", len);
2786 } else if (SvTYPE(referent) == SVt_REGEXP &&
2787 (!(PL_curcop->cop_hints & HINT_NO_AMAGIC) ||
2788 amagic_is_enabled(string_amg))) {
2789 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2793 /* If the regex is UTF-8 we want the containing scalar to
2794 have an UTF-8 flag too */
2801 *lp = RX_WRAPLEN(re);
2803 return RX_WRAPPED(re);
2805 const char *const typestr = sv_reftype(referent, 0);
2806 const STRLEN typelen = strlen(typestr);
2807 UV addr = PTR2UV(referent);
2808 const char *stashname = NULL;
2809 STRLEN stashnamelen = 0; /* hush, gcc */
2810 const char *buffer_end;
2812 if (SvOBJECT(referent)) {
2813 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2816 stashname = HEK_KEY(name);
2817 stashnamelen = HEK_LEN(name);
2819 if (HEK_UTF8(name)) {
2825 stashname = "__ANON__";
2828 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2829 + 2 * sizeof(UV) + 2 /* )\0 */;
2831 len = typelen + 3 /* (0x */
2832 + 2 * sizeof(UV) + 2 /* )\0 */;
2835 Newx(buffer, len, char);
2836 buffer_end = retval = buffer + len;
2838 /* Working backwards */
2842 *--retval = PL_hexdigit[addr & 15];
2843 } while (addr >>= 4);
2849 memcpy(retval, typestr, typelen);
2853 retval -= stashnamelen;
2854 memcpy(retval, stashname, stashnamelen);
2856 /* retval may not necessarily have reached the start of the
2858 assert (retval >= buffer);
2860 len = buffer_end - retval - 1; /* -1 for that \0 */
2872 if (flags & SV_MUTABLE_RETURN)
2873 return SvPVX_mutable(sv);
2874 if (flags & SV_CONST_RETURN)
2875 return (char *)SvPVX_const(sv);
2880 /* I'm assuming that if both IV and NV are equally valid then
2881 converting the IV is going to be more efficient */
2882 const U32 isUIOK = SvIsUV(sv);
2883 char buf[TYPE_CHARS(UV)];
2887 if (SvTYPE(sv) < SVt_PVIV)
2888 sv_upgrade(sv, SVt_PVIV);
2889 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2891 /* inlined from sv_setpvn */
2892 s = SvGROW_mutable(sv, len + 1);
2893 Move(ptr, s, len, char);
2897 else if (SvNOK(sv)) {
2898 if (SvTYPE(sv) < SVt_PVNV)
2899 sv_upgrade(sv, SVt_PVNV);
2900 if (SvNVX(sv) == 0.0) {
2901 s = SvGROW_mutable(sv, 2);
2906 /* The +20 is pure guesswork. Configure test needed. --jhi */
2907 s = SvGROW_mutable(sv, NV_DIG + 20);
2908 /* some Xenix systems wipe out errno here */
2909 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2918 else if (isGV_with_GP(sv)) {
2919 GV *const gv = MUTABLE_GV(sv);
2920 SV *const buffer = sv_newmortal();
2922 gv_efullname3(buffer, gv, "*");
2924 assert(SvPOK(buffer));
2928 *lp = SvCUR(buffer);
2929 return SvPVX(buffer);
2931 else if (isREGEXP(sv)) {
2932 if (lp) *lp = RX_WRAPLEN((REGEXP *)sv);
2933 return RX_WRAPPED((REGEXP *)sv);
2938 if (flags & SV_UNDEF_RETURNS_NULL)
2940 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2942 /* Typically the caller expects that sv_any is not NULL now. */
2943 if (!SvREADONLY(sv) && SvTYPE(sv) < SVt_PV)
2944 sv_upgrade(sv, SVt_PV);
2949 const STRLEN len = s - SvPVX_const(sv);
2955 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2956 PTR2UV(sv),SvPVX_const(sv)));
2957 if (flags & SV_CONST_RETURN)
2958 return (char *)SvPVX_const(sv);
2959 if (flags & SV_MUTABLE_RETURN)
2960 return SvPVX_mutable(sv);
2965 =for apidoc sv_copypv
2967 Copies a stringified representation of the source SV into the
2968 destination SV. Automatically performs any necessary mg_get and
2969 coercion of numeric values into strings. Guaranteed to preserve
2970 UTF8 flag even from overloaded objects. Similar in nature to
2971 sv_2pv[_flags] but operates directly on an SV instead of just the
2972 string. Mostly uses sv_2pv_flags to do its work, except when that
2973 would lose the UTF-8'ness of the PV.
2975 =for apidoc sv_copypv_nomg
2977 Like sv_copypv, but doesn't invoke get magic first.
2979 =for apidoc sv_copypv_flags
2981 Implementation of sv_copypv and sv_copypv_nomg. Calls get magic iff flags
2988 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
2990 PERL_ARGS_ASSERT_SV_COPYPV;
2992 sv_copypv_flags(dsv, ssv, 0);
2996 Perl_sv_copypv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
3001 PERL_ARGS_ASSERT_SV_COPYPV_FLAGS;
3003 if ((flags & SV_GMAGIC) && SvGMAGICAL(ssv))
3005 s = SvPV_nomg_const(ssv,len);
3006 sv_setpvn(dsv,s,len);
3014 =for apidoc sv_2pvbyte
3016 Return a pointer to the byte-encoded representation of the SV, and set *lp
3017 to its length. May cause the SV to be downgraded from UTF-8 as a
3020 Usually accessed via the C<SvPVbyte> macro.
3026 Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *const lp)
3028 PERL_ARGS_ASSERT_SV_2PVBYTE;
3030 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3031 || isGV_with_GP(sv) || SvROK(sv)) {
3032 SV *sv2 = sv_newmortal();
3036 else SvGETMAGIC(sv);
3037 sv_utf8_downgrade(sv,0);
3038 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3042 =for apidoc sv_2pvutf8
3044 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3045 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3047 Usually accessed via the C<SvPVutf8> macro.
3053 Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *const lp)
3055 PERL_ARGS_ASSERT_SV_2PVUTF8;
3057 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3058 || isGV_with_GP(sv) || SvROK(sv))
3059 sv = sv_mortalcopy(sv);
3062 sv_utf8_upgrade_nomg(sv);
3063 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3068 =for apidoc sv_2bool
3070 This macro is only used by sv_true() or its macro equivalent, and only if
3071 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3072 It calls sv_2bool_flags with the SV_GMAGIC flag.
3074 =for apidoc sv_2bool_flags
3076 This function is only used by sv_true() and friends, and only if
3077 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3078 contain SV_GMAGIC, then it does an mg_get() first.
3085 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3089 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3091 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3097 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3098 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3099 return cBOOL(SvTRUE(tmpsv));
3101 return SvRV(sv) != 0;
3103 return SvTRUE_common(sv, isGV_with_GP(sv) ? 1 : 0);
3107 =for apidoc sv_utf8_upgrade
3109 Converts the PV of an SV to its UTF-8-encoded form.
3110 Forces the SV to string form if it is not already.
3111 Will C<mg_get> on C<sv> if appropriate.
3112 Always sets the SvUTF8 flag to avoid future validity checks even
3113 if the whole string is the same in UTF-8 as not.
3114 Returns the number of bytes in the converted string
3116 This is not a general purpose byte encoding to Unicode interface:
3117 use the Encode extension for that.
3119 =for apidoc sv_utf8_upgrade_nomg
3121 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3123 =for apidoc sv_utf8_upgrade_flags
3125 Converts the PV of an SV to its UTF-8-encoded form.
3126 Forces the SV to string form if it is not already.
3127 Always sets the SvUTF8 flag to avoid future validity checks even
3128 if all the bytes are invariant in UTF-8.
3129 If C<flags> has C<SV_GMAGIC> bit set,
3130 will C<mg_get> on C<sv> if appropriate, else not.
3131 Returns the number of bytes in the converted string
3132 C<sv_utf8_upgrade> and
3133 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3135 This is not a general purpose byte encoding to Unicode interface:
3136 use the Encode extension for that.
3140 The grow version is currently not externally documented. It adds a parameter,
3141 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3142 have free after it upon return. This allows the caller to reserve extra space
3143 that it intends to fill, to avoid extra grows.
3145 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3146 which can be used to tell this function to not first check to see if there are
3147 any characters that are different in UTF-8 (variant characters) which would
3148 force it to allocate a new string to sv, but to assume there are. Typically
3149 this flag is used by a routine that has already parsed the string to find that
3150 there are such characters, and passes this information on so that the work
3151 doesn't have to be repeated.
3153 (One might think that the calling routine could pass in the position of the
3154 first such variant, so it wouldn't have to be found again. But that is not the
3155 case, because typically when the caller is likely to use this flag, it won't be
3156 calling this routine unless it finds something that won't fit into a byte.
3157 Otherwise it tries to not upgrade and just use bytes. But some things that
3158 do fit into a byte are variants in utf8, and the caller may not have been
3159 keeping track of these.)
3161 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3162 isn't guaranteed due to having other routines do the work in some input cases,
3163 or if the input is already flagged as being in utf8.
3165 The speed of this could perhaps be improved for many cases if someone wanted to
3166 write a fast function that counts the number of variant characters in a string,
3167 especially if it could return the position of the first one.
3172 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3176 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3178 if (sv == &PL_sv_undef)
3180 if (!SvPOK_nog(sv)) {
3182 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3183 (void) sv_2pv_flags(sv,&len, flags);
3185 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3189 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3194 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3199 sv_force_normal_flags(sv, 0);
3202 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3203 sv_recode_to_utf8(sv, PL_encoding);
3204 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3208 if (SvCUR(sv) == 0) {
3209 if (extra) SvGROW(sv, extra);
3210 } else { /* Assume Latin-1/EBCDIC */
3211 /* This function could be much more efficient if we
3212 * had a FLAG in SVs to signal if there are any variant
3213 * chars in the PV. Given that there isn't such a flag
3214 * make the loop as fast as possible (although there are certainly ways
3215 * to speed this up, eg. through vectorization) */
3216 U8 * s = (U8 *) SvPVX_const(sv);
3217 U8 * e = (U8 *) SvEND(sv);
3219 STRLEN two_byte_count = 0;
3221 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3223 /* See if really will need to convert to utf8. We mustn't rely on our
3224 * incoming SV being well formed and having a trailing '\0', as certain
3225 * code in pp_formline can send us partially built SVs. */
3229 if (NATIVE_IS_INVARIANT(ch)) continue;
3231 t--; /* t already incremented; re-point to first variant */
3236 /* utf8 conversion not needed because all are invariants. Mark as
3237 * UTF-8 even if no variant - saves scanning loop */
3239 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3244 /* Here, the string should be converted to utf8, either because of an
3245 * input flag (two_byte_count = 0), or because a character that
3246 * requires 2 bytes was found (two_byte_count = 1). t points either to
3247 * the beginning of the string (if we didn't examine anything), or to
3248 * the first variant. In either case, everything from s to t - 1 will
3249 * occupy only 1 byte each on output.
3251 * There are two main ways to convert. One is to create a new string
3252 * and go through the input starting from the beginning, appending each
3253 * converted value onto the new string as we go along. It's probably
3254 * best to allocate enough space in the string for the worst possible
3255 * case rather than possibly running out of space and having to
3256 * reallocate and then copy what we've done so far. Since everything
3257 * from s to t - 1 is invariant, the destination can be initialized
3258 * with these using a fast memory copy
3260 * The other way is to figure out exactly how big the string should be
3261 * by parsing the entire input. Then you don't have to make it big
3262 * enough to handle the worst possible case, and more importantly, if
3263 * the string you already have is large enough, you don't have to
3264 * allocate a new string, you can copy the last character in the input
3265 * string to the final position(s) that will be occupied by the
3266 * converted string and go backwards, stopping at t, since everything
3267 * before that is invariant.
3269 * There are advantages and disadvantages to each method.
3271 * In the first method, we can allocate a new string, do the memory
3272 * copy from the s to t - 1, and then proceed through the rest of the
3273 * string byte-by-byte.
3275 * In the second method, we proceed through the rest of the input
3276 * string just calculating how big the converted string will be. Then
3277 * there are two cases:
3278 * 1) if the string has enough extra space to handle the converted
3279 * value. We go backwards through the string, converting until we
3280 * get to the position we are at now, and then stop. If this
3281 * position is far enough along in the string, this method is
3282 * faster than the other method. If the memory copy were the same
3283 * speed as the byte-by-byte loop, that position would be about
3284 * half-way, as at the half-way mark, parsing to the end and back
3285 * is one complete string's parse, the same amount as starting
3286 * over and going all the way through. Actually, it would be
3287 * somewhat less than half-way, as it's faster to just count bytes
3288 * than to also copy, and we don't have the overhead of allocating
3289 * a new string, changing the scalar to use it, and freeing the
3290 * existing one. But if the memory copy is fast, the break-even
3291 * point is somewhere after half way. The counting loop could be
3292 * sped up by vectorization, etc, to move the break-even point
3293 * further towards the beginning.
3294 * 2) if the string doesn't have enough space to handle the converted
3295 * value. A new string will have to be allocated, and one might
3296 * as well, given that, start from the beginning doing the first
3297 * method. We've spent extra time parsing the string and in
3298 * exchange all we've gotten is that we know precisely how big to
3299 * make the new one. Perl is more optimized for time than space,
3300 * so this case is a loser.
3301 * So what I've decided to do is not use the 2nd method unless it is
3302 * guaranteed that a new string won't have to be allocated, assuming
3303 * the worst case. I also decided not to put any more conditions on it
3304 * than this, for now. It seems likely that, since the worst case is
3305 * twice as big as the unknown portion of the string (plus 1), we won't
3306 * be guaranteed enough space, causing us to go to the first method,
3307 * unless the string is short, or the first variant character is near
3308 * the end of it. In either of these cases, it seems best to use the
3309 * 2nd method. The only circumstance I can think of where this would
3310 * be really slower is if the string had once had much more data in it
3311 * than it does now, but there is still a substantial amount in it */
3314 STRLEN invariant_head = t - s;
3315 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3316 if (SvLEN(sv) < size) {
3318 /* Here, have decided to allocate a new string */
3323 Newx(dst, size, U8);
3325 /* If no known invariants at the beginning of the input string,
3326 * set so starts from there. Otherwise, can use memory copy to
3327 * get up to where we are now, and then start from here */
3329 if (invariant_head <= 0) {
3332 Copy(s, dst, invariant_head, char);
3333 d = dst + invariant_head;
3337 const UV uv = NATIVE8_TO_UNI(*t++);
3338 if (UNI_IS_INVARIANT(uv))
3339 *d++ = (U8)UNI_TO_NATIVE(uv);
3341 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3342 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3346 SvPV_free(sv); /* No longer using pre-existing string */
3347 SvPV_set(sv, (char*)dst);
3348 SvCUR_set(sv, d - dst);
3349 SvLEN_set(sv, size);
3352 /* Here, have decided to get the exact size of the string.
3353 * Currently this happens only when we know that there is
3354 * guaranteed enough space to fit the converted string, so
3355 * don't have to worry about growing. If two_byte_count is 0,
3356 * then t points to the first byte of the string which hasn't
3357 * been examined yet. Otherwise two_byte_count is 1, and t
3358 * points to the first byte in the string that will expand to
3359 * two. Depending on this, start examining at t or 1 after t.
3362 U8 *d = t + two_byte_count;
3365 /* Count up the remaining bytes that expand to two */
3368 const U8 chr = *d++;
3369 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3372 /* The string will expand by just the number of bytes that
3373 * occupy two positions. But we are one afterwards because of
3374 * the increment just above. This is the place to put the
3375 * trailing NUL, and to set the length before we decrement */
3377 d += two_byte_count;
3378 SvCUR_set(sv, d - s);
3382 /* Having decremented d, it points to the position to put the
3383 * very last byte of the expanded string. Go backwards through
3384 * the string, copying and expanding as we go, stopping when we
3385 * get to the part that is invariant the rest of the way down */
3389 const U8 ch = NATIVE8_TO_UNI(*e--);
3390 if (UNI_IS_INVARIANT(ch)) {
3391 *d-- = UNI_TO_NATIVE(ch);
3393 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3394 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3399 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3400 /* Update pos. We do it at the end rather than during
3401 * the upgrade, to avoid slowing down the common case
3402 * (upgrade without pos) */
3403 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3405 I32 pos = mg->mg_len;
3406 if (pos > 0 && (U32)pos > invariant_head) {
3407 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3408 STRLEN n = (U32)pos - invariant_head;
3410 if (UTF8_IS_START(*d))
3415 mg->mg_len = d - (U8*)SvPVX(sv);
3418 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3419 magic_setutf8(sv,mg); /* clear UTF8 cache */
3424 /* Mark as UTF-8 even if no variant - saves scanning loop */
3430 =for apidoc sv_utf8_downgrade
3432 Attempts to convert the PV of an SV from characters to bytes.
3433 If the PV contains a character that cannot fit
3434 in a byte, this conversion will fail;
3435 in this case, either returns false or, if C<fail_ok> is not
3438 This is not a general purpose Unicode to byte encoding interface:
3439 use the Encode extension for that.
3445 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3449 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3451 if (SvPOKp(sv) && SvUTF8(sv)) {
3455 int mg_flags = SV_GMAGIC;
3458 sv_force_normal_flags(sv, 0);
3460 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3462 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3464 I32 pos = mg->mg_len;
3466 sv_pos_b2u(sv, &pos);
3467 mg_flags = 0; /* sv_pos_b2u does get magic */
3471 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3472 magic_setutf8(sv,mg); /* clear UTF8 cache */
3475 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3477 if (!utf8_to_bytes(s, &len)) {
3482 Perl_croak(aTHX_ "Wide character in %s",
3485 Perl_croak(aTHX_ "Wide character");
3496 =for apidoc sv_utf8_encode
3498 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3499 flag off so that it looks like octets again.
3505 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3507 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3509 if (SvREADONLY(sv)) {
3510 sv_force_normal_flags(sv, 0);
3512 (void) sv_utf8_upgrade(sv);
3517 =for apidoc sv_utf8_decode
3519 If the PV of the SV is an octet sequence in UTF-8
3520 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3521 so that it looks like a character. If the PV contains only single-byte
3522 characters, the C<SvUTF8> flag stays off.
3523 Scans PV for validity and returns false if the PV is invalid UTF-8.
3529 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3531 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3534 const U8 *start, *c;
3537 /* The octets may have got themselves encoded - get them back as
3540 if (!sv_utf8_downgrade(sv, TRUE))
3543 /* it is actually just a matter of turning the utf8 flag on, but
3544 * we want to make sure everything inside is valid utf8 first.
3546 c = start = (const U8 *) SvPVX_const(sv);
3547 if (!is_utf8_string(c, SvCUR(sv)))
3549 e = (const U8 *) SvEND(sv);
3552 if (!UTF8_IS_INVARIANT(ch)) {
3557 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3558 /* adjust pos to the start of a UTF8 char sequence */
3559 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3561 I32 pos = mg->mg_len;
3563 for (c = start + pos; c > start; c--) {
3564 if (UTF8_IS_START(*c))
3567 mg->mg_len = c - start;
3570 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3571 magic_setutf8(sv,mg); /* clear UTF8 cache */
3578 =for apidoc sv_setsv
3580 Copies the contents of the source SV C<ssv> into the destination SV
3581 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3582 function if the source SV needs to be reused. Does not handle 'set' magic.
3583 Loosely speaking, it performs a copy-by-value, obliterating any previous
3584 content of the destination.
3586 You probably want to use one of the assortment of wrappers, such as
3587 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3588 C<SvSetMagicSV_nosteal>.
3590 =for apidoc sv_setsv_flags
3592 Copies the contents of the source SV C<ssv> into the destination SV
3593 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3594 function if the source SV needs to be reused. Does not handle 'set' magic.
3595 Loosely speaking, it performs a copy-by-value, obliterating any previous
3596 content of the destination.
3597 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3598 C<ssv> if appropriate, else not. If the C<flags>
3599 parameter has the C<NOSTEAL> bit set then the
3600 buffers of temps will not be stolen. <sv_setsv>
3601 and C<sv_setsv_nomg> are implemented in terms of this function.
3603 You probably want to use one of the assortment of wrappers, such as
3604 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3605 C<SvSetMagicSV_nosteal>.
3607 This is the primary function for copying scalars, and most other
3608 copy-ish functions and macros use this underneath.
3614 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3616 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3617 HV *old_stash = NULL;
3619 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3621 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3622 const char * const name = GvNAME(sstr);
3623 const STRLEN len = GvNAMELEN(sstr);
3625 if (dtype >= SVt_PV) {
3631 SvUPGRADE(dstr, SVt_PVGV);
3632 (void)SvOK_off(dstr);
3633 /* We have to turn this on here, even though we turn it off
3634 below, as GvSTASH will fail an assertion otherwise. */
3635 isGV_with_GP_on(dstr);
3637 GvSTASH(dstr) = GvSTASH(sstr);
3639 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3640 gv_name_set(MUTABLE_GV(dstr), name, len,
3641 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3642 SvFAKE_on(dstr); /* can coerce to non-glob */
3645 if(GvGP(MUTABLE_GV(sstr))) {
3646 /* If source has method cache entry, clear it */
3648 SvREFCNT_dec(GvCV(sstr));
3649 GvCV_set(sstr, NULL);
3652 /* If source has a real method, then a method is
3655 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3661 /* If dest already had a real method, that's a change as well */
3663 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3664 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3669 /* We don't need to check the name of the destination if it was not a
3670 glob to begin with. */
3671 if(dtype == SVt_PVGV) {
3672 const char * const name = GvNAME((const GV *)dstr);
3675 /* The stash may have been detached from the symbol table, so
3677 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3681 const STRLEN len = GvNAMELEN(dstr);
3682 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3683 || (len == 1 && name[0] == ':')) {
3686 /* Set aside the old stash, so we can reset isa caches on
3688 if((old_stash = GvHV(dstr)))
3689 /* Make sure we do not lose it early. */
3690 SvREFCNT_inc_simple_void_NN(
3691 sv_2mortal((SV *)old_stash)
3697 gp_free(MUTABLE_GV(dstr));
3698 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3699 (void)SvOK_off(dstr);
3700 isGV_with_GP_on(dstr);
3701 GvINTRO_off(dstr); /* one-shot flag */
3702 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3703 if (SvTAINTED(sstr))
3705 if (GvIMPORTED(dstr) != GVf_IMPORTED
3706 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3708 GvIMPORTED_on(dstr);
3711 if(mro_changes == 2) {
3712 if (GvAV((const GV *)sstr)) {
3714 SV * const sref = (SV *)GvAV((const GV *)dstr);
3715 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3716 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3717 AV * const ary = newAV();
3718 av_push(ary, mg->mg_obj); /* takes the refcount */
3719 mg->mg_obj = (SV *)ary;
3721 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3723 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3725 mro_isa_changed_in(GvSTASH(dstr));
3727 else if(mro_changes == 3) {
3728 HV * const stash = GvHV(dstr);
3729 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3735 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3740 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3742 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3744 const int intro = GvINTRO(dstr);
3747 const U32 stype = SvTYPE(sref);
3749 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3752 GvINTRO_off(dstr); /* one-shot flag */
3753 GvLINE(dstr) = CopLINE(PL_curcop);
3754 GvEGV(dstr) = MUTABLE_GV(dstr);
3759 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3760 import_flag = GVf_IMPORTED_CV;
3763 location = (SV **) &GvHV(dstr);
3764 import_flag = GVf_IMPORTED_HV;
3767 location = (SV **) &GvAV(dstr);
3768 import_flag = GVf_IMPORTED_AV;
3771 location = (SV **) &GvIOp(dstr);
3774 location = (SV **) &GvFORM(dstr);
3777 location = &GvSV(dstr);
3778 import_flag = GVf_IMPORTED_SV;
3781 if (stype == SVt_PVCV) {
3782 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3783 if (GvCVGEN(dstr)) {
3784 SvREFCNT_dec(GvCV(dstr));
3785 GvCV_set(dstr, NULL);
3786 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3789 SAVEGENERICSV(*location);
3793 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3794 CV* const cv = MUTABLE_CV(*location);
3796 if (!GvCVGEN((const GV *)dstr) &&
3797 (CvROOT(cv) || CvXSUB(cv)) &&
3798 /* redundant check that avoids creating the extra SV
3799 most of the time: */
3800 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3802 SV * const new_const_sv =
3803 CvCONST((const CV *)sref)
3804 ? cv_const_sv((const CV *)sref)
3806 report_redefined_cv(
3807 sv_2mortal(Perl_newSVpvf(aTHX_
3810 HvNAME_HEK(GvSTASH((const GV *)dstr))
3812 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3815 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3819 cv_ckproto_len_flags(cv, (const GV *)dstr,
3820 SvPOK(sref) ? CvPROTO(sref) : NULL,
3821 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3822 SvPOK(sref) ? SvUTF8(sref) : 0);
3824 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3825 GvASSUMECV_on(dstr);
3826 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3829 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3830 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3831 GvFLAGS(dstr) |= import_flag;
3833 if (stype == SVt_PVHV) {
3834 const char * const name = GvNAME((GV*)dstr);
3835 const STRLEN len = GvNAMELEN(dstr);
3838 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3839 || (len == 1 && name[0] == ':')
3841 && (!dref || HvENAME_get(dref))
3844 (HV *)sref, (HV *)dref,
3850 stype == SVt_PVAV && sref != dref
3851 && strEQ(GvNAME((GV*)dstr), "ISA")
3852 /* The stash may have been detached from the symbol table, so
3853 check its name before doing anything. */
3854 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3857 MAGIC * const omg = dref && SvSMAGICAL(dref)
3858 ? mg_find(dref, PERL_MAGIC_isa)
3860 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3861 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3862 AV * const ary = newAV();
3863 av_push(ary, mg->mg_obj); /* takes the refcount */
3864 mg->mg_obj = (SV *)ary;
3867 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3868 SV **svp = AvARRAY((AV *)omg->mg_obj);
3869 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3873 SvREFCNT_inc_simple_NN(*svp++)
3879 SvREFCNT_inc_simple_NN(omg->mg_obj)
3883 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3888 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3890 mg = mg_find(sref, PERL_MAGIC_isa);
3892 /* Since the *ISA assignment could have affected more than
3893 one stash, don't call mro_isa_changed_in directly, but let
3894 magic_clearisa do it for us, as it already has the logic for
3895 dealing with globs vs arrays of globs. */
3897 Perl_magic_clearisa(aTHX_ NULL, mg);
3899 else if (stype == SVt_PVIO) {
3900 DEBUG_o(Perl_deb(aTHX_ "glob_assign_ref clearing PL_stashcache\n"));
3901 /* It's a cache. It will rebuild itself quite happily.
3902 It's a lot of effort to work out exactly which key (or keys)
3903 might be invalidated by the creation of the this file handle.
3905 hv_clear(PL_stashcache);
3910 if (SvTAINTED(sstr))
3916 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3923 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3928 if (SvIS_FREED(dstr)) {
3929 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3930 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3932 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3934 sstr = &PL_sv_undef;
3935 if (SvIS_FREED(sstr)) {
3936 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3937 (void*)sstr, (void*)dstr);
3939 stype = SvTYPE(sstr);
3940 dtype = SvTYPE(dstr);
3942 /* There's a lot of redundancy below but we're going for speed here */
3947 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3948 (void)SvOK_off(dstr);
3956 sv_upgrade(dstr, SVt_IV);
3960 sv_upgrade(dstr, SVt_PVIV);
3964 goto end_of_first_switch;
3966 (void)SvIOK_only(dstr);
3967 SvIV_set(dstr, SvIVX(sstr));
3970 /* SvTAINTED can only be true if the SV has taint magic, which in
3971 turn means that the SV type is PVMG (or greater). This is the
3972 case statement for SVt_IV, so this cannot be true (whatever gcov
3974 assert(!SvTAINTED(sstr));
3979 if (dtype < SVt_PV && dtype != SVt_IV)
3980 sv_upgrade(dstr, SVt_IV);
3988 sv_upgrade(dstr, SVt_NV);
3992 sv_upgrade(dstr, SVt_PVNV);
3996 goto end_of_first_switch;
3998 SvNV_set(dstr, SvNVX(sstr));
3999 (void)SvNOK_only(dstr);
4000 /* SvTAINTED can only be true if the SV has taint magic, which in
4001 turn means that the SV type is PVMG (or greater). This is the
4002 case statement for SVt_NV, so this cannot be true (whatever gcov
4004 assert(!SvTAINTED(sstr));
4011 sv_upgrade(dstr, SVt_PV);
4014 if (dtype < SVt_PVIV)
4015 sv_upgrade(dstr, SVt_PVIV);
4018 if (dtype < SVt_PVNV)
4019 sv_upgrade(dstr, SVt_PVNV);
4023 const char * const type = sv_reftype(sstr,0);
4025 /* diag_listed_as: Bizarre copy of %s */
4026 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4028 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4034 if (dtype < SVt_REGEXP)
4036 if (dtype >= SVt_PV) {
4042 sv_upgrade(dstr, SVt_REGEXP);
4046 /* case SVt_BIND: */
4050 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4052 if (SvTYPE(sstr) != stype)
4053 stype = SvTYPE(sstr);
4055 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4056 glob_assign_glob(dstr, sstr, dtype);
4059 if (stype == SVt_PVLV)
4061 if (isREGEXP(sstr)) goto upgregexp;
4062 SvUPGRADE(dstr, SVt_PVNV);
4065 SvUPGRADE(dstr, (svtype)stype);
4067 end_of_first_switch:
4069 /* dstr may have been upgraded. */
4070 dtype = SvTYPE(dstr);
4071 sflags = SvFLAGS(sstr);
4073 if (dtype == SVt_PVCV) {
4074 /* Assigning to a subroutine sets the prototype. */
4077 const char *const ptr = SvPV_const(sstr, len);
4079 SvGROW(dstr, len + 1);
4080 Copy(ptr, SvPVX(dstr), len + 1, char);
4081 SvCUR_set(dstr, len);
4083 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4084 CvAUTOLOAD_off(dstr);
4089 else if (dtype == SVt_PVAV || dtype == SVt_PVHV || dtype == SVt_PVFM) {
4090 const char * const type = sv_reftype(dstr,0);
4092 /* diag_listed_as: Cannot copy to %s */
4093 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4095 Perl_croak(aTHX_ "Cannot copy to %s", type);
4096 } else if (sflags & SVf_ROK) {
4097 if (isGV_with_GP(dstr)
4098 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4101 if (GvIMPORTED(dstr) != GVf_IMPORTED
4102 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4104 GvIMPORTED_on(dstr);
4109 glob_assign_glob(dstr, sstr, dtype);
4113 if (dtype >= SVt_PV) {
4114 if (isGV_with_GP(dstr)) {
4115 glob_assign_ref(dstr, sstr);
4118 if (SvPVX_const(dstr)) {
4124 (void)SvOK_off(dstr);
4125 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4126 SvFLAGS(dstr) |= sflags & SVf_ROK;
4127 assert(!(sflags & SVp_NOK));
4128 assert(!(sflags & SVp_IOK));
4129 assert(!(sflags & SVf_NOK));
4130 assert(!(sflags & SVf_IOK));
4132 else if (isGV_with_GP(dstr)) {
4133 if (!(sflags & SVf_OK)) {
4134 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4135 "Undefined value assigned to typeglob");
4138 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4139 if (dstr != (const SV *)gv) {
4140 const char * const name = GvNAME((const GV *)dstr);
4141 const STRLEN len = GvNAMELEN(dstr);
4142 HV *old_stash = NULL;
4143 bool reset_isa = FALSE;
4144 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4145 || (len == 1 && name[0] == ':')) {
4146 /* Set aside the old stash, so we can reset isa caches
4147 on its subclasses. */
4148 if((old_stash = GvHV(dstr))) {
4149 /* Make sure we do not lose it early. */
4150 SvREFCNT_inc_simple_void_NN(
4151 sv_2mortal((SV *)old_stash)
4158 gp_free(MUTABLE_GV(dstr));
4159 GvGP_set(dstr, gp_ref(GvGP(gv)));
4162 HV * const stash = GvHV(dstr);
4164 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4174 else if ((dtype == SVt_REGEXP || dtype == SVt_PVLV)
4175 && (stype == SVt_REGEXP || isREGEXP(sstr))) {
4176 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4178 else if (sflags & SVp_POK) {
4182 * Check to see if we can just swipe the string. If so, it's a
4183 * possible small lose on short strings, but a big win on long ones.
4184 * It might even be a win on short strings if SvPVX_const(dstr)
4185 * has to be allocated and SvPVX_const(sstr) has to be freed.
4186 * Likewise if we can set up COW rather than doing an actual copy, we
4187 * drop to the else clause, as the swipe code and the COW setup code
4188 * have much in common.
4191 /* Whichever path we take through the next code, we want this true,
4192 and doing it now facilitates the COW check. */
4193 (void)SvPOK_only(dstr);
4196 /* If we're already COW then this clause is not true, and if COW
4197 is allowed then we drop down to the else and make dest COW
4198 with us. If caller hasn't said that we're allowed to COW
4199 shared hash keys then we don't do the COW setup, even if the
4200 source scalar is a shared hash key scalar. */
4201 (((flags & SV_COW_SHARED_HASH_KEYS)
4202 ? !(sflags & SVf_IsCOW)
4203 : 1 /* If making a COW copy is forbidden then the behaviour we
4204 desire is as if the source SV isn't actually already
4205 COW, even if it is. So we act as if the source flags
4206 are not COW, rather than actually testing them. */
4208 #ifndef PERL_OLD_COPY_ON_WRITE
4209 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4210 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4211 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4212 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4213 but in turn, it's somewhat dead code, never expected to go
4214 live, but more kept as a placeholder on how to do it better
4215 in a newer implementation. */
4216 /* If we are COW and dstr is a suitable target then we drop down
4217 into the else and make dest a COW of us. */
4218 || (SvFLAGS(dstr) & SVf_BREAK)
4223 (sflags & SVs_TEMP) && /* slated for free anyway? */
4224 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4225 (!(flags & SV_NOSTEAL)) &&
4226 /* and we're allowed to steal temps */
4227 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4228 SvLEN(sstr)) /* and really is a string */
4229 #ifdef PERL_OLD_COPY_ON_WRITE
4230 && ((flags & SV_COW_SHARED_HASH_KEYS)
4231 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4232 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4233 && SvTYPE(sstr) >= SVt_PVIV))
4237 /* Failed the swipe test, and it's not a shared hash key either.
4238 Have to copy the string. */
4239 STRLEN len = SvCUR(sstr);
4240 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4241 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4242 SvCUR_set(dstr, len);
4243 *SvEND(dstr) = '\0';
4245 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4247 /* Either it's a shared hash key, or it's suitable for
4248 copy-on-write or we can swipe the string. */
4250 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4254 #ifdef PERL_OLD_COPY_ON_WRITE
4256 if (!(sflags & SVf_IsCOW)) {
4258 /* Make the source SV into a loop of 1.
4259 (about to become 2) */
4260 SV_COW_NEXT_SV_SET(sstr, sstr);
4264 /* Initial code is common. */
4265 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4270 /* making another shared SV. */
4271 STRLEN cur = SvCUR(sstr);
4272 STRLEN len = SvLEN(sstr);
4273 #ifdef PERL_OLD_COPY_ON_WRITE
4275 assert (SvTYPE(dstr) >= SVt_PVIV);
4276 /* SvIsCOW_normal */
4277 /* splice us in between source and next-after-source. */
4278 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4279 SV_COW_NEXT_SV_SET(sstr, dstr);
4280 SvPV_set(dstr, SvPVX_mutable(sstr));
4284 /* SvIsCOW_shared_hash */
4285 DEBUG_C(PerlIO_printf(Perl_debug_log,
4286 "Copy on write: Sharing hash\n"));
4288 assert (SvTYPE(dstr) >= SVt_PV);
4290 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4292 SvLEN_set(dstr, len);
4293 SvCUR_set(dstr, cur);
4297 { /* Passes the swipe test. */
4298 SvPV_set(dstr, SvPVX_mutable(sstr));
4299 SvLEN_set(dstr, SvLEN(sstr));
4300 SvCUR_set(dstr, SvCUR(sstr));
4303 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4304 SvPV_set(sstr, NULL);
4310 if (sflags & SVp_NOK) {
4311 SvNV_set(dstr, SvNVX(sstr));
4313 if (sflags & SVp_IOK) {
4314 SvIV_set(dstr, SvIVX(sstr));
4315 /* Must do this otherwise some other overloaded use of 0x80000000
4316 gets confused. I guess SVpbm_VALID */
4317 if (sflags & SVf_IVisUV)
4320 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4322 const MAGIC * const smg = SvVSTRING_mg(sstr);
4324 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4325 smg->mg_ptr, smg->mg_len);
4326 SvRMAGICAL_on(dstr);
4330 else if (sflags & (SVp_IOK|SVp_NOK)) {
4331 (void)SvOK_off(dstr);
4332 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4333 if (sflags & SVp_IOK) {
4334 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4335 SvIV_set(dstr, SvIVX(sstr));
4337 if (sflags & SVp_NOK) {
4338 SvNV_set(dstr, SvNVX(sstr));
4342 if (isGV_with_GP(sstr)) {
4343 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4346 (void)SvOK_off(dstr);
4348 if (SvTAINTED(sstr))
4353 =for apidoc sv_setsv_mg
4355 Like C<sv_setsv>, but also handles 'set' magic.
4361 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4363 PERL_ARGS_ASSERT_SV_SETSV_MG;
4365 sv_setsv(dstr,sstr);
4369 #ifdef PERL_OLD_COPY_ON_WRITE
4371 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4373 STRLEN cur = SvCUR(sstr);
4374 STRLEN len = SvLEN(sstr);
4377 PERL_ARGS_ASSERT_SV_SETSV_COW;
4380 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4381 (void*)sstr, (void*)dstr);
4388 if (SvTHINKFIRST(dstr))
4389 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4390 else if (SvPVX_const(dstr))
4391 Safefree(SvPVX_mutable(dstr));
4395 SvUPGRADE(dstr, SVt_PVIV);
4397 assert (SvPOK(sstr));
4398 assert (SvPOKp(sstr));
4399 assert (!SvIOK(sstr));
4400 assert (!SvIOKp(sstr));
4401 assert (!SvNOK(sstr));
4402 assert (!SvNOKp(sstr));
4404 if (SvIsCOW(sstr)) {
4406 if (SvLEN(sstr) == 0) {
4407 /* source is a COW shared hash key. */
4408 DEBUG_C(PerlIO_printf(Perl_debug_log,
4409 "Fast copy on write: Sharing hash\n"));
4410 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4413 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4415 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4416 SvUPGRADE(sstr, SVt_PVIV);
4418 DEBUG_C(PerlIO_printf(Perl_debug_log,
4419 "Fast copy on write: Converting sstr to COW\n"));
4420 SV_COW_NEXT_SV_SET(dstr, sstr);
4422 SV_COW_NEXT_SV_SET(sstr, dstr);
4423 new_pv = SvPVX_mutable(sstr);
4426 SvPV_set(dstr, new_pv);
4427 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_IsCOW);
4430 SvLEN_set(dstr, len);
4431 SvCUR_set(dstr, cur);
4440 =for apidoc sv_setpvn
4442 Copies a string into an SV. The C<len> parameter indicates the number of
4443 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4444 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4450 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4455 PERL_ARGS_ASSERT_SV_SETPVN;
4457 SV_CHECK_THINKFIRST_COW_DROP(sv);
4463 /* len is STRLEN which is unsigned, need to copy to signed */
4466 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4469 SvUPGRADE(sv, SVt_PV);
4471 dptr = SvGROW(sv, len + 1);
4472 Move(ptr,dptr,len,char);
4475 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4477 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4481 =for apidoc sv_setpvn_mg
4483 Like C<sv_setpvn>, but also handles 'set' magic.
4489 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4491 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4493 sv_setpvn(sv,ptr,len);
4498 =for apidoc sv_setpv
4500 Copies a string into an SV. The string must be null-terminated. Does not
4501 handle 'set' magic. See C<sv_setpv_mg>.
4507 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4512 PERL_ARGS_ASSERT_SV_SETPV;
4514 SV_CHECK_THINKFIRST_COW_DROP(sv);
4520 SvUPGRADE(sv, SVt_PV);
4522 SvGROW(sv, len + 1);
4523 Move(ptr,SvPVX(sv),len+1,char);
4525 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4527 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4531 =for apidoc sv_setpv_mg
4533 Like C<sv_setpv>, but also handles 'set' magic.
4539 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4541 PERL_ARGS_ASSERT_SV_SETPV_MG;
4548 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4552 PERL_ARGS_ASSERT_SV_SETHEK;
4558 if (HEK_LEN(hek) == HEf_SVKEY) {
4559 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4562 const int flags = HEK_FLAGS(hek);
4563 if (flags & HVhek_WASUTF8) {
4564 STRLEN utf8_len = HEK_LEN(hek);
4565 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4566 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4569 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4570 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4573 else SvUTF8_off(sv);
4577 SV_CHECK_THINKFIRST_COW_DROP(sv);
4578 SvUPGRADE(sv, SVt_PV);
4579 Safefree(SvPVX(sv));
4580 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4581 SvCUR_set(sv, HEK_LEN(hek));
4587 else SvUTF8_off(sv);
4595 =for apidoc sv_usepvn_flags
4597 Tells an SV to use C<ptr> to find its string value. Normally the
4598 string is stored inside the SV but sv_usepvn allows the SV to use an
4599 outside string. The C<ptr> should point to memory that was allocated
4600 by C<malloc>. It must be the start of a mallocked block
4601 of memory, and not a pointer to the middle of it. The
4602 string length, C<len>, must be supplied. By default
4603 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4604 so that pointer should not be freed or used by the programmer after
4605 giving it to sv_usepvn, and neither should any pointers from "behind"
4606 that pointer (e.g. ptr + 1) be used.
4608 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4609 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4610 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4611 C<len>, and already meets the requirements for storing in C<SvPVX>).
4617 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4622 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4624 SV_CHECK_THINKFIRST_COW_DROP(sv);
4625 SvUPGRADE(sv, SVt_PV);
4628 if (flags & SV_SMAGIC)
4632 if (SvPVX_const(sv))
4636 if (flags & SV_HAS_TRAILING_NUL)
4637 assert(ptr[len] == '\0');
4640 allocate = (flags & SV_HAS_TRAILING_NUL)
4642 #ifdef Perl_safesysmalloc_size
4645 PERL_STRLEN_ROUNDUP(len + 1);
4647 if (flags & SV_HAS_TRAILING_NUL) {
4648 /* It's long enough - do nothing.
4649 Specifically Perl_newCONSTSUB is relying on this. */
4652 /* Force a move to shake out bugs in callers. */
4653 char *new_ptr = (char*)safemalloc(allocate);
4654 Copy(ptr, new_ptr, len, char);
4655 PoisonFree(ptr,len,char);
4659 ptr = (char*) saferealloc (ptr, allocate);
4662 #ifdef Perl_safesysmalloc_size
4663 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4665 SvLEN_set(sv, allocate);
4669 if (!(flags & SV_HAS_TRAILING_NUL)) {
4672 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4674 if (flags & SV_SMAGIC)
4678 #ifdef PERL_OLD_COPY_ON_WRITE
4679 /* Need to do this *after* making the SV normal, as we need the buffer
4680 pointer to remain valid until after we've copied it. If we let go too early,
4681 another thread could invalidate it by unsharing last of the same hash key
4682 (which it can do by means other than releasing copy-on-write Svs)
4683 or by changing the other copy-on-write SVs in the loop. */
4685 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4687 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4689 { /* this SV was SvIsCOW_normal(sv) */
4690 /* we need to find the SV pointing to us. */
4691 SV *current = SV_COW_NEXT_SV(after);
4693 if (current == sv) {
4694 /* The SV we point to points back to us (there were only two of us
4696 Hence other SV is no longer copy on write either. */
4699 /* We need to follow the pointers around the loop. */
4701 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4704 /* don't loop forever if the structure is bust, and we have
4705 a pointer into a closed loop. */
4706 assert (current != after);
4707 assert (SvPVX_const(current) == pvx);
4709 /* Make the SV before us point to the SV after us. */
4710 SV_COW_NEXT_SV_SET(current, after);
4716 =for apidoc sv_force_normal_flags
4718 Undo various types of fakery on an SV, where fakery means
4719 "more than" a string: if the PV is a shared string, make
4720 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4721 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4722 we do the copy, and is also used locally; if this is a
4723 vstring, drop the vstring magic. If C<SV_COW_DROP_PV> is set
4724 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4725 SvPOK_off rather than making a copy. (Used where this
4726 scalar is about to be set to some other value.) In addition,
4727 the C<flags> parameter gets passed to C<sv_unref_flags()>
4728 when unreffing. C<sv_force_normal> calls this function
4729 with flags set to 0.
4735 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4739 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4741 #ifdef PERL_OLD_COPY_ON_WRITE
4742 if (SvREADONLY(sv)) {
4743 if (IN_PERL_RUNTIME)
4744 Perl_croak_no_modify(aTHX);
4748 const char * const pvx = SvPVX_const(sv);
4749 const STRLEN len = SvLEN(sv);
4750 const STRLEN cur = SvCUR(sv);
4751 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4752 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4753 we'll fail an assertion. */
4754 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4757 PerlIO_printf(Perl_debug_log,
4758 "Copy on write: Force normal %ld\n",
4763 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4766 if (flags & SV_COW_DROP_PV) {
4767 /* OK, so we don't need to copy our buffer. */
4770 SvGROW(sv, cur + 1);
4771 Move(pvx,SvPVX(sv),cur,char);
4776 sv_release_COW(sv, pvx, next);
4778 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4785 if (SvREADONLY(sv)) {
4786 if (IN_PERL_RUNTIME)
4787 Perl_croak_no_modify();
4791 const char * const pvx = SvPVX_const(sv);
4792 const STRLEN len = SvCUR(sv);
4796 if (flags & SV_COW_DROP_PV) {
4797 /* OK, so we don't need to copy our buffer. */
4800 SvGROW(sv, len + 1);
4801 Move(pvx,SvPVX(sv),len,char);
4804 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4808 sv_unref_flags(sv, flags);
4809 else if (SvFAKE(sv) && isGV_with_GP(sv))
4810 sv_unglob(sv, flags);
4811 else if (SvFAKE(sv) && isREGEXP(sv)) {
4812 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4813 to sv_unglob. We only need it here, so inline it. */
4814 const bool islv = SvTYPE(sv) == SVt_PVLV;
4815 const svtype new_type =
4816 islv ? SVt_NULL : SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4817 SV *const temp = newSV_type(new_type);
4818 regexp *const temp_p = ReANY((REGEXP *)sv);
4820 if (new_type == SVt_PVMG) {
4821 SvMAGIC_set(temp, SvMAGIC(sv));
4822 SvMAGIC_set(sv, NULL);
4823 SvSTASH_set(temp, SvSTASH(sv));
4824 SvSTASH_set(sv, NULL);
4826 if (!islv) SvCUR_set(temp, SvCUR(sv));
4827 /* Remember that SvPVX is in the head, not the body. But
4828 RX_WRAPPED is in the body. */
4829 assert(ReANY((REGEXP *)sv)->mother_re);
4830 /* Their buffer is already owned by someone else. */
4831 if (flags & SV_COW_DROP_PV) {
4832 /* SvLEN is already 0. For SVt_REGEXP, we have a brand new
4833 zeroed body. For SVt_PVLV, it should have been set to 0
4834 before turning into a regexp. */
4835 assert(!SvLEN(islv ? sv : temp));
4836 sv->sv_u.svu_pv = 0;
4839 sv->sv_u.svu_pv = savepvn(RX_WRAPPED((REGEXP *)sv), SvCUR(sv));
4840 SvLEN_set(islv ? sv : temp, SvCUR(sv)+1);
4844 /* Now swap the rest of the bodies. */
4848 SvFLAGS(sv) &= ~SVTYPEMASK;
4849 SvFLAGS(sv) |= new_type;
4850 SvANY(sv) = SvANY(temp);
4853 SvFLAGS(temp) &= ~(SVTYPEMASK);
4854 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4855 SvANY(temp) = temp_p;
4856 temp->sv_u.svu_rx = (regexp *)temp_p;
4860 else if (SvVOK(sv)) sv_unmagic(sv, PERL_MAGIC_vstring);
4866 Efficient removal of characters from the beginning of the string buffer.
4867 SvPOK(sv), or at least SvPOKp(sv), must be true and the C<ptr> must be a
4868 pointer to somewhere inside the string buffer. The C<ptr> becomes the first
4869 character of the adjusted string. Uses the "OOK hack". On return, only
4870 SvPOK(sv) and SvPOKp(sv) among the OK flags will be true.
4872 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4873 refer to the same chunk of data.
4875 The unfortunate similarity of this function's name to that of Perl's C<chop>
4876 operator is strictly coincidental. This function works from the left;
4877 C<chop> works from the right.
4883 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4894 PERL_ARGS_ASSERT_SV_CHOP;
4896 if (!ptr || !SvPOKp(sv))
4898 delta = ptr - SvPVX_const(sv);
4900 /* Nothing to do. */
4903 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4904 if (delta > max_delta)
4905 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4906 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4907 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4908 SV_CHECK_THINKFIRST(sv);
4909 SvPOK_only_UTF8(sv);
4912 if (!SvLEN(sv)) { /* make copy of shared string */
4913 const char *pvx = SvPVX_const(sv);
4914 const STRLEN len = SvCUR(sv);
4915 SvGROW(sv, len + 1);
4916 Move(pvx,SvPVX(sv),len,char);
4922 SvOOK_offset(sv, old_delta);
4924 SvLEN_set(sv, SvLEN(sv) - delta);
4925 SvCUR_set(sv, SvCUR(sv) - delta);
4926 SvPV_set(sv, SvPVX(sv) + delta);
4928 p = (U8 *)SvPVX_const(sv);
4931 /* how many bytes were evacuated? we will fill them with sentinel
4932 bytes, except for the part holding the new offset of course. */
4935 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4937 assert(evacn <= delta + old_delta);
4943 if (delta < 0x100) {
4947 p -= sizeof(STRLEN);
4948 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4952 /* Fill the preceding buffer with sentinals to verify that no-one is
4962 =for apidoc sv_catpvn
4964 Concatenates the string onto the end of the string which is in the SV. The
4965 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4966 status set, then the bytes appended should be valid UTF-8.
4967 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4969 =for apidoc sv_catpvn_flags
4971 Concatenates the string onto the end of the string which is in the SV. The
4972 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4973 status set, then the bytes appended should be valid UTF-8.
4974 If C<flags> has the C<SV_SMAGIC> bit set, will
4975 C<mg_set> on C<dsv> afterwards if appropriate.
4976 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4977 in terms of this function.
4983 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4987 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4989 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4990 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
4992 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
4993 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
4994 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
4997 else SvGROW(dsv, dlen + slen + 1);
4999 sstr = SvPVX_const(dsv);
5000 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5001 SvCUR_set(dsv, SvCUR(dsv) + slen);
5004 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5005 const char * const send = sstr + slen;
5008 /* Something this code does not account for, which I think is
5009 impossible; it would require the same pv to be treated as
5010 bytes *and* utf8, which would indicate a bug elsewhere. */
5011 assert(sstr != dstr);
5013 SvGROW(dsv, dlen + slen * 2 + 1);
5014 d = (U8 *)SvPVX(dsv) + dlen;
5016 while (sstr < send) {
5017 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5018 if (UNI_IS_INVARIANT(uv))
5019 *d++ = (U8)UTF_TO_NATIVE(uv);
5021 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5022 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5025 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5028 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5030 if (flags & SV_SMAGIC)
5035 =for apidoc sv_catsv
5037 Concatenates the string from SV C<ssv> onto the end of the string in SV
5038 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5039 Handles 'get' magic on both SVs, but no 'set' magic. See C<sv_catsv_mg> and
5042 =for apidoc sv_catsv_flags
5044 Concatenates the string from SV C<ssv> onto the end of the string in SV
5045 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5046 If C<flags> include C<SV_GMAGIC> bit set, will call C<mg_get> on both SVs if
5047 appropriate. If C<flags> include C<SV_SMAGIC>, C<mg_set> will be called on
5048 the modified SV afterward, if appropriate. C<sv_catsv>, C<sv_catsv_nomg>,
5049 and C<sv_catsv_mg> are implemented in terms of this function.
5054 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5058 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5062 const char *spv = SvPV_flags_const(ssv, slen, flags);
5064 if (flags & SV_GMAGIC)
5066 sv_catpvn_flags(dsv, spv, slen,
5067 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5068 if (flags & SV_SMAGIC)
5075 =for apidoc sv_catpv
5077 Concatenates the string onto the end of the string which is in the SV.
5078 If the SV has the UTF-8 status set, then the bytes appended should be
5079 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5084 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5091 PERL_ARGS_ASSERT_SV_CATPV;
5095 junk = SvPV_force(sv, tlen);
5097 SvGROW(sv, tlen + len + 1);
5099 ptr = SvPVX_const(sv);
5100 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5101 SvCUR_set(sv, SvCUR(sv) + len);
5102 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5107 =for apidoc sv_catpv_flags
5109 Concatenates the string onto the end of the string which is in the SV.
5110 If the SV has the UTF-8 status set, then the bytes appended should
5111 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5112 on the modified SV if appropriate.
5118 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5120 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5121 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5125 =for apidoc sv_catpv_mg
5127 Like C<sv_catpv>, but also handles 'set' magic.
5133 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5135 PERL_ARGS_ASSERT_SV_CATPV_MG;
5144 Creates a new SV. A non-zero C<len> parameter indicates the number of
5145 bytes of preallocated string space the SV should have. An extra byte for a
5146 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5147 space is allocated.) The reference count for the new SV is set to 1.
5149 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5150 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5151 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5152 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5153 modules supporting older perls.
5159 Perl_newSV(pTHX_ const STRLEN len)
5166 sv_upgrade(sv, SVt_PV);
5167 SvGROW(sv, len + 1);
5172 =for apidoc sv_magicext
5174 Adds magic to an SV, upgrading it if necessary. Applies the
5175 supplied vtable and returns a pointer to the magic added.
5177 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5178 In particular, you can add magic to SvREADONLY SVs, and add more than
5179 one instance of the same 'how'.
5181 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5182 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5183 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5184 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5186 (This is now used as a subroutine by C<sv_magic>.)
5191 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5192 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5197 PERL_ARGS_ASSERT_SV_MAGICEXT;
5199 SvUPGRADE(sv, SVt_PVMG);
5200 Newxz(mg, 1, MAGIC);
5201 mg->mg_moremagic = SvMAGIC(sv);
5202 SvMAGIC_set(sv, mg);
5204 /* Sometimes a magic contains a reference loop, where the sv and
5205 object refer to each other. To prevent a reference loop that
5206 would prevent such objects being freed, we look for such loops
5207 and if we find one we avoid incrementing the object refcount.
5209 Note we cannot do this to avoid self-tie loops as intervening RV must
5210 have its REFCNT incremented to keep it in existence.
5213 if (!obj || obj == sv ||
5214 how == PERL_MAGIC_arylen ||
5215 how == PERL_MAGIC_symtab ||
5216 (SvTYPE(obj) == SVt_PVGV &&
5217 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5218 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5219 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5224 mg->mg_obj = SvREFCNT_inc_simple(obj);
5225 mg->mg_flags |= MGf_REFCOUNTED;
5228 /* Normal self-ties simply pass a null object, and instead of
5229 using mg_obj directly, use the SvTIED_obj macro to produce a
5230 new RV as needed. For glob "self-ties", we are tieing the PVIO
5231 with an RV obj pointing to the glob containing the PVIO. In
5232 this case, to avoid a reference loop, we need to weaken the
5236 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5237 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5243 mg->mg_len = namlen;
5246 mg->mg_ptr = savepvn(name, namlen);
5247 else if (namlen == HEf_SVKEY) {
5248 /* Yes, this is casting away const. This is only for the case of
5249 HEf_SVKEY. I think we need to document this aberation of the
5250 constness of the API, rather than making name non-const, as
5251 that change propagating outwards a long way. */
5252 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5254 mg->mg_ptr = (char *) name;
5256 mg->mg_virtual = (MGVTBL *) vtable;
5263 =for apidoc sv_magic
5265 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5266 necessary, then adds a new magic item of type C<how> to the head of the
5269 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5270 handling of the C<name> and C<namlen> arguments.
5272 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5273 to add more than one instance of the same 'how'.
5279 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5280 const char *const name, const I32 namlen)
5283 const MGVTBL *vtable;
5286 unsigned int vtable_index;
5288 PERL_ARGS_ASSERT_SV_MAGIC;
5290 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5291 || ((flags = PL_magic_data[how]),
5292 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5293 > magic_vtable_max))
5294 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5296 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5297 Useful for attaching extension internal data to perl vars.
5298 Note that multiple extensions may clash if magical scalars
5299 etc holding private data from one are passed to another. */
5301 vtable = (vtable_index == magic_vtable_max)
5302 ? NULL : PL_magic_vtables + vtable_index;
5304 #ifdef PERL_OLD_COPY_ON_WRITE
5306 sv_force_normal_flags(sv, 0);
5308 if (SvREADONLY(sv)) {
5310 /* its okay to attach magic to shared strings */
5314 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5317 Perl_croak_no_modify();
5320 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5321 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5322 /* sv_magic() refuses to add a magic of the same 'how' as an
5325 if (how == PERL_MAGIC_taint)
5331 /* Rest of work is done else where */
5332 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5335 case PERL_MAGIC_taint:
5338 case PERL_MAGIC_ext:
5339 case PERL_MAGIC_dbfile:
5346 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5353 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5355 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5356 for (mg = *mgp; mg; mg = *mgp) {
5357 const MGVTBL* const virt = mg->mg_virtual;
5358 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5359 *mgp = mg->mg_moremagic;
5360 if (virt && virt->svt_free)
5361 virt->svt_free(aTHX_ sv, mg);
5362 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5364 Safefree(mg->mg_ptr);
5365 else if (mg->mg_len == HEf_SVKEY)
5366 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5367 else if (mg->mg_type == PERL_MAGIC_utf8)
5368 Safefree(mg->mg_ptr);
5370 if (mg->mg_flags & MGf_REFCOUNTED)
5371 SvREFCNT_dec(mg->mg_obj);
5375 mgp = &mg->mg_moremagic;
5378 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5379 mg_magical(sv); /* else fix the flags now */
5383 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5389 =for apidoc sv_unmagic
5391 Removes all magic of type C<type> from an SV.
5397 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5399 PERL_ARGS_ASSERT_SV_UNMAGIC;
5400 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5404 =for apidoc sv_unmagicext
5406 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5412 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5414 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5415 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5419 =for apidoc sv_rvweaken
5421 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5422 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5423 push a back-reference to this RV onto the array of backreferences
5424 associated with that magic. If the RV is magical, set magic will be
5425 called after the RV is cleared.
5431 Perl_sv_rvweaken(pTHX_ SV *const sv)
5435 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5437 if (!SvOK(sv)) /* let undefs pass */
5440 Perl_croak(aTHX_ "Can't weaken a nonreference");
5441 else if (SvWEAKREF(sv)) {
5442 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5445 else if (SvREADONLY(sv)) croak_no_modify();
5447 Perl_sv_add_backref(aTHX_ tsv, sv);
5453 /* Give tsv backref magic if it hasn't already got it, then push a
5454 * back-reference to sv onto the array associated with the backref magic.
5456 * As an optimisation, if there's only one backref and it's not an AV,
5457 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5458 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5462 /* A discussion about the backreferences array and its refcount:
5464 * The AV holding the backreferences is pointed to either as the mg_obj of
5465 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5466 * xhv_backreferences field. The array is created with a refcount
5467 * of 2. This means that if during global destruction the array gets
5468 * picked on before its parent to have its refcount decremented by the
5469 * random zapper, it won't actually be freed, meaning it's still there for
5470 * when its parent gets freed.
5472 * When the parent SV is freed, the extra ref is killed by
5473 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5474 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5476 * When a single backref SV is stored directly, it is not reference
5481 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5488 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5490 /* find slot to store array or singleton backref */
5492 if (SvTYPE(tsv) == SVt_PVHV) {
5493 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5496 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5498 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5499 mg = mg_find(tsv, PERL_MAGIC_backref);
5501 svp = &(mg->mg_obj);
5504 /* create or retrieve the array */
5506 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5507 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5512 SvREFCNT_inc_simple_void(av);
5513 /* av now has a refcnt of 2; see discussion above */
5515 /* move single existing backref to the array */
5517 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5521 mg->mg_flags |= MGf_REFCOUNTED;
5524 av = MUTABLE_AV(*svp);
5527 /* optimisation: store single backref directly in HvAUX or mg_obj */
5531 /* push new backref */
5532 assert(SvTYPE(av) == SVt_PVAV);
5533 if (AvFILLp(av) >= AvMAX(av)) {
5534 av_extend(av, AvFILLp(av)+1);
5536 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5539 /* delete a back-reference to ourselves from the backref magic associated
5540 * with the SV we point to.
5544 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5549 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5551 if (SvTYPE(tsv) == SVt_PVHV) {
5553 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5555 else if (SvIS_FREED(tsv) && PL_phase == PERL_PHASE_DESTRUCT) {
5556 /* It's possible for the the last (strong) reference to tsv to have
5557 become freed *before* the last thing holding a weak reference.
5558 If both survive longer than the backreferences array, then when
5559 the referent's reference count drops to 0 and it is freed, it's
5560 not able to chase the backreferences, so they aren't NULLed.
5562 For example, a CV holds a weak reference to its stash. If both the
5563 CV and the stash survive longer than the backreferences array,
5564 and the CV gets picked for the SvBREAK() treatment first,
5565 *and* it turns out that the stash is only being kept alive because
5566 of an our variable in the pad of the CV, then midway during CV
5567 destruction the stash gets freed, but CvSTASH() isn't set to NULL.
5568 It ends up pointing to the freed HV. Hence it's chased in here, and
5569 if this block wasn't here, it would hit the !svp panic just below.
5571 I don't believe that "better" destruction ordering is going to help
5572 here - during global destruction there's always going to be the
5573 chance that something goes out of order. We've tried to make it
5574 foolproof before, and it only resulted in evolutionary pressure on
5575 fools. Which made us look foolish for our hubris. :-(
5581 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5582 svp = mg ? &(mg->mg_obj) : NULL;
5586 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5588 /* It's possible that sv is being freed recursively part way through the
5589 freeing of tsv. If this happens, the backreferences array of tsv has
5590 already been freed, and so svp will be NULL. If this is the case,
5591 we should not panic. Instead, nothing needs doing, so return. */
5592 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5594 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5595 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5598 if (SvTYPE(*svp) == SVt_PVAV) {
5602 AV * const av = (AV*)*svp;
5604 assert(!SvIS_FREED(av));
5608 /* for an SV with N weak references to it, if all those
5609 * weak refs are deleted, then sv_del_backref will be called
5610 * N times and O(N^2) compares will be done within the backref
5611 * array. To ameliorate this potential slowness, we:
5612 * 1) make sure this code is as tight as possible;
5613 * 2) when looking for SV, look for it at both the head and tail of the
5614 * array first before searching the rest, since some create/destroy
5615 * patterns will cause the backrefs to be freed in order.
5622 SV **p = &svp[fill];
5623 SV *const topsv = *p;
5630 /* We weren't the last entry.
5631 An unordered list has this property that you
5632 can take the last element off the end to fill
5633 the hole, and it's still an unordered list :-)
5639 break; /* should only be one */
5646 AvFILLp(av) = fill-1;
5648 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5649 /* freed AV; skip */
5652 /* optimisation: only a single backref, stored directly */
5654 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5661 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5667 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5672 /* after multiple passes through Perl_sv_clean_all() for a thingy
5673 * that has badly leaked, the backref array may have gotten freed,
5674 * since we only protect it against 1 round of cleanup */
5675 if (SvIS_FREED(av)) {
5676 if (PL_in_clean_all) /* All is fair */
5679 "panic: magic_killbackrefs (freed backref AV/SV)");
5683 is_array = (SvTYPE(av) == SVt_PVAV);
5685 assert(!SvIS_FREED(av));
5688 last = svp + AvFILLp(av);
5691 /* optimisation: only a single backref, stored directly */
5697 while (svp <= last) {
5699 SV *const referrer = *svp;
5700 if (SvWEAKREF(referrer)) {
5701 /* XXX Should we check that it hasn't changed? */
5702 assert(SvROK(referrer));
5703 SvRV_set(referrer, 0);
5705 SvWEAKREF_off(referrer);
5706 SvSETMAGIC(referrer);
5707 } else if (SvTYPE(referrer) == SVt_PVGV ||
5708 SvTYPE(referrer) == SVt_PVLV) {
5709 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5710 /* You lookin' at me? */
5711 assert(GvSTASH(referrer));
5712 assert(GvSTASH(referrer) == (const HV *)sv);
5713 GvSTASH(referrer) = 0;
5714 } else if (SvTYPE(referrer) == SVt_PVCV ||
5715 SvTYPE(referrer) == SVt_PVFM) {
5716 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5717 /* You lookin' at me? */
5718 assert(CvSTASH(referrer));
5719 assert(CvSTASH(referrer) == (const HV *)sv);
5720 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5723 assert(SvTYPE(sv) == SVt_PVGV);
5724 /* You lookin' at me? */
5725 assert(CvGV(referrer));
5726 assert(CvGV(referrer) == (const GV *)sv);
5727 anonymise_cv_maybe(MUTABLE_GV(sv),
5728 MUTABLE_CV(referrer));
5733 "panic: magic_killbackrefs (flags=%"UVxf")",
5734 (UV)SvFLAGS(referrer));
5745 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5751 =for apidoc sv_insert
5753 Inserts a string at the specified offset/length within the SV. Similar to
5754 the Perl substr() function. Handles get magic.
5756 =for apidoc sv_insert_flags
5758 Same as C<sv_insert>, but the extra C<flags> are passed to the
5759 C<SvPV_force_flags> that applies to C<bigstr>.
5765 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5772 SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5775 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5778 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5779 SvPV_force_flags(bigstr, curlen, flags);
5780 (void)SvPOK_only_UTF8(bigstr);
5781 if (offset + len > curlen) {
5782 SvGROW(bigstr, offset+len+1);
5783 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5784 SvCUR_set(bigstr, offset+len);
5788 i = littlelen - len;
5789 if (i > 0) { /* string might grow */
5790 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5791 mid = big + offset + len;
5792 midend = bigend = big + SvCUR(bigstr);
5795 while (midend > mid) /* shove everything down */
5796 *--bigend = *--midend;
5797 Move(little,big+offset,littlelen,char);
5798 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5803 Move(little,SvPVX(bigstr)+offset,len,char);
5808 big = SvPVX(bigstr);
5811 bigend = big + SvCUR(bigstr);
5813 if (midend > bigend)
5814 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
5817 if (mid - big > bigend - midend) { /* faster to shorten from end */
5819 Move(little, mid, littlelen,char);
5822 i = bigend - midend;
5824 Move(midend, mid, i,char);
5828 SvCUR_set(bigstr, mid - big);
5830 else if ((i = mid - big)) { /* faster from front */
5831 midend -= littlelen;
5833 Move(big, midend - i, i, char);
5834 sv_chop(bigstr,midend-i);
5836 Move(little, mid, littlelen,char);
5838 else if (littlelen) {
5839 midend -= littlelen;
5840 sv_chop(bigstr,midend);
5841 Move(little,midend,littlelen,char);
5844 sv_chop(bigstr,midend);
5850 =for apidoc sv_replace
5852 Make the first argument a copy of the second, then delete the original.
5853 The target SV physically takes over ownership of the body of the source SV
5854 and inherits its flags; however, the target keeps any magic it owns,
5855 and any magic in the source is discarded.
5856 Note that this is a rather specialist SV copying operation; most of the
5857 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5863 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5866 const U32 refcnt = SvREFCNT(sv);
5868 PERL_ARGS_ASSERT_SV_REPLACE;
5870 SV_CHECK_THINKFIRST_COW_DROP(sv);
5871 if (SvREFCNT(nsv) != 1) {
5872 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5873 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5875 if (SvMAGICAL(sv)) {
5879 sv_upgrade(nsv, SVt_PVMG);
5880 SvMAGIC_set(nsv, SvMAGIC(sv));
5881 SvFLAGS(nsv) |= SvMAGICAL(sv);
5883 SvMAGIC_set(sv, NULL);
5887 assert(!SvREFCNT(sv));
5888 #ifdef DEBUG_LEAKING_SCALARS
5889 sv->sv_flags = nsv->sv_flags;
5890 sv->sv_any = nsv->sv_any;
5891 sv->sv_refcnt = nsv->sv_refcnt;
5892 sv->sv_u = nsv->sv_u;
5894 StructCopy(nsv,sv,SV);
5896 if(SvTYPE(sv) == SVt_IV) {
5898 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5902 #ifdef PERL_OLD_COPY_ON_WRITE
5903 if (SvIsCOW_normal(nsv)) {
5904 /* We need to follow the pointers around the loop to make the
5905 previous SV point to sv, rather than nsv. */
5908 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5911 assert(SvPVX_const(current) == SvPVX_const(nsv));
5913 /* Make the SV before us point to the SV after us. */
5915 PerlIO_printf(Perl_debug_log, "previous is\n");
5917 PerlIO_printf(Perl_debug_log,
5918 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5919 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5921 SV_COW_NEXT_SV_SET(current, sv);
5924 SvREFCNT(sv) = refcnt;
5925 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5930 /* We're about to free a GV which has a CV that refers back to us.
5931 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5935 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5940 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5943 assert(SvREFCNT(gv) == 0);
5944 assert(isGV(gv) && isGV_with_GP(gv));
5946 assert(!CvANON(cv));
5947 assert(CvGV(cv) == gv);
5948 assert(!CvNAMED(cv));
5950 /* will the CV shortly be freed by gp_free() ? */
5951 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5952 SvANY(cv)->xcv_gv_u.xcv_gv = NULL;
5956 /* if not, anonymise: */
5957 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5958 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5959 : newSVpvn_flags( "__ANON__", 8, 0 );
5960 sv_catpvs(gvname, "::__ANON__");
5961 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5962 SvREFCNT_dec(gvname);
5966 SvANY(cv)->xcv_gv_u.xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5971 =for apidoc sv_clear
5973 Clear an SV: call any destructors, free up any memory used by the body,
5974 and free the body itself. The SV's head is I<not> freed, although
5975 its type is set to all 1's so that it won't inadvertently be assumed
5976 to be live during global destruction etc.
5977 This function should only be called when REFCNT is zero. Most of the time
5978 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5985 Perl_sv_clear(pTHX_ SV *const orig_sv)
5990 const struct body_details *sv_type_details;
5996 PERL_ARGS_ASSERT_SV_CLEAR;
5998 /* within this loop, sv is the SV currently being freed, and
5999 * iter_sv is the most recent AV or whatever that's being iterated
6000 * over to provide more SVs */
6006 assert(SvREFCNT(sv) == 0);
6007 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
6009 if (type <= SVt_IV) {
6010 /* See the comment in sv.h about the collusion between this
6011 * early return and the overloading of the NULL slots in the
6015 SvFLAGS(sv) &= SVf_BREAK;
6016 SvFLAGS(sv) |= SVTYPEMASK;
6020 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6022 if (type >= SVt_PVMG) {
6024 if (!curse(sv, 1)) goto get_next_sv;
6025 type = SvTYPE(sv); /* destructor may have changed it */
6027 /* Free back-references before magic, in case the magic calls
6028 * Perl code that has weak references to sv. */
6029 if (type == SVt_PVHV) {
6030 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6034 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6035 SvREFCNT_dec(SvOURSTASH(sv));
6036 } else if (SvMAGIC(sv)) {
6037 /* Free back-references before other types of magic. */
6038 sv_unmagic(sv, PERL_MAGIC_backref);
6042 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6043 SvREFCNT_dec(SvSTASH(sv));
6046 /* case SVt_BIND: */
6049 IoIFP(sv) != PerlIO_stdin() &&
6050 IoIFP(sv) != PerlIO_stdout() &&
6051 IoIFP(sv) != PerlIO_stderr() &&
6052 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6054 io_close(MUTABLE_IO(sv), FALSE);
6056 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6057 PerlDir_close(IoDIRP(sv));
6058 IoDIRP(sv) = (DIR*)NULL;
6059 Safefree(IoTOP_NAME(sv));
6060 Safefree(IoFMT_NAME(sv));
6061 Safefree(IoBOTTOM_NAME(sv));
6062 if ((const GV *)sv == PL_statgv)
6066 /* FIXME for plugins */
6068 pregfree2((REGEXP*) sv);
6072 cv_undef(MUTABLE_CV(sv));
6073 /* If we're in a stash, we don't own a reference to it.
6074 * However it does have a back reference to us, which needs to
6076 if ((stash = CvSTASH(sv)))
6077 sv_del_backref(MUTABLE_SV(stash), sv);
6080 if (PL_last_swash_hv == (const HV *)sv) {
6081 PL_last_swash_hv = NULL;
6083 if (HvTOTALKEYS((HV*)sv) > 0) {
6085 /* this statement should match the one at the beginning of
6086 * hv_undef_flags() */
6087 if ( PL_phase != PERL_PHASE_DESTRUCT
6088 && (name = HvNAME((HV*)sv)))
6090 if (PL_stashcache) {
6091 DEBUG_o(Perl_deb(aTHX_ "sv_clear clearing PL_stashcache for '%"SVf"'\n",
6093 (void)hv_delete(PL_stashcache, name,
6094 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6096 hv_name_set((HV*)sv, NULL, 0, 0);
6099 /* save old iter_sv in unused SvSTASH field */
6100 assert(!SvOBJECT(sv));
6101 SvSTASH(sv) = (HV*)iter_sv;
6104 /* save old hash_index in unused SvMAGIC field */
6105 assert(!SvMAGICAL(sv));
6106 assert(!SvMAGIC(sv));
6107 ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index = hash_index;
6110 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6111 goto get_next_sv; /* process this new sv */
6113 /* free empty hash */
6114 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6115 assert(!HvARRAY((HV*)sv));
6119 AV* av = MUTABLE_AV(sv);
6120 if (PL_comppad == av) {
6124 if (AvREAL(av) && AvFILLp(av) > -1) {
6125 next_sv = AvARRAY(av)[AvFILLp(av)--];
6126 /* save old iter_sv in top-most slot of AV,
6127 * and pray that it doesn't get wiped in the meantime */
6128 AvARRAY(av)[AvMAX(av)] = iter_sv;
6130 goto get_next_sv; /* process this new sv */
6132 Safefree(AvALLOC(av));
6137 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6138 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6139 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6140 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6142 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6143 SvREFCNT_dec(LvTARG(sv));
6144 if (isREGEXP(sv)) goto freeregexp;
6146 if (isGV_with_GP(sv)) {
6147 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6148 && HvENAME_get(stash))
6149 mro_method_changed_in(stash);
6150 gp_free(MUTABLE_GV(sv));
6152 unshare_hek(GvNAME_HEK(sv));
6153 /* If we're in a stash, we don't own a reference to it.
6154 * However it does have a back reference to us, which
6155 * needs to be cleared. */
6156 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6157 sv_del_backref(MUTABLE_SV(stash), sv);
6159 /* FIXME. There are probably more unreferenced pointers to SVs
6160 * in the interpreter struct that we should check and tidy in
6161 * a similar fashion to this: */
6162 /* See also S_sv_unglob, which does the same thing. */
6163 if ((const GV *)sv == PL_last_in_gv)
6164 PL_last_in_gv = NULL;
6165 else if ((const GV *)sv == PL_statgv)
6172 /* Don't bother with SvOOK_off(sv); as we're only going to
6176 SvOOK_offset(sv, offset);
6177 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6178 /* Don't even bother with turning off the OOK flag. */
6183 SV * const target = SvRV(sv);
6185 sv_del_backref(target, sv);
6190 #ifdef PERL_OLD_COPY_ON_WRITE
6191 else if (SvPVX_const(sv)
6192 && !(SvTYPE(sv) == SVt_PVIO
6193 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6197 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6201 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6203 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6206 } else if (SvLEN(sv)) {
6207 Safefree(SvPVX_mutable(sv));
6211 else if (SvPVX_const(sv) && SvLEN(sv)
6212 && !(SvTYPE(sv) == SVt_PVIO
6213 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6214 Safefree(SvPVX_mutable(sv));
6215 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6216 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6226 SvFLAGS(sv) &= SVf_BREAK;
6227 SvFLAGS(sv) |= SVTYPEMASK;
6229 sv_type_details = bodies_by_type + type;
6230 if (sv_type_details->arena) {
6231 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6232 &PL_body_roots[type]);
6234 else if (sv_type_details->body_size) {
6235 safefree(SvANY(sv));
6239 /* caller is responsible for freeing the head of the original sv */
6240 if (sv != orig_sv && !SvREFCNT(sv))
6243 /* grab and free next sv, if any */
6251 else if (!iter_sv) {
6253 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6254 AV *const av = (AV*)iter_sv;
6255 if (AvFILLp(av) > -1) {
6256 sv = AvARRAY(av)[AvFILLp(av)--];
6258 else { /* no more elements of current AV to free */
6261 /* restore previous value, squirrelled away */
6262 iter_sv = AvARRAY(av)[AvMAX(av)];
6263 Safefree(AvALLOC(av));
6266 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6267 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6268 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6269 /* no more elements of current HV to free */
6272 /* Restore previous values of iter_sv and hash_index,
6273 * squirrelled away */
6274 assert(!SvOBJECT(sv));
6275 iter_sv = (SV*)SvSTASH(sv);
6276 assert(!SvMAGICAL(sv));
6277 hash_index = ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index;
6279 /* perl -DA does not like rubbish in SvMAGIC. */
6283 /* free any remaining detritus from the hash struct */
6284 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6285 assert(!HvARRAY((HV*)sv));
6290 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6294 if (!SvREFCNT(sv)) {
6298 if (--(SvREFCNT(sv)))
6302 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6303 "Attempt to free temp prematurely: SV 0x%"UVxf
6304 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6308 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6309 /* make sure SvREFCNT(sv)==0 happens very seldom */
6310 SvREFCNT(sv) = (~(U32)0)/2;
6319 /* This routine curses the sv itself, not the object referenced by sv. So
6320 sv does not have to be ROK. */
6323 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6326 PERL_ARGS_ASSERT_CURSE;
6327 assert(SvOBJECT(sv));
6329 if (PL_defstash && /* Still have a symbol table? */
6336 stash = SvSTASH(sv);
6337 destructor = StashHANDLER(stash,DESTROY);
6339 /* A constant subroutine can have no side effects, so
6340 don't bother calling it. */
6341 && !CvCONST(destructor)
6342 /* Don't bother calling an empty destructor or one that
6343 returns immediately. */
6344 && (CvISXSUB(destructor)
6345 || (CvSTART(destructor)
6346 && (CvSTART(destructor)->op_next->op_type
6348 && (CvSTART(destructor)->op_next->op_type
6350 || CvSTART(destructor)->op_next->op_next->op_type
6356 SV* const tmpref = newRV(sv);
6357 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6359 PUSHSTACKi(PERLSI_DESTROY);
6364 call_sv(MUTABLE_SV(destructor),
6365 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6369 if(SvREFCNT(tmpref) < 2) {
6370 /* tmpref is not kept alive! */
6372 SvRV_set(tmpref, NULL);
6375 SvREFCNT_dec(tmpref);
6377 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6380 if (check_refcnt && SvREFCNT(sv)) {
6381 if (PL_in_clean_objs)
6383 "DESTROY created new reference to dead object '%"HEKf"'",
6384 HEKfARG(HvNAME_HEK(stash)));
6385 /* DESTROY gave object new lease on life */
6391 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6392 SvOBJECT_off(sv); /* Curse the object. */
6393 if (SvTYPE(sv) != SVt_PVIO)
6394 --PL_sv_objcount;/* XXX Might want something more general */
6400 =for apidoc sv_newref
6402 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6409 Perl_sv_newref(pTHX_ SV *const sv)
6411 PERL_UNUSED_CONTEXT;
6420 Decrement an SV's reference count, and if it drops to zero, call
6421 C<sv_clear> to invoke destructors and free up any memory used by
6422 the body; finally, deallocate the SV's head itself.
6423 Normally called via a wrapper macro C<SvREFCNT_dec>.
6429 Perl_sv_free(pTHX_ SV *const sv)
6434 if (SvREFCNT(sv) == 0) {
6435 if (SvFLAGS(sv) & SVf_BREAK)
6436 /* this SV's refcnt has been artificially decremented to
6437 * trigger cleanup */
6439 if (PL_in_clean_all) /* All is fair */
6441 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6442 /* make sure SvREFCNT(sv)==0 happens very seldom */
6443 SvREFCNT(sv) = (~(U32)0)/2;
6446 if (ckWARN_d(WARN_INTERNAL)) {
6447 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6448 Perl_dump_sv_child(aTHX_ sv);
6450 #ifdef DEBUG_LEAKING_SCALARS
6453 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6454 if (PL_warnhook == PERL_WARNHOOK_FATAL
6455 || ckDEAD(packWARN(WARN_INTERNAL))) {
6456 /* Don't let Perl_warner cause us to escape our fate: */
6460 /* This may not return: */
6461 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6462 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6463 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6466 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6471 if (--(SvREFCNT(sv)) > 0)
6473 Perl_sv_free2(aTHX_ sv);
6477 Perl_sv_free2(pTHX_ SV *const sv)
6481 PERL_ARGS_ASSERT_SV_FREE2;
6485 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6486 "Attempt to free temp prematurely: SV 0x%"UVxf
6487 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6491 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6492 /* make sure SvREFCNT(sv)==0 happens very seldom */
6493 SvREFCNT(sv) = (~(U32)0)/2;
6504 Returns the length of the string in the SV. Handles magic and type
6505 coercion and sets the UTF8 flag appropriately. See also C<SvCUR>, which
6506 gives raw access to the xpv_cur slot.
6512 Perl_sv_len(pTHX_ register SV *const sv)
6519 (void)SvPV_const(sv, len);
6524 =for apidoc sv_len_utf8
6526 Returns the number of characters in the string in an SV, counting wide
6527 UTF-8 bytes as a single character. Handles magic and type coercion.
6533 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6534 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6535 * (Note that the mg_len is not the length of the mg_ptr field.
6536 * This allows the cache to store the character length of the string without
6537 * needing to malloc() extra storage to attach to the mg_ptr.)
6542 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6548 return sv_len_utf8_nomg(sv);
6552 Perl_sv_len_utf8_nomg(pTHX_ SV * const sv)
6556 const U8 *s = (U8*)SvPV_nomg_const(sv, len);
6558 PERL_ARGS_ASSERT_SV_LEN_UTF8_NOMG;
6560 if (PL_utf8cache && SvUTF8(sv)) {
6562 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6564 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6565 if (mg->mg_len != -1)
6568 /* We can use the offset cache for a headstart.
6569 The longer value is stored in the first pair. */
6570 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6572 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6576 if (PL_utf8cache < 0) {
6577 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6578 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6582 ulen = Perl_utf8_length(aTHX_ s, s + len);
6583 utf8_mg_len_cache_update(sv, &mg, ulen);
6587 return SvUTF8(sv) ? Perl_utf8_length(aTHX_ s, s + len) : len;
6590 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6593 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6594 STRLEN *const uoffset_p, bool *const at_end)
6596 const U8 *s = start;
6597 STRLEN uoffset = *uoffset_p;
6599 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6601 while (s < send && uoffset) {
6608 else if (s > send) {
6610 /* This is the existing behaviour. Possibly it should be a croak, as
6611 it's actually a bounds error */
6614 *uoffset_p -= uoffset;
6618 /* Given the length of the string in both bytes and UTF-8 characters, decide
6619 whether to walk forwards or backwards to find the byte corresponding to
6620 the passed in UTF-8 offset. */
6622 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6623 STRLEN uoffset, const STRLEN uend)
6625 STRLEN backw = uend - uoffset;
6627 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6629 if (uoffset < 2 * backw) {
6630 /* The assumption is that going forwards is twice the speed of going
6631 forward (that's where the 2 * backw comes from).
6632 (The real figure of course depends on the UTF-8 data.) */
6633 const U8 *s = start;
6635 while (s < send && uoffset--)
6645 while (UTF8_IS_CONTINUATION(*send))
6648 return send - start;
6651 /* For the string representation of the given scalar, find the byte
6652 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6653 give another position in the string, *before* the sought offset, which
6654 (which is always true, as 0, 0 is a valid pair of positions), which should
6655 help reduce the amount of linear searching.
6656 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6657 will be used to reduce the amount of linear searching. The cache will be
6658 created if necessary, and the found value offered to it for update. */
6660 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6661 const U8 *const send, STRLEN uoffset,
6662 STRLEN uoffset0, STRLEN boffset0)
6664 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6666 bool at_end = FALSE;
6668 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6670 assert (uoffset >= uoffset0);
6675 if (!SvREADONLY(sv) && !SvGMAGICAL(sv) && SvPOK(sv)
6677 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6678 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6679 if ((*mgp)->mg_ptr) {
6680 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6681 if (cache[0] == uoffset) {
6682 /* An exact match. */
6685 if (cache[2] == uoffset) {
6686 /* An exact match. */
6690 if (cache[0] < uoffset) {
6691 /* The cache already knows part of the way. */
6692 if (cache[0] > uoffset0) {
6693 /* The cache knows more than the passed in pair */
6694 uoffset0 = cache[0];
6695 boffset0 = cache[1];
6697 if ((*mgp)->mg_len != -1) {
6698 /* And we know the end too. */
6700 + sv_pos_u2b_midway(start + boffset0, send,
6702 (*mgp)->mg_len - uoffset0);
6704 uoffset -= uoffset0;
6706 + sv_pos_u2b_forwards(start + boffset0,
6707 send, &uoffset, &at_end);
6708 uoffset += uoffset0;
6711 else if (cache[2] < uoffset) {
6712 /* We're between the two cache entries. */
6713 if (cache[2] > uoffset0) {
6714 /* and the cache knows more than the passed in pair */
6715 uoffset0 = cache[2];
6716 boffset0 = cache[3];
6720 + sv_pos_u2b_midway(start + boffset0,
6723 cache[0] - uoffset0);
6726 + sv_pos_u2b_midway(start + boffset0,
6729 cache[2] - uoffset0);
6733 else if ((*mgp)->mg_len != -1) {
6734 /* If we can take advantage of a passed in offset, do so. */
6735 /* In fact, offset0 is either 0, or less than offset, so don't
6736 need to worry about the other possibility. */
6738 + sv_pos_u2b_midway(start + boffset0, send,
6740 (*mgp)->mg_len - uoffset0);
6745 if (!found || PL_utf8cache < 0) {
6746 STRLEN real_boffset;
6747 uoffset -= uoffset0;
6748 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6749 send, &uoffset, &at_end);
6750 uoffset += uoffset0;
6752 if (found && PL_utf8cache < 0)
6753 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6755 boffset = real_boffset;
6758 if (PL_utf8cache && !SvGMAGICAL(sv) && SvPOK(sv)) {
6760 utf8_mg_len_cache_update(sv, mgp, uoffset);
6762 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6769 =for apidoc sv_pos_u2b_flags
6771 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6772 the start of the string, to a count of the equivalent number of bytes; if
6773 lenp is non-zero, it does the same to lenp, but this time starting from
6774 the offset, rather than from the start
6775 of the string. Handles type coercion.
6776 I<flags> is passed to C<SvPV_flags>, and usually should be
6777 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6783 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6784 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6785 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6790 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6797 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6799 start = (U8*)SvPV_flags(sv, len, flags);
6801 const U8 * const send = start + len;
6803 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6806 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6807 is 0, and *lenp is already set to that. */) {
6808 /* Convert the relative offset to absolute. */
6809 const STRLEN uoffset2 = uoffset + *lenp;
6810 const STRLEN boffset2
6811 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6812 uoffset, boffset) - boffset;
6826 =for apidoc sv_pos_u2b
6828 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6829 the start of the string, to a count of the equivalent number of bytes; if
6830 lenp is non-zero, it does the same to lenp, but this time starting from
6831 the offset, rather than from the start of the string. Handles magic and
6834 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6841 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6842 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6843 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6847 /* This function is subject to size and sign problems */
6850 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6852 PERL_ARGS_ASSERT_SV_POS_U2B;
6855 STRLEN ulen = (STRLEN)*lenp;
6856 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6857 SV_GMAGIC|SV_CONST_RETURN);
6860 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6861 SV_GMAGIC|SV_CONST_RETURN);
6866 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6869 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6870 if (SvREADONLY(sv) || SvGMAGICAL(sv) || !SvPOK(sv))
6873 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6874 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6875 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6879 (*mgp)->mg_len = ulen;
6880 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6881 if (ulen != (STRLEN) (*mgp)->mg_len)
6882 (*mgp)->mg_len = -1;
6885 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6886 byte length pairing. The (byte) length of the total SV is passed in too,
6887 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6888 may not have updated SvCUR, so we can't rely on reading it directly.
6890 The proffered utf8/byte length pairing isn't used if the cache already has
6891 two pairs, and swapping either for the proffered pair would increase the
6892 RMS of the intervals between known byte offsets.
6894 The cache itself consists of 4 STRLEN values
6895 0: larger UTF-8 offset
6896 1: corresponding byte offset
6897 2: smaller UTF-8 offset
6898 3: corresponding byte offset
6900 Unused cache pairs have the value 0, 0.
6901 Keeping the cache "backwards" means that the invariant of
6902 cache[0] >= cache[2] is maintained even with empty slots, which means that
6903 the code that uses it doesn't need to worry if only 1 entry has actually
6904 been set to non-zero. It also makes the "position beyond the end of the
6905 cache" logic much simpler, as the first slot is always the one to start
6909 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6910 const STRLEN utf8, const STRLEN blen)
6914 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6919 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6920 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6921 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6923 (*mgp)->mg_len = -1;
6927 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6928 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6929 (*mgp)->mg_ptr = (char *) cache;
6933 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6934 /* SvPOKp() because it's possible that sv has string overloading, and
6935 therefore is a reference, hence SvPVX() is actually a pointer.
6936 This cures the (very real) symptoms of RT 69422, but I'm not actually
6937 sure whether we should even be caching the results of UTF-8
6938 operations on overloading, given that nothing stops overloading
6939 returning a different value every time it's called. */
6940 const U8 *start = (const U8 *) SvPVX_const(sv);
6941 const STRLEN realutf8 = utf8_length(start, start + byte);
6943 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6947 /* Cache is held with the later position first, to simplify the code
6948 that deals with unbounded ends. */
6950 ASSERT_UTF8_CACHE(cache);
6951 if (cache[1] == 0) {
6952 /* Cache is totally empty */
6955 } else if (cache[3] == 0) {
6956 if (byte > cache[1]) {
6957 /* New one is larger, so goes first. */
6958 cache[2] = cache[0];
6959 cache[3] = cache[1];
6967 #define THREEWAY_SQUARE(a,b,c,d) \
6968 ((float)((d) - (c))) * ((float)((d) - (c))) \
6969 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6970 + ((float)((b) - (a))) * ((float)((b) - (a)))
6972 /* Cache has 2 slots in use, and we know three potential pairs.
6973 Keep the two that give the lowest RMS distance. Do the
6974 calculation in bytes simply because we always know the byte
6975 length. squareroot has the same ordering as the positive value,
6976 so don't bother with the actual square root. */
6977 if (byte > cache[1]) {
6978 /* New position is after the existing pair of pairs. */
6979 const float keep_earlier
6980 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6981 const float keep_later
6982 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6984 if (keep_later < keep_earlier) {
6985 cache[2] = cache[0];
6986 cache[3] = cache[1];
6995 else if (byte > cache[3]) {
6996 /* New position is between the existing pair of pairs. */
6997 const float keep_earlier
6998 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6999 const float keep_later
7000 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7002 if (keep_later < keep_earlier) {
7012 /* New position is before the existing pair of pairs. */
7013 const float keep_earlier
7014 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7015 const float keep_later
7016 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7018 if (keep_later < keep_earlier) {
7023 cache[0] = cache[2];
7024 cache[1] = cache[3];
7030 ASSERT_UTF8_CACHE(cache);
7033 /* We already know all of the way, now we may be able to walk back. The same
7034 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7035 backward is half the speed of walking forward. */
7037 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7038 const U8 *end, STRLEN endu)
7040 const STRLEN forw = target - s;
7041 STRLEN backw = end - target;
7043 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7045 if (forw < 2 * backw) {
7046 return utf8_length(s, target);
7049 while (end > target) {
7051 while (UTF8_IS_CONTINUATION(*end)) {
7060 =for apidoc sv_pos_b2u
7062 Converts the value pointed to by offsetp from a count of bytes from the
7063 start of the string, to a count of the equivalent number of UTF-8 chars.
7064 Handles magic and type coercion.
7070 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7071 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7076 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7079 const STRLEN byte = *offsetp;
7080 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7086 PERL_ARGS_ASSERT_SV_POS_B2U;
7091 s = (const U8*)SvPV_const(sv, blen);
7094 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7095 ", byte=%"UVuf, (UV)blen, (UV)byte);
7101 && SvTYPE(sv) >= SVt_PVMG
7102 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7105 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7106 if (cache[1] == byte) {
7107 /* An exact match. */
7108 *offsetp = cache[0];
7111 if (cache[3] == byte) {
7112 /* An exact match. */
7113 *offsetp = cache[2];
7117 if (cache[1] < byte) {
7118 /* We already know part of the way. */
7119 if (mg->mg_len != -1) {
7120 /* Actually, we know the end too. */
7122 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7123 s + blen, mg->mg_len - cache[0]);
7125 len = cache[0] + utf8_length(s + cache[1], send);
7128 else if (cache[3] < byte) {
7129 /* We're between the two cached pairs, so we do the calculation
7130 offset by the byte/utf-8 positions for the earlier pair,
7131 then add the utf-8 characters from the string start to
7133 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7134 s + cache[1], cache[0] - cache[2])
7138 else { /* cache[3] > byte */
7139 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7143 ASSERT_UTF8_CACHE(cache);
7145 } else if (mg->mg_len != -1) {
7146 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7150 if (!found || PL_utf8cache < 0) {
7151 const STRLEN real_len = utf8_length(s, send);
7153 if (found && PL_utf8cache < 0)
7154 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7161 utf8_mg_len_cache_update(sv, &mg, len);
7163 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7168 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7169 STRLEN real, SV *const sv)
7171 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7173 /* As this is debugging only code, save space by keeping this test here,
7174 rather than inlining it in all the callers. */
7175 if (from_cache == real)
7178 /* Need to turn the assertions off otherwise we may recurse infinitely
7179 while printing error messages. */
7180 SAVEI8(PL_utf8cache);
7182 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7183 func, (UV) from_cache, (UV) real, SVfARG(sv));
7189 Returns a boolean indicating whether the strings in the two SVs are
7190 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7191 coerce its args to strings if necessary.
7193 =for apidoc sv_eq_flags
7195 Returns a boolean indicating whether the strings in the two SVs are
7196 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7197 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7203 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7211 SV* svrecode = NULL;
7218 /* if pv1 and pv2 are the same, second SvPV_const call may
7219 * invalidate pv1 (if we are handling magic), so we may need to
7221 if (sv1 == sv2 && flags & SV_GMAGIC
7222 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7223 pv1 = SvPV_const(sv1, cur1);
7224 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7226 pv1 = SvPV_flags_const(sv1, cur1, flags);
7234 pv2 = SvPV_flags_const(sv2, cur2, flags);
7236 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7237 /* Differing utf8ness.
7238 * Do not UTF8size the comparands as a side-effect. */
7241 svrecode = newSVpvn(pv2, cur2);
7242 sv_recode_to_utf8(svrecode, PL_encoding);
7243 pv2 = SvPV_const(svrecode, cur2);
7246 svrecode = newSVpvn(pv1, cur1);
7247 sv_recode_to_utf8(svrecode, PL_encoding);
7248 pv1 = SvPV_const(svrecode, cur1);
7250 /* Now both are in UTF-8. */
7252 SvREFCNT_dec(svrecode);
7258 /* sv1 is the UTF-8 one */
7259 return bytes_cmp_utf8((const U8*)pv2, cur2,
7260 (const U8*)pv1, cur1) == 0;
7263 /* sv2 is the UTF-8 one */
7264 return bytes_cmp_utf8((const U8*)pv1, cur1,
7265 (const U8*)pv2, cur2) == 0;
7271 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7273 SvREFCNT_dec(svrecode);
7281 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7282 string in C<sv1> is less than, equal to, or greater than the string in
7283 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7284 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7286 =for apidoc sv_cmp_flags
7288 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7289 string in C<sv1> is less than, equal to, or greater than the string in
7290 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7291 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7292 also C<sv_cmp_locale_flags>.
7298 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7300 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7304 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7309 const char *pv1, *pv2;
7312 SV *svrecode = NULL;
7319 pv1 = SvPV_flags_const(sv1, cur1, flags);
7326 pv2 = SvPV_flags_const(sv2, cur2, flags);
7328 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7329 /* Differing utf8ness.
7330 * Do not UTF8size the comparands as a side-effect. */
7333 svrecode = newSVpvn(pv2, cur2);
7334 sv_recode_to_utf8(svrecode, PL_encoding);
7335 pv2 = SvPV_const(svrecode, cur2);
7338 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7339 (const U8*)pv1, cur1);
7340 return retval ? retval < 0 ? -1 : +1 : 0;
7345 svrecode = newSVpvn(pv1, cur1);
7346 sv_recode_to_utf8(svrecode, PL_encoding);
7347 pv1 = SvPV_const(svrecode, cur1);
7350 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7351 (const U8*)pv2, cur2);
7352 return retval ? retval < 0 ? -1 : +1 : 0;
7358 cmp = cur2 ? -1 : 0;
7362 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7365 cmp = retval < 0 ? -1 : 1;
7366 } else if (cur1 == cur2) {
7369 cmp = cur1 < cur2 ? -1 : 1;
7373 SvREFCNT_dec(svrecode);
7381 =for apidoc sv_cmp_locale
7383 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7384 'use bytes' aware, handles get magic, and will coerce its args to strings
7385 if necessary. See also C<sv_cmp>.
7387 =for apidoc sv_cmp_locale_flags
7389 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7390 'use bytes' aware and will coerce its args to strings if necessary. If the
7391 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7397 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7399 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7403 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7407 #ifdef USE_LOCALE_COLLATE
7413 if (PL_collation_standard)
7417 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7419 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7421 if (!pv1 || !len1) {
7432 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7435 return retval < 0 ? -1 : 1;
7438 * When the result of collation is equality, that doesn't mean
7439 * that there are no differences -- some locales exclude some
7440 * characters from consideration. So to avoid false equalities,
7441 * we use the raw string as a tiebreaker.
7447 #endif /* USE_LOCALE_COLLATE */
7449 return sv_cmp(sv1, sv2);
7453 #ifdef USE_LOCALE_COLLATE
7456 =for apidoc sv_collxfrm
7458 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7459 C<sv_collxfrm_flags>.
7461 =for apidoc sv_collxfrm_flags
7463 Add Collate Transform magic to an SV if it doesn't already have it. If the
7464 flags contain SV_GMAGIC, it handles get-magic.
7466 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7467 scalar data of the variable, but transformed to such a format that a normal
7468 memory comparison can be used to compare the data according to the locale
7475 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7480 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7482 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7483 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7489 Safefree(mg->mg_ptr);
7490 s = SvPV_flags_const(sv, len, flags);
7491 if ((xf = mem_collxfrm(s, len, &xlen))) {
7493 #ifdef PERL_OLD_COPY_ON_WRITE
7495 sv_force_normal_flags(sv, 0);
7497 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7511 if (mg && mg->mg_ptr) {
7513 return mg->mg_ptr + sizeof(PL_collation_ix);
7521 #endif /* USE_LOCALE_COLLATE */
7524 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7526 SV * const tsv = newSV(0);
7529 sv_gets(tsv, fp, 0);
7530 sv_utf8_upgrade_nomg(tsv);
7531 SvCUR_set(sv,append);
7534 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7538 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7541 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7542 /* Grab the size of the record we're getting */
7543 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7550 /* VMS wants read instead of fread, because fread doesn't respect */
7551 /* RMS record boundaries. This is not necessarily a good thing to be */
7552 /* doing, but we've got no other real choice - except avoid stdio
7553 as implementation - perhaps write a :vms layer ?
7555 fd = PerlIO_fileno(fp);
7557 bytesread = PerlLIO_read(fd, buffer, recsize);
7559 else /* in-memory file from PerlIO::Scalar */
7562 bytesread = PerlIO_read(fp, buffer, recsize);
7567 SvCUR_set(sv, bytesread + append);
7568 buffer[bytesread] = '\0';
7569 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7575 Get a line from the filehandle and store it into the SV, optionally
7576 appending to the currently-stored string. If C<append> is not 0, the
7577 line is appended to the SV instead of overwriting it. C<append> should
7578 be set to the byte offset that the appended string should start at
7579 in the SV (typically, C<SvCUR(sv)> is a suitable choice).
7585 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7596 PERL_ARGS_ASSERT_SV_GETS;
7598 if (SvTHINKFIRST(sv))
7599 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7600 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7602 However, perlbench says it's slower, because the existing swipe code
7603 is faster than copy on write.
7604 Swings and roundabouts. */
7605 SvUPGRADE(sv, SVt_PV);
7608 if (PerlIO_isutf8(fp)) {
7610 sv_utf8_upgrade_nomg(sv);
7611 sv_pos_u2b(sv,&append,0);
7613 } else if (SvUTF8(sv)) {
7614 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7622 if (PerlIO_isutf8(fp))
7625 if (IN_PERL_COMPILETIME) {
7626 /* we always read code in line mode */
7630 else if (RsSNARF(PL_rs)) {
7631 /* If it is a regular disk file use size from stat() as estimate
7632 of amount we are going to read -- may result in mallocing
7633 more memory than we really need if the layers below reduce
7634 the size we read (e.g. CRLF or a gzip layer).
7637 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7638 const Off_t offset = PerlIO_tell(fp);
7639 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7640 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7646 else if (RsRECORD(PL_rs)) {
7647 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7649 else if (RsPARA(PL_rs)) {
7655 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7656 if (PerlIO_isutf8(fp)) {
7657 rsptr = SvPVutf8(PL_rs, rslen);
7660 if (SvUTF8(PL_rs)) {
7661 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7662 Perl_croak(aTHX_ "Wide character in $/");
7665 rsptr = SvPV_const(PL_rs, rslen);
7669 rslast = rslen ? rsptr[rslen - 1] : '\0';
7671 if (rspara) { /* have to do this both before and after */
7672 do { /* to make sure file boundaries work right */
7675 i = PerlIO_getc(fp);
7679 PerlIO_ungetc(fp,i);
7685 /* See if we know enough about I/O mechanism to cheat it ! */
7687 /* This used to be #ifdef test - it is made run-time test for ease
7688 of abstracting out stdio interface. One call should be cheap
7689 enough here - and may even be a macro allowing compile
7693 if (PerlIO_fast_gets(fp)) {
7696 * We're going to steal some values from the stdio struct
7697 * and put EVERYTHING in the innermost loop into registers.
7703 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7704 /* An ungetc()d char is handled separately from the regular
7705 * buffer, so we getc() it back out and stuff it in the buffer.
7707 i = PerlIO_getc(fp);
7708 if (i == EOF) return 0;
7709 *(--((*fp)->_ptr)) = (unsigned char) i;
7713 /* Here is some breathtakingly efficient cheating */
7715 cnt = PerlIO_get_cnt(fp); /* get count into register */
7716 /* make sure we have the room */
7717 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7718 /* Not room for all of it
7719 if we are looking for a separator and room for some
7721 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7722 /* just process what we have room for */
7723 shortbuffered = cnt - SvLEN(sv) + append + 1;
7724 cnt -= shortbuffered;
7728 /* remember that cnt can be negative */
7729 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7734 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7735 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7736 DEBUG_P(PerlIO_printf(Perl_debug_log,
7737 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7738 DEBUG_P(PerlIO_printf(Perl_debug_log,
7739 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7740 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7741 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7746 while (cnt > 0) { /* this | eat */
7748 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7749 goto thats_all_folks; /* screams | sed :-) */
7753 Copy(ptr, bp, cnt, char); /* this | eat */
7754 bp += cnt; /* screams | dust */
7755 ptr += cnt; /* louder | sed :-) */
7757 assert (!shortbuffered);
7758 goto cannot_be_shortbuffered;
7762 if (shortbuffered) { /* oh well, must extend */
7763 cnt = shortbuffered;
7765 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7767 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7768 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7772 cannot_be_shortbuffered:
7773 DEBUG_P(PerlIO_printf(Perl_debug_log,
7774 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7775 PTR2UV(ptr),(long)cnt));
7776 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7778 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7779 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7780 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7781 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7783 /* This used to call 'filbuf' in stdio form, but as that behaves like
7784 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7785 another abstraction. */
7786 i = PerlIO_getc(fp); /* get more characters */
7788 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7789 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7790 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7791 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7793 cnt = PerlIO_get_cnt(fp);
7794 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7795 DEBUG_P(PerlIO_printf(Perl_debug_log,
7796 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7798 if (i == EOF) /* all done for ever? */
7799 goto thats_really_all_folks;
7801 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7803 SvGROW(sv, bpx + cnt + 2);
7804 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7806 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7808 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7809 goto thats_all_folks;
7813 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7814 memNE((char*)bp - rslen, rsptr, rslen))
7815 goto screamer; /* go back to the fray */
7816 thats_really_all_folks:
7818 cnt += shortbuffered;
7819 DEBUG_P(PerlIO_printf(Perl_debug_log,
7820 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7821 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7822 DEBUG_P(PerlIO_printf(Perl_debug_log,
7823 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7824 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7825 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7827 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7828 DEBUG_P(PerlIO_printf(Perl_debug_log,
7829 "Screamer: done, len=%ld, string=|%.*s|\n",
7830 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7834 /*The big, slow, and stupid way. */
7835 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7836 STDCHAR *buf = NULL;
7837 Newx(buf, 8192, STDCHAR);
7845 const STDCHAR * const bpe = buf + sizeof(buf);
7847 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7848 ; /* keep reading */
7852 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7853 /* Accommodate broken VAXC compiler, which applies U8 cast to
7854 * both args of ?: operator, causing EOF to change into 255
7857 i = (U8)buf[cnt - 1];
7863 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7865 sv_catpvn_nomg(sv, (char *) buf, cnt);
7867 sv_setpvn(sv, (char *) buf, cnt); /* "nomg" is implied */
7869 if (i != EOF && /* joy */
7871 SvCUR(sv) < rslen ||
7872 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7876 * If we're reading from a TTY and we get a short read,
7877 * indicating that the user hit his EOF character, we need
7878 * to notice it now, because if we try to read from the TTY
7879 * again, the EOF condition will disappear.
7881 * The comparison of cnt to sizeof(buf) is an optimization
7882 * that prevents unnecessary calls to feof().
7886 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7890 #ifdef USE_HEAP_INSTEAD_OF_STACK
7895 if (rspara) { /* have to do this both before and after */
7896 while (i != EOF) { /* to make sure file boundaries work right */
7897 i = PerlIO_getc(fp);
7899 PerlIO_ungetc(fp,i);
7905 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7911 Auto-increment of the value in the SV, doing string to numeric conversion
7912 if necessary. Handles 'get' magic and operator overloading.
7918 Perl_sv_inc(pTHX_ register SV *const sv)
7927 =for apidoc sv_inc_nomg
7929 Auto-increment of the value in the SV, doing string to numeric conversion
7930 if necessary. Handles operator overloading. Skips handling 'get' magic.
7936 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7944 if (SvTHINKFIRST(sv)) {
7945 if (SvIsCOW(sv) || isGV_with_GP(sv))
7946 sv_force_normal_flags(sv, 0);
7947 if (SvREADONLY(sv)) {
7948 if (IN_PERL_RUNTIME)
7949 Perl_croak_no_modify();
7953 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7955 i = PTR2IV(SvRV(sv));
7960 flags = SvFLAGS(sv);
7961 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7962 /* It's (privately or publicly) a float, but not tested as an
7963 integer, so test it to see. */
7965 flags = SvFLAGS(sv);
7967 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7968 /* It's publicly an integer, or privately an integer-not-float */
7969 #ifdef PERL_PRESERVE_IVUV
7973 if (SvUVX(sv) == UV_MAX)
7974 sv_setnv(sv, UV_MAX_P1);
7976 (void)SvIOK_only_UV(sv);
7977 SvUV_set(sv, SvUVX(sv) + 1);
7979 if (SvIVX(sv) == IV_MAX)
7980 sv_setuv(sv, (UV)IV_MAX + 1);
7982 (void)SvIOK_only(sv);
7983 SvIV_set(sv, SvIVX(sv) + 1);
7988 if (flags & SVp_NOK) {
7989 const NV was = SvNVX(sv);
7990 if (NV_OVERFLOWS_INTEGERS_AT &&
7991 was >= NV_OVERFLOWS_INTEGERS_AT) {
7992 /* diag_listed_as: Lost precision when %s %f by 1 */
7993 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7994 "Lost precision when incrementing %" NVff " by 1",
7997 (void)SvNOK_only(sv);
7998 SvNV_set(sv, was + 1.0);
8002 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
8003 if ((flags & SVTYPEMASK) < SVt_PVIV)
8004 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
8005 (void)SvIOK_only(sv);
8010 while (isALPHA(*d)) d++;
8011 while (isDIGIT(*d)) d++;
8012 if (d < SvEND(sv)) {
8013 #ifdef PERL_PRESERVE_IVUV
8014 /* Got to punt this as an integer if needs be, but we don't issue
8015 warnings. Probably ought to make the sv_iv_please() that does
8016 the conversion if possible, and silently. */
8017 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8018 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8019 /* Need to try really hard to see if it's an integer.
8020 9.22337203685478e+18 is an integer.
8021 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8022 so $a="9.22337203685478e+18"; $a+0; $a++
8023 needs to be the same as $a="9.22337203685478e+18"; $a++
8030 /* sv_2iv *should* have made this an NV */
8031 if (flags & SVp_NOK) {
8032 (void)SvNOK_only(sv);
8033 SvNV_set(sv, SvNVX(sv) + 1.0);
8036 /* I don't think we can get here. Maybe I should assert this
8037 And if we do get here I suspect that sv_setnv will croak. NWC
8039 #if defined(USE_LONG_DOUBLE)
8040 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",
8041 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8043 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8044 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8047 #endif /* PERL_PRESERVE_IVUV */
8048 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8052 while (d >= SvPVX_const(sv)) {
8060 /* MKS: The original code here died if letters weren't consecutive.
8061 * at least it didn't have to worry about non-C locales. The
8062 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8063 * arranged in order (although not consecutively) and that only
8064 * [A-Za-z] are accepted by isALPHA in the C locale.
8066 if (*d != 'z' && *d != 'Z') {
8067 do { ++*d; } while (!isALPHA(*d));
8070 *(d--) -= 'z' - 'a';
8075 *(d--) -= 'z' - 'a' + 1;
8079 /* oh,oh, the number grew */
8080 SvGROW(sv, SvCUR(sv) + 2);
8081 SvCUR_set(sv, SvCUR(sv) + 1);
8082 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8093 Auto-decrement of the value in the SV, doing string to numeric conversion
8094 if necessary. Handles 'get' magic and operator overloading.
8100 Perl_sv_dec(pTHX_ register SV *const sv)
8110 =for apidoc sv_dec_nomg
8112 Auto-decrement of the value in the SV, doing string to numeric conversion
8113 if necessary. Handles operator overloading. Skips handling 'get' magic.
8119 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8126 if (SvTHINKFIRST(sv)) {
8127 if (SvIsCOW(sv) || isGV_with_GP(sv))
8128 sv_force_normal_flags(sv, 0);
8129 if (SvREADONLY(sv)) {
8130 if (IN_PERL_RUNTIME)
8131 Perl_croak_no_modify();
8135 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8137 i = PTR2IV(SvRV(sv));
8142 /* Unlike sv_inc we don't have to worry about string-never-numbers
8143 and keeping them magic. But we mustn't warn on punting */
8144 flags = SvFLAGS(sv);
8145 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8146 /* It's publicly an integer, or privately an integer-not-float */
8147 #ifdef PERL_PRESERVE_IVUV
8151 if (SvUVX(sv) == 0) {
8152 (void)SvIOK_only(sv);
8156 (void)SvIOK_only_UV(sv);
8157 SvUV_set(sv, SvUVX(sv) - 1);
8160 if (SvIVX(sv) == IV_MIN) {
8161 sv_setnv(sv, (NV)IV_MIN);
8165 (void)SvIOK_only(sv);
8166 SvIV_set(sv, SvIVX(sv) - 1);
8171 if (flags & SVp_NOK) {
8174 const NV was = SvNVX(sv);
8175 if (NV_OVERFLOWS_INTEGERS_AT &&
8176 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8177 /* diag_listed_as: Lost precision when %s %f by 1 */
8178 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8179 "Lost precision when decrementing %" NVff " by 1",
8182 (void)SvNOK_only(sv);
8183 SvNV_set(sv, was - 1.0);
8187 if (!(flags & SVp_POK)) {
8188 if ((flags & SVTYPEMASK) < SVt_PVIV)
8189 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8191 (void)SvIOK_only(sv);
8194 #ifdef PERL_PRESERVE_IVUV
8196 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8197 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8198 /* Need to try really hard to see if it's an integer.
8199 9.22337203685478e+18 is an integer.
8200 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8201 so $a="9.22337203685478e+18"; $a+0; $a--
8202 needs to be the same as $a="9.22337203685478e+18"; $a--
8209 /* sv_2iv *should* have made this an NV */
8210 if (flags & SVp_NOK) {
8211 (void)SvNOK_only(sv);
8212 SvNV_set(sv, SvNVX(sv) - 1.0);
8215 /* I don't think we can get here. Maybe I should assert this
8216 And if we do get here I suspect that sv_setnv will croak. NWC
8218 #if defined(USE_LONG_DOUBLE)
8219 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",
8220 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8222 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8223 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8227 #endif /* PERL_PRESERVE_IVUV */
8228 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8231 /* this define is used to eliminate a chunk of duplicated but shared logic
8232 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8233 * used anywhere but here - yves
8235 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8238 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8242 =for apidoc sv_mortalcopy
8244 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8245 The new SV is marked as mortal. It will be destroyed "soon", either by an
8246 explicit call to FREETMPS, or by an implicit call at places such as
8247 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8252 /* Make a string that will exist for the duration of the expression
8253 * evaluation. Actually, it may have to last longer than that, but
8254 * hopefully we won't free it until it has been assigned to a
8255 * permanent location. */
8258 Perl_sv_mortalcopy_flags(pTHX_ SV *const oldstr, U32 flags)
8263 if (flags & SV_GMAGIC)
8264 SvGETMAGIC(oldstr); /* before new_SV, in case it dies */
8266 sv_setsv_flags(sv,oldstr,flags & ~SV_GMAGIC);
8267 PUSH_EXTEND_MORTAL__SV_C(sv);
8273 =for apidoc sv_newmortal
8275 Creates a new null SV which is mortal. The reference count of the SV is
8276 set to 1. It will be destroyed "soon", either by an explicit call to
8277 FREETMPS, or by an implicit call at places such as statement boundaries.
8278 See also C<sv_mortalcopy> and C<sv_2mortal>.
8284 Perl_sv_newmortal(pTHX)
8290 SvFLAGS(sv) = SVs_TEMP;
8291 PUSH_EXTEND_MORTAL__SV_C(sv);
8297 =for apidoc newSVpvn_flags
8299 Creates a new SV and copies a string into it. The reference count for the
8300 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8301 string. You are responsible for ensuring that the source string is at least
8302 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8303 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8304 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8305 returning. If C<SVf_UTF8> is set, C<s>
8306 is considered to be in UTF-8 and the
8307 C<SVf_UTF8> flag will be set on the new SV.
8308 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8310 #define newSVpvn_utf8(s, len, u) \
8311 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8317 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8322 /* All the flags we don't support must be zero.
8323 And we're new code so I'm going to assert this from the start. */
8324 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8326 sv_setpvn(sv,s,len);
8328 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8329 * and do what it does ourselves here.
8330 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8331 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8332 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8333 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8336 SvFLAGS(sv) |= flags;
8338 if(flags & SVs_TEMP){
8339 PUSH_EXTEND_MORTAL__SV_C(sv);
8346 =for apidoc sv_2mortal
8348 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8349 by an explicit call to FREETMPS, or by an implicit call at places such as
8350 statement boundaries. SvTEMP() is turned on which means that the SV's
8351 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8352 and C<sv_mortalcopy>.
8358 Perl_sv_2mortal(pTHX_ register SV *const sv)
8363 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8365 PUSH_EXTEND_MORTAL__SV_C(sv);
8373 Creates a new SV and copies a string into it. The reference count for the
8374 SV is set to 1. If C<len> is zero, Perl will compute the length using
8375 strlen(). For efficiency, consider using C<newSVpvn> instead.
8381 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8387 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8392 =for apidoc newSVpvn
8394 Creates a new SV and copies a buffer into it, which may contain NUL characters
8395 (C<\0>) and other binary data. The reference count for the SV is set to 1.
8396 Note that if C<len> is zero, Perl will create a zero length (Perl) string. You
8397 are responsible for ensuring that the source buffer is at least
8398 C<len> bytes long. If the C<buffer> argument is NULL the new SV will be
8405 Perl_newSVpvn(pTHX_ const char *const buffer, const STRLEN len)
8411 sv_setpvn(sv,buffer,len);
8416 =for apidoc newSVhek
8418 Creates a new SV from the hash key structure. It will generate scalars that
8419 point to the shared string table where possible. Returns a new (undefined)
8420 SV if the hek is NULL.
8426 Perl_newSVhek(pTHX_ const HEK *const hek)
8436 if (HEK_LEN(hek) == HEf_SVKEY) {
8437 return newSVsv(*(SV**)HEK_KEY(hek));
8439 const int flags = HEK_FLAGS(hek);
8440 if (flags & HVhek_WASUTF8) {
8442 Andreas would like keys he put in as utf8 to come back as utf8
8444 STRLEN utf8_len = HEK_LEN(hek);
8445 SV * const sv = newSV_type(SVt_PV);
8446 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8447 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8448 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8451 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8452 /* We don't have a pointer to the hv, so we have to replicate the
8453 flag into every HEK. This hv is using custom a hasing
8454 algorithm. Hence we can't return a shared string scalar, as
8455 that would contain the (wrong) hash value, and might get passed
8456 into an hv routine with a regular hash.
8457 Similarly, a hash that isn't using shared hash keys has to have
8458 the flag in every key so that we know not to try to call
8459 share_hek_hek on it. */
8461 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8466 /* This will be overwhelminly the most common case. */
8468 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8469 more efficient than sharepvn(). */
8473 sv_upgrade(sv, SVt_PV);
8474 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8475 SvCUR_set(sv, HEK_LEN(hek));
8487 =for apidoc newSVpvn_share
8489 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8490 table. If the string does not already exist in the table, it is
8491 created first. Turns on READONLY and FAKE. If the C<hash> parameter
8492 is non-zero, that value is used; otherwise the hash is computed.
8493 The string's hash can later be retrieved from the SV
8494 with the C<SvSHARED_HASH()> macro. The idea here is
8495 that as the string table is used for shared hash keys these strings will have
8496 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8502 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8506 bool is_utf8 = FALSE;
8507 const char *const orig_src = src;
8510 STRLEN tmplen = -len;
8512 /* See the note in hv.c:hv_fetch() --jhi */
8513 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8517 PERL_HASH(hash, src, len);
8519 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8520 changes here, update it there too. */
8521 sv_upgrade(sv, SVt_PV);
8522 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8529 if (src != orig_src)
8535 =for apidoc newSVpv_share
8537 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8544 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8546 return newSVpvn_share(src, strlen(src), hash);
8549 #if defined(PERL_IMPLICIT_CONTEXT)
8551 /* pTHX_ magic can't cope with varargs, so this is a no-context
8552 * version of the main function, (which may itself be aliased to us).
8553 * Don't access this version directly.
8557 Perl_newSVpvf_nocontext(const char *const pat, ...)
8563 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8565 va_start(args, pat);
8566 sv = vnewSVpvf(pat, &args);
8573 =for apidoc newSVpvf
8575 Creates a new SV and initializes it with the string formatted like
8582 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8587 PERL_ARGS_ASSERT_NEWSVPVF;
8589 va_start(args, pat);
8590 sv = vnewSVpvf(pat, &args);
8595 /* backend for newSVpvf() and newSVpvf_nocontext() */
8598 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8603 PERL_ARGS_ASSERT_VNEWSVPVF;
8606 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8613 Creates a new SV and copies a floating point value into it.
8614 The reference count for the SV is set to 1.
8620 Perl_newSVnv(pTHX_ const NV n)
8633 Creates a new SV and copies an integer into it. The reference count for the
8640 Perl_newSViv(pTHX_ const IV i)
8653 Creates a new SV and copies an unsigned integer into it.
8654 The reference count for the SV is set to 1.
8660 Perl_newSVuv(pTHX_ const UV u)
8671 =for apidoc newSV_type
8673 Creates a new SV, of the type specified. The reference count for the new SV
8680 Perl_newSV_type(pTHX_ const svtype type)
8685 sv_upgrade(sv, type);
8690 =for apidoc newRV_noinc
8692 Creates an RV wrapper for an SV. The reference count for the original
8693 SV is B<not> incremented.
8699 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8702 SV *sv = newSV_type(SVt_IV);
8704 PERL_ARGS_ASSERT_NEWRV_NOINC;
8707 SvRV_set(sv, tmpRef);
8712 /* newRV_inc is the official function name to use now.
8713 * newRV_inc is in fact #defined to newRV in sv.h
8717 Perl_newRV(pTHX_ SV *const sv)
8721 PERL_ARGS_ASSERT_NEWRV;
8723 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8729 Creates a new SV which is an exact duplicate of the original SV.
8736 Perl_newSVsv(pTHX_ register SV *const old)
8743 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8744 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8747 /* Do this here, otherwise we leak the new SV if this croaks. */
8750 /* SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8751 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8752 sv_setsv_flags(sv, old, SV_NOSTEAL);
8757 =for apidoc sv_reset
8759 Underlying implementation for the C<reset> Perl function.
8760 Note that the perl-level function is vaguely deprecated.
8766 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8768 PERL_ARGS_ASSERT_SV_RESET;
8770 sv_resetpvn(*s ? s : NULL, strlen(s), stash);
8774 Perl_sv_resetpvn(pTHX_ const char *s, STRLEN len, HV * const stash)
8777 char todo[PERL_UCHAR_MAX+1];
8783 if (!s) { /* reset ?? searches */
8784 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8786 const U32 count = mg->mg_len / sizeof(PMOP**);
8787 PMOP **pmp = (PMOP**) mg->mg_ptr;
8788 PMOP *const *const end = pmp + count;
8792 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8794 (*pmp)->op_pmflags &= ~PMf_USED;
8802 /* reset variables */
8804 if (!HvARRAY(stash))
8807 Zero(todo, 256, char);
8811 I32 i = (unsigned char)*s;
8815 max = (unsigned char)*s++;
8816 for ( ; i <= max; i++) {
8819 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8821 for (entry = HvARRAY(stash)[i];
8823 entry = HeNEXT(entry))
8828 if (!todo[(U8)*HeKEY(entry)])
8830 gv = MUTABLE_GV(HeVAL(entry));
8833 if (SvTHINKFIRST(sv)) {
8834 if (!SvREADONLY(sv) && SvROK(sv))
8836 /* XXX Is this continue a bug? Why should THINKFIRST
8837 exempt us from resetting arrays and hashes? */
8841 if (SvTYPE(sv) >= SVt_PV) {
8843 if (SvPVX_const(sv) != NULL)
8851 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8853 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8856 # if defined(USE_ENVIRON_ARRAY)
8859 # endif /* USE_ENVIRON_ARRAY */
8870 Using various gambits, try to get an IO from an SV: the IO slot if its a
8871 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8872 named after the PV if we're a string.
8874 'Get' magic is ignored on the sv passed in, but will be called on
8875 C<SvRV(sv)> if sv is an RV.
8881 Perl_sv_2io(pTHX_ SV *const sv)
8886 PERL_ARGS_ASSERT_SV_2IO;
8888 switch (SvTYPE(sv)) {
8890 io = MUTABLE_IO(sv);
8894 if (isGV_with_GP(sv)) {
8895 gv = MUTABLE_GV(sv);
8898 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8899 HEKfARG(GvNAME_HEK(gv)));
8905 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8907 SvGETMAGIC(SvRV(sv));
8908 return sv_2io(SvRV(sv));
8910 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
8917 if (SvGMAGICAL(sv)) {
8918 newsv = sv_newmortal();
8919 sv_setsv_nomg(newsv, sv);
8921 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
8931 Using various gambits, try to get a CV from an SV; in addition, try if
8932 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8933 The flags in C<lref> are passed to gv_fetchsv.
8939 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8945 PERL_ARGS_ASSERT_SV_2CV;
8952 switch (SvTYPE(sv)) {
8956 return MUTABLE_CV(sv);
8966 sv = amagic_deref_call(sv, to_cv_amg);
8969 if (SvTYPE(sv) == SVt_PVCV) {
8970 cv = MUTABLE_CV(sv);
8975 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
8976 gv = MUTABLE_GV(sv);
8978 Perl_croak(aTHX_ "Not a subroutine reference");
8980 else if (isGV_with_GP(sv)) {
8981 gv = MUTABLE_GV(sv);
8984 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8991 /* Some flags to gv_fetchsv mean don't really create the GV */
8992 if (!isGV_with_GP(gv)) {
8997 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8998 /* XXX this is probably not what they think they're getting.
8999 * It has the same effect as "sub name;", i.e. just a forward
9010 Returns true if the SV has a true value by Perl's rules.
9011 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9012 instead use an in-line version.
9018 Perl_sv_true(pTHX_ register SV *const sv)
9023 const XPV* const tXpv = (XPV*)SvANY(sv);
9025 (tXpv->xpv_cur > 1 ||
9026 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9033 return SvIVX(sv) != 0;
9036 return SvNVX(sv) != 0.0;
9038 return sv_2bool(sv);
9044 =for apidoc sv_pvn_force
9046 Get a sensible string out of the SV somehow.
9047 A private implementation of the C<SvPV_force> macro for compilers which
9048 can't cope with complex macro expressions. Always use the macro instead.
9050 =for apidoc sv_pvn_force_flags
9052 Get a sensible string out of the SV somehow.
9053 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9054 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9055 implemented in terms of this function.
9056 You normally want to use the various wrapper macros instead: see
9057 C<SvPV_force> and C<SvPV_force_nomg>
9063 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9067 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9069 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9070 if (SvTHINKFIRST(sv) && !SvROK(sv))
9071 sv_force_normal_flags(sv, 0);
9081 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9082 const char * const ref = sv_reftype(sv,0);
9084 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9085 ref, OP_DESC(PL_op));
9087 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9089 if (SvTYPE(sv) > SVt_PVLV
9090 || isGV_with_GP(sv))
9091 /* diag_listed_as: Can't coerce %s to %s in %s */
9092 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9094 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9101 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9104 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9105 SvGROW(sv, len + 1);
9106 Move(s,SvPVX(sv),len,char);
9108 SvPVX(sv)[len] = '\0';
9111 SvPOK_on(sv); /* validate pointer */
9113 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9114 PTR2UV(sv),SvPVX_const(sv)));
9117 (void)SvPOK_only_UTF8(sv);
9118 return SvPVX_mutable(sv);
9122 =for apidoc sv_pvbyten_force
9124 The backend for the C<SvPVbytex_force> macro. Always use the macro
9131 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9133 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9135 sv_pvn_force(sv,lp);
9136 sv_utf8_downgrade(sv,0);
9142 =for apidoc sv_pvutf8n_force
9144 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9151 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9153 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9156 sv_utf8_upgrade_nomg(sv);
9162 =for apidoc sv_reftype
9164 Returns a string describing what the SV is a reference to.
9170 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9172 PERL_ARGS_ASSERT_SV_REFTYPE;
9173 if (ob && SvOBJECT(sv)) {
9174 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9177 switch (SvTYPE(sv)) {
9192 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9193 /* tied lvalues should appear to be
9194 * scalars for backwards compatibility */
9195 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9196 ? "SCALAR" : "LVALUE");
9197 case SVt_PVAV: return "ARRAY";
9198 case SVt_PVHV: return "HASH";
9199 case SVt_PVCV: return "CODE";
9200 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9201 ? "GLOB" : "SCALAR");
9202 case SVt_PVFM: return "FORMAT";
9203 case SVt_PVIO: return "IO";
9204 case SVt_BIND: return "BIND";
9205 case SVt_REGEXP: return "REGEXP";
9206 default: return "UNKNOWN";
9214 Returns a SV describing what the SV passed in is a reference to.
9220 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9222 PERL_ARGS_ASSERT_SV_REF;
9225 dst = sv_newmortal();
9227 if (ob && SvOBJECT(sv)) {
9228 HvNAME_get(SvSTASH(sv))
9229 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9230 : sv_setpvn(dst, "__ANON__", 8);
9233 const char * reftype = sv_reftype(sv, 0);
9234 sv_setpv(dst, reftype);
9240 =for apidoc sv_isobject
9242 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9243 object. If the SV is not an RV, or if the object is not blessed, then this
9250 Perl_sv_isobject(pTHX_ SV *sv)
9266 Returns a boolean indicating whether the SV is blessed into the specified
9267 class. This does not check for subtypes; use C<sv_derived_from> to verify
9268 an inheritance relationship.
9274 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9278 PERL_ARGS_ASSERT_SV_ISA;
9288 hvname = HvNAME_get(SvSTASH(sv));
9292 return strEQ(hvname, name);
9298 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9299 it will be upgraded to one. If C<classname> is non-null then the new SV will
9300 be blessed in the specified package. The new SV is returned and its
9301 reference count is 1.
9307 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9312 PERL_ARGS_ASSERT_NEWSVRV;
9316 SV_CHECK_THINKFIRST_COW_DROP(rv);
9318 if (SvTYPE(rv) >= SVt_PVMG) {
9319 const U32 refcnt = SvREFCNT(rv);
9323 SvREFCNT(rv) = refcnt;
9325 sv_upgrade(rv, SVt_IV);
9326 } else if (SvROK(rv)) {
9327 SvREFCNT_dec(SvRV(rv));
9329 prepare_SV_for_RV(rv);
9337 HV* const stash = gv_stashpv(classname, GV_ADD);
9338 (void)sv_bless(rv, stash);
9344 =for apidoc sv_setref_pv
9346 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9347 argument will be upgraded to an RV. That RV will be modified to point to
9348 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9349 into the SV. The C<classname> argument indicates the package for the
9350 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9351 will have a reference count of 1, and the RV will be returned.
9353 Do not use with other Perl types such as HV, AV, SV, CV, because those
9354 objects will become corrupted by the pointer copy process.
9356 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9362 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9366 PERL_ARGS_ASSERT_SV_SETREF_PV;
9369 sv_setsv(rv, &PL_sv_undef);
9373 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9378 =for apidoc sv_setref_iv
9380 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9381 argument will be upgraded to an RV. That RV will be modified to point to
9382 the new SV. The C<classname> argument indicates the package for the
9383 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9384 will have a reference count of 1, and the RV will be returned.
9390 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9392 PERL_ARGS_ASSERT_SV_SETREF_IV;
9394 sv_setiv(newSVrv(rv,classname), iv);
9399 =for apidoc sv_setref_uv
9401 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9402 argument will be upgraded to an RV. That RV will be modified to point to
9403 the new SV. The C<classname> argument indicates the package for the
9404 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9405 will have a reference count of 1, and the RV will be returned.
9411 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9413 PERL_ARGS_ASSERT_SV_SETREF_UV;
9415 sv_setuv(newSVrv(rv,classname), uv);
9420 =for apidoc sv_setref_nv
9422 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9423 argument will be upgraded to an RV. That RV will be modified to point to
9424 the new SV. The C<classname> argument indicates the package for the
9425 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9426 will have a reference count of 1, and the RV will be returned.
9432 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9434 PERL_ARGS_ASSERT_SV_SETREF_NV;
9436 sv_setnv(newSVrv(rv,classname), nv);
9441 =for apidoc sv_setref_pvn
9443 Copies a string into a new SV, optionally blessing the SV. The length of the
9444 string must be specified with C<n>. The C<rv> argument will be upgraded to
9445 an RV. That RV will be modified to point to the new SV. The C<classname>
9446 argument indicates the package for the blessing. Set C<classname> to
9447 C<NULL> to avoid the blessing. The new SV will have a reference count
9448 of 1, and the RV will be returned.
9450 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9456 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9457 const char *const pv, const STRLEN n)
9459 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9461 sv_setpvn(newSVrv(rv,classname), pv, n);
9466 =for apidoc sv_bless
9468 Blesses an SV into a specified package. The SV must be an RV. The package
9469 must be designated by its stash (see C<gv_stashpv()>). The reference count
9470 of the SV is unaffected.
9476 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9481 PERL_ARGS_ASSERT_SV_BLESS;
9484 Perl_croak(aTHX_ "Can't bless non-reference value");
9486 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9487 if (SvREADONLY(tmpRef) && !SvIsCOW(tmpRef))
9488 Perl_croak_no_modify();
9489 if (SvOBJECT(tmpRef)) {
9490 if (SvTYPE(tmpRef) != SVt_PVIO)
9492 SvREFCNT_dec(SvSTASH(tmpRef));
9495 SvOBJECT_on(tmpRef);
9496 if (SvTYPE(tmpRef) != SVt_PVIO)
9498 SvUPGRADE(tmpRef, SVt_PVMG);
9499 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9501 if(SvSMAGICAL(tmpRef))
9502 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9510 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9511 * as it is after unglobbing it.
9514 PERL_STATIC_INLINE void
9515 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9520 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9522 PERL_ARGS_ASSERT_SV_UNGLOB;
9524 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9526 if (!(flags & SV_COW_DROP_PV))
9527 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9530 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9531 && HvNAME_get(stash))
9532 mro_method_changed_in(stash);
9533 gp_free(MUTABLE_GV(sv));
9536 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9540 if (GvNAME_HEK(sv)) {
9541 unshare_hek(GvNAME_HEK(sv));
9543 isGV_with_GP_off(sv);
9545 if(SvTYPE(sv) == SVt_PVGV) {
9546 /* need to keep SvANY(sv) in the right arena */
9547 xpvmg = new_XPVMG();
9548 StructCopy(SvANY(sv), xpvmg, XPVMG);
9549 del_XPVGV(SvANY(sv));
9552 SvFLAGS(sv) &= ~SVTYPEMASK;
9553 SvFLAGS(sv) |= SVt_PVMG;
9556 /* Intentionally not calling any local SET magic, as this isn't so much a
9557 set operation as merely an internal storage change. */
9558 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9559 else sv_setsv_flags(sv, temp, 0);
9561 if ((const GV *)sv == PL_last_in_gv)
9562 PL_last_in_gv = NULL;
9563 else if ((const GV *)sv == PL_statgv)
9568 =for apidoc sv_unref_flags
9570 Unsets the RV status of the SV, and decrements the reference count of
9571 whatever was being referenced by the RV. This can almost be thought of
9572 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9573 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9574 (otherwise the decrementing is conditional on the reference count being
9575 different from one or the reference being a readonly SV).
9582 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9584 SV* const target = SvRV(ref);
9586 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9588 if (SvWEAKREF(ref)) {
9589 sv_del_backref(target, ref);
9591 SvRV_set(ref, NULL);
9594 SvRV_set(ref, NULL);
9596 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9597 assigned to as BEGIN {$a = \"Foo"} will fail. */
9598 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9599 SvREFCNT_dec(target);
9600 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9601 sv_2mortal(target); /* Schedule for freeing later */
9605 =for apidoc sv_untaint
9607 Untaint an SV. Use C<SvTAINTED_off> instead.
9613 Perl_sv_untaint(pTHX_ SV *const sv)
9615 PERL_ARGS_ASSERT_SV_UNTAINT;
9617 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9618 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9625 =for apidoc sv_tainted
9627 Test an SV for taintedness. Use C<SvTAINTED> instead.
9633 Perl_sv_tainted(pTHX_ SV *const sv)
9635 PERL_ARGS_ASSERT_SV_TAINTED;
9637 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9638 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9639 if (mg && (mg->mg_len & 1) )
9646 =for apidoc sv_setpviv
9648 Copies an integer into the given SV, also updating its string value.
9649 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9655 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9657 char buf[TYPE_CHARS(UV)];
9659 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9661 PERL_ARGS_ASSERT_SV_SETPVIV;
9663 sv_setpvn(sv, ptr, ebuf - ptr);
9667 =for apidoc sv_setpviv_mg
9669 Like C<sv_setpviv>, but also handles 'set' magic.
9675 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9677 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9683 #if defined(PERL_IMPLICIT_CONTEXT)
9685 /* pTHX_ magic can't cope with varargs, so this is a no-context
9686 * version of the main function, (which may itself be aliased to us).
9687 * Don't access this version directly.
9691 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9696 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9698 va_start(args, pat);
9699 sv_vsetpvf(sv, pat, &args);
9703 /* pTHX_ magic can't cope with varargs, so this is a no-context
9704 * version of the main function, (which may itself be aliased to us).
9705 * Don't access this version directly.
9709 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9714 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9716 va_start(args, pat);
9717 sv_vsetpvf_mg(sv, pat, &args);
9723 =for apidoc sv_setpvf
9725 Works like C<sv_catpvf> but copies the text into the SV instead of
9726 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9732 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9736 PERL_ARGS_ASSERT_SV_SETPVF;
9738 va_start(args, pat);
9739 sv_vsetpvf(sv, pat, &args);
9744 =for apidoc sv_vsetpvf
9746 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9747 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9749 Usually used via its frontend C<sv_setpvf>.
9755 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9757 PERL_ARGS_ASSERT_SV_VSETPVF;
9759 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9763 =for apidoc sv_setpvf_mg
9765 Like C<sv_setpvf>, but also handles 'set' magic.
9771 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9775 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9777 va_start(args, pat);
9778 sv_vsetpvf_mg(sv, pat, &args);
9783 =for apidoc sv_vsetpvf_mg
9785 Like C<sv_vsetpvf>, but also handles 'set' magic.
9787 Usually used via its frontend C<sv_setpvf_mg>.
9793 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9795 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9797 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9801 #if defined(PERL_IMPLICIT_CONTEXT)
9803 /* pTHX_ magic can't cope with varargs, so this is a no-context
9804 * version of the main function, (which may itself be aliased to us).
9805 * Don't access this version directly.
9809 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9814 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9816 va_start(args, pat);
9817 sv_vcatpvf(sv, pat, &args);
9821 /* pTHX_ magic can't cope with varargs, so this is a no-context
9822 * version of the main function, (which may itself be aliased to us).
9823 * Don't access this version directly.
9827 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9832 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9834 va_start(args, pat);
9835 sv_vcatpvf_mg(sv, pat, &args);
9841 =for apidoc sv_catpvf
9843 Processes its arguments like C<sprintf> and appends the formatted
9844 output to an SV. If the appended data contains "wide" characters
9845 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9846 and characters >255 formatted with %c), the original SV might get
9847 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9848 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9849 valid UTF-8; if the original SV was bytes, the pattern should be too.
9854 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9858 PERL_ARGS_ASSERT_SV_CATPVF;
9860 va_start(args, pat);
9861 sv_vcatpvf(sv, pat, &args);
9866 =for apidoc sv_vcatpvf
9868 Processes its arguments like C<vsprintf> and appends the formatted output
9869 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9871 Usually used via its frontend C<sv_catpvf>.
9877 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9879 PERL_ARGS_ASSERT_SV_VCATPVF;
9881 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9885 =for apidoc sv_catpvf_mg
9887 Like C<sv_catpvf>, but also handles 'set' magic.
9893 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9897 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9899 va_start(args, pat);
9900 sv_vcatpvf_mg(sv, pat, &args);
9905 =for apidoc sv_vcatpvf_mg
9907 Like C<sv_vcatpvf>, but also handles 'set' magic.
9909 Usually used via its frontend C<sv_catpvf_mg>.
9915 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9917 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9919 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9924 =for apidoc sv_vsetpvfn
9926 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9929 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9935 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9936 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9938 PERL_ARGS_ASSERT_SV_VSETPVFN;
9941 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, 0);
9946 * Warn of missing argument to sprintf, and then return a defined value
9947 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9949 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9951 S_vcatpvfn_missing_argument(pTHX) {
9952 if (ckWARN(WARN_MISSING)) {
9953 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9954 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9961 S_expect_number(pTHX_ char **const pattern)
9966 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9968 switch (**pattern) {
9969 case '1': case '2': case '3':
9970 case '4': case '5': case '6':
9971 case '7': case '8': case '9':
9972 var = *(*pattern)++ - '0';
9973 while (isDIGIT(**pattern)) {
9974 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9976 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9984 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9986 const int neg = nv < 0;
9989 PERL_ARGS_ASSERT_F0CONVERT;
9997 if (uv & 1 && uv == nv)
9998 uv--; /* Round to even */
10000 const unsigned dig = uv % 10;
10002 } while (uv /= 10);
10013 =for apidoc sv_vcatpvfn
10015 =for apidoc sv_vcatpvfn_flags
10017 Processes its arguments like C<vsprintf> and appends the formatted output
10018 to an SV. Uses an array of SVs if the C style variable argument list is
10019 missing (NULL). When running with taint checks enabled, indicates via
10020 C<maybe_tainted> if results are untrustworthy (often due to the use of
10023 If called as C<sv_vcatpvfn> or flags include C<SV_GMAGIC>, calls get magic.
10025 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10030 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10031 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10032 vec_utf8 = DO_UTF8(vecsv);
10034 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10037 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10038 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10040 PERL_ARGS_ASSERT_SV_VCATPVFN;
10042 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, SV_GMAGIC|SV_SMAGIC);
10046 Perl_sv_vcatpvfn_flags(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10047 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted,
10053 const char *patend;
10056 static const char nullstr[] = "(null)";
10058 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10059 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10061 /* Times 4: a decimal digit takes more than 3 binary digits.
10062 * NV_DIG: mantissa takes than many decimal digits.
10063 * Plus 32: Playing safe. */
10064 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10065 /* large enough for "%#.#f" --chip */
10066 /* what about long double NVs? --jhi */
10068 PERL_ARGS_ASSERT_SV_VCATPVFN_FLAGS;
10069 PERL_UNUSED_ARG(maybe_tainted);
10071 if (flags & SV_GMAGIC)
10074 /* no matter what, this is a string now */
10075 (void)SvPV_force_nomg(sv, origlen);
10077 /* special-case "", "%s", and "%-p" (SVf - see below) */
10080 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10082 const char * const s = va_arg(*args, char*);
10083 sv_catpv_nomg(sv, s ? s : nullstr);
10085 else if (svix < svmax) {
10086 /* we want get magic on the source but not the target. sv_catsv can't do that, though */
10087 SvGETMAGIC(*svargs);
10088 sv_catsv_nomg(sv, *svargs);
10091 S_vcatpvfn_missing_argument(aTHX);
10094 if (args && patlen == 3 && pat[0] == '%' &&
10095 pat[1] == '-' && pat[2] == 'p') {
10096 argsv = MUTABLE_SV(va_arg(*args, void*));
10097 sv_catsv_nomg(sv, argsv);
10101 #ifndef USE_LONG_DOUBLE
10102 /* special-case "%.<number>[gf]" */
10103 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10104 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10105 unsigned digits = 0;
10109 while (*pp >= '0' && *pp <= '9')
10110 digits = 10 * digits + (*pp++ - '0');
10111 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10112 const NV nv = SvNV(*svargs);
10114 /* Add check for digits != 0 because it seems that some
10115 gconverts are buggy in this case, and we don't yet have
10116 a Configure test for this. */
10117 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10118 /* 0, point, slack */
10119 Gconvert(nv, (int)digits, 0, ebuf);
10120 sv_catpv_nomg(sv, ebuf);
10121 if (*ebuf) /* May return an empty string for digits==0 */
10124 } else if (!digits) {
10127 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10128 sv_catpvn_nomg(sv, p, l);
10134 #endif /* !USE_LONG_DOUBLE */
10136 if (!args && svix < svmax && DO_UTF8(*svargs))
10139 patend = (char*)pat + patlen;
10140 for (p = (char*)pat; p < patend; p = q) {
10143 bool vectorize = FALSE;
10144 bool vectorarg = FALSE;
10145 bool vec_utf8 = FALSE;
10151 bool has_precis = FALSE;
10153 const I32 osvix = svix;
10154 bool is_utf8 = FALSE; /* is this item utf8? */
10155 #ifdef HAS_LDBL_SPRINTF_BUG
10156 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10157 with sfio - Allen <allens@cpan.org> */
10158 bool fix_ldbl_sprintf_bug = FALSE;
10162 U8 utf8buf[UTF8_MAXBYTES+1];
10163 STRLEN esignlen = 0;
10165 const char *eptr = NULL;
10166 const char *fmtstart;
10169 const U8 *vecstr = NULL;
10176 /* we need a long double target in case HAS_LONG_DOUBLE but
10177 not USE_LONG_DOUBLE
10179 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10187 const char *dotstr = ".";
10188 STRLEN dotstrlen = 1;
10189 I32 efix = 0; /* explicit format parameter index */
10190 I32 ewix = 0; /* explicit width index */
10191 I32 epix = 0; /* explicit precision index */
10192 I32 evix = 0; /* explicit vector index */
10193 bool asterisk = FALSE;
10195 /* echo everything up to the next format specification */
10196 for (q = p; q < patend && *q != '%'; ++q) ;
10198 if (has_utf8 && !pat_utf8)
10199 sv_catpvn_nomg_utf8_upgrade(sv, p, q - p, nsv);
10201 sv_catpvn_nomg(sv, p, q - p);
10210 We allow format specification elements in this order:
10211 \d+\$ explicit format parameter index
10213 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10214 0 flag (as above): repeated to allow "v02"
10215 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10216 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10218 [%bcdefginopsuxDFOUX] format (mandatory)
10223 As of perl5.9.3, printf format checking is on by default.
10224 Internally, perl uses %p formats to provide an escape to
10225 some extended formatting. This block deals with those
10226 extensions: if it does not match, (char*)q is reset and
10227 the normal format processing code is used.
10229 Currently defined extensions are:
10230 %p include pointer address (standard)
10231 %-p (SVf) include an SV (previously %_)
10232 %-<num>p include an SV with precision <num>
10234 %3p include a HEK with precision of 256
10235 %<num>p (where num != 2 or 3) reserved for future
10238 Robin Barker 2005-07-14 (but modified since)
10240 %1p (VDf) removed. RMB 2007-10-19
10247 n = expect_number(&q);
10249 if (sv) { /* SVf */
10254 argsv = MUTABLE_SV(va_arg(*args, void*));
10255 eptr = SvPV_const(argsv, elen);
10256 if (DO_UTF8(argsv))
10260 else if (n==2 || n==3) { /* HEKf */
10261 HEK * const hek = va_arg(*args, HEK *);
10262 eptr = HEK_KEY(hek);
10263 elen = HEK_LEN(hek);
10264 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10265 if (n==3) precis = 256, has_precis = TRUE;
10269 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10270 "internal %%<num>p might conflict with future printf extensions");
10276 if ( (width = expect_number(&q)) ) {
10291 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10320 if ( (ewix = expect_number(&q)) )
10329 if ((vectorarg = asterisk)) {
10342 width = expect_number(&q);
10345 if (vectorize && vectorarg) {
10346 /* vectorizing, but not with the default "." */
10348 vecsv = va_arg(*args, SV*);
10350 vecsv = (evix > 0 && evix <= svmax)
10351 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10353 vecsv = svix < svmax
10354 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10356 dotstr = SvPV_const(vecsv, dotstrlen);
10357 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10358 bad with tied or overloaded values that return UTF8. */
10359 if (DO_UTF8(vecsv))
10361 else if (has_utf8) {
10362 vecsv = sv_mortalcopy(vecsv);
10363 sv_utf8_upgrade(vecsv);
10364 dotstr = SvPV_const(vecsv, dotstrlen);
10371 i = va_arg(*args, int);
10373 i = (ewix ? ewix <= svmax : svix < svmax) ?
10374 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10376 width = (i < 0) ? -i : i;
10386 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10388 /* XXX: todo, support specified precision parameter */
10392 i = va_arg(*args, int);
10394 i = (ewix ? ewix <= svmax : svix < svmax)
10395 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10397 has_precis = !(i < 0);
10401 while (isDIGIT(*q))
10402 precis = precis * 10 + (*q++ - '0');
10411 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10412 vecsv = svargs[efix ? efix-1 : svix++];
10413 vecstr = (U8*)SvPV_const(vecsv,veclen);
10414 vec_utf8 = DO_UTF8(vecsv);
10416 /* if this is a version object, we need to convert
10417 * back into v-string notation and then let the
10418 * vectorize happen normally
10420 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10421 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10422 Perl_ck_warner_d(aTHX_ packWARN(WARN_PRINTF),
10423 "vector argument not supported with alpha versions");
10426 vecsv = sv_newmortal();
10427 scan_vstring((char *)vecstr, (char *)vecstr + veclen,
10429 vecstr = (U8*)SvPV_const(vecsv, veclen);
10430 vec_utf8 = DO_UTF8(vecsv);
10444 case 'I': /* Ix, I32x, and I64x */
10445 # ifdef USE_64_BIT_INT
10446 if (q[1] == '6' && q[2] == '4') {
10452 if (q[1] == '3' && q[2] == '2') {
10456 # ifdef USE_64_BIT_INT
10462 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10474 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10475 if (*q == 'l') { /* lld, llf */
10484 if (*++q == 'h') { /* hhd, hhu */
10513 if (!vectorize && !args) {
10515 const I32 i = efix-1;
10516 argsv = (i >= 0 && i < svmax)
10517 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10519 argsv = (svix >= 0 && svix < svmax)
10520 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10524 switch (c = *q++) {
10531 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10533 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10535 eptr = (char*)utf8buf;
10536 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10550 eptr = va_arg(*args, char*);
10552 elen = strlen(eptr);
10554 eptr = (char *)nullstr;
10555 elen = sizeof nullstr - 1;
10559 eptr = SvPV_const(argsv, elen);
10560 if (DO_UTF8(argsv)) {
10561 STRLEN old_precis = precis;
10562 if (has_precis && precis < elen) {
10563 STRLEN ulen = sv_or_pv_len_utf8(argsv, eptr, elen);
10564 STRLEN p = precis > ulen ? ulen : precis;
10565 precis = sv_or_pv_pos_u2b(argsv, eptr, p, 0);
10566 /* sticks at end */
10568 if (width) { /* fudge width (can't fudge elen) */
10569 if (has_precis && precis < elen)
10570 width += precis - old_precis;
10573 elen - sv_or_pv_len_utf8(argsv,eptr,elen);
10580 if (has_precis && precis < elen)
10587 if (alt || vectorize)
10589 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10610 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10619 esignbuf[esignlen++] = plus;
10623 case 'c': iv = (char)va_arg(*args, int); break;
10624 case 'h': iv = (short)va_arg(*args, int); break;
10625 case 'l': iv = va_arg(*args, long); break;
10626 case 'V': iv = va_arg(*args, IV); break;
10627 case 'z': iv = va_arg(*args, SSize_t); break;
10628 case 't': iv = va_arg(*args, ptrdiff_t); break;
10629 default: iv = va_arg(*args, int); break;
10631 case 'j': iv = va_arg(*args, intmax_t); break;
10635 iv = va_arg(*args, Quad_t); break;
10642 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10644 case 'c': iv = (char)tiv; break;
10645 case 'h': iv = (short)tiv; break;
10646 case 'l': iv = (long)tiv; break;
10648 default: iv = tiv; break;
10651 iv = (Quad_t)tiv; break;
10657 if ( !vectorize ) /* we already set uv above */
10662 esignbuf[esignlen++] = plus;
10666 esignbuf[esignlen++] = '-';
10710 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10721 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10722 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10723 case 'l': uv = va_arg(*args, unsigned long); break;
10724 case 'V': uv = va_arg(*args, UV); break;
10725 case 'z': uv = va_arg(*args, Size_t); break;
10726 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10728 case 'j': uv = va_arg(*args, uintmax_t); break;
10730 default: uv = va_arg(*args, unsigned); break;
10733 uv = va_arg(*args, Uquad_t); break;
10740 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10742 case 'c': uv = (unsigned char)tuv; break;
10743 case 'h': uv = (unsigned short)tuv; break;
10744 case 'l': uv = (unsigned long)tuv; break;
10746 default: uv = tuv; break;
10749 uv = (Uquad_t)tuv; break;
10758 char *ptr = ebuf + sizeof ebuf;
10759 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10765 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10769 } while (uv >>= 4);
10771 esignbuf[esignlen++] = '0';
10772 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10778 *--ptr = '0' + dig;
10779 } while (uv >>= 3);
10780 if (alt && *ptr != '0')
10786 *--ptr = '0' + dig;
10787 } while (uv >>= 1);
10789 esignbuf[esignlen++] = '0';
10790 esignbuf[esignlen++] = c;
10793 default: /* it had better be ten or less */
10796 *--ptr = '0' + dig;
10797 } while (uv /= base);
10800 elen = (ebuf + sizeof ebuf) - ptr;
10804 zeros = precis - elen;
10805 else if (precis == 0 && elen == 1 && *eptr == '0'
10806 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10809 /* a precision nullifies the 0 flag. */
10816 /* FLOATING POINT */
10819 c = 'f'; /* maybe %F isn't supported here */
10821 case 'e': case 'E':
10823 case 'g': case 'G':
10827 /* This is evil, but floating point is even more evil */
10829 /* for SV-style calling, we can only get NV
10830 for C-style calling, we assume %f is double;
10831 for simplicity we allow any of %Lf, %llf, %qf for long double
10835 #if defined(USE_LONG_DOUBLE)
10839 /* [perl #20339] - we should accept and ignore %lf rather than die */
10843 #if defined(USE_LONG_DOUBLE)
10844 intsize = args ? 0 : 'q';
10848 #if defined(HAS_LONG_DOUBLE)
10861 /* now we need (long double) if intsize == 'q', else (double) */
10863 #if LONG_DOUBLESIZE > DOUBLESIZE
10865 va_arg(*args, long double) :
10866 va_arg(*args, double)
10868 va_arg(*args, double)
10873 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10874 else. frexp() has some unspecified behaviour for those three */
10875 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10877 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10878 will cast our (long double) to (double) */
10879 (void)Perl_frexp(nv, &i);
10880 if (i == PERL_INT_MIN)
10881 Perl_die(aTHX_ "panic: frexp");
10883 need = BIT_DIGITS(i);
10885 need += has_precis ? precis : 6; /* known default */
10890 #ifdef HAS_LDBL_SPRINTF_BUG
10891 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10892 with sfio - Allen <allens@cpan.org> */
10895 # define MY_DBL_MAX DBL_MAX
10896 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10897 # if DOUBLESIZE >= 8
10898 # define MY_DBL_MAX 1.7976931348623157E+308L
10900 # define MY_DBL_MAX 3.40282347E+38L
10904 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10905 # define MY_DBL_MAX_BUG 1L
10907 # define MY_DBL_MAX_BUG MY_DBL_MAX
10911 # define MY_DBL_MIN DBL_MIN
10912 # else /* XXX guessing! -Allen */
10913 # if DOUBLESIZE >= 8
10914 # define MY_DBL_MIN 2.2250738585072014E-308L
10916 # define MY_DBL_MIN 1.17549435E-38L
10920 if ((intsize == 'q') && (c == 'f') &&
10921 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10922 (need < DBL_DIG)) {
10923 /* it's going to be short enough that
10924 * long double precision is not needed */
10926 if ((nv <= 0L) && (nv >= -0L))
10927 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10929 /* would use Perl_fp_class as a double-check but not
10930 * functional on IRIX - see perl.h comments */
10932 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10933 /* It's within the range that a double can represent */
10934 #if defined(DBL_MAX) && !defined(DBL_MIN)
10935 if ((nv >= ((long double)1/DBL_MAX)) ||
10936 (nv <= (-(long double)1/DBL_MAX)))
10938 fix_ldbl_sprintf_bug = TRUE;
10941 if (fix_ldbl_sprintf_bug == TRUE) {
10951 # undef MY_DBL_MAX_BUG
10954 #endif /* HAS_LDBL_SPRINTF_BUG */
10956 need += 20; /* fudge factor */
10957 if (PL_efloatsize < need) {
10958 Safefree(PL_efloatbuf);
10959 PL_efloatsize = need + 20; /* more fudge */
10960 Newx(PL_efloatbuf, PL_efloatsize, char);
10961 PL_efloatbuf[0] = '\0';
10964 if ( !(width || left || plus || alt) && fill != '0'
10965 && has_precis && intsize != 'q' ) { /* Shortcuts */
10966 /* See earlier comment about buggy Gconvert when digits,
10968 if ( c == 'g' && precis) {
10969 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10970 /* May return an empty string for digits==0 */
10971 if (*PL_efloatbuf) {
10972 elen = strlen(PL_efloatbuf);
10973 goto float_converted;
10975 } else if ( c == 'f' && !precis) {
10976 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10981 char *ptr = ebuf + sizeof ebuf;
10984 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10985 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10986 if (intsize == 'q') {
10987 /* Copy the one or more characters in a long double
10988 * format before the 'base' ([efgEFG]) character to
10989 * the format string. */
10990 static char const prifldbl[] = PERL_PRIfldbl;
10991 char const *p = prifldbl + sizeof(prifldbl) - 3;
10992 while (p >= prifldbl) { *--ptr = *p--; }
10997 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11002 do { *--ptr = '0' + (base % 10); } while (base /= 10);
11014 /* No taint. Otherwise we are in the strange situation
11015 * where printf() taints but print($float) doesn't.
11017 #if defined(HAS_LONG_DOUBLE)
11018 elen = ((intsize == 'q')
11019 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
11020 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
11022 elen = my_sprintf(PL_efloatbuf, ptr, nv);
11026 eptr = PL_efloatbuf;
11034 i = SvCUR(sv) - origlen;
11037 case 'c': *(va_arg(*args, char*)) = i; break;
11038 case 'h': *(va_arg(*args, short*)) = i; break;
11039 default: *(va_arg(*args, int*)) = i; break;
11040 case 'l': *(va_arg(*args, long*)) = i; break;
11041 case 'V': *(va_arg(*args, IV*)) = i; break;
11042 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11043 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11045 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11049 *(va_arg(*args, Quad_t*)) = i; break;
11056 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11057 continue; /* not "break" */
11064 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11065 && ckWARN(WARN_PRINTF))
11067 SV * const msg = sv_newmortal();
11068 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11069 (PL_op->op_type == OP_PRTF) ? "" : "s");
11070 if (fmtstart < patend) {
11071 const char * const fmtend = q < patend ? q : patend;
11073 sv_catpvs(msg, "\"%");
11074 for (f = fmtstart; f < fmtend; f++) {
11076 sv_catpvn_nomg(msg, f, 1);
11078 Perl_sv_catpvf(aTHX_ msg,
11079 "\\%03"UVof, (UV)*f & 0xFF);
11082 sv_catpvs(msg, "\"");
11084 sv_catpvs(msg, "end of string");
11086 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11089 /* output mangled stuff ... */
11095 /* ... right here, because formatting flags should not apply */
11096 SvGROW(sv, SvCUR(sv) + elen + 1);
11098 Copy(eptr, p, elen, char);
11101 SvCUR_set(sv, p - SvPVX_const(sv));
11103 continue; /* not "break" */
11106 if (is_utf8 != has_utf8) {
11109 sv_utf8_upgrade(sv);
11112 const STRLEN old_elen = elen;
11113 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11114 sv_utf8_upgrade(nsv);
11115 eptr = SvPVX_const(nsv);
11118 if (width) { /* fudge width (can't fudge elen) */
11119 width += elen - old_elen;
11125 have = esignlen + zeros + elen;
11127 croak_memory_wrap();
11129 need = (have > width ? have : width);
11132 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11133 croak_memory_wrap();
11134 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11136 if (esignlen && fill == '0') {
11138 for (i = 0; i < (int)esignlen; i++)
11139 *p++ = esignbuf[i];
11141 if (gap && !left) {
11142 memset(p, fill, gap);
11145 if (esignlen && fill != '0') {
11147 for (i = 0; i < (int)esignlen; i++)
11148 *p++ = esignbuf[i];
11152 for (i = zeros; i; i--)
11156 Copy(eptr, p, elen, char);
11160 memset(p, ' ', gap);
11165 Copy(dotstr, p, dotstrlen, char);
11169 vectorize = FALSE; /* done iterating over vecstr */
11176 SvCUR_set(sv, p - SvPVX_const(sv));
11185 /* =========================================================================
11187 =head1 Cloning an interpreter
11189 All the macros and functions in this section are for the private use of
11190 the main function, perl_clone().
11192 The foo_dup() functions make an exact copy of an existing foo thingy.
11193 During the course of a cloning, a hash table is used to map old addresses
11194 to new addresses. The table is created and manipulated with the
11195 ptr_table_* functions.
11199 * =========================================================================*/
11202 #if defined(USE_ITHREADS)
11204 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11205 #ifndef GpREFCNT_inc
11206 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11210 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11211 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11212 If this changes, please unmerge ss_dup.
11213 Likewise, sv_dup_inc_multiple() relies on this fact. */
11214 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11215 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11216 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11217 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11218 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11219 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11220 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11221 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11222 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11223 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11224 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11225 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11226 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11228 /* clone a parser */
11231 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11235 PERL_ARGS_ASSERT_PARSER_DUP;
11240 /* look for it in the table first */
11241 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11245 /* create anew and remember what it is */
11246 Newxz(parser, 1, yy_parser);
11247 ptr_table_store(PL_ptr_table, proto, parser);
11249 /* XXX these not yet duped */
11250 parser->old_parser = NULL;
11251 parser->stack = NULL;
11253 parser->stack_size = 0;
11254 /* XXX parser->stack->state = 0; */
11256 /* XXX eventually, just Copy() most of the parser struct ? */
11258 parser->lex_brackets = proto->lex_brackets;
11259 parser->lex_casemods = proto->lex_casemods;
11260 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11261 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11262 parser->lex_casestack = savepvn(proto->lex_casestack,
11263 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11264 parser->lex_defer = proto->lex_defer;
11265 parser->lex_dojoin = proto->lex_dojoin;
11266 parser->lex_expect = proto->lex_expect;
11267 parser->lex_formbrack = proto->lex_formbrack;
11268 parser->lex_inpat = proto->lex_inpat;
11269 parser->lex_inwhat = proto->lex_inwhat;
11270 parser->lex_op = proto->lex_op;
11271 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11272 parser->lex_starts = proto->lex_starts;
11273 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11274 parser->multi_close = proto->multi_close;
11275 parser->multi_open = proto->multi_open;
11276 parser->multi_start = proto->multi_start;
11277 parser->multi_end = proto->multi_end;
11278 parser->preambled = proto->preambled;
11279 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11280 parser->linestr = sv_dup_inc(proto->linestr, param);
11281 parser->expect = proto->expect;
11282 parser->copline = proto->copline;
11283 parser->last_lop_op = proto->last_lop_op;
11284 parser->lex_state = proto->lex_state;
11285 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11286 /* rsfp_filters entries have fake IoDIRP() */
11287 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11288 parser->in_my = proto->in_my;
11289 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11290 parser->error_count = proto->error_count;
11293 parser->linestr = sv_dup_inc(proto->linestr, param);
11296 char * const ols = SvPVX(proto->linestr);
11297 char * const ls = SvPVX(parser->linestr);
11299 parser->bufptr = ls + (proto->bufptr >= ols ?
11300 proto->bufptr - ols : 0);
11301 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11302 proto->oldbufptr - ols : 0);
11303 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11304 proto->oldoldbufptr - ols : 0);
11305 parser->linestart = ls + (proto->linestart >= ols ?
11306 proto->linestart - ols : 0);
11307 parser->last_uni = ls + (proto->last_uni >= ols ?
11308 proto->last_uni - ols : 0);
11309 parser->last_lop = ls + (proto->last_lop >= ols ?
11310 proto->last_lop - ols : 0);
11312 parser->bufend = ls + SvCUR(parser->linestr);
11315 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11319 parser->endwhite = proto->endwhite;
11320 parser->faketokens = proto->faketokens;
11321 parser->lasttoke = proto->lasttoke;
11322 parser->nextwhite = proto->nextwhite;
11323 parser->realtokenstart = proto->realtokenstart;
11324 parser->skipwhite = proto->skipwhite;
11325 parser->thisclose = proto->thisclose;
11326 parser->thismad = proto->thismad;
11327 parser->thisopen = proto->thisopen;
11328 parser->thisstuff = proto->thisstuff;
11329 parser->thistoken = proto->thistoken;
11330 parser->thiswhite = proto->thiswhite;
11332 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11333 parser->curforce = proto->curforce;
11335 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11336 Copy(proto->nexttype, parser->nexttype, 5, I32);
11337 parser->nexttoke = proto->nexttoke;
11340 /* XXX should clone saved_curcop here, but we aren't passed
11341 * proto_perl; so do it in perl_clone_using instead */
11347 /* duplicate a file handle */
11350 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11354 PERL_ARGS_ASSERT_FP_DUP;
11355 PERL_UNUSED_ARG(type);
11358 return (PerlIO*)NULL;
11360 /* look for it in the table first */
11361 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11365 /* create anew and remember what it is */
11366 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11367 ptr_table_store(PL_ptr_table, fp, ret);
11371 /* duplicate a directory handle */
11374 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11380 const Direntry_t *dirent;
11381 char smallbuf[256];
11387 PERL_UNUSED_CONTEXT;
11388 PERL_ARGS_ASSERT_DIRP_DUP;
11393 /* look for it in the table first */
11394 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11400 PERL_UNUSED_ARG(param);
11404 /* open the current directory (so we can switch back) */
11405 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11407 /* chdir to our dir handle and open the present working directory */
11408 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11409 PerlDir_close(pwd);
11410 return (DIR *)NULL;
11412 /* Now we should have two dir handles pointing to the same dir. */
11414 /* Be nice to the calling code and chdir back to where we were. */
11415 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11417 /* We have no need of the pwd handle any more. */
11418 PerlDir_close(pwd);
11421 # define d_namlen(d) (d)->d_namlen
11423 # define d_namlen(d) strlen((d)->d_name)
11425 /* Iterate once through dp, to get the file name at the current posi-
11426 tion. Then step back. */
11427 pos = PerlDir_tell(dp);
11428 if ((dirent = PerlDir_read(dp))) {
11429 len = d_namlen(dirent);
11430 if (len <= sizeof smallbuf) name = smallbuf;
11431 else Newx(name, len, char);
11432 Move(dirent->d_name, name, len, char);
11434 PerlDir_seek(dp, pos);
11436 /* Iterate through the new dir handle, till we find a file with the
11438 if (!dirent) /* just before the end */
11440 pos = PerlDir_tell(ret);
11441 if (PerlDir_read(ret)) continue; /* not there yet */
11442 PerlDir_seek(ret, pos); /* step back */
11446 const long pos0 = PerlDir_tell(ret);
11448 pos = PerlDir_tell(ret);
11449 if ((dirent = PerlDir_read(ret))) {
11450 if (len == d_namlen(dirent)
11451 && memEQ(name, dirent->d_name, len)) {
11453 PerlDir_seek(ret, pos); /* step back */
11456 /* else we are not there yet; keep iterating */
11458 else { /* This is not meant to happen. The best we can do is
11459 reset the iterator to the beginning. */
11460 PerlDir_seek(ret, pos0);
11467 if (name && name != smallbuf)
11472 ret = win32_dirp_dup(dp, param);
11475 /* pop it in the pointer table */
11477 ptr_table_store(PL_ptr_table, dp, ret);
11482 /* duplicate a typeglob */
11485 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11489 PERL_ARGS_ASSERT_GP_DUP;
11493 /* look for it in the table first */
11494 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11498 /* create anew and remember what it is */
11500 ptr_table_store(PL_ptr_table, gp, ret);
11503 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11504 on Newxz() to do this for us. */
11505 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11506 ret->gp_io = io_dup_inc(gp->gp_io, param);
11507 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11508 ret->gp_av = av_dup_inc(gp->gp_av, param);
11509 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11510 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11511 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11512 ret->gp_cvgen = gp->gp_cvgen;
11513 ret->gp_line = gp->gp_line;
11514 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11518 /* duplicate a chain of magic */
11521 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11523 MAGIC *mgret = NULL;
11524 MAGIC **mgprev_p = &mgret;
11526 PERL_ARGS_ASSERT_MG_DUP;
11528 for (; mg; mg = mg->mg_moremagic) {
11531 if ((param->flags & CLONEf_JOIN_IN)
11532 && mg->mg_type == PERL_MAGIC_backref)
11533 /* when joining, we let the individual SVs add themselves to
11534 * backref as needed. */
11537 Newx(nmg, 1, MAGIC);
11539 mgprev_p = &(nmg->mg_moremagic);
11541 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11542 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11543 from the original commit adding Perl_mg_dup() - revision 4538.
11544 Similarly there is the annotation "XXX random ptr?" next to the
11545 assignment to nmg->mg_ptr. */
11548 /* FIXME for plugins
11549 if (nmg->mg_type == PERL_MAGIC_qr) {
11550 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11554 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11555 ? nmg->mg_type == PERL_MAGIC_backref
11556 /* The backref AV has its reference
11557 * count deliberately bumped by 1 */
11558 ? SvREFCNT_inc(av_dup_inc((const AV *)
11559 nmg->mg_obj, param))
11560 : sv_dup_inc(nmg->mg_obj, param)
11561 : sv_dup(nmg->mg_obj, param);
11563 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11564 if (nmg->mg_len > 0) {
11565 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11566 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11567 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11569 AMT * const namtp = (AMT*)nmg->mg_ptr;
11570 sv_dup_inc_multiple((SV**)(namtp->table),
11571 (SV**)(namtp->table), NofAMmeth, param);
11574 else if (nmg->mg_len == HEf_SVKEY)
11575 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11577 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11578 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11584 #endif /* USE_ITHREADS */
11586 struct ptr_tbl_arena {
11587 struct ptr_tbl_arena *next;
11588 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11591 /* create a new pointer-mapping table */
11594 Perl_ptr_table_new(pTHX)
11597 PERL_UNUSED_CONTEXT;
11599 Newx(tbl, 1, PTR_TBL_t);
11600 tbl->tbl_max = 511;
11601 tbl->tbl_items = 0;
11602 tbl->tbl_arena = NULL;
11603 tbl->tbl_arena_next = NULL;
11604 tbl->tbl_arena_end = NULL;
11605 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11609 #define PTR_TABLE_HASH(ptr) \
11610 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11612 /* map an existing pointer using a table */
11614 STATIC PTR_TBL_ENT_t *
11615 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11617 PTR_TBL_ENT_t *tblent;
11618 const UV hash = PTR_TABLE_HASH(sv);
11620 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11622 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11623 for (; tblent; tblent = tblent->next) {
11624 if (tblent->oldval == sv)
11631 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11633 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11635 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11636 PERL_UNUSED_CONTEXT;
11638 return tblent ? tblent->newval : NULL;
11641 /* add a new entry to a pointer-mapping table */
11644 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11646 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11648 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11649 PERL_UNUSED_CONTEXT;
11652 tblent->newval = newsv;
11654 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11656 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11657 struct ptr_tbl_arena *new_arena;
11659 Newx(new_arena, 1, struct ptr_tbl_arena);
11660 new_arena->next = tbl->tbl_arena;
11661 tbl->tbl_arena = new_arena;
11662 tbl->tbl_arena_next = new_arena->array;
11663 tbl->tbl_arena_end = new_arena->array
11664 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11667 tblent = tbl->tbl_arena_next++;
11669 tblent->oldval = oldsv;
11670 tblent->newval = newsv;
11671 tblent->next = tbl->tbl_ary[entry];
11672 tbl->tbl_ary[entry] = tblent;
11674 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11675 ptr_table_split(tbl);
11679 /* double the hash bucket size of an existing ptr table */
11682 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11684 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11685 const UV oldsize = tbl->tbl_max + 1;
11686 UV newsize = oldsize * 2;
11689 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11690 PERL_UNUSED_CONTEXT;
11692 Renew(ary, newsize, PTR_TBL_ENT_t*);
11693 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11694 tbl->tbl_max = --newsize;
11695 tbl->tbl_ary = ary;
11696 for (i=0; i < oldsize; i++, ary++) {
11697 PTR_TBL_ENT_t **entp = ary;
11698 PTR_TBL_ENT_t *ent = *ary;
11699 PTR_TBL_ENT_t **curentp;
11702 curentp = ary + oldsize;
11704 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11706 ent->next = *curentp;
11716 /* remove all the entries from a ptr table */
11717 /* Deprecated - will be removed post 5.14 */
11720 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11722 if (tbl && tbl->tbl_items) {
11723 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11725 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11728 struct ptr_tbl_arena *next = arena->next;
11734 tbl->tbl_items = 0;
11735 tbl->tbl_arena = NULL;
11736 tbl->tbl_arena_next = NULL;
11737 tbl->tbl_arena_end = NULL;
11741 /* clear and free a ptr table */
11744 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11746 struct ptr_tbl_arena *arena;
11752 arena = tbl->tbl_arena;
11755 struct ptr_tbl_arena *next = arena->next;
11761 Safefree(tbl->tbl_ary);
11765 #if defined(USE_ITHREADS)
11768 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11770 PERL_ARGS_ASSERT_RVPV_DUP;
11772 assert(!isREGEXP(sstr));
11774 if (SvWEAKREF(sstr)) {
11775 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11776 if (param->flags & CLONEf_JOIN_IN) {
11777 /* if joining, we add any back references individually rather
11778 * than copying the whole backref array */
11779 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11783 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11785 else if (SvPVX_const(sstr)) {
11786 /* Has something there */
11788 /* Normal PV - clone whole allocated space */
11789 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11790 /* sstr may not be that normal, but actually copy on write.
11791 But we are a true, independent SV, so: */
11795 /* Special case - not normally malloced for some reason */
11796 if (isGV_with_GP(sstr)) {
11797 /* Don't need to do anything here. */
11799 else if ((SvIsCOW(sstr))) {
11800 /* A "shared" PV - clone it as "shared" PV */
11802 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11806 /* Some other special case - random pointer */
11807 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11812 /* Copy the NULL */
11813 SvPV_set(dstr, NULL);
11817 /* duplicate a list of SVs. source and dest may point to the same memory. */
11819 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11820 SSize_t items, CLONE_PARAMS *const param)
11822 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11824 while (items-- > 0) {
11825 *dest++ = sv_dup_inc(*source++, param);
11831 /* duplicate an SV of any type (including AV, HV etc) */
11834 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11839 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11841 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11842 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11847 /* look for it in the table first */
11848 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11852 if(param->flags & CLONEf_JOIN_IN) {
11853 /** We are joining here so we don't want do clone
11854 something that is bad **/
11855 if (SvTYPE(sstr) == SVt_PVHV) {
11856 const HEK * const hvname = HvNAME_HEK(sstr);
11858 /** don't clone stashes if they already exist **/
11859 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11860 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11861 ptr_table_store(PL_ptr_table, sstr, dstr);
11865 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11866 HV *stash = GvSTASH(sstr);
11867 const HEK * hvname;
11868 if (stash && (hvname = HvNAME_HEK(stash))) {
11869 /** don't clone GVs if they already exist **/
11871 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11872 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11874 stash, GvNAME(sstr),
11880 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11881 ptr_table_store(PL_ptr_table, sstr, *svp);
11888 /* create anew and remember what it is */
11891 #ifdef DEBUG_LEAKING_SCALARS
11892 dstr->sv_debug_optype = sstr->sv_debug_optype;
11893 dstr->sv_debug_line = sstr->sv_debug_line;
11894 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11895 dstr->sv_debug_parent = (SV*)sstr;
11896 FREE_SV_DEBUG_FILE(dstr);
11897 dstr->sv_debug_file = savesharedpv(sstr->sv_debug_file);
11900 ptr_table_store(PL_ptr_table, sstr, dstr);
11903 SvFLAGS(dstr) = SvFLAGS(sstr);
11904 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11905 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11908 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11909 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11910 (void*)PL_watch_pvx, SvPVX_const(sstr));
11913 /* don't clone objects whose class has asked us not to */
11914 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11919 switch (SvTYPE(sstr)) {
11921 SvANY(dstr) = NULL;
11924 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11926 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11928 SvIV_set(dstr, SvIVX(sstr));
11932 SvANY(dstr) = new_XNV();
11933 SvNV_set(dstr, SvNVX(sstr));
11935 /* case SVt_BIND: */
11938 /* These are all the types that need complex bodies allocating. */
11940 const svtype sv_type = SvTYPE(sstr);
11941 const struct body_details *const sv_type_details
11942 = bodies_by_type + sv_type;
11946 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11961 assert(sv_type_details->body_size);
11962 if (sv_type_details->arena) {
11963 new_body_inline(new_body, sv_type);
11965 = (void*)((char*)new_body - sv_type_details->offset);
11967 new_body = new_NOARENA(sv_type_details);
11971 SvANY(dstr) = new_body;
11974 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11975 ((char*)SvANY(dstr)) + sv_type_details->offset,
11976 sv_type_details->copy, char);
11978 Copy(((char*)SvANY(sstr)),
11979 ((char*)SvANY(dstr)),
11980 sv_type_details->body_size + sv_type_details->offset, char);
11983 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11984 && !isGV_with_GP(dstr)
11986 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11987 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11989 /* The Copy above means that all the source (unduplicated) pointers
11990 are now in the destination. We can check the flags and the
11991 pointers in either, but it's possible that there's less cache
11992 missing by always going for the destination.
11993 FIXME - instrument and check that assumption */
11994 if (sv_type >= SVt_PVMG) {
11995 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11996 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11997 } else if (SvMAGIC(dstr))
11998 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
12000 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
12003 /* The cast silences a GCC warning about unhandled types. */
12004 switch ((int)sv_type) {
12015 /* FIXME for plugins */
12016 dstr->sv_u.svu_rx = ((REGEXP *)dstr)->sv_any;
12017 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
12020 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
12021 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
12022 LvTARG(dstr) = dstr;
12023 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
12024 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
12026 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
12027 if (isREGEXP(sstr)) goto duprex;
12029 /* non-GP case already handled above */
12030 if(isGV_with_GP(sstr)) {
12031 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
12032 /* Don't call sv_add_backref here as it's going to be
12033 created as part of the magic cloning of the symbol
12034 table--unless this is during a join and the stash
12035 is not actually being cloned. */
12036 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12037 at the point of this comment. */
12038 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12039 if (param->flags & CLONEf_JOIN_IN)
12040 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12041 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12042 (void)GpREFCNT_inc(GvGP(dstr));
12046 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12047 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12048 /* I have no idea why fake dirp (rsfps)
12049 should be treated differently but otherwise
12050 we end up with leaks -- sky*/
12051 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12052 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12053 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12055 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12056 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12057 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12058 if (IoDIRP(dstr)) {
12059 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12062 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12064 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12066 if (IoOFP(dstr) == IoIFP(sstr))
12067 IoOFP(dstr) = IoIFP(dstr);
12069 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12070 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12071 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12072 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12075 /* avoid cloning an empty array */
12076 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12077 SV **dst_ary, **src_ary;
12078 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12080 src_ary = AvARRAY((const AV *)sstr);
12081 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12082 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12083 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12084 AvALLOC((const AV *)dstr) = dst_ary;
12085 if (AvREAL((const AV *)sstr)) {
12086 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12090 while (items-- > 0)
12091 *dst_ary++ = sv_dup(*src_ary++, param);
12093 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12094 while (items-- > 0) {
12095 *dst_ary++ = &PL_sv_undef;
12099 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12100 AvALLOC((const AV *)dstr) = (SV**)NULL;
12101 AvMAX( (const AV *)dstr) = -1;
12102 AvFILLp((const AV *)dstr) = -1;
12106 if (HvARRAY((const HV *)sstr)) {
12108 const bool sharekeys = !!HvSHAREKEYS(sstr);
12109 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12110 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12112 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12113 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12115 HvARRAY(dstr) = (HE**)darray;
12116 while (i <= sxhv->xhv_max) {
12117 const HE * const source = HvARRAY(sstr)[i];
12118 HvARRAY(dstr)[i] = source
12119 ? he_dup(source, sharekeys, param) : 0;
12123 const struct xpvhv_aux * const saux = HvAUX(sstr);
12124 struct xpvhv_aux * const daux = HvAUX(dstr);
12125 /* This flag isn't copied. */
12128 if (saux->xhv_name_count) {
12129 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12131 = saux->xhv_name_count < 0
12132 ? -saux->xhv_name_count
12133 : saux->xhv_name_count;
12134 HEK **shekp = sname + count;
12136 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12137 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12138 while (shekp-- > sname) {
12140 *dhekp = hek_dup(*shekp, param);
12144 daux->xhv_name_u.xhvnameu_name
12145 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12148 daux->xhv_name_count = saux->xhv_name_count;
12150 daux->xhv_riter = saux->xhv_riter;
12151 daux->xhv_eiter = saux->xhv_eiter
12152 ? he_dup(saux->xhv_eiter,
12153 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12154 /* backref array needs refcnt=2; see sv_add_backref */
12155 daux->xhv_backreferences =
12156 (param->flags & CLONEf_JOIN_IN)
12157 /* when joining, we let the individual GVs and
12158 * CVs add themselves to backref as
12159 * needed. This avoids pulling in stuff
12160 * that isn't required, and simplifies the
12161 * case where stashes aren't cloned back
12162 * if they already exist in the parent
12165 : saux->xhv_backreferences
12166 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12167 ? MUTABLE_AV(SvREFCNT_inc(
12168 sv_dup_inc((const SV *)
12169 saux->xhv_backreferences, param)))
12170 : MUTABLE_AV(sv_dup((const SV *)
12171 saux->xhv_backreferences, param))
12174 daux->xhv_mro_meta = saux->xhv_mro_meta
12175 ? mro_meta_dup(saux->xhv_mro_meta, param)
12177 daux->xhv_super = NULL;
12179 /* Record stashes for possible cloning in Perl_clone(). */
12181 av_push(param->stashes, dstr);
12185 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12188 if (!(param->flags & CLONEf_COPY_STACKS)) {
12193 /* NOTE: not refcounted */
12194 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12195 hv_dup(CvSTASH(dstr), param);
12196 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12197 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12198 if (!CvISXSUB(dstr)) {
12200 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12202 CvSLABBED_off(dstr);
12203 } else if (CvCONST(dstr)) {
12204 CvXSUBANY(dstr).any_ptr =
12205 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12207 assert(!CvSLABBED(dstr));
12208 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12210 SvANY((CV *)dstr)->xcv_gv_u.xcv_hek =
12211 share_hek_hek(CvNAME_HEK((CV *)sstr));
12212 /* don't dup if copying back - CvGV isn't refcounted, so the
12213 * duped GV may never be freed. A bit of a hack! DAPM */
12215 SvANY(MUTABLE_CV(dstr))->xcv_gv_u.xcv_gv =
12217 ? gv_dup_inc(CvGV(sstr), param)
12218 : (param->flags & CLONEf_JOIN_IN)
12220 : gv_dup(CvGV(sstr), param);
12222 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12224 CvWEAKOUTSIDE(sstr)
12225 ? cv_dup( CvOUTSIDE(dstr), param)
12226 : cv_dup_inc(CvOUTSIDE(dstr), param);
12232 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12239 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12241 PERL_ARGS_ASSERT_SV_DUP_INC;
12242 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12246 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12248 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12249 PERL_ARGS_ASSERT_SV_DUP;
12251 /* Track every SV that (at least initially) had a reference count of 0.
12252 We need to do this by holding an actual reference to it in this array.
12253 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12254 (akin to the stashes hash, and the perl stack), we come unstuck if
12255 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12256 thread) is manipulated in a CLONE method, because CLONE runs before the
12257 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12258 (and fix things up by giving each a reference via the temps stack).
12259 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12260 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12261 before the walk of unreferenced happens and a reference to that is SV
12262 added to the temps stack. At which point we have the same SV considered
12263 to be in use, and free to be re-used. Not good.
12265 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12266 assert(param->unreferenced);
12267 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12273 /* duplicate a context */
12276 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12278 PERL_CONTEXT *ncxs;
12280 PERL_ARGS_ASSERT_CX_DUP;
12283 return (PERL_CONTEXT*)NULL;
12285 /* look for it in the table first */
12286 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12290 /* create anew and remember what it is */
12291 Newx(ncxs, max + 1, PERL_CONTEXT);
12292 ptr_table_store(PL_ptr_table, cxs, ncxs);
12293 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12296 PERL_CONTEXT * const ncx = &ncxs[ix];
12297 if (CxTYPE(ncx) == CXt_SUBST) {
12298 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12301 ncx->blk_oldcop = (COP*)any_dup(ncx->blk_oldcop, param->proto_perl);
12302 switch (CxTYPE(ncx)) {
12304 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12305 ? cv_dup_inc(ncx->blk_sub.cv, param)
12306 : cv_dup(ncx->blk_sub.cv,param));
12307 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12308 ? av_dup_inc(ncx->blk_sub.argarray,
12311 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12313 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12314 ncx->blk_sub.oldcomppad);
12317 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12319 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12320 ncx->blk_eval.cv = cv_dup(ncx->blk_eval.cv, param);
12322 case CXt_LOOP_LAZYSV:
12323 ncx->blk_loop.state_u.lazysv.end
12324 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12325 /* We are taking advantage of av_dup_inc and sv_dup_inc
12326 actually being the same function, and order equivalence of
12328 We can assert the later [but only at run time :-(] */
12329 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12330 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12332 ncx->blk_loop.state_u.ary.ary
12333 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12334 case CXt_LOOP_LAZYIV:
12335 case CXt_LOOP_PLAIN:
12336 if (CxPADLOOP(ncx)) {
12337 ncx->blk_loop.itervar_u.oldcomppad
12338 = (PAD*)ptr_table_fetch(PL_ptr_table,
12339 ncx->blk_loop.itervar_u.oldcomppad);
12341 ncx->blk_loop.itervar_u.gv
12342 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12347 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12348 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12349 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12364 /* duplicate a stack info structure */
12367 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12371 PERL_ARGS_ASSERT_SI_DUP;
12374 return (PERL_SI*)NULL;
12376 /* look for it in the table first */
12377 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12381 /* create anew and remember what it is */
12382 Newxz(nsi, 1, PERL_SI);
12383 ptr_table_store(PL_ptr_table, si, nsi);
12385 nsi->si_stack = av_dup_inc(si->si_stack, param);
12386 nsi->si_cxix = si->si_cxix;
12387 nsi->si_cxmax = si->si_cxmax;
12388 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12389 nsi->si_type = si->si_type;
12390 nsi->si_prev = si_dup(si->si_prev, param);
12391 nsi->si_next = si_dup(si->si_next, param);
12392 nsi->si_markoff = si->si_markoff;
12397 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12398 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12399 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12400 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12401 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12402 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12403 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12404 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12405 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12406 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12407 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12408 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12409 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12410 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12411 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12412 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12415 #define pv_dup_inc(p) SAVEPV(p)
12416 #define pv_dup(p) SAVEPV(p)
12417 #define svp_dup_inc(p,pp) any_dup(p,pp)
12419 /* map any object to the new equivent - either something in the
12420 * ptr table, or something in the interpreter structure
12424 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12428 PERL_ARGS_ASSERT_ANY_DUP;
12431 return (void*)NULL;
12433 /* look for it in the table first */
12434 ret = ptr_table_fetch(PL_ptr_table, v);
12438 /* see if it is part of the interpreter structure */
12439 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12440 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12448 /* duplicate the save stack */
12451 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12454 ANY * const ss = proto_perl->Isavestack;
12455 const I32 max = proto_perl->Isavestack_max;
12456 I32 ix = proto_perl->Isavestack_ix;
12469 void (*dptr) (void*);
12470 void (*dxptr) (pTHX_ void*);
12472 PERL_ARGS_ASSERT_SS_DUP;
12474 Newxz(nss, max, ANY);
12477 const UV uv = POPUV(ss,ix);
12478 const U8 type = (U8)uv & SAVE_MASK;
12480 TOPUV(nss,ix) = uv;
12482 case SAVEt_CLEARSV:
12483 case SAVEt_CLEARPADRANGE:
12485 case SAVEt_HELEM: /* hash element */
12486 sv = (const SV *)POPPTR(ss,ix);
12487 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12489 case SAVEt_ITEM: /* normal string */
12490 case SAVEt_GVSV: /* scalar slot in GV */
12491 case SAVEt_SV: /* scalar reference */
12492 sv = (const SV *)POPPTR(ss,ix);
12493 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12496 case SAVEt_MORTALIZESV:
12497 sv = (const SV *)POPPTR(ss,ix);
12498 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12500 case SAVEt_SHARED_PVREF: /* char* in shared space */
12501 c = (char*)POPPTR(ss,ix);
12502 TOPPTR(nss,ix) = savesharedpv(c);
12503 ptr = POPPTR(ss,ix);
12504 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12506 case SAVEt_GENERIC_SVREF: /* generic sv */
12507 case SAVEt_SVREF: /* scalar reference */
12508 sv = (const SV *)POPPTR(ss,ix);
12509 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12510 ptr = POPPTR(ss,ix);
12511 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12513 case SAVEt_HV: /* hash reference */
12514 case SAVEt_AV: /* array reference */
12515 sv = (const SV *) POPPTR(ss,ix);
12516 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12518 case SAVEt_COMPPAD:
12520 sv = (const SV *) POPPTR(ss,ix);
12521 TOPPTR(nss,ix) = sv_dup(sv, param);
12523 case SAVEt_INT: /* int reference */
12524 ptr = POPPTR(ss,ix);
12525 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12526 intval = (int)POPINT(ss,ix);
12527 TOPINT(nss,ix) = intval;
12529 case SAVEt_LONG: /* long reference */
12530 ptr = POPPTR(ss,ix);
12531 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12532 longval = (long)POPLONG(ss,ix);
12533 TOPLONG(nss,ix) = longval;
12535 case SAVEt_I32: /* I32 reference */
12536 ptr = POPPTR(ss,ix);
12537 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12539 TOPINT(nss,ix) = i;
12541 case SAVEt_IV: /* IV reference */
12542 ptr = POPPTR(ss,ix);
12543 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12545 TOPIV(nss,ix) = iv;
12547 case SAVEt_HPTR: /* HV* reference */
12548 case SAVEt_APTR: /* AV* reference */
12549 case SAVEt_SPTR: /* SV* reference */
12550 ptr = POPPTR(ss,ix);
12551 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12552 sv = (const SV *)POPPTR(ss,ix);
12553 TOPPTR(nss,ix) = sv_dup(sv, param);
12555 case SAVEt_VPTR: /* random* reference */
12556 ptr = POPPTR(ss,ix);
12557 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12559 case SAVEt_INT_SMALL:
12560 case SAVEt_I32_SMALL:
12561 case SAVEt_I16: /* I16 reference */
12562 case SAVEt_I8: /* I8 reference */
12564 ptr = POPPTR(ss,ix);
12565 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12567 case SAVEt_GENERIC_PVREF: /* generic char* */
12568 case SAVEt_PPTR: /* char* reference */
12569 ptr = POPPTR(ss,ix);
12570 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12571 c = (char*)POPPTR(ss,ix);
12572 TOPPTR(nss,ix) = pv_dup(c);
12574 case SAVEt_GP: /* scalar reference */
12575 gp = (GP*)POPPTR(ss,ix);
12576 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12577 (void)GpREFCNT_inc(gp);
12578 gv = (const GV *)POPPTR(ss,ix);
12579 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12582 ptr = POPPTR(ss,ix);
12583 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12584 /* these are assumed to be refcounted properly */
12586 switch (((OP*)ptr)->op_type) {
12588 case OP_LEAVESUBLV:
12592 case OP_LEAVEWRITE:
12593 TOPPTR(nss,ix) = ptr;
12596 (void) OpREFCNT_inc(o);
12600 TOPPTR(nss,ix) = NULL;
12605 TOPPTR(nss,ix) = NULL;
12607 case SAVEt_FREECOPHH:
12608 ptr = POPPTR(ss,ix);
12609 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12612 hv = (const HV *)POPPTR(ss,ix);
12613 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12615 TOPINT(nss,ix) = i;
12618 c = (char*)POPPTR(ss,ix);
12619 TOPPTR(nss,ix) = pv_dup_inc(c);
12621 case SAVEt_STACK_POS: /* Position on Perl stack */
12623 TOPINT(nss,ix) = i;
12625 case SAVEt_DESTRUCTOR:
12626 ptr = POPPTR(ss,ix);
12627 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12628 dptr = POPDPTR(ss,ix);
12629 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12630 any_dup(FPTR2DPTR(void *, dptr),
12633 case SAVEt_DESTRUCTOR_X:
12634 ptr = POPPTR(ss,ix);
12635 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12636 dxptr = POPDXPTR(ss,ix);
12637 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12638 any_dup(FPTR2DPTR(void *, dxptr),
12641 case SAVEt_REGCONTEXT:
12643 ix -= uv >> SAVE_TIGHT_SHIFT;
12645 case SAVEt_AELEM: /* array element */
12646 sv = (const SV *)POPPTR(ss,ix);
12647 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12649 TOPINT(nss,ix) = i;
12650 av = (const AV *)POPPTR(ss,ix);
12651 TOPPTR(nss,ix) = av_dup_inc(av, param);
12654 ptr = POPPTR(ss,ix);
12655 TOPPTR(nss,ix) = ptr;
12658 ptr = POPPTR(ss,ix);
12659 ptr = cophh_copy((COPHH*)ptr);
12660 TOPPTR(nss,ix) = ptr;
12662 TOPINT(nss,ix) = i;
12663 if (i & HINT_LOCALIZE_HH) {
12664 hv = (const HV *)POPPTR(ss,ix);
12665 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12668 case SAVEt_PADSV_AND_MORTALIZE:
12669 longval = (long)POPLONG(ss,ix);
12670 TOPLONG(nss,ix) = longval;
12671 ptr = POPPTR(ss,ix);
12672 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12673 sv = (const SV *)POPPTR(ss,ix);
12674 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12676 case SAVEt_SET_SVFLAGS:
12678 TOPINT(nss,ix) = i;
12680 TOPINT(nss,ix) = i;
12681 sv = (const SV *)POPPTR(ss,ix);
12682 TOPPTR(nss,ix) = sv_dup(sv, param);
12684 case SAVEt_RE_STATE:
12686 const struct re_save_state *const old_state
12687 = (struct re_save_state *)
12688 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12689 struct re_save_state *const new_state
12690 = (struct re_save_state *)
12691 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12693 Copy(old_state, new_state, 1, struct re_save_state);
12694 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12696 new_state->re_state_bostr
12697 = pv_dup(old_state->re_state_bostr);
12698 new_state->re_state_regeol
12699 = pv_dup(old_state->re_state_regeol);
12700 #ifdef PERL_OLD_COPY_ON_WRITE
12701 new_state->re_state_nrs
12702 = sv_dup(old_state->re_state_nrs, param);
12704 new_state->re_state_reg_magic
12705 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12707 new_state->re_state_reg_oldcurpm
12708 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12710 new_state->re_state_reg_curpm
12711 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12713 new_state->re_state_reg_oldsaved
12714 = pv_dup(old_state->re_state_reg_oldsaved);
12715 new_state->re_state_reg_poscache
12716 = pv_dup(old_state->re_state_reg_poscache);
12717 new_state->re_state_reg_starttry
12718 = pv_dup(old_state->re_state_reg_starttry);
12721 case SAVEt_COMPILE_WARNINGS:
12722 ptr = POPPTR(ss,ix);
12723 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12726 ptr = POPPTR(ss,ix);
12727 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12731 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12739 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12740 * flag to the result. This is done for each stash before cloning starts,
12741 * so we know which stashes want their objects cloned */
12744 do_mark_cloneable_stash(pTHX_ SV *const sv)
12746 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12748 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12749 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12750 if (cloner && GvCV(cloner)) {
12757 mXPUSHs(newSVhek(hvname));
12759 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12766 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12774 =for apidoc perl_clone
12776 Create and return a new interpreter by cloning the current one.
12778 perl_clone takes these flags as parameters:
12780 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12781 without it we only clone the data and zero the stacks,
12782 with it we copy the stacks and the new perl interpreter is
12783 ready to run at the exact same point as the previous one.
12784 The pseudo-fork code uses COPY_STACKS while the
12785 threads->create doesn't.
12787 CLONEf_KEEP_PTR_TABLE -
12788 perl_clone keeps a ptr_table with the pointer of the old
12789 variable as a key and the new variable as a value,
12790 this allows it to check if something has been cloned and not
12791 clone it again but rather just use the value and increase the
12792 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12793 the ptr_table using the function
12794 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12795 reason to keep it around is if you want to dup some of your own
12796 variable who are outside the graph perl scans, example of this
12797 code is in threads.xs create.
12799 CLONEf_CLONE_HOST -
12800 This is a win32 thing, it is ignored on unix, it tells perls
12801 win32host code (which is c++) to clone itself, this is needed on
12802 win32 if you want to run two threads at the same time,
12803 if you just want to do some stuff in a separate perl interpreter
12804 and then throw it away and return to the original one,
12805 you don't need to do anything.
12810 /* XXX the above needs expanding by someone who actually understands it ! */
12811 EXTERN_C PerlInterpreter *
12812 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12815 perl_clone(PerlInterpreter *proto_perl, UV flags)
12818 #ifdef PERL_IMPLICIT_SYS
12820 PERL_ARGS_ASSERT_PERL_CLONE;
12822 /* perlhost.h so we need to call into it
12823 to clone the host, CPerlHost should have a c interface, sky */
12825 if (flags & CLONEf_CLONE_HOST) {
12826 return perl_clone_host(proto_perl,flags);
12828 return perl_clone_using(proto_perl, flags,
12830 proto_perl->IMemShared,
12831 proto_perl->IMemParse,
12833 proto_perl->IStdIO,
12837 proto_perl->IProc);
12841 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12842 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12843 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12844 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12845 struct IPerlDir* ipD, struct IPerlSock* ipS,
12846 struct IPerlProc* ipP)
12848 /* XXX many of the string copies here can be optimized if they're
12849 * constants; they need to be allocated as common memory and just
12850 * their pointers copied. */
12853 CLONE_PARAMS clone_params;
12854 CLONE_PARAMS* const param = &clone_params;
12856 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12858 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12859 #else /* !PERL_IMPLICIT_SYS */
12861 CLONE_PARAMS clone_params;
12862 CLONE_PARAMS* param = &clone_params;
12863 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12865 PERL_ARGS_ASSERT_PERL_CLONE;
12866 #endif /* PERL_IMPLICIT_SYS */
12868 /* for each stash, determine whether its objects should be cloned */
12869 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12870 PERL_SET_THX(my_perl);
12873 PoisonNew(my_perl, 1, PerlInterpreter);
12876 PL_defstash = NULL; /* may be used by perl malloc() */
12879 PL_scopestack_name = 0;
12881 PL_savestack_ix = 0;
12882 PL_savestack_max = -1;
12883 PL_sig_pending = 0;
12885 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12886 # ifdef DEBUG_LEAKING_SCALARS
12887 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12889 #else /* !DEBUGGING */
12890 Zero(my_perl, 1, PerlInterpreter);
12891 #endif /* DEBUGGING */
12893 #ifdef PERL_IMPLICIT_SYS
12894 /* host pointers */
12896 PL_MemShared = ipMS;
12897 PL_MemParse = ipMP;
12904 #endif /* PERL_IMPLICIT_SYS */
12906 param->flags = flags;
12907 /* Nothing in the core code uses this, but we make it available to
12908 extensions (using mg_dup). */
12909 param->proto_perl = proto_perl;
12910 /* Likely nothing will use this, but it is initialised to be consistent
12911 with Perl_clone_params_new(). */
12912 param->new_perl = my_perl;
12913 param->unreferenced = NULL;
12915 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12917 PL_body_arenas = NULL;
12918 Zero(&PL_body_roots, 1, PL_body_roots);
12921 PL_sv_objcount = 0;
12923 PL_sv_arenaroot = NULL;
12925 PL_debug = proto_perl->Idebug;
12927 PL_hash_seed = proto_perl->Ihash_seed;
12928 PL_rehash_seed = proto_perl->Irehash_seed;
12930 /* dbargs array probably holds garbage */
12933 PL_compiling = proto_perl->Icompiling;
12935 /* pseudo environmental stuff */
12936 PL_origargc = proto_perl->Iorigargc;
12937 PL_origargv = proto_perl->Iorigargv;
12939 #if !NO_TAINT_SUPPORT
12940 /* Set tainting stuff before PerlIO_debug can possibly get called */
12941 PL_tainting = proto_perl->Itainting;
12942 PL_taint_warn = proto_perl->Itaint_warn;
12944 PL_tainting = FALSE;
12945 PL_taint_warn = FALSE;
12948 PL_minus_c = proto_perl->Iminus_c;
12950 PL_localpatches = proto_perl->Ilocalpatches;
12951 PL_splitstr = proto_perl->Isplitstr;
12952 PL_minus_n = proto_perl->Iminus_n;
12953 PL_minus_p = proto_perl->Iminus_p;
12954 PL_minus_l = proto_perl->Iminus_l;
12955 PL_minus_a = proto_perl->Iminus_a;
12956 PL_minus_E = proto_perl->Iminus_E;
12957 PL_minus_F = proto_perl->Iminus_F;
12958 PL_doswitches = proto_perl->Idoswitches;
12959 PL_dowarn = proto_perl->Idowarn;
12960 PL_sawampersand = proto_perl->Isawampersand;
12961 PL_unsafe = proto_perl->Iunsafe;
12962 PL_perldb = proto_perl->Iperldb;
12963 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12964 PL_exit_flags = proto_perl->Iexit_flags;
12966 /* XXX time(&PL_basetime) when asked for? */
12967 PL_basetime = proto_perl->Ibasetime;
12969 PL_maxsysfd = proto_perl->Imaxsysfd;
12970 PL_statusvalue = proto_perl->Istatusvalue;
12972 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12974 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12977 /* RE engine related */
12978 Zero(&PL_reg_state, 1, struct re_save_state);
12979 PL_regmatch_slab = NULL;
12981 PL_sub_generation = proto_perl->Isub_generation;
12983 /* funky return mechanisms */
12984 PL_forkprocess = proto_perl->Iforkprocess;
12986 /* internal state */
12987 PL_maxo = proto_perl->Imaxo;
12989 PL_main_start = proto_perl->Imain_start;
12990 PL_eval_root = proto_perl->Ieval_root;
12991 PL_eval_start = proto_perl->Ieval_start;
12993 PL_filemode = proto_perl->Ifilemode;
12994 PL_lastfd = proto_perl->Ilastfd;
12995 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12998 PL_gensym = proto_perl->Igensym;
13000 PL_laststatval = proto_perl->Ilaststatval;
13001 PL_laststype = proto_perl->Ilaststype;
13004 PL_profiledata = NULL;
13006 PL_generation = proto_perl->Igeneration;
13008 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13009 PL_in_clean_all = proto_perl->Iin_clean_all;
13011 PL_delaymagic_uid = proto_perl->Idelaymagic_uid;
13012 PL_delaymagic_euid = proto_perl->Idelaymagic_euid;
13013 PL_delaymagic_gid = proto_perl->Idelaymagic_gid;
13014 PL_delaymagic_egid = proto_perl->Idelaymagic_egid;
13015 PL_nomemok = proto_perl->Inomemok;
13016 PL_an = proto_perl->Ian;
13017 PL_evalseq = proto_perl->Ievalseq;
13018 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13019 PL_origalen = proto_perl->Iorigalen;
13021 PL_sighandlerp = proto_perl->Isighandlerp;
13023 PL_runops = proto_perl->Irunops;
13025 PL_subline = proto_perl->Isubline;
13028 PL_cryptseen = proto_perl->Icryptseen;
13031 PL_hints = proto_perl->Ihints;
13033 #ifdef USE_LOCALE_COLLATE
13034 PL_collation_ix = proto_perl->Icollation_ix;
13035 PL_collation_standard = proto_perl->Icollation_standard;
13036 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13037 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13038 #endif /* USE_LOCALE_COLLATE */
13040 #ifdef USE_LOCALE_NUMERIC
13041 PL_numeric_standard = proto_perl->Inumeric_standard;
13042 PL_numeric_local = proto_perl->Inumeric_local;
13043 #endif /* !USE_LOCALE_NUMERIC */
13045 /* Did the locale setup indicate UTF-8? */
13046 PL_utf8locale = proto_perl->Iutf8locale;
13047 /* Unicode features (see perlrun/-C) */
13048 PL_unicode = proto_perl->Iunicode;
13050 /* Pre-5.8 signals control */
13051 PL_signals = proto_perl->Isignals;
13053 /* times() ticks per second */
13054 PL_clocktick = proto_perl->Iclocktick;
13056 /* Recursion stopper for PerlIO_find_layer */
13057 PL_in_load_module = proto_perl->Iin_load_module;
13059 /* sort() routine */
13060 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13062 /* Not really needed/useful since the reenrant_retint is "volatile",
13063 * but do it for consistency's sake. */
13064 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13066 /* Hooks to shared SVs and locks. */
13067 PL_sharehook = proto_perl->Isharehook;
13068 PL_lockhook = proto_perl->Ilockhook;
13069 PL_unlockhook = proto_perl->Iunlockhook;
13070 PL_threadhook = proto_perl->Ithreadhook;
13071 PL_destroyhook = proto_perl->Idestroyhook;
13072 PL_signalhook = proto_perl->Isignalhook;
13074 PL_globhook = proto_perl->Iglobhook;
13077 PL_last_swash_hv = NULL; /* reinits on demand */
13078 PL_last_swash_klen = 0;
13079 PL_last_swash_key[0]= '\0';
13080 PL_last_swash_tmps = (U8*)NULL;
13081 PL_last_swash_slen = 0;
13083 PL_glob_index = proto_perl->Iglob_index;
13084 PL_srand_called = proto_perl->Isrand_called;
13086 if (flags & CLONEf_COPY_STACKS) {
13087 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13088 PL_tmps_ix = proto_perl->Itmps_ix;
13089 PL_tmps_max = proto_perl->Itmps_max;
13090 PL_tmps_floor = proto_perl->Itmps_floor;
13092 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13093 * NOTE: unlike the others! */
13094 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13095 PL_scopestack_max = proto_perl->Iscopestack_max;
13097 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13098 * NOTE: unlike the others! */
13099 PL_savestack_ix = proto_perl->Isavestack_ix;
13100 PL_savestack_max = proto_perl->Isavestack_max;
13103 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13104 PL_top_env = &PL_start_env;
13106 PL_op = proto_perl->Iop;
13109 PL_Xpv = (XPV*)NULL;
13110 my_perl->Ina = proto_perl->Ina;
13112 PL_statbuf = proto_perl->Istatbuf;
13113 PL_statcache = proto_perl->Istatcache;
13116 PL_timesbuf = proto_perl->Itimesbuf;
13119 #if !NO_TAINT_SUPPORT
13120 PL_tainted = proto_perl->Itainted;
13122 PL_tainted = FALSE;
13124 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13126 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13128 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13129 PL_restartop = proto_perl->Irestartop;
13130 PL_in_eval = proto_perl->Iin_eval;
13131 PL_delaymagic = proto_perl->Idelaymagic;
13132 PL_phase = proto_perl->Iphase;
13133 PL_localizing = proto_perl->Ilocalizing;
13135 PL_hv_fetch_ent_mh = NULL;
13136 PL_modcount = proto_perl->Imodcount;
13137 PL_lastgotoprobe = NULL;
13138 PL_dumpindent = proto_perl->Idumpindent;
13140 PL_efloatbuf = NULL; /* reinits on demand */
13141 PL_efloatsize = 0; /* reinits on demand */
13145 PL_regdummy = proto_perl->Iregdummy;
13146 PL_colorset = 0; /* reinits PL_colors[] */
13147 /*PL_colors[6] = {0,0,0,0,0,0};*/
13149 /* Pluggable optimizer */
13150 PL_peepp = proto_perl->Ipeepp;
13151 PL_rpeepp = proto_perl->Irpeepp;
13152 /* op_free() hook */
13153 PL_opfreehook = proto_perl->Iopfreehook;
13155 #ifdef USE_REENTRANT_API
13156 /* XXX: things like -Dm will segfault here in perlio, but doing
13157 * PERL_SET_CONTEXT(proto_perl);
13158 * breaks too many other things
13160 Perl_reentrant_init(aTHX);
13163 /* create SV map for pointer relocation */
13164 PL_ptr_table = ptr_table_new();
13166 /* initialize these special pointers as early as possible */
13168 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13169 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13170 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13172 /* create (a non-shared!) shared string table */
13173 PL_strtab = newHV();
13174 HvSHAREKEYS_off(PL_strtab);
13175 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13176 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13178 /* This PV will be free'd special way so must set it same way op.c does */
13179 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13180 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13182 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13183 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13184 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13185 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13187 param->stashes = newAV(); /* Setup array of objects to call clone on */
13188 /* This makes no difference to the implementation, as it always pushes
13189 and shifts pointers to other SVs without changing their reference
13190 count, with the array becoming empty before it is freed. However, it
13191 makes it conceptually clear what is going on, and will avoid some
13192 work inside av.c, filling slots between AvFILL() and AvMAX() with
13193 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13194 AvREAL_off(param->stashes);
13196 if (!(flags & CLONEf_COPY_STACKS)) {
13197 param->unreferenced = newAV();
13200 #ifdef PERLIO_LAYERS
13201 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13202 PerlIO_clone(aTHX_ proto_perl, param);
13205 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13206 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13207 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13208 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13209 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13210 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13213 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13214 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13215 PL_inplace = SAVEPV(proto_perl->Iinplace);
13216 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13218 /* magical thingies */
13220 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13222 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13223 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13224 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13227 /* Clone the regex array */
13228 /* ORANGE FIXME for plugins, probably in the SV dup code.
13229 newSViv(PTR2IV(CALLREGDUPE(
13230 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13232 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13233 PL_regex_pad = AvARRAY(PL_regex_padav);
13235 PL_stashpadmax = proto_perl->Istashpadmax;
13236 PL_stashpadix = proto_perl->Istashpadix ;
13237 Newx(PL_stashpad, PL_stashpadmax, HV *);
13240 for (; o < PL_stashpadmax; ++o)
13241 PL_stashpad[o] = hv_dup(proto_perl->Istashpad[o], param);
13244 /* shortcuts to various I/O objects */
13245 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13246 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13247 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13248 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13249 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13250 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13251 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13253 /* shortcuts to regexp stuff */
13254 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13256 /* shortcuts to misc objects */
13257 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13259 /* shortcuts to debugging objects */
13260 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13261 PL_DBline = gv_dup(proto_perl->IDBline, param);
13262 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13263 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13264 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13265 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13267 /* symbol tables */
13268 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13269 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13270 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13271 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13272 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13274 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13275 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13276 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13277 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13278 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13279 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13280 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13281 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13283 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13285 /* subprocess state */
13286 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13288 if (proto_perl->Iop_mask)
13289 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13292 /* PL_asserting = proto_perl->Iasserting; */
13294 /* current interpreter roots */
13295 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13297 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13300 /* runtime control stuff */
13301 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13303 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13305 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13307 /* interpreter atexit processing */
13308 PL_exitlistlen = proto_perl->Iexitlistlen;
13309 if (PL_exitlistlen) {
13310 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13311 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13314 PL_exitlist = (PerlExitListEntry*)NULL;
13316 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13317 if (PL_my_cxt_size) {
13318 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13319 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13320 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13321 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13322 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13326 PL_my_cxt_list = (void**)NULL;
13327 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13328 PL_my_cxt_keys = (const char**)NULL;
13331 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13332 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13333 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13334 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13336 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13338 PAD_CLONE_VARS(proto_perl, param);
13340 #ifdef HAVE_INTERP_INTERN
13341 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13344 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13346 #ifdef PERL_USES_PL_PIDSTATUS
13347 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13349 PL_osname = SAVEPV(proto_perl->Iosname);
13350 PL_parser = parser_dup(proto_perl->Iparser, param);
13352 /* XXX this only works if the saved cop has already been cloned */
13353 if (proto_perl->Iparser) {
13354 PL_parser->saved_curcop = (COP*)any_dup(
13355 proto_perl->Iparser->saved_curcop,
13359 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13361 #ifdef USE_LOCALE_COLLATE
13362 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13363 #endif /* USE_LOCALE_COLLATE */
13365 #ifdef USE_LOCALE_NUMERIC
13366 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13367 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13368 #endif /* !USE_LOCALE_NUMERIC */
13370 /* Unicode inversion lists */
13371 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13372 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13374 PL_PerlSpace = sv_dup_inc(proto_perl->IPerlSpace, param);
13375 PL_XPerlSpace = sv_dup_inc(proto_perl->IXPerlSpace, param);
13377 PL_L1PosixAlnum = sv_dup_inc(proto_perl->IL1PosixAlnum, param);
13378 PL_PosixAlnum = sv_dup_inc(proto_perl->IPosixAlnum, param);
13380 PL_L1PosixAlpha = sv_dup_inc(proto_perl->IL1PosixAlpha, param);
13381 PL_PosixAlpha = sv_dup_inc(proto_perl->IPosixAlpha, param);
13383 PL_PosixBlank = sv_dup_inc(proto_perl->IPosixBlank, param);
13384 PL_XPosixBlank = sv_dup_inc(proto_perl->IXPosixBlank, param);
13386 PL_L1Cased = sv_dup_inc(proto_perl->IL1Cased, param);
13388 PL_PosixCntrl = sv_dup_inc(proto_perl->IPosixCntrl, param);
13389 PL_XPosixCntrl = sv_dup_inc(proto_perl->IXPosixCntrl, param);
13391 PL_PosixDigit = sv_dup_inc(proto_perl->IPosixDigit, param);
13393 PL_L1PosixGraph = sv_dup_inc(proto_perl->IL1PosixGraph, param);
13394 PL_PosixGraph = sv_dup_inc(proto_perl->IPosixGraph, param);
13396 PL_L1PosixLower = sv_dup_inc(proto_perl->IL1PosixLower, param);
13397 PL_PosixLower = sv_dup_inc(proto_perl->IPosixLower, param);
13399 PL_L1PosixPrint = sv_dup_inc(proto_perl->IL1PosixPrint, param);
13400 PL_PosixPrint = sv_dup_inc(proto_perl->IPosixPrint, param);
13402 PL_L1PosixPunct = sv_dup_inc(proto_perl->IL1PosixPunct, param);
13403 PL_PosixPunct = sv_dup_inc(proto_perl->IPosixPunct, param);
13405 PL_PosixSpace = sv_dup_inc(proto_perl->IPosixSpace, param);
13406 PL_XPosixSpace = sv_dup_inc(proto_perl->IXPosixSpace, param);
13408 PL_L1PosixUpper = sv_dup_inc(proto_perl->IL1PosixUpper, param);
13409 PL_PosixUpper = sv_dup_inc(proto_perl->IPosixUpper, param);
13411 PL_L1PosixWord = sv_dup_inc(proto_perl->IL1PosixWord, param);
13412 PL_PosixWord = sv_dup_inc(proto_perl->IPosixWord, param);
13414 PL_PosixXDigit = sv_dup_inc(proto_perl->IPosixXDigit, param);
13415 PL_XPosixXDigit = sv_dup_inc(proto_perl->IXPosixXDigit, param);
13417 PL_VertSpace = sv_dup_inc(proto_perl->IVertSpace, param);
13419 PL_NonL1NonFinalFold = sv_dup_inc(proto_perl->INonL1NonFinalFold, param);
13420 PL_HasMultiCharFold= sv_dup_inc(proto_perl->IHasMultiCharFold, param);
13422 /* utf8 character class swashes */
13423 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13424 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13425 PL_utf8_blank = sv_dup_inc(proto_perl->Iutf8_blank, param);
13426 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13427 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13428 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13429 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13430 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13431 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13432 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13433 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13434 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13435 PL_utf8_X_regular_begin = sv_dup_inc(proto_perl->Iutf8_X_regular_begin, param);
13436 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13437 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13438 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13439 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13440 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13441 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13442 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13443 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13444 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13445 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13446 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13447 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13448 PL_utf8_charname_begin = sv_dup_inc(proto_perl->Iutf8_charname_begin, param);
13449 PL_utf8_charname_continue = sv_dup_inc(proto_perl->Iutf8_charname_continue, param);
13450 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13451 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13452 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13454 if (proto_perl->Ipsig_pend) {
13455 Newxz(PL_psig_pend, SIG_SIZE, int);
13458 PL_psig_pend = (int*)NULL;
13461 if (proto_perl->Ipsig_name) {
13462 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13463 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13465 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13468 PL_psig_ptr = (SV**)NULL;
13469 PL_psig_name = (SV**)NULL;
13472 if (flags & CLONEf_COPY_STACKS) {
13473 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13474 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13475 PL_tmps_ix+1, param);
13477 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13478 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13479 Newxz(PL_markstack, i, I32);
13480 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13481 - proto_perl->Imarkstack);
13482 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13483 - proto_perl->Imarkstack);
13484 Copy(proto_perl->Imarkstack, PL_markstack,
13485 PL_markstack_ptr - PL_markstack + 1, I32);
13487 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13488 * NOTE: unlike the others! */
13489 Newxz(PL_scopestack, PL_scopestack_max, I32);
13490 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13493 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13494 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13496 /* NOTE: si_dup() looks at PL_markstack */
13497 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13499 /* PL_curstack = PL_curstackinfo->si_stack; */
13500 PL_curstack = av_dup(proto_perl->Icurstack, param);
13501 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13503 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13504 PL_stack_base = AvARRAY(PL_curstack);
13505 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13506 - proto_perl->Istack_base);
13507 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13509 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13510 PL_savestack = ss_dup(proto_perl, param);
13514 ENTER; /* perl_destruct() wants to LEAVE; */
13517 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13518 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13520 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13521 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13522 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13523 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13524 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13525 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13527 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13529 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13530 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13531 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13532 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13534 PL_stashcache = newHV();
13536 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13537 proto_perl->Iwatchaddr);
13538 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13539 if (PL_debug && PL_watchaddr) {
13540 PerlIO_printf(Perl_debug_log,
13541 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13542 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13543 PTR2UV(PL_watchok));
13546 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13547 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13548 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13550 /* Call the ->CLONE method, if it exists, for each of the stashes
13551 identified by sv_dup() above.
13553 while(av_len(param->stashes) != -1) {
13554 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13555 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13556 if (cloner && GvCV(cloner)) {
13561 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13563 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13569 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13570 ptr_table_free(PL_ptr_table);
13571 PL_ptr_table = NULL;
13574 if (!(flags & CLONEf_COPY_STACKS)) {
13575 unreferenced_to_tmp_stack(param->unreferenced);
13578 SvREFCNT_dec(param->stashes);
13580 /* orphaned? eg threads->new inside BEGIN or use */
13581 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13582 SvREFCNT_inc_simple_void(PL_compcv);
13583 SAVEFREESV(PL_compcv);
13590 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13592 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13594 if (AvFILLp(unreferenced) > -1) {
13595 SV **svp = AvARRAY(unreferenced);
13596 SV **const last = svp + AvFILLp(unreferenced);
13600 if (SvREFCNT(*svp) == 1)
13602 } while (++svp <= last);
13604 EXTEND_MORTAL(count);
13605 svp = AvARRAY(unreferenced);
13608 if (SvREFCNT(*svp) == 1) {
13609 /* Our reference is the only one to this SV. This means that
13610 in this thread, the scalar effectively has a 0 reference.
13611 That doesn't work (cleanup never happens), so donate our
13612 reference to it onto the save stack. */
13613 PL_tmps_stack[++PL_tmps_ix] = *svp;
13615 /* As an optimisation, because we are already walking the
13616 entire array, instead of above doing either
13617 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13618 release our reference to the scalar, so that at the end of
13619 the array owns zero references to the scalars it happens to
13620 point to. We are effectively converting the array from
13621 AvREAL() on to AvREAL() off. This saves the av_clear()
13622 (triggered by the SvREFCNT_dec(unreferenced) below) from
13623 walking the array a second time. */
13624 SvREFCNT_dec(*svp);
13627 } while (++svp <= last);
13628 AvREAL_off(unreferenced);
13630 SvREFCNT_dec(unreferenced);
13634 Perl_clone_params_del(CLONE_PARAMS *param)
13636 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13638 PerlInterpreter *const to = param->new_perl;
13640 PerlInterpreter *const was = PERL_GET_THX;
13642 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13648 SvREFCNT_dec(param->stashes);
13649 if (param->unreferenced)
13650 unreferenced_to_tmp_stack(param->unreferenced);
13660 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13663 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13664 does a dTHX; to get the context from thread local storage.
13665 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13666 a version that passes in my_perl. */
13667 PerlInterpreter *const was = PERL_GET_THX;
13668 CLONE_PARAMS *param;
13670 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13676 /* Given that we've set the context, we can do this unshared. */
13677 Newx(param, 1, CLONE_PARAMS);
13680 param->proto_perl = from;
13681 param->new_perl = to;
13682 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13683 AvREAL_off(param->stashes);
13684 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13692 #endif /* USE_ITHREADS */
13695 Perl_init_constants(pTHX)
13697 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
13698 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
13699 SvANY(&PL_sv_undef) = NULL;
13701 SvANY(&PL_sv_no) = new_XPVNV();
13702 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
13703 SvFLAGS(&PL_sv_no) = SVt_PVNV|SVf_READONLY
13704 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13707 SvANY(&PL_sv_yes) = new_XPVNV();
13708 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
13709 SvFLAGS(&PL_sv_yes) = SVt_PVNV|SVf_READONLY
13710 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13713 SvPV_set(&PL_sv_no, (char*)PL_No);
13714 SvCUR_set(&PL_sv_no, 0);
13715 SvLEN_set(&PL_sv_no, 0);
13716 SvIV_set(&PL_sv_no, 0);
13717 SvNV_set(&PL_sv_no, 0);
13719 SvPV_set(&PL_sv_yes, (char*)PL_Yes);
13720 SvCUR_set(&PL_sv_yes, 1);
13721 SvLEN_set(&PL_sv_yes, 0);
13722 SvIV_set(&PL_sv_yes, 1);
13723 SvNV_set(&PL_sv_yes, 1);
13727 =head1 Unicode Support
13729 =for apidoc sv_recode_to_utf8
13731 The encoding is assumed to be an Encode object, on entry the PV
13732 of the sv is assumed to be octets in that encoding, and the sv
13733 will be converted into Unicode (and UTF-8).
13735 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13736 is not a reference, nothing is done to the sv. If the encoding is not
13737 an C<Encode::XS> Encoding object, bad things will happen.
13738 (See F<lib/encoding.pm> and L<Encode>.)
13740 The PV of the sv is returned.
13745 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13749 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13751 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13765 Passing sv_yes is wrong - it needs to be or'ed set of constants
13766 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13767 remove converted chars from source.
13769 Both will default the value - let them.
13771 XPUSHs(&PL_sv_yes);
13774 call_method("decode", G_SCALAR);
13778 s = SvPV_const(uni, len);
13779 if (s != SvPVX_const(sv)) {
13780 SvGROW(sv, len + 1);
13781 Move(s, SvPVX(sv), len + 1, char);
13782 SvCUR_set(sv, len);
13786 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13787 /* clear pos and any utf8 cache */
13788 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13791 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13792 magic_setutf8(sv,mg); /* clear UTF8 cache */
13797 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13801 =for apidoc sv_cat_decode
13803 The encoding is assumed to be an Encode object, the PV of the ssv is
13804 assumed to be octets in that encoding and decoding the input starts
13805 from the position which (PV + *offset) pointed to. The dsv will be
13806 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13807 when the string tstr appears in decoding output or the input ends on
13808 the PV of the ssv. The value which the offset points will be modified
13809 to the last input position on the ssv.
13811 Returns TRUE if the terminator was found, else returns FALSE.
13816 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13817 SV *ssv, int *offset, char *tstr, int tlen)
13822 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13824 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13835 offsv = newSViv(*offset);
13837 mPUSHp(tstr, tlen);
13839 call_method("cat_decode", G_SCALAR);
13841 ret = SvTRUE(TOPs);
13842 *offset = SvIV(offsv);
13848 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13853 /* ---------------------------------------------------------------------
13855 * support functions for report_uninit()
13858 /* the maxiumum size of array or hash where we will scan looking
13859 * for the undefined element that triggered the warning */
13861 #define FUV_MAX_SEARCH_SIZE 1000
13863 /* Look for an entry in the hash whose value has the same SV as val;
13864 * If so, return a mortal copy of the key. */
13867 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13873 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13875 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13876 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13879 array = HvARRAY(hv);
13881 for (i=HvMAX(hv); i>=0; i--) {
13883 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13884 if (HeVAL(entry) != val)
13886 if ( HeVAL(entry) == &PL_sv_undef ||
13887 HeVAL(entry) == &PL_sv_placeholder)
13891 if (HeKLEN(entry) == HEf_SVKEY)
13892 return sv_mortalcopy(HeKEY_sv(entry));
13893 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13899 /* Look for an entry in the array whose value has the same SV as val;
13900 * If so, return the index, otherwise return -1. */
13903 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13907 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13909 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13910 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13913 if (val != &PL_sv_undef) {
13914 SV ** const svp = AvARRAY(av);
13917 for (i=AvFILLp(av); i>=0; i--)
13924 /* varname(): return the name of a variable, optionally with a subscript.
13925 * If gv is non-zero, use the name of that global, along with gvtype (one
13926 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13927 * targ. Depending on the value of the subscript_type flag, return:
13930 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13931 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13932 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13933 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13936 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13937 const SV *const keyname, I32 aindex, int subscript_type)
13940 SV * const name = sv_newmortal();
13941 if (gv && isGV(gv)) {
13943 buffer[0] = gvtype;
13946 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13948 gv_fullname4(name, gv, buffer, 0);
13950 if ((unsigned int)SvPVX(name)[1] <= 26) {
13952 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13954 /* Swap the 1 unprintable control character for the 2 byte pretty
13955 version - ie substr($name, 1, 1) = $buffer; */
13956 sv_insert(name, 1, 1, buffer, 2);
13960 CV * const cv = gv ? ((CV *)gv) : find_runcv(NULL);
13964 assert(!cv || SvTYPE(cv) == SVt_PVCV || SvTYPE(cv) == SVt_PVFM);
13966 if (!cv || !CvPADLIST(cv))
13968 av = *PadlistARRAY(CvPADLIST(cv));
13969 sv = *av_fetch(av, targ, FALSE);
13970 sv_setsv_flags(name, sv, 0);
13973 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13974 SV * const sv = newSV(0);
13975 *SvPVX(name) = '$';
13976 Perl_sv_catpvf(aTHX_ name, "{%s}",
13977 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
13978 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
13981 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13982 *SvPVX(name) = '$';
13983 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13985 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13986 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13987 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13995 =for apidoc find_uninit_var
13997 Find the name of the undefined variable (if any) that caused the operator
13998 to issue a "Use of uninitialized value" warning.
13999 If match is true, only return a name if its value matches uninit_sv.
14000 So roughly speaking, if a unary operator (such as OP_COS) generates a
14001 warning, then following the direct child of the op may yield an
14002 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
14003 other hand, with OP_ADD there are two branches to follow, so we only print
14004 the variable name if we get an exact match.
14006 The name is returned as a mortal SV.
14008 Assumes that PL_op is the op that originally triggered the error, and that
14009 PL_comppad/PL_curpad points to the currently executing pad.
14015 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
14021 const OP *o, *o2, *kid;
14023 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
14024 uninit_sv == &PL_sv_placeholder)))
14027 switch (obase->op_type) {
14034 const bool pad = ( obase->op_type == OP_PADAV
14035 || obase->op_type == OP_PADHV
14036 || obase->op_type == OP_PADRANGE
14039 const bool hash = ( obase->op_type == OP_PADHV
14040 || obase->op_type == OP_RV2HV
14041 || (obase->op_type == OP_PADRANGE
14042 && SvTYPE(PAD_SVl(obase->op_targ)) == SVt_PVHV)
14046 int subscript_type = FUV_SUBSCRIPT_WITHIN;
14048 if (pad) { /* @lex, %lex */
14049 sv = PAD_SVl(obase->op_targ);
14053 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14054 /* @global, %global */
14055 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14058 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14060 else if (obase == PL_op) /* @{expr}, %{expr} */
14061 return find_uninit_var(cUNOPx(obase)->op_first,
14063 else /* @{expr}, %{expr} as a sub-expression */
14067 /* attempt to find a match within the aggregate */
14069 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14071 subscript_type = FUV_SUBSCRIPT_HASH;
14074 index = find_array_subscript((const AV *)sv, uninit_sv);
14076 subscript_type = FUV_SUBSCRIPT_ARRAY;
14079 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14082 return varname(gv, hash ? '%' : '@', obase->op_targ,
14083 keysv, index, subscript_type);
14087 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14089 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14090 if (!gv || !GvSTASH(gv))
14092 if (match && (GvSV(gv) != uninit_sv))
14094 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14097 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14100 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14102 return varname(NULL, '$', obase->op_targ,
14103 NULL, 0, FUV_SUBSCRIPT_NONE);
14106 gv = cGVOPx_gv(obase);
14107 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14109 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14111 case OP_AELEMFAST_LEX:
14114 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14115 if (!av || SvRMAGICAL(av))
14117 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14118 if (!svp || *svp != uninit_sv)
14121 return varname(NULL, '$', obase->op_targ,
14122 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14125 gv = cGVOPx_gv(obase);
14130 AV *const av = GvAV(gv);
14131 if (!av || SvRMAGICAL(av))
14133 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14134 if (!svp || *svp != uninit_sv)
14137 return varname(gv, '$', 0,
14138 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14143 o = cUNOPx(obase)->op_first;
14144 if (!o || o->op_type != OP_NULL ||
14145 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14147 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14152 bool negate = FALSE;
14154 if (PL_op == obase)
14155 /* $a[uninit_expr] or $h{uninit_expr} */
14156 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14159 o = cBINOPx(obase)->op_first;
14160 kid = cBINOPx(obase)->op_last;
14162 /* get the av or hv, and optionally the gv */
14164 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14165 sv = PAD_SV(o->op_targ);
14167 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14168 && cUNOPo->op_first->op_type == OP_GV)
14170 gv = cGVOPx_gv(cUNOPo->op_first);
14174 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14179 if (kid && kid->op_type == OP_NEGATE) {
14181 kid = cUNOPx(kid)->op_first;
14184 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14185 /* index is constant */
14188 kidsv = sv_2mortal(newSVpvs("-"));
14189 sv_catsv(kidsv, cSVOPx_sv(kid));
14192 kidsv = cSVOPx_sv(kid);
14196 if (obase->op_type == OP_HELEM) {
14197 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14198 if (!he || HeVAL(he) != uninit_sv)
14202 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14203 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14205 if (!svp || *svp != uninit_sv)
14209 if (obase->op_type == OP_HELEM)
14210 return varname(gv, '%', o->op_targ,
14211 kidsv, 0, FUV_SUBSCRIPT_HASH);
14213 return varname(gv, '@', o->op_targ, NULL,
14214 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14215 FUV_SUBSCRIPT_ARRAY);
14218 /* index is an expression;
14219 * attempt to find a match within the aggregate */
14220 if (obase->op_type == OP_HELEM) {
14221 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14223 return varname(gv, '%', o->op_targ,
14224 keysv, 0, FUV_SUBSCRIPT_HASH);
14228 = find_array_subscript((const AV *)sv, uninit_sv);
14230 return varname(gv, '@', o->op_targ,
14231 NULL, index, FUV_SUBSCRIPT_ARRAY);
14236 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14238 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14244 /* only examine RHS */
14245 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14248 o = cUNOPx(obase)->op_first;
14249 if ( o->op_type == OP_PUSHMARK
14250 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)
14254 if (!o->op_sibling) {
14255 /* one-arg version of open is highly magical */
14257 if (o->op_type == OP_GV) { /* open FOO; */
14259 if (match && GvSV(gv) != uninit_sv)
14261 return varname(gv, '$', 0,
14262 NULL, 0, FUV_SUBSCRIPT_NONE);
14264 /* other possibilities not handled are:
14265 * open $x; or open my $x; should return '${*$x}'
14266 * open expr; should return '$'.expr ideally
14272 /* ops where $_ may be an implicit arg */
14277 if ( !(obase->op_flags & OPf_STACKED)) {
14278 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14279 ? PAD_SVl(obase->op_targ)
14282 sv = sv_newmortal();
14283 sv_setpvs(sv, "$_");
14292 match = 1; /* print etc can return undef on defined args */
14293 /* skip filehandle as it can't produce 'undef' warning */
14294 o = cUNOPx(obase)->op_first;
14295 if ((obase->op_flags & OPf_STACKED)
14297 ( o->op_type == OP_PUSHMARK
14298 || (o->op_type == OP_NULL && o->op_targ == OP_PUSHMARK)))
14299 o = o->op_sibling->op_sibling;
14303 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14304 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14306 /* the following ops are capable of returning PL_sv_undef even for
14307 * defined arg(s) */
14326 case OP_GETPEERNAME:
14374 case OP_SMARTMATCH:
14383 /* XXX tmp hack: these two may call an XS sub, and currently
14384 XS subs don't have a SUB entry on the context stack, so CV and
14385 pad determination goes wrong, and BAD things happen. So, just
14386 don't try to determine the value under those circumstances.
14387 Need a better fix at dome point. DAPM 11/2007 */
14393 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14394 if (gv && GvSV(gv) == uninit_sv)
14395 return newSVpvs_flags("$.", SVs_TEMP);
14400 /* def-ness of rval pos() is independent of the def-ness of its arg */
14401 if ( !(obase->op_flags & OPf_MOD))
14406 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14407 return newSVpvs_flags("${$/}", SVs_TEMP);
14412 if (!(obase->op_flags & OPf_KIDS))
14414 o = cUNOPx(obase)->op_first;
14420 /* This loop checks all the kid ops, skipping any that cannot pos-
14421 * sibly be responsible for the uninitialized value; i.e., defined
14422 * constants and ops that return nothing. If there is only one op
14423 * left that is not skipped, then we *know* it is responsible for
14424 * the uninitialized value. If there is more than one op left, we
14425 * have to look for an exact match in the while() loop below.
14426 * Note that we skip padrange, because the individual pad ops that
14427 * it replaced are still in the tree, so we work on them instead.
14430 for (kid=o; kid; kid = kid->op_sibling) {
14432 const OPCODE type = kid->op_type;
14433 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14434 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14435 || (type == OP_PUSHMARK)
14436 || (type == OP_PADRANGE)
14440 if (o2) { /* more than one found */
14447 return find_uninit_var(o2, uninit_sv, match);
14449 /* scan all args */
14451 sv = find_uninit_var(o, uninit_sv, 1);
14463 =for apidoc report_uninit
14465 Print appropriate "Use of uninitialized variable" warning.
14471 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14475 SV* varname = NULL;
14476 if (uninit_sv && PL_curpad) {
14477 varname = find_uninit_var(PL_op, uninit_sv,0);
14479 sv_insert(varname, 0, 0, " ", 1);
14481 /* diag_listed_as: Use of uninitialized value%s */
14482 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14483 SVfARG(varname ? varname : &PL_sv_no),
14484 " in ", OP_DESC(PL_op));
14487 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14493 * c-indentation-style: bsd
14494 * c-basic-offset: 4
14495 * indent-tabs-mode: nil
14498 * ex: set ts=8 sts=4 sw=4 et: