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
38 /* Missing proto on LynxOS */
39 char *gconvert(double, int, int, char *);
42 #ifdef PERL_UTF8_CACHE_ASSERT
43 /* if adding more checks watch out for the following tests:
44 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
45 * lib/utf8.t lib/Unicode/Collate/t/index.t
48 # define ASSERT_UTF8_CACHE(cache) \
49 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
50 assert((cache)[2] <= (cache)[3]); \
51 assert((cache)[3] <= (cache)[1]);} \
54 # define ASSERT_UTF8_CACHE(cache) NOOP
57 #ifdef PERL_OLD_COPY_ON_WRITE
58 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
59 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
60 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
64 /* ============================================================================
66 =head1 Allocation and deallocation of SVs.
68 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
69 sv, av, hv...) contains type and reference count information, and for
70 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
71 contains fields specific to each type. Some types store all they need
72 in the head, so don't have a body.
74 In all but the most memory-paranoid configuations (ex: PURIFY), heads
75 and bodies are allocated out of arenas, which by default are
76 approximately 4K chunks of memory parcelled up into N heads or bodies.
77 Sv-bodies are allocated by their sv-type, guaranteeing size
78 consistency needed to allocate safely from arrays.
80 For SV-heads, the first slot in each arena is reserved, and holds a
81 link to the next arena, some flags, and a note of the number of slots.
82 Snaked through each arena chain is a linked list of free items; when
83 this becomes empty, an extra arena is allocated and divided up into N
84 items which are threaded into the free list.
86 SV-bodies are similar, but they use arena-sets by default, which
87 separate the link and info from the arena itself, and reclaim the 1st
88 slot in the arena. SV-bodies are further described later.
90 The following global variables are associated with arenas:
92 PL_sv_arenaroot pointer to list of SV arenas
93 PL_sv_root pointer to list of free SV structures
95 PL_body_arenas head of linked-list of body arenas
96 PL_body_roots[] array of pointers to list of free bodies of svtype
97 arrays are indexed by the svtype needed
99 A few special SV heads are not allocated from an arena, but are
100 instead directly created in the interpreter structure, eg PL_sv_undef.
101 The size of arenas can be changed from the default by setting
102 PERL_ARENA_SIZE appropriately at compile time.
104 The SV arena serves the secondary purpose of allowing still-live SVs
105 to be located and destroyed during final cleanup.
107 At the lowest level, the macros new_SV() and del_SV() grab and free
108 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
109 to return the SV to the free list with error checking.) new_SV() calls
110 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
111 SVs in the free list have their SvTYPE field set to all ones.
113 At the time of very final cleanup, sv_free_arenas() is called from
114 perl_destruct() to physically free all the arenas allocated since the
115 start of the interpreter.
117 The function visit() scans the SV arenas list, and calls a specified
118 function for each SV it finds which is still live - ie which has an SvTYPE
119 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
120 following functions (specified as [function that calls visit()] / [function
121 called by visit() for each SV]):
123 sv_report_used() / do_report_used()
124 dump all remaining SVs (debugging aid)
126 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
127 Attempt to free all objects pointed to by RVs,
128 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
129 try to do the same for all objects indirectly
130 referenced by typeglobs too. Called once from
131 perl_destruct(), prior to calling sv_clean_all()
134 sv_clean_all() / do_clean_all()
135 SvREFCNT_dec(sv) each remaining SV, possibly
136 triggering an sv_free(). It also sets the
137 SVf_BREAK flag on the SV to indicate that the
138 refcnt has been artificially lowered, and thus
139 stopping sv_free() from giving spurious warnings
140 about SVs which unexpectedly have a refcnt
141 of zero. called repeatedly from perl_destruct()
142 until there are no SVs left.
144 =head2 Arena allocator API Summary
146 Private API to rest of sv.c
150 new_XIV(), del_XIV(),
151 new_XNV(), del_XNV(),
156 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
160 * ========================================================================= */
163 * "A time to plant, and a time to uproot what was planted..."
167 Perl_offer_nice_chunk(pTHX_ void *const chunk, const U32 chunk_size)
173 PERL_ARGS_ASSERT_OFFER_NICE_CHUNK;
175 new_chunk = (void *)(chunk);
176 new_chunk_size = (chunk_size);
177 if (new_chunk_size > PL_nice_chunk_size) {
178 Safefree(PL_nice_chunk);
179 PL_nice_chunk = (char *) new_chunk;
180 PL_nice_chunk_size = new_chunk_size;
187 # define MEM_LOG_NEW_SV(sv, file, line, func) \
188 Perl_mem_log_new_sv(sv, file, line, func)
189 # define MEM_LOG_DEL_SV(sv, file, line, func) \
190 Perl_mem_log_del_sv(sv, file, line, func)
192 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
193 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
196 #ifdef DEBUG_LEAKING_SCALARS
197 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
198 # define DEBUG_SV_SERIAL(sv) \
199 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
200 PTR2UV(sv), (long)(sv)->sv_debug_serial))
202 # define FREE_SV_DEBUG_FILE(sv)
203 # define DEBUG_SV_SERIAL(sv) NOOP
207 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
208 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
209 /* Whilst I'd love to do this, it seems that things like to check on
211 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
213 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
214 PoisonNew(&SvREFCNT(sv), 1, U32)
216 # define SvARENA_CHAIN(sv) SvANY(sv)
217 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
218 # define POSION_SV_HEAD(sv)
221 /* Mark an SV head as unused, and add to free list.
223 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
224 * its refcount artificially decremented during global destruction, so
225 * there may be dangling pointers to it. The last thing we want in that
226 * case is for it to be reused. */
228 #define plant_SV(p) \
230 const U32 old_flags = SvFLAGS(p); \
231 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
232 DEBUG_SV_SERIAL(p); \
233 FREE_SV_DEBUG_FILE(p); \
235 SvFLAGS(p) = SVTYPEMASK; \
236 if (!(old_flags & SVf_BREAK)) { \
237 SvARENA_CHAIN_SET(p, PL_sv_root); \
243 #define uproot_SV(p) \
246 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
251 /* make some more SVs by adding another arena */
260 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
261 PL_nice_chunk = NULL;
262 PL_nice_chunk_size = 0;
265 char *chunk; /* must use New here to match call to */
266 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
267 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
273 /* new_SV(): return a new, empty SV head */
275 #ifdef DEBUG_LEAKING_SCALARS
276 /* provide a real function for a debugger to play with */
278 S_new_SV(pTHX_ const char *file, int line, const char *func)
285 sv = S_more_sv(aTHX);
289 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
290 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
296 sv->sv_debug_inpad = 0;
297 sv->sv_debug_cloned = 0;
298 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
300 sv->sv_debug_serial = PL_sv_serial++;
302 MEM_LOG_NEW_SV(sv, file, line, func);
303 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
304 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
308 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
316 (p) = S_more_sv(aTHX); \
320 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
325 /* del_SV(): return an empty SV head to the free list */
338 S_del_sv(pTHX_ SV *p)
342 PERL_ARGS_ASSERT_DEL_SV;
347 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
348 const SV * const sv = sva + 1;
349 const SV * const svend = &sva[SvREFCNT(sva)];
350 if (p >= sv && p < svend) {
356 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
357 "Attempt to free non-arena SV: 0x%"UVxf
358 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
365 #else /* ! DEBUGGING */
367 #define del_SV(p) plant_SV(p)
369 #endif /* DEBUGGING */
373 =head1 SV Manipulation Functions
375 =for apidoc sv_add_arena
377 Given a chunk of memory, link it to the head of the list of arenas,
378 and split it into a list of free SVs.
384 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
387 SV *const sva = MUTABLE_SV(ptr);
391 PERL_ARGS_ASSERT_SV_ADD_ARENA;
393 /* The first SV in an arena isn't an SV. */
394 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
395 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
396 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
398 PL_sv_arenaroot = sva;
399 PL_sv_root = sva + 1;
401 svend = &sva[SvREFCNT(sva) - 1];
404 SvARENA_CHAIN_SET(sv, (sv + 1));
408 /* Must always set typemask because it's always checked in on cleanup
409 when the arenas are walked looking for objects. */
410 SvFLAGS(sv) = SVTYPEMASK;
413 SvARENA_CHAIN_SET(sv, 0);
417 SvFLAGS(sv) = SVTYPEMASK;
420 /* visit(): call the named function for each non-free SV in the arenas
421 * whose flags field matches the flags/mask args. */
424 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
430 PERL_ARGS_ASSERT_VISIT;
432 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
433 register const SV * const svend = &sva[SvREFCNT(sva)];
435 for (sv = sva + 1; sv < svend; ++sv) {
436 if (SvTYPE(sv) != SVTYPEMASK
437 && (sv->sv_flags & mask) == flags
450 /* called by sv_report_used() for each live SV */
453 do_report_used(pTHX_ SV *const sv)
455 if (SvTYPE(sv) != SVTYPEMASK) {
456 PerlIO_printf(Perl_debug_log, "****\n");
463 =for apidoc sv_report_used
465 Dump the contents of all SVs not yet freed. (Debugging aid).
471 Perl_sv_report_used(pTHX)
474 visit(do_report_used, 0, 0);
480 /* called by sv_clean_objs() for each live SV */
483 do_clean_objs(pTHX_ SV *const ref)
488 SV * const target = SvRV(ref);
489 if (SvOBJECT(target)) {
490 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
491 if (SvWEAKREF(ref)) {
492 sv_del_backref(target, ref);
498 SvREFCNT_dec(target);
503 /* XXX Might want to check arrays, etc. */
506 /* called by sv_clean_objs() for each live SV */
508 #ifndef DISABLE_DESTRUCTOR_KLUDGE
510 do_clean_named_objs(pTHX_ SV *const sv)
513 assert(SvTYPE(sv) == SVt_PVGV);
514 assert(isGV_with_GP(sv));
517 #ifdef PERL_DONT_CREATE_GVSV
520 SvOBJECT(GvSV(sv))) ||
521 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
522 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
523 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
524 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
525 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
527 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
528 SvFLAGS(sv) |= SVf_BREAK;
536 =for apidoc sv_clean_objs
538 Attempt to destroy all objects not yet freed
544 Perl_sv_clean_objs(pTHX)
547 PL_in_clean_objs = TRUE;
548 visit(do_clean_objs, SVf_ROK, SVf_ROK);
549 #ifndef DISABLE_DESTRUCTOR_KLUDGE
550 /* some barnacles may yet remain, clinging to typeglobs */
551 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
553 PL_in_clean_objs = FALSE;
556 /* called by sv_clean_all() for each live SV */
559 do_clean_all(pTHX_ SV *const sv)
562 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
563 /* don't clean pid table and strtab */
566 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
567 SvFLAGS(sv) |= SVf_BREAK;
572 =for apidoc sv_clean_all
574 Decrement the refcnt of each remaining SV, possibly triggering a
575 cleanup. This function may have to be called multiple times to free
576 SVs which are in complex self-referential hierarchies.
582 Perl_sv_clean_all(pTHX)
586 PL_in_clean_all = TRUE;
587 cleaned = visit(do_clean_all, 0,0);
588 PL_in_clean_all = FALSE;
593 ARENASETS: a meta-arena implementation which separates arena-info
594 into struct arena_set, which contains an array of struct
595 arena_descs, each holding info for a single arena. By separating
596 the meta-info from the arena, we recover the 1st slot, formerly
597 borrowed for list management. The arena_set is about the size of an
598 arena, avoiding the needless malloc overhead of a naive linked-list.
600 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
601 memory in the last arena-set (1/2 on average). In trade, we get
602 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
603 smaller types). The recovery of the wasted space allows use of
604 small arenas for large, rare body types, by changing array* fields
605 in body_details_by_type[] below.
608 char *arena; /* the raw storage, allocated aligned */
609 size_t size; /* its size ~4k typ */
610 U32 misc; /* type, and in future other things. */
615 /* Get the maximum number of elements in set[] such that struct arena_set
616 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
617 therefore likely to be 1 aligned memory page. */
619 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
620 - 2 * sizeof(int)) / sizeof (struct arena_desc))
623 struct arena_set* next;
624 unsigned int set_size; /* ie ARENAS_PER_SET */
625 unsigned int curr; /* index of next available arena-desc */
626 struct arena_desc set[ARENAS_PER_SET];
630 =for apidoc sv_free_arenas
632 Deallocate the memory used by all arenas. Note that all the individual SV
633 heads and bodies within the arenas must already have been freed.
638 Perl_sv_free_arenas(pTHX)
645 /* Free arenas here, but be careful about fake ones. (We assume
646 contiguity of the fake ones with the corresponding real ones.) */
648 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
649 svanext = MUTABLE_SV(SvANY(sva));
650 while (svanext && SvFAKE(svanext))
651 svanext = MUTABLE_SV(SvANY(svanext));
658 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
661 struct arena_set *current = aroot;
664 assert(aroot->set[i].arena);
665 Safefree(aroot->set[i].arena);
673 i = PERL_ARENA_ROOTS_SIZE;
675 PL_body_roots[i] = 0;
677 Safefree(PL_nice_chunk);
678 PL_nice_chunk = NULL;
679 PL_nice_chunk_size = 0;
685 Here are mid-level routines that manage the allocation of bodies out
686 of the various arenas. There are 5 kinds of arenas:
688 1. SV-head arenas, which are discussed and handled above
689 2. regular body arenas
690 3. arenas for reduced-size bodies
692 5. pte arenas (thread related)
694 Arena types 2 & 3 are chained by body-type off an array of
695 arena-root pointers, which is indexed by svtype. Some of the
696 larger/less used body types are malloced singly, since a large
697 unused block of them is wasteful. Also, several svtypes dont have
698 bodies; the data fits into the sv-head itself. The arena-root
699 pointer thus has a few unused root-pointers (which may be hijacked
700 later for arena types 4,5)
702 3 differs from 2 as an optimization; some body types have several
703 unused fields in the front of the structure (which are kept in-place
704 for consistency). These bodies can be allocated in smaller chunks,
705 because the leading fields arent accessed. Pointers to such bodies
706 are decremented to point at the unused 'ghost' memory, knowing that
707 the pointers are used with offsets to the real memory.
709 HE, HEK arenas are managed separately, with separate code, but may
710 be merge-able later..
712 PTE arenas are not sv-bodies, but they share these mid-level
713 mechanics, so are considered here. The new mid-level mechanics rely
714 on the sv_type of the body being allocated, so we just reserve one
715 of the unused body-slots for PTEs, then use it in those (2) PTE
716 contexts below (line ~10k)
719 /* get_arena(size): this creates custom-sized arenas
720 TBD: export properly for hv.c: S_more_he().
723 Perl_get_arena(pTHX_ const size_t arena_size, const U32 misc)
726 struct arena_desc* adesc;
727 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
730 /* shouldnt need this
731 if (!arena_size) arena_size = PERL_ARENA_SIZE;
734 /* may need new arena-set to hold new arena */
735 if (!aroot || aroot->curr >= aroot->set_size) {
736 struct arena_set *newroot;
737 Newxz(newroot, 1, struct arena_set);
738 newroot->set_size = ARENAS_PER_SET;
739 newroot->next = aroot;
741 PL_body_arenas = (void *) newroot;
742 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
745 /* ok, now have arena-set with at least 1 empty/available arena-desc */
746 curr = aroot->curr++;
747 adesc = &(aroot->set[curr]);
748 assert(!adesc->arena);
750 Newx(adesc->arena, arena_size, char);
751 adesc->size = arena_size;
753 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
754 curr, (void*)adesc->arena, (UV)arena_size));
760 /* return a thing to the free list */
762 #define del_body(thing, root) \
764 void ** const thing_copy = (void **)thing;\
765 *thing_copy = *root; \
766 *root = (void*)thing_copy; \
771 =head1 SV-Body Allocation
773 Allocation of SV-bodies is similar to SV-heads, differing as follows;
774 the allocation mechanism is used for many body types, so is somewhat
775 more complicated, it uses arena-sets, and has no need for still-live
778 At the outermost level, (new|del)_X*V macros return bodies of the
779 appropriate type. These macros call either (new|del)_body_type or
780 (new|del)_body_allocated macro pairs, depending on specifics of the
781 type. Most body types use the former pair, the latter pair is used to
782 allocate body types with "ghost fields".
784 "ghost fields" are fields that are unused in certain types, and
785 consequently don't need to actually exist. They are declared because
786 they're part of a "base type", which allows use of functions as
787 methods. The simplest examples are AVs and HVs, 2 aggregate types
788 which don't use the fields which support SCALAR semantics.
790 For these types, the arenas are carved up into appropriately sized
791 chunks, we thus avoid wasted memory for those unaccessed members.
792 When bodies are allocated, we adjust the pointer back in memory by the
793 size of the part not allocated, so it's as if we allocated the full
794 structure. (But things will all go boom if you write to the part that
795 is "not there", because you'll be overwriting the last members of the
796 preceding structure in memory.)
798 We calculate the correction using the STRUCT_OFFSET macro on the first
799 member present. If the allocated structure is smaller (no initial NV
800 actually allocated) then the net effect is to subtract the size of the NV
801 from the pointer, to return a new pointer as if an initial NV were actually
802 allocated. (We were using structures named *_allocated for this, but
803 this turned out to be a subtle bug, because a structure without an NV
804 could have a lower alignment constraint, but the compiler is allowed to
805 optimised accesses based on the alignment constraint of the actual pointer
806 to the full structure, for example, using a single 64 bit load instruction
807 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
809 This is the same trick as was used for NV and IV bodies. Ironically it
810 doesn't need to be used for NV bodies any more, because NV is now at
811 the start of the structure. IV bodies don't need it either, because
812 they are no longer allocated.
814 In turn, the new_body_* allocators call S_new_body(), which invokes
815 new_body_inline macro, which takes a lock, and takes a body off the
816 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
817 necessary to refresh an empty list. Then the lock is released, and
818 the body is returned.
820 S_more_bodies calls get_arena(), and carves it up into an array of N
821 bodies, which it strings into a linked list. It looks up arena-size
822 and body-size from the body_details table described below, thus
823 supporting the multiple body-types.
825 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
826 the (new|del)_X*V macros are mapped directly to malloc/free.
832 For each sv-type, struct body_details bodies_by_type[] carries
833 parameters which control these aspects of SV handling:
835 Arena_size determines whether arenas are used for this body type, and if
836 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
837 zero, forcing individual mallocs and frees.
839 Body_size determines how big a body is, and therefore how many fit into
840 each arena. Offset carries the body-pointer adjustment needed for
841 "ghost fields", and is used in *_allocated macros.
843 But its main purpose is to parameterize info needed in
844 Perl_sv_upgrade(). The info here dramatically simplifies the function
845 vs the implementation in 5.8.8, making it table-driven. All fields
846 are used for this, except for arena_size.
848 For the sv-types that have no bodies, arenas are not used, so those
849 PL_body_roots[sv_type] are unused, and can be overloaded. In
850 something of a special case, SVt_NULL is borrowed for HE arenas;
851 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
852 bodies_by_type[SVt_NULL] slot is not used, as the table is not
855 PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
856 they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
857 just use the same allocation semantics. At first, PTEs were also
858 overloaded to a non-body sv-type, but this yielded hard-to-find malloc
859 bugs, so was simplified by claiming a new slot. This choice has no
860 consequence at this time.
864 struct body_details {
865 U8 body_size; /* Size to allocate */
866 U8 copy; /* Size of structure to copy (may be shorter) */
868 unsigned int type : 4; /* We have space for a sanity check. */
869 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
870 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
871 unsigned int arena : 1; /* Allocated from an arena */
872 size_t arena_size; /* Size of arena to allocate */
880 /* With -DPURFIY we allocate everything directly, and don't use arenas.
881 This seems a rather elegant way to simplify some of the code below. */
882 #define HASARENA FALSE
884 #define HASARENA TRUE
886 #define NOARENA FALSE
888 /* Size the arenas to exactly fit a given number of bodies. A count
889 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
890 simplifying the default. If count > 0, the arena is sized to fit
891 only that many bodies, allowing arenas to be used for large, rare
892 bodies (XPVFM, XPVIO) without undue waste. The arena size is
893 limited by PERL_ARENA_SIZE, so we can safely oversize the
896 #define FIT_ARENA0(body_size) \
897 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
898 #define FIT_ARENAn(count,body_size) \
899 ( count * body_size <= PERL_ARENA_SIZE) \
900 ? count * body_size \
901 : FIT_ARENA0 (body_size)
902 #define FIT_ARENA(count,body_size) \
904 ? FIT_ARENAn (count, body_size) \
905 : FIT_ARENA0 (body_size)
907 /* Calculate the length to copy. Specifically work out the length less any
908 final padding the compiler needed to add. See the comment in sv_upgrade
909 for why copying the padding proved to be a bug. */
911 #define copy_length(type, last_member) \
912 STRUCT_OFFSET(type, last_member) \
913 + sizeof (((type*)SvANY((const SV *)0))->last_member)
915 static const struct body_details bodies_by_type[] = {
916 { sizeof(HE), 0, 0, SVt_NULL,
917 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
919 /* The bind placeholder pretends to be an RV for now.
920 Also it's marked as "can't upgrade" to stop anyone using it before it's
922 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
924 /* IVs are in the head, so the allocation size is 0.
925 However, the slot is overloaded for PTEs. */
926 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
927 sizeof(IV), /* This is used to copy out the IV body. */
928 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
929 NOARENA /* IVS don't need an arena */,
930 /* But PTEs need to know the size of their arena */
931 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
934 /* 8 bytes on most ILP32 with IEEE doubles */
935 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
936 FIT_ARENA(0, sizeof(NV)) },
938 /* 8 bytes on most ILP32 with IEEE doubles */
939 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
940 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
941 + STRUCT_OFFSET(XPV, xpv_cur),
942 SVt_PV, FALSE, NONV, HASARENA,
943 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
946 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
947 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
948 + STRUCT_OFFSET(XPVIV, xpv_cur),
949 SVt_PVIV, FALSE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
953 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
954 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
957 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
958 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
961 { sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
962 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
963 + STRUCT_OFFSET(regexp, xpv_cur),
964 SVt_REGEXP, FALSE, NONV, HASARENA,
965 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
969 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
970 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
973 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
974 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
976 { sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill),
977 copy_length(XPVAV, xmg_stash) - STRUCT_OFFSET(XPVAV, xav_fill),
978 + STRUCT_OFFSET(XPVAV, xav_fill),
979 SVt_PVAV, TRUE, NONV, HASARENA,
980 FIT_ARENA(0, sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill)) },
982 { sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill),
983 copy_length(XPVHV, xmg_stash) - STRUCT_OFFSET(XPVHV, xhv_fill),
984 + STRUCT_OFFSET(XPVHV, xhv_fill),
985 SVt_PVHV, TRUE, NONV, HASARENA,
986 FIT_ARENA(0, sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill)) },
989 { sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
990 sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
991 + STRUCT_OFFSET(XPVCV, xpv_cur),
992 SVt_PVCV, TRUE, NONV, HASARENA,
993 FIT_ARENA(0, sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur)) },
995 { sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
996 sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
997 + STRUCT_OFFSET(XPVFM, xpv_cur),
998 SVt_PVFM, TRUE, NONV, NOARENA,
999 FIT_ARENA(20, sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur)) },
1001 /* XPVIO is 84 bytes, fits 48x */
1002 { sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1003 sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1004 + STRUCT_OFFSET(XPVIO, xpv_cur),
1005 SVt_PVIO, TRUE, NONV, HASARENA,
1006 FIT_ARENA(24, sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur)) },
1009 #define new_body_type(sv_type) \
1010 (void *)((char *)S_new_body(aTHX_ sv_type))
1012 #define del_body_type(p, sv_type) \
1013 del_body(p, &PL_body_roots[sv_type])
1016 #define new_body_allocated(sv_type) \
1017 (void *)((char *)S_new_body(aTHX_ sv_type) \
1018 - bodies_by_type[sv_type].offset)
1020 #define del_body_allocated(p, sv_type) \
1021 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1024 #define my_safemalloc(s) (void*)safemalloc(s)
1025 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1026 #define my_safefree(p) safefree((char*)p)
1030 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1031 #define del_XNV(p) my_safefree(p)
1033 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1034 #define del_XPVNV(p) my_safefree(p)
1036 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1037 #define del_XPVAV(p) my_safefree(p)
1039 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1040 #define del_XPVHV(p) my_safefree(p)
1042 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1043 #define del_XPVMG(p) my_safefree(p)
1045 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1046 #define del_XPVGV(p) my_safefree(p)
1050 #define new_XNV() new_body_type(SVt_NV)
1051 #define del_XNV(p) del_body_type(p, SVt_NV)
1053 #define new_XPVNV() new_body_type(SVt_PVNV)
1054 #define del_XPVNV(p) del_body_type(p, SVt_PVNV)
1056 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1057 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1059 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1060 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1062 #define new_XPVMG() new_body_type(SVt_PVMG)
1063 #define del_XPVMG(p) del_body_type(p, SVt_PVMG)
1065 #define new_XPVGV() new_body_type(SVt_PVGV)
1066 #define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1070 /* no arena for you! */
1072 #define new_NOARENA(details) \
1073 my_safemalloc((details)->body_size + (details)->offset)
1074 #define new_NOARENAZ(details) \
1075 my_safecalloc((details)->body_size + (details)->offset)
1078 S_more_bodies (pTHX_ const svtype sv_type)
1081 void ** const root = &PL_body_roots[sv_type];
1082 const struct body_details * const bdp = &bodies_by_type[sv_type];
1083 const size_t body_size = bdp->body_size;
1086 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1087 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1088 static bool done_sanity_check;
1090 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1091 * variables like done_sanity_check. */
1092 if (!done_sanity_check) {
1093 unsigned int i = SVt_LAST;
1095 done_sanity_check = TRUE;
1098 assert (bodies_by_type[i].type == i);
1102 assert(bdp->arena_size);
1104 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1106 end = start + arena_size - 2 * body_size;
1108 /* computed count doesnt reflect the 1st slot reservation */
1109 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1110 DEBUG_m(PerlIO_printf(Perl_debug_log,
1111 "arena %p end %p arena-size %d (from %d) type %d "
1113 (void*)start, (void*)end, (int)arena_size,
1114 (int)bdp->arena_size, sv_type, (int)body_size,
1115 (int)arena_size / (int)body_size));
1117 DEBUG_m(PerlIO_printf(Perl_debug_log,
1118 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1119 (void*)start, (void*)end,
1120 (int)bdp->arena_size, sv_type, (int)body_size,
1121 (int)bdp->arena_size / (int)body_size));
1123 *root = (void *)start;
1125 while (start <= end) {
1126 char * const next = start + body_size;
1127 *(void**) start = (void *)next;
1130 *(void **)start = 0;
1135 /* grab a new thing from the free list, allocating more if necessary.
1136 The inline version is used for speed in hot routines, and the
1137 function using it serves the rest (unless PURIFY).
1139 #define new_body_inline(xpv, sv_type) \
1141 void ** const r3wt = &PL_body_roots[sv_type]; \
1142 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1143 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1144 *(r3wt) = *(void**)(xpv); \
1150 S_new_body(pTHX_ const svtype sv_type)
1154 new_body_inline(xpv, sv_type);
1160 static const struct body_details fake_rv =
1161 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1164 =for apidoc sv_upgrade
1166 Upgrade an SV to a more complex form. Generally adds a new body type to the
1167 SV, then copies across as much information as possible from the old body.
1168 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1174 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1179 const svtype old_type = SvTYPE(sv);
1180 const struct body_details *new_type_details;
1181 const struct body_details *old_type_details
1182 = bodies_by_type + old_type;
1183 SV *referant = NULL;
1185 PERL_ARGS_ASSERT_SV_UPGRADE;
1187 if (old_type == new_type)
1190 /* This clause was purposefully added ahead of the early return above to
1191 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1192 inference by Nick I-S that it would fix other troublesome cases. See
1193 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1195 Given that shared hash key scalars are no longer PVIV, but PV, there is
1196 no longer need to unshare so as to free up the IVX slot for its proper
1197 purpose. So it's safe to move the early return earlier. */
1199 if (new_type != SVt_PV && SvIsCOW(sv)) {
1200 sv_force_normal_flags(sv, 0);
1203 old_body = SvANY(sv);
1205 /* Copying structures onto other structures that have been neatly zeroed
1206 has a subtle gotcha. Consider XPVMG
1208 +------+------+------+------+------+-------+-------+
1209 | NV | CUR | LEN | IV | MAGIC | STASH |
1210 +------+------+------+------+------+-------+-------+
1211 0 4 8 12 16 20 24 28
1213 where NVs are aligned to 8 bytes, so that sizeof that structure is
1214 actually 32 bytes long, with 4 bytes of padding at the end:
1216 +------+------+------+------+------+-------+-------+------+
1217 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1218 +------+------+------+------+------+-------+-------+------+
1219 0 4 8 12 16 20 24 28 32
1221 so what happens if you allocate memory for this structure:
1223 +------+------+------+------+------+-------+-------+------+------+...
1224 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1225 +------+------+------+------+------+-------+-------+------+------+...
1226 0 4 8 12 16 20 24 28 32 36
1228 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1229 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1230 started out as zero once, but it's quite possible that it isn't. So now,
1231 rather than a nicely zeroed GP, you have it pointing somewhere random.
1234 (In fact, GP ends up pointing at a previous GP structure, because the
1235 principle cause of the padding in XPVMG getting garbage is a copy of
1236 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1237 this happens to be moot because XPVGV has been re-ordered, with GP
1238 no longer after STASH)
1240 So we are careful and work out the size of used parts of all the
1248 referant = SvRV(sv);
1249 old_type_details = &fake_rv;
1250 if (new_type == SVt_NV)
1251 new_type = SVt_PVNV;
1253 if (new_type < SVt_PVIV) {
1254 new_type = (new_type == SVt_NV)
1255 ? SVt_PVNV : SVt_PVIV;
1260 if (new_type < SVt_PVNV) {
1261 new_type = SVt_PVNV;
1265 assert(new_type > SVt_PV);
1266 assert(SVt_IV < SVt_PV);
1267 assert(SVt_NV < SVt_PV);
1274 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1275 there's no way that it can be safely upgraded, because perl.c
1276 expects to Safefree(SvANY(PL_mess_sv)) */
1277 assert(sv != PL_mess_sv);
1278 /* This flag bit is used to mean other things in other scalar types.
1279 Given that it only has meaning inside the pad, it shouldn't be set
1280 on anything that can get upgraded. */
1281 assert(!SvPAD_TYPED(sv));
1284 if (old_type_details->cant_upgrade)
1285 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1286 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1289 if (old_type > new_type)
1290 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1291 (int)old_type, (int)new_type);
1293 new_type_details = bodies_by_type + new_type;
1295 SvFLAGS(sv) &= ~SVTYPEMASK;
1296 SvFLAGS(sv) |= new_type;
1298 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1299 the return statements above will have triggered. */
1300 assert (new_type != SVt_NULL);
1303 assert(old_type == SVt_NULL);
1304 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1308 assert(old_type == SVt_NULL);
1309 SvANY(sv) = new_XNV();
1314 assert(new_type_details->body_size);
1317 assert(new_type_details->arena);
1318 assert(new_type_details->arena_size);
1319 /* This points to the start of the allocated area. */
1320 new_body_inline(new_body, new_type);
1321 Zero(new_body, new_type_details->body_size, char);
1322 new_body = ((char *)new_body) - new_type_details->offset;
1324 /* We always allocated the full length item with PURIFY. To do this
1325 we fake things so that arena is false for all 16 types.. */
1326 new_body = new_NOARENAZ(new_type_details);
1328 SvANY(sv) = new_body;
1329 if (new_type == SVt_PVAV) {
1333 if (old_type_details->body_size) {
1336 /* It will have been zeroed when the new body was allocated.
1337 Lets not write to it, in case it confuses a write-back
1343 #ifndef NODEFAULT_SHAREKEYS
1344 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1346 HvMAX(sv) = 7; /* (start with 8 buckets) */
1347 if (old_type_details->body_size) {
1350 /* It will have been zeroed when the new body was allocated.
1351 Lets not write to it, in case it confuses a write-back
1356 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1357 The target created by newSVrv also is, and it can have magic.
1358 However, it never has SvPVX set.
1360 if (old_type == SVt_IV) {
1362 } else if (old_type >= SVt_PV) {
1363 assert(SvPVX_const(sv) == 0);
1366 if (old_type >= SVt_PVMG) {
1367 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1368 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1370 sv->sv_u.svu_array = NULL; /* or svu_hash */
1376 /* XXX Is this still needed? Was it ever needed? Surely as there is
1377 no route from NV to PVIV, NOK can never be true */
1378 assert(!SvNOKp(sv));
1390 assert(new_type_details->body_size);
1391 /* We always allocated the full length item with PURIFY. To do this
1392 we fake things so that arena is false for all 16 types.. */
1393 if(new_type_details->arena) {
1394 /* This points to the start of the allocated area. */
1395 new_body_inline(new_body, new_type);
1396 Zero(new_body, new_type_details->body_size, char);
1397 new_body = ((char *)new_body) - new_type_details->offset;
1399 new_body = new_NOARENAZ(new_type_details);
1401 SvANY(sv) = new_body;
1403 if (old_type_details->copy) {
1404 /* There is now the potential for an upgrade from something without
1405 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1406 int offset = old_type_details->offset;
1407 int length = old_type_details->copy;
1409 if (new_type_details->offset > old_type_details->offset) {
1410 const int difference
1411 = new_type_details->offset - old_type_details->offset;
1412 offset += difference;
1413 length -= difference;
1415 assert (length >= 0);
1417 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1421 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1422 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1423 * correct 0.0 for us. Otherwise, if the old body didn't have an
1424 * NV slot, but the new one does, then we need to initialise the
1425 * freshly created NV slot with whatever the correct bit pattern is
1427 if (old_type_details->zero_nv && !new_type_details->zero_nv
1428 && !isGV_with_GP(sv))
1432 if (new_type == SVt_PVIO) {
1433 IO * const io = MUTABLE_IO(sv);
1434 GV *iogv = gv_fetchpvs("IO::Handle::", GV_ADD, SVt_PVHV);
1437 /* Clear the stashcache because a new IO could overrule a package
1439 hv_clear(PL_stashcache);
1441 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1442 IoPAGE_LEN(sv) = 60;
1444 if (old_type < SVt_PV) {
1445 /* referant will be NULL unless the old type was SVt_IV emulating
1447 sv->sv_u.svu_rv = referant;
1451 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1452 (unsigned long)new_type);
1455 if (old_type > SVt_IV) { /* SVt_IVs are overloaded for PTEs */
1457 my_safefree(old_body);
1459 /* Note that there is an assumption that all bodies of types that
1460 can be upgraded came from arenas. Only the more complex non-
1461 upgradable types are allowed to be directly malloc()ed. */
1462 assert(old_type_details->arena);
1463 del_body((void*)((char*)old_body + old_type_details->offset),
1464 &PL_body_roots[old_type]);
1470 =for apidoc sv_backoff
1472 Remove any string offset. You should normally use the C<SvOOK_off> macro
1479 Perl_sv_backoff(pTHX_ register SV *const sv)
1482 const char * const s = SvPVX_const(sv);
1484 PERL_ARGS_ASSERT_SV_BACKOFF;
1485 PERL_UNUSED_CONTEXT;
1488 assert(SvTYPE(sv) != SVt_PVHV);
1489 assert(SvTYPE(sv) != SVt_PVAV);
1491 SvOOK_offset(sv, delta);
1493 SvLEN_set(sv, SvLEN(sv) + delta);
1494 SvPV_set(sv, SvPVX(sv) - delta);
1495 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1496 SvFLAGS(sv) &= ~SVf_OOK;
1503 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1504 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1505 Use the C<SvGROW> wrapper instead.
1511 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1515 PERL_ARGS_ASSERT_SV_GROW;
1517 if (PL_madskills && newlen >= 0x100000) {
1518 PerlIO_printf(Perl_debug_log,
1519 "Allocation too large: %"UVxf"\n", (UV)newlen);
1521 #ifdef HAS_64K_LIMIT
1522 if (newlen >= 0x10000) {
1523 PerlIO_printf(Perl_debug_log,
1524 "Allocation too large: %"UVxf"\n", (UV)newlen);
1527 #endif /* HAS_64K_LIMIT */
1530 if (SvTYPE(sv) < SVt_PV) {
1531 sv_upgrade(sv, SVt_PV);
1532 s = SvPVX_mutable(sv);
1534 else if (SvOOK(sv)) { /* pv is offset? */
1536 s = SvPVX_mutable(sv);
1537 if (newlen > SvLEN(sv))
1538 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1539 #ifdef HAS_64K_LIMIT
1540 if (newlen >= 0x10000)
1545 s = SvPVX_mutable(sv);
1547 if (newlen > SvLEN(sv)) { /* need more room? */
1548 #ifndef Perl_safesysmalloc_size
1549 newlen = PERL_STRLEN_ROUNDUP(newlen);
1551 if (SvLEN(sv) && s) {
1552 s = (char*)saferealloc(s, newlen);
1555 s = (char*)safemalloc(newlen);
1556 if (SvPVX_const(sv) && SvCUR(sv)) {
1557 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1561 #ifdef Perl_safesysmalloc_size
1562 /* Do this here, do it once, do it right, and then we will never get
1563 called back into sv_grow() unless there really is some growing
1565 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1567 SvLEN_set(sv, newlen);
1574 =for apidoc sv_setiv
1576 Copies an integer into the given SV, upgrading first if necessary.
1577 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1583 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1587 PERL_ARGS_ASSERT_SV_SETIV;
1589 SV_CHECK_THINKFIRST_COW_DROP(sv);
1590 switch (SvTYPE(sv)) {
1593 sv_upgrade(sv, SVt_IV);
1596 sv_upgrade(sv, SVt_PVIV);
1600 if (!isGV_with_GP(sv))
1607 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1611 (void)SvIOK_only(sv); /* validate number */
1617 =for apidoc sv_setiv_mg
1619 Like C<sv_setiv>, but also handles 'set' magic.
1625 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1627 PERL_ARGS_ASSERT_SV_SETIV_MG;
1634 =for apidoc sv_setuv
1636 Copies an unsigned integer into the given SV, upgrading first if necessary.
1637 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1643 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1645 PERL_ARGS_ASSERT_SV_SETUV;
1647 /* With these two if statements:
1648 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1651 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1653 If you wish to remove them, please benchmark to see what the effect is
1655 if (u <= (UV)IV_MAX) {
1656 sv_setiv(sv, (IV)u);
1665 =for apidoc sv_setuv_mg
1667 Like C<sv_setuv>, but also handles 'set' magic.
1673 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1675 PERL_ARGS_ASSERT_SV_SETUV_MG;
1682 =for apidoc sv_setnv
1684 Copies a double into the given SV, upgrading first if necessary.
1685 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1691 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1695 PERL_ARGS_ASSERT_SV_SETNV;
1697 SV_CHECK_THINKFIRST_COW_DROP(sv);
1698 switch (SvTYPE(sv)) {
1701 sv_upgrade(sv, SVt_NV);
1705 sv_upgrade(sv, SVt_PVNV);
1709 if (!isGV_with_GP(sv))
1716 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1721 (void)SvNOK_only(sv); /* validate number */
1726 =for apidoc sv_setnv_mg
1728 Like C<sv_setnv>, but also handles 'set' magic.
1734 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1736 PERL_ARGS_ASSERT_SV_SETNV_MG;
1742 /* Print an "isn't numeric" warning, using a cleaned-up,
1743 * printable version of the offending string
1747 S_not_a_number(pTHX_ SV *const sv)
1754 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1757 dsv = newSVpvs_flags("", SVs_TEMP);
1758 pv = sv_uni_display(dsv, sv, 10, 0);
1761 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1762 /* each *s can expand to 4 chars + "...\0",
1763 i.e. need room for 8 chars */
1765 const char *s = SvPVX_const(sv);
1766 const char * const end = s + SvCUR(sv);
1767 for ( ; s < end && d < limit; s++ ) {
1769 if (ch & 128 && !isPRINT_LC(ch)) {
1778 else if (ch == '\r') {
1782 else if (ch == '\f') {
1786 else if (ch == '\\') {
1790 else if (ch == '\0') {
1794 else if (isPRINT_LC(ch))
1811 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1812 "Argument \"%s\" isn't numeric in %s", pv,
1815 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1816 "Argument \"%s\" isn't numeric", pv);
1820 =for apidoc looks_like_number
1822 Test if the content of an SV looks like a number (or is a number).
1823 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1824 non-numeric warning), even if your atof() doesn't grok them.
1830 Perl_looks_like_number(pTHX_ SV *const sv)
1832 register const char *sbegin;
1835 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1838 sbegin = SvPVX_const(sv);
1841 else if (SvPOKp(sv))
1842 sbegin = SvPV_const(sv, len);
1844 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1845 return grok_number(sbegin, len, NULL);
1849 S_glob_2number(pTHX_ GV * const gv)
1851 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1852 SV *const buffer = sv_newmortal();
1854 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1856 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1859 gv_efullname3(buffer, gv, "*");
1860 SvFLAGS(gv) |= wasfake;
1862 /* We know that all GVs stringify to something that is not-a-number,
1863 so no need to test that. */
1864 if (ckWARN(WARN_NUMERIC))
1865 not_a_number(buffer);
1866 /* We just want something true to return, so that S_sv_2iuv_common
1867 can tail call us and return true. */
1871 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1872 until proven guilty, assume that things are not that bad... */
1877 As 64 bit platforms often have an NV that doesn't preserve all bits of
1878 an IV (an assumption perl has been based on to date) it becomes necessary
1879 to remove the assumption that the NV always carries enough precision to
1880 recreate the IV whenever needed, and that the NV is the canonical form.
1881 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1882 precision as a side effect of conversion (which would lead to insanity
1883 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1884 1) to distinguish between IV/UV/NV slots that have cached a valid
1885 conversion where precision was lost and IV/UV/NV slots that have a
1886 valid conversion which has lost no precision
1887 2) to ensure that if a numeric conversion to one form is requested that
1888 would lose precision, the precise conversion (or differently
1889 imprecise conversion) is also performed and cached, to prevent
1890 requests for different numeric formats on the same SV causing
1891 lossy conversion chains. (lossless conversion chains are perfectly
1896 SvIOKp is true if the IV slot contains a valid value
1897 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1898 SvNOKp is true if the NV slot contains a valid value
1899 SvNOK is true only if the NV value is accurate
1902 while converting from PV to NV, check to see if converting that NV to an
1903 IV(or UV) would lose accuracy over a direct conversion from PV to
1904 IV(or UV). If it would, cache both conversions, return NV, but mark
1905 SV as IOK NOKp (ie not NOK).
1907 While converting from PV to IV, check to see if converting that IV to an
1908 NV would lose accuracy over a direct conversion from PV to NV. If it
1909 would, cache both conversions, flag similarly.
1911 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1912 correctly because if IV & NV were set NV *always* overruled.
1913 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1914 changes - now IV and NV together means that the two are interchangeable:
1915 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1917 The benefit of this is that operations such as pp_add know that if
1918 SvIOK is true for both left and right operands, then integer addition
1919 can be used instead of floating point (for cases where the result won't
1920 overflow). Before, floating point was always used, which could lead to
1921 loss of precision compared with integer addition.
1923 * making IV and NV equal status should make maths accurate on 64 bit
1925 * may speed up maths somewhat if pp_add and friends start to use
1926 integers when possible instead of fp. (Hopefully the overhead in
1927 looking for SvIOK and checking for overflow will not outweigh the
1928 fp to integer speedup)
1929 * will slow down integer operations (callers of SvIV) on "inaccurate"
1930 values, as the change from SvIOK to SvIOKp will cause a call into
1931 sv_2iv each time rather than a macro access direct to the IV slot
1932 * should speed up number->string conversion on integers as IV is
1933 favoured when IV and NV are equally accurate
1935 ####################################################################
1936 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1937 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1938 On the other hand, SvUOK is true iff UV.
1939 ####################################################################
1941 Your mileage will vary depending your CPU's relative fp to integer
1945 #ifndef NV_PRESERVES_UV
1946 # define IS_NUMBER_UNDERFLOW_IV 1
1947 # define IS_NUMBER_UNDERFLOW_UV 2
1948 # define IS_NUMBER_IV_AND_UV 2
1949 # define IS_NUMBER_OVERFLOW_IV 4
1950 # define IS_NUMBER_OVERFLOW_UV 5
1952 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1954 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1956 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1964 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1966 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1967 if (SvNVX(sv) < (NV)IV_MIN) {
1968 (void)SvIOKp_on(sv);
1970 SvIV_set(sv, IV_MIN);
1971 return IS_NUMBER_UNDERFLOW_IV;
1973 if (SvNVX(sv) > (NV)UV_MAX) {
1974 (void)SvIOKp_on(sv);
1977 SvUV_set(sv, UV_MAX);
1978 return IS_NUMBER_OVERFLOW_UV;
1980 (void)SvIOKp_on(sv);
1982 /* Can't use strtol etc to convert this string. (See truth table in
1984 if (SvNVX(sv) <= (UV)IV_MAX) {
1985 SvIV_set(sv, I_V(SvNVX(sv)));
1986 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1987 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1989 /* Integer is imprecise. NOK, IOKp */
1991 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1994 SvUV_set(sv, U_V(SvNVX(sv)));
1995 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1996 if (SvUVX(sv) == UV_MAX) {
1997 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1998 possibly be preserved by NV. Hence, it must be overflow.
2000 return IS_NUMBER_OVERFLOW_UV;
2002 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2004 /* Integer is imprecise. NOK, IOKp */
2006 return IS_NUMBER_OVERFLOW_IV;
2008 #endif /* !NV_PRESERVES_UV*/
2011 S_sv_2iuv_common(pTHX_ SV *const sv)
2015 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2018 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2019 * without also getting a cached IV/UV from it at the same time
2020 * (ie PV->NV conversion should detect loss of accuracy and cache
2021 * IV or UV at same time to avoid this. */
2022 /* IV-over-UV optimisation - choose to cache IV if possible */
2024 if (SvTYPE(sv) == SVt_NV)
2025 sv_upgrade(sv, SVt_PVNV);
2027 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2028 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2029 certainly cast into the IV range at IV_MAX, whereas the correct
2030 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2032 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2033 if (Perl_isnan(SvNVX(sv))) {
2039 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2040 SvIV_set(sv, I_V(SvNVX(sv)));
2041 if (SvNVX(sv) == (NV) SvIVX(sv)
2042 #ifndef NV_PRESERVES_UV
2043 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2044 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2045 /* Don't flag it as "accurately an integer" if the number
2046 came from a (by definition imprecise) NV operation, and
2047 we're outside the range of NV integer precision */
2051 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2053 /* scalar has trailing garbage, eg "42a" */
2055 DEBUG_c(PerlIO_printf(Perl_debug_log,
2056 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2062 /* IV not precise. No need to convert from PV, as NV
2063 conversion would already have cached IV if it detected
2064 that PV->IV would be better than PV->NV->IV
2065 flags already correct - don't set public IOK. */
2066 DEBUG_c(PerlIO_printf(Perl_debug_log,
2067 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2072 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2073 but the cast (NV)IV_MIN rounds to a the value less (more
2074 negative) than IV_MIN which happens to be equal to SvNVX ??
2075 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2076 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2077 (NV)UVX == NVX are both true, but the values differ. :-(
2078 Hopefully for 2s complement IV_MIN is something like
2079 0x8000000000000000 which will be exact. NWC */
2082 SvUV_set(sv, U_V(SvNVX(sv)));
2084 (SvNVX(sv) == (NV) SvUVX(sv))
2085 #ifndef NV_PRESERVES_UV
2086 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2087 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2088 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2089 /* Don't flag it as "accurately an integer" if the number
2090 came from a (by definition imprecise) NV operation, and
2091 we're outside the range of NV integer precision */
2097 DEBUG_c(PerlIO_printf(Perl_debug_log,
2098 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2104 else if (SvPOKp(sv) && SvLEN(sv)) {
2106 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2107 /* We want to avoid a possible problem when we cache an IV/ a UV which
2108 may be later translated to an NV, and the resulting NV is not
2109 the same as the direct translation of the initial string
2110 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2111 be careful to ensure that the value with the .456 is around if the
2112 NV value is requested in the future).
2114 This means that if we cache such an IV/a UV, we need to cache the
2115 NV as well. Moreover, we trade speed for space, and do not
2116 cache the NV if we are sure it's not needed.
2119 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2120 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2121 == IS_NUMBER_IN_UV) {
2122 /* It's definitely an integer, only upgrade to PVIV */
2123 if (SvTYPE(sv) < SVt_PVIV)
2124 sv_upgrade(sv, SVt_PVIV);
2126 } else if (SvTYPE(sv) < SVt_PVNV)
2127 sv_upgrade(sv, SVt_PVNV);
2129 /* If NVs preserve UVs then we only use the UV value if we know that
2130 we aren't going to call atof() below. If NVs don't preserve UVs
2131 then the value returned may have more precision than atof() will
2132 return, even though value isn't perfectly accurate. */
2133 if ((numtype & (IS_NUMBER_IN_UV
2134 #ifdef NV_PRESERVES_UV
2137 )) == IS_NUMBER_IN_UV) {
2138 /* This won't turn off the public IOK flag if it was set above */
2139 (void)SvIOKp_on(sv);
2141 if (!(numtype & IS_NUMBER_NEG)) {
2143 if (value <= (UV)IV_MAX) {
2144 SvIV_set(sv, (IV)value);
2146 /* it didn't overflow, and it was positive. */
2147 SvUV_set(sv, value);
2151 /* 2s complement assumption */
2152 if (value <= (UV)IV_MIN) {
2153 SvIV_set(sv, -(IV)value);
2155 /* Too negative for an IV. This is a double upgrade, but
2156 I'm assuming it will be rare. */
2157 if (SvTYPE(sv) < SVt_PVNV)
2158 sv_upgrade(sv, SVt_PVNV);
2162 SvNV_set(sv, -(NV)value);
2163 SvIV_set(sv, IV_MIN);
2167 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2168 will be in the previous block to set the IV slot, and the next
2169 block to set the NV slot. So no else here. */
2171 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2172 != IS_NUMBER_IN_UV) {
2173 /* It wasn't an (integer that doesn't overflow the UV). */
2174 SvNV_set(sv, Atof(SvPVX_const(sv)));
2176 if (! numtype && ckWARN(WARN_NUMERIC))
2179 #if defined(USE_LONG_DOUBLE)
2180 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2181 PTR2UV(sv), SvNVX(sv)));
2183 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2184 PTR2UV(sv), SvNVX(sv)));
2187 #ifdef NV_PRESERVES_UV
2188 (void)SvIOKp_on(sv);
2190 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2195 NOOP; /* Integer is imprecise. NOK, IOKp */
2197 /* UV will not work better than IV */
2199 if (SvNVX(sv) > (NV)UV_MAX) {
2201 /* Integer is inaccurate. NOK, IOKp, is UV */
2202 SvUV_set(sv, UV_MAX);
2204 SvUV_set(sv, U_V(SvNVX(sv)));
2205 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2206 NV preservse UV so can do correct comparison. */
2207 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2210 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2215 #else /* NV_PRESERVES_UV */
2216 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2217 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2218 /* The IV/UV slot will have been set from value returned by
2219 grok_number above. The NV slot has just been set using
2222 assert (SvIOKp(sv));
2224 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2225 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2226 /* Small enough to preserve all bits. */
2227 (void)SvIOKp_on(sv);
2229 SvIV_set(sv, I_V(SvNVX(sv)));
2230 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2232 /* Assumption: first non-preserved integer is < IV_MAX,
2233 this NV is in the preserved range, therefore: */
2234 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2236 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2240 0 0 already failed to read UV.
2241 0 1 already failed to read UV.
2242 1 0 you won't get here in this case. IV/UV
2243 slot set, public IOK, Atof() unneeded.
2244 1 1 already read UV.
2245 so there's no point in sv_2iuv_non_preserve() attempting
2246 to use atol, strtol, strtoul etc. */
2248 sv_2iuv_non_preserve (sv, numtype);
2250 sv_2iuv_non_preserve (sv);
2254 #endif /* NV_PRESERVES_UV */
2255 /* It might be more code efficient to go through the entire logic above
2256 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2257 gets complex and potentially buggy, so more programmer efficient
2258 to do it this way, by turning off the public flags: */
2260 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2264 if (isGV_with_GP(sv))
2265 return glob_2number(MUTABLE_GV(sv));
2267 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2268 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2271 if (SvTYPE(sv) < SVt_IV)
2272 /* Typically the caller expects that sv_any is not NULL now. */
2273 sv_upgrade(sv, SVt_IV);
2274 /* Return 0 from the caller. */
2281 =for apidoc sv_2iv_flags
2283 Return the integer value of an SV, doing any necessary string
2284 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2285 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2291 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2296 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2297 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2298 cache IVs just in case. In practice it seems that they never
2299 actually anywhere accessible by user Perl code, let alone get used
2300 in anything other than a string context. */
2301 if (flags & SV_GMAGIC)
2306 return I_V(SvNVX(sv));
2308 if (SvPOKp(sv) && SvLEN(sv)) {
2311 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2313 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2314 == IS_NUMBER_IN_UV) {
2315 /* It's definitely an integer */
2316 if (numtype & IS_NUMBER_NEG) {
2317 if (value < (UV)IV_MIN)
2320 if (value < (UV)IV_MAX)
2325 if (ckWARN(WARN_NUMERIC))
2328 return I_V(Atof(SvPVX_const(sv)));
2333 assert(SvTYPE(sv) >= SVt_PVMG);
2334 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2335 } else if (SvTHINKFIRST(sv)) {
2339 SV * const tmpstr=AMG_CALLun(sv,numer);
2340 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2341 return SvIV(tmpstr);
2344 return PTR2IV(SvRV(sv));
2347 sv_force_normal_flags(sv, 0);
2349 if (SvREADONLY(sv) && !SvOK(sv)) {
2350 if (ckWARN(WARN_UNINITIALIZED))
2356 if (S_sv_2iuv_common(aTHX_ sv))
2359 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2360 PTR2UV(sv),SvIVX(sv)));
2361 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2365 =for apidoc sv_2uv_flags
2367 Return the unsigned integer value of an SV, doing any necessary string
2368 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2369 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2375 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2380 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2381 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2382 cache IVs just in case. */
2383 if (flags & SV_GMAGIC)
2388 return U_V(SvNVX(sv));
2389 if (SvPOKp(sv) && SvLEN(sv)) {
2392 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2394 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2395 == IS_NUMBER_IN_UV) {
2396 /* It's definitely an integer */
2397 if (!(numtype & IS_NUMBER_NEG))
2401 if (ckWARN(WARN_NUMERIC))
2404 return U_V(Atof(SvPVX_const(sv)));
2409 assert(SvTYPE(sv) >= SVt_PVMG);
2410 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2411 } else if (SvTHINKFIRST(sv)) {
2415 SV *const tmpstr = AMG_CALLun(sv,numer);
2416 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2417 return SvUV(tmpstr);
2420 return PTR2UV(SvRV(sv));
2423 sv_force_normal_flags(sv, 0);
2425 if (SvREADONLY(sv) && !SvOK(sv)) {
2426 if (ckWARN(WARN_UNINITIALIZED))
2432 if (S_sv_2iuv_common(aTHX_ sv))
2436 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2437 PTR2UV(sv),SvUVX(sv)));
2438 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2444 Return the num value of an SV, doing any necessary string or integer
2445 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2452 Perl_sv_2nv(pTHX_ register SV *const sv)
2457 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2458 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2459 cache IVs just in case. */
2463 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2464 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2465 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2467 return Atof(SvPVX_const(sv));
2471 return (NV)SvUVX(sv);
2473 return (NV)SvIVX(sv);
2478 assert(SvTYPE(sv) >= SVt_PVMG);
2479 /* This falls through to the report_uninit near the end of the
2481 } else if (SvTHINKFIRST(sv)) {
2485 SV *const tmpstr = AMG_CALLun(sv,numer);
2486 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2487 return SvNV(tmpstr);
2490 return PTR2NV(SvRV(sv));
2493 sv_force_normal_flags(sv, 0);
2495 if (SvREADONLY(sv) && !SvOK(sv)) {
2496 if (ckWARN(WARN_UNINITIALIZED))
2501 if (SvTYPE(sv) < SVt_NV) {
2502 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2503 sv_upgrade(sv, SVt_NV);
2504 #ifdef USE_LONG_DOUBLE
2506 STORE_NUMERIC_LOCAL_SET_STANDARD();
2507 PerlIO_printf(Perl_debug_log,
2508 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2509 PTR2UV(sv), SvNVX(sv));
2510 RESTORE_NUMERIC_LOCAL();
2514 STORE_NUMERIC_LOCAL_SET_STANDARD();
2515 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2516 PTR2UV(sv), SvNVX(sv));
2517 RESTORE_NUMERIC_LOCAL();
2521 else if (SvTYPE(sv) < SVt_PVNV)
2522 sv_upgrade(sv, SVt_PVNV);
2527 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2528 #ifdef NV_PRESERVES_UV
2534 /* Only set the public NV OK flag if this NV preserves the IV */
2535 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2537 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2538 : (SvIVX(sv) == I_V(SvNVX(sv))))
2544 else if (SvPOKp(sv) && SvLEN(sv)) {
2546 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2547 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2549 #ifdef NV_PRESERVES_UV
2550 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2551 == IS_NUMBER_IN_UV) {
2552 /* It's definitely an integer */
2553 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2555 SvNV_set(sv, Atof(SvPVX_const(sv)));
2561 SvNV_set(sv, Atof(SvPVX_const(sv)));
2562 /* Only set the public NV OK flag if this NV preserves the value in
2563 the PV at least as well as an IV/UV would.
2564 Not sure how to do this 100% reliably. */
2565 /* if that shift count is out of range then Configure's test is
2566 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2568 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2569 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2570 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2571 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2572 /* Can't use strtol etc to convert this string, so don't try.
2573 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2576 /* value has been set. It may not be precise. */
2577 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2578 /* 2s complement assumption for (UV)IV_MIN */
2579 SvNOK_on(sv); /* Integer is too negative. */
2584 if (numtype & IS_NUMBER_NEG) {
2585 SvIV_set(sv, -(IV)value);
2586 } else if (value <= (UV)IV_MAX) {
2587 SvIV_set(sv, (IV)value);
2589 SvUV_set(sv, value);
2593 if (numtype & IS_NUMBER_NOT_INT) {
2594 /* I believe that even if the original PV had decimals,
2595 they are lost beyond the limit of the FP precision.
2596 However, neither is canonical, so both only get p
2597 flags. NWC, 2000/11/25 */
2598 /* Both already have p flags, so do nothing */
2600 const NV nv = SvNVX(sv);
2601 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2602 if (SvIVX(sv) == I_V(nv)) {
2605 /* It had no "." so it must be integer. */
2609 /* between IV_MAX and NV(UV_MAX).
2610 Could be slightly > UV_MAX */
2612 if (numtype & IS_NUMBER_NOT_INT) {
2613 /* UV and NV both imprecise. */
2615 const UV nv_as_uv = U_V(nv);
2617 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2626 /* It might be more code efficient to go through the entire logic above
2627 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2628 gets complex and potentially buggy, so more programmer efficient
2629 to do it this way, by turning off the public flags: */
2631 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2632 #endif /* NV_PRESERVES_UV */
2635 if (isGV_with_GP(sv)) {
2636 glob_2number(MUTABLE_GV(sv));
2640 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2642 assert (SvTYPE(sv) >= SVt_NV);
2643 /* Typically the caller expects that sv_any is not NULL now. */
2644 /* XXX Ilya implies that this is a bug in callers that assume this
2645 and ideally should be fixed. */
2648 #if defined(USE_LONG_DOUBLE)
2650 STORE_NUMERIC_LOCAL_SET_STANDARD();
2651 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2652 PTR2UV(sv), SvNVX(sv));
2653 RESTORE_NUMERIC_LOCAL();
2657 STORE_NUMERIC_LOCAL_SET_STANDARD();
2658 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2659 PTR2UV(sv), SvNVX(sv));
2660 RESTORE_NUMERIC_LOCAL();
2669 Return an SV with the numeric value of the source SV, doing any necessary
2670 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2671 access this function.
2677 Perl_sv_2num(pTHX_ register SV *const sv)
2679 PERL_ARGS_ASSERT_SV_2NUM;
2684 SV * const tmpsv = AMG_CALLun(sv,numer);
2685 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2686 return sv_2num(tmpsv);
2688 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2691 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2692 * UV as a string towards the end of buf, and return pointers to start and
2695 * We assume that buf is at least TYPE_CHARS(UV) long.
2699 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2701 char *ptr = buf + TYPE_CHARS(UV);
2702 char * const ebuf = ptr;
2705 PERL_ARGS_ASSERT_UIV_2BUF;
2717 *--ptr = '0' + (char)(uv % 10);
2726 =for apidoc sv_2pv_flags
2728 Returns a pointer to the string value of an SV, and sets *lp to its length.
2729 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2731 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2732 usually end up here too.
2738 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2748 if (SvGMAGICAL(sv)) {
2749 if (flags & SV_GMAGIC)
2754 if (flags & SV_MUTABLE_RETURN)
2755 return SvPVX_mutable(sv);
2756 if (flags & SV_CONST_RETURN)
2757 return (char *)SvPVX_const(sv);
2760 if (SvIOKp(sv) || SvNOKp(sv)) {
2761 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2766 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2767 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2769 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2776 #ifdef FIXNEGATIVEZERO
2777 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2783 SvUPGRADE(sv, SVt_PV);
2786 s = SvGROW_mutable(sv, len + 1);
2789 return (char*)memcpy(s, tbuf, len + 1);
2795 assert(SvTYPE(sv) >= SVt_PVMG);
2796 /* This falls through to the report_uninit near the end of the
2798 } else if (SvTHINKFIRST(sv)) {
2802 SV *const tmpstr = AMG_CALLun(sv,string);
2803 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2805 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2809 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2810 if (flags & SV_CONST_RETURN) {
2811 pv = (char *) SvPVX_const(tmpstr);
2813 pv = (flags & SV_MUTABLE_RETURN)
2814 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2817 *lp = SvCUR(tmpstr);
2819 pv = sv_2pv_flags(tmpstr, lp, flags);
2832 SV *const referent = SvRV(sv);
2836 retval = buffer = savepvn("NULLREF", len);
2837 } else if (SvTYPE(referent) == SVt_REGEXP) {
2838 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2843 /* If the regex is UTF-8 we want the containing scalar to
2844 have an UTF-8 flag too */
2850 if ((seen_evals = RX_SEEN_EVALS(re)))
2851 PL_reginterp_cnt += seen_evals;
2854 *lp = RX_WRAPLEN(re);
2856 return RX_WRAPPED(re);
2858 const char *const typestr = sv_reftype(referent, 0);
2859 const STRLEN typelen = strlen(typestr);
2860 UV addr = PTR2UV(referent);
2861 const char *stashname = NULL;
2862 STRLEN stashnamelen = 0; /* hush, gcc */
2863 const char *buffer_end;
2865 if (SvOBJECT(referent)) {
2866 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2869 stashname = HEK_KEY(name);
2870 stashnamelen = HEK_LEN(name);
2872 if (HEK_UTF8(name)) {
2878 stashname = "__ANON__";
2881 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2882 + 2 * sizeof(UV) + 2 /* )\0 */;
2884 len = typelen + 3 /* (0x */
2885 + 2 * sizeof(UV) + 2 /* )\0 */;
2888 Newx(buffer, len, char);
2889 buffer_end = retval = buffer + len;
2891 /* Working backwards */
2895 *--retval = PL_hexdigit[addr & 15];
2896 } while (addr >>= 4);
2902 memcpy(retval, typestr, typelen);
2906 retval -= stashnamelen;
2907 memcpy(retval, stashname, stashnamelen);
2909 /* retval may not neccesarily have reached the start of the
2911 assert (retval >= buffer);
2913 len = buffer_end - retval - 1; /* -1 for that \0 */
2921 if (SvREADONLY(sv) && !SvOK(sv)) {
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2926 if (ckWARN(WARN_UNINITIALIZED))
2931 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2932 /* I'm assuming that if both IV and NV are equally valid then
2933 converting the IV is going to be more efficient */
2934 const U32 isUIOK = SvIsUV(sv);
2935 char buf[TYPE_CHARS(UV)];
2939 if (SvTYPE(sv) < SVt_PVIV)
2940 sv_upgrade(sv, SVt_PVIV);
2941 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2943 /* inlined from sv_setpvn */
2944 s = SvGROW_mutable(sv, len + 1);
2945 Move(ptr, s, len, char);
2949 else if (SvNOKp(sv)) {
2951 if (SvTYPE(sv) < SVt_PVNV)
2952 sv_upgrade(sv, SVt_PVNV);
2953 /* The +20 is pure guesswork. Configure test needed. --jhi */
2954 s = SvGROW_mutable(sv, NV_DIG + 20);
2955 /* some Xenix systems wipe out errno here */
2957 if (SvNVX(sv) == 0.0)
2958 my_strlcpy(s, "0", SvLEN(sv));
2962 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2965 #ifdef FIXNEGATIVEZERO
2966 if (*s == '-' && s[1] == '0' && !s[2]) {
2978 if (isGV_with_GP(sv)) {
2979 GV *const gv = MUTABLE_GV(sv);
2980 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2981 SV *const buffer = sv_newmortal();
2983 /* FAKE globs can get coerced, so need to turn this off temporarily
2986 gv_efullname3(buffer, gv, "*");
2987 SvFLAGS(gv) |= wasfake;
2989 assert(SvPOK(buffer));
2991 *lp = SvCUR(buffer);
2993 return SvPVX(buffer);
2998 if (flags & SV_UNDEF_RETURNS_NULL)
3000 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3002 if (SvTYPE(sv) < SVt_PV)
3003 /* Typically the caller expects that sv_any is not NULL now. */
3004 sv_upgrade(sv, SVt_PV);
3008 const STRLEN len = s - SvPVX_const(sv);
3014 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3015 PTR2UV(sv),SvPVX_const(sv)));
3016 if (flags & SV_CONST_RETURN)
3017 return (char *)SvPVX_const(sv);
3018 if (flags & SV_MUTABLE_RETURN)
3019 return SvPVX_mutable(sv);
3024 =for apidoc sv_copypv
3026 Copies a stringified representation of the source SV into the
3027 destination SV. Automatically performs any necessary mg_get and
3028 coercion of numeric values into strings. Guaranteed to preserve
3029 UTF8 flag even from overloaded objects. Similar in nature to
3030 sv_2pv[_flags] but operates directly on an SV instead of just the
3031 string. Mostly uses sv_2pv_flags to do its work, except when that
3032 would lose the UTF-8'ness of the PV.
3038 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3041 const char * const s = SvPV_const(ssv,len);
3043 PERL_ARGS_ASSERT_SV_COPYPV;
3045 sv_setpvn(dsv,s,len);
3053 =for apidoc sv_2pvbyte
3055 Return a pointer to the byte-encoded representation of the SV, and set *lp
3056 to its length. May cause the SV to be downgraded from UTF-8 as a
3059 Usually accessed via the C<SvPVbyte> macro.
3065 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3067 PERL_ARGS_ASSERT_SV_2PVBYTE;
3069 sv_utf8_downgrade(sv,0);
3070 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3074 =for apidoc sv_2pvutf8
3076 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3077 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3079 Usually accessed via the C<SvPVutf8> macro.
3085 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3087 PERL_ARGS_ASSERT_SV_2PVUTF8;
3089 sv_utf8_upgrade(sv);
3090 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3095 =for apidoc sv_2bool
3097 This function is only called on magical items, and is only used by
3098 sv_true() or its macro equivalent.
3104 Perl_sv_2bool(pTHX_ register SV *const sv)
3108 PERL_ARGS_ASSERT_SV_2BOOL;
3116 SV * const tmpsv = AMG_CALLun(sv,bool_);
3117 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3118 return (bool)SvTRUE(tmpsv);
3120 return SvRV(sv) != 0;
3123 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3125 (*sv->sv_u.svu_pv > '0' ||
3126 Xpvtmp->xpv_cur > 1 ||
3127 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3134 return SvIVX(sv) != 0;
3137 return SvNVX(sv) != 0.0;
3139 if (isGV_with_GP(sv))
3149 =for apidoc sv_utf8_upgrade
3151 Converts the PV of an SV to its UTF-8-encoded form.
3152 Forces the SV to string form if it is not already.
3153 Will C<mg_get> on C<sv> if appropriate.
3154 Always sets the SvUTF8 flag to avoid future validity checks even
3155 if the whole string is the same in UTF-8 as not.
3156 Returns the number of bytes in the converted string
3158 This is not as a general purpose byte encoding to Unicode interface:
3159 use the Encode extension for that.
3161 =for apidoc sv_utf8_upgrade_nomg
3163 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3165 =for apidoc sv_utf8_upgrade_flags
3167 Converts the PV of an SV to its UTF-8-encoded form.
3168 Forces the SV to string form if it is not already.
3169 Always sets the SvUTF8 flag to avoid future validity checks even
3170 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3171 will C<mg_get> on C<sv> if appropriate, else not.
3172 Returns the number of bytes in the converted string
3173 C<sv_utf8_upgrade> and
3174 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3176 This is not as a general purpose byte encoding to Unicode interface:
3177 use the Encode extension for that.
3181 The grow version is currently not externally documented. It adds a parameter,
3182 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3183 have free after it upon return. This allows the caller to reserve extra space
3184 that it intends to fill, to avoid extra grows.
3186 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3187 which can be used to tell this function to not first check to see if there are
3188 any characters that are different in UTF-8 (variant characters) which would
3189 force it to allocate a new string to sv, but to assume there are. Typically
3190 this flag is used by a routine that has already parsed the string to find that
3191 there are such characters, and passes this information on so that the work
3192 doesn't have to be repeated.
3194 (One might think that the calling routine could pass in the position of the
3195 first such variant, so it wouldn't have to be found again. But that is not the
3196 case, because typically when the caller is likely to use this flag, it won't be
3197 calling this routine unless it finds something that won't fit into a byte.
3198 Otherwise it tries to not upgrade and just use bytes. But some things that
3199 do fit into a byte are variants in utf8, and the caller may not have been
3200 keeping track of these.)
3202 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3203 isn't guaranteed due to having other routines do the work in some input cases,
3204 or if the input is already flagged as being in utf8.
3206 The speed of this could perhaps be improved for many cases if someone wanted to
3207 write a fast function that counts the number of variant characters in a string,
3208 especially if it could return the position of the first one.
3213 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3217 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3219 if (sv == &PL_sv_undef)
3223 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3224 (void) sv_2pv_flags(sv,&len, flags);
3226 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3230 (void) SvPV_force(sv,len);
3235 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3240 sv_force_normal_flags(sv, 0);
3243 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3244 sv_recode_to_utf8(sv, PL_encoding);
3245 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3249 if (SvCUR(sv) == 0) {
3250 if (extra) SvGROW(sv, extra);
3251 } else { /* Assume Latin-1/EBCDIC */
3252 /* This function could be much more efficient if we
3253 * had a FLAG in SVs to signal if there are any variant
3254 * chars in the PV. Given that there isn't such a flag
3255 * make the loop as fast as possible (although there are certainly ways
3256 * to speed this up, eg. through vectorization) */
3257 U8 * s = (U8 *) SvPVX_const(sv);
3258 U8 * e = (U8 *) SvEND(sv);
3260 STRLEN two_byte_count = 0;
3262 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3264 /* See if really will need to convert to utf8. We mustn't rely on our
3265 * incoming SV being well formed and having a trailing '\0', as certain
3266 * code in pp_formline can send us partially built SVs. */
3270 if (NATIVE_IS_INVARIANT(ch)) continue;
3272 t--; /* t already incremented; re-point to first variant */
3277 /* utf8 conversion not needed because all are invariants. Mark as
3278 * UTF-8 even if no variant - saves scanning loop */
3284 /* Here, the string should be converted to utf8, either because of an
3285 * input flag (two_byte_count = 0), or because a character that
3286 * requires 2 bytes was found (two_byte_count = 1). t points either to
3287 * the beginning of the string (if we didn't examine anything), or to
3288 * the first variant. In either case, everything from s to t - 1 will
3289 * occupy only 1 byte each on output.
3291 * There are two main ways to convert. One is to create a new string
3292 * and go through the input starting from the beginning, appending each
3293 * converted value onto the new string as we go along. It's probably
3294 * best to allocate enough space in the string for the worst possible
3295 * case rather than possibly running out of space and having to
3296 * reallocate and then copy what we've done so far. Since everything
3297 * from s to t - 1 is invariant, the destination can be initialized
3298 * with these using a fast memory copy
3300 * The other way is to figure out exactly how big the string should be
3301 * by parsing the entire input. Then you don't have to make it big
3302 * enough to handle the worst possible case, and more importantly, if
3303 * the string you already have is large enough, you don't have to
3304 * allocate a new string, you can copy the last character in the input
3305 * string to the final position(s) that will be occupied by the
3306 * converted string and go backwards, stopping at t, since everything
3307 * before that is invariant.
3309 * There are advantages and disadvantages to each method.
3311 * In the first method, we can allocate a new string, do the memory
3312 * copy from the s to t - 1, and then proceed through the rest of the
3313 * string byte-by-byte.
3315 * In the second method, we proceed through the rest of the input
3316 * string just calculating how big the converted string will be. Then
3317 * there are two cases:
3318 * 1) if the string has enough extra space to handle the converted
3319 * value. We go backwards through the string, converting until we
3320 * get to the position we are at now, and then stop. If this
3321 * position is far enough along in the string, this method is
3322 * faster than the other method. If the memory copy were the same
3323 * speed as the byte-by-byte loop, that position would be about
3324 * half-way, as at the half-way mark, parsing to the end and back
3325 * is one complete string's parse, the same amount as starting
3326 * over and going all the way through. Actually, it would be
3327 * somewhat less than half-way, as it's faster to just count bytes
3328 * than to also copy, and we don't have the overhead of allocating
3329 * a new string, changing the scalar to use it, and freeing the
3330 * existing one. But if the memory copy is fast, the break-even
3331 * point is somewhere after half way. The counting loop could be
3332 * sped up by vectorization, etc, to move the break-even point
3333 * further towards the beginning.
3334 * 2) if the string doesn't have enough space to handle the converted
3335 * value. A new string will have to be allocated, and one might
3336 * as well, given that, start from the beginning doing the first
3337 * method. We've spent extra time parsing the string and in
3338 * exchange all we've gotten is that we know precisely how big to
3339 * make the new one. Perl is more optimized for time than space,
3340 * so this case is a loser.
3341 * So what I've decided to do is not use the 2nd method unless it is
3342 * guaranteed that a new string won't have to be allocated, assuming
3343 * the worst case. I also decided not to put any more conditions on it
3344 * than this, for now. It seems likely that, since the worst case is
3345 * twice as big as the unknown portion of the string (plus 1), we won't
3346 * be guaranteed enough space, causing us to go to the first method,
3347 * unless the string is short, or the first variant character is near
3348 * the end of it. In either of these cases, it seems best to use the
3349 * 2nd method. The only circumstance I can think of where this would
3350 * be really slower is if the string had once had much more data in it
3351 * than it does now, but there is still a substantial amount in it */
3354 STRLEN invariant_head = t - s;
3355 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3356 if (SvLEN(sv) < size) {
3358 /* Here, have decided to allocate a new string */
3363 Newx(dst, size, U8);
3365 /* If no known invariants at the beginning of the input string,
3366 * set so starts from there. Otherwise, can use memory copy to
3367 * get up to where we are now, and then start from here */
3369 if (invariant_head <= 0) {
3372 Copy(s, dst, invariant_head, char);
3373 d = dst + invariant_head;
3377 const UV uv = NATIVE8_TO_UNI(*t++);
3378 if (UNI_IS_INVARIANT(uv))
3379 *d++ = (U8)UNI_TO_NATIVE(uv);
3381 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3382 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3386 SvPV_free(sv); /* No longer using pre-existing string */
3387 SvPV_set(sv, (char*)dst);
3388 SvCUR_set(sv, d - dst);
3389 SvLEN_set(sv, size);
3392 /* Here, have decided to get the exact size of the string.
3393 * Currently this happens only when we know that there is
3394 * guaranteed enough space to fit the converted string, so
3395 * don't have to worry about growing. If two_byte_count is 0,
3396 * then t points to the first byte of the string which hasn't
3397 * been examined yet. Otherwise two_byte_count is 1, and t
3398 * points to the first byte in the string that will expand to
3399 * two. Depending on this, start examining at t or 1 after t.
3402 U8 *d = t + two_byte_count;
3405 /* Count up the remaining bytes that expand to two */
3408 const U8 chr = *d++;
3409 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3412 /* The string will expand by just the number of bytes that
3413 * occupy two positions. But we are one afterwards because of
3414 * the increment just above. This is the place to put the
3415 * trailing NUL, and to set the length before we decrement */
3417 d += two_byte_count;
3418 SvCUR_set(sv, d - s);
3422 /* Having decremented d, it points to the position to put the
3423 * very last byte of the expanded string. Go backwards through
3424 * the string, copying and expanding as we go, stopping when we
3425 * get to the part that is invariant the rest of the way down */
3429 const U8 ch = NATIVE8_TO_UNI(*e--);
3430 if (UNI_IS_INVARIANT(ch)) {
3431 *d-- = UNI_TO_NATIVE(ch);
3433 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3434 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3441 /* Mark as UTF-8 even if no variant - saves scanning loop */
3447 =for apidoc sv_utf8_downgrade
3449 Attempts to convert the PV of an SV from characters to bytes.
3450 If the PV contains a character that cannot fit
3451 in a byte, this conversion will fail;
3452 in this case, either returns false or, if C<fail_ok> is not
3455 This is not as a general purpose Unicode to byte encoding interface:
3456 use the Encode extension for that.
3462 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3466 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3468 if (SvPOKp(sv) && SvUTF8(sv)) {
3474 sv_force_normal_flags(sv, 0);
3476 s = (U8 *) SvPV(sv, len);
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;
3510 sv_force_normal_flags(sv, 0);
3512 if (SvREADONLY(sv)) {
3513 Perl_croak(aTHX_ "%s", PL_no_modify);
3515 (void) sv_utf8_upgrade(sv);
3520 =for apidoc sv_utf8_decode
3522 If the PV of the SV is an octet sequence in UTF-8
3523 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3524 so that it looks like a character. If the PV contains only single-byte
3525 characters, the C<SvUTF8> flag stays being off.
3526 Scans PV for validity and returns false if the PV is invalid UTF-8.
3532 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3534 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3540 /* The octets may have got themselves encoded - get them back as
3543 if (!sv_utf8_downgrade(sv, TRUE))
3546 /* it is actually just a matter of turning the utf8 flag on, but
3547 * we want to make sure everything inside is valid utf8 first.
3549 c = (const U8 *) SvPVX_const(sv);
3550 if (!is_utf8_string(c, SvCUR(sv)+1))
3552 e = (const U8 *) SvEND(sv);
3555 if (!UTF8_IS_INVARIANT(ch)) {
3565 =for apidoc sv_setsv
3567 Copies the contents of the source SV C<ssv> into the destination SV
3568 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3569 function if the source SV needs to be reused. Does not handle 'set' magic.
3570 Loosely speaking, it performs a copy-by-value, obliterating any previous
3571 content of the destination.
3573 You probably want to use one of the assortment of wrappers, such as
3574 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3575 C<SvSetMagicSV_nosteal>.
3577 =for apidoc sv_setsv_flags
3579 Copies the contents of the source SV C<ssv> into the destination SV
3580 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3581 function if the source SV needs to be reused. Does not handle 'set' magic.
3582 Loosely speaking, it performs a copy-by-value, obliterating any previous
3583 content of the destination.
3584 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3585 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3586 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3587 and C<sv_setsv_nomg> are implemented in terms of this function.
3589 You probably want to use one of the assortment of wrappers, such as
3590 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3591 C<SvSetMagicSV_nosteal>.
3593 This is the primary function for copying scalars, and most other
3594 copy-ish functions and macros use this underneath.
3600 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3602 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3604 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3606 if (dtype != SVt_PVGV) {
3607 const char * const name = GvNAME(sstr);
3608 const STRLEN len = GvNAMELEN(sstr);
3610 if (dtype >= SVt_PV) {
3616 SvUPGRADE(dstr, SVt_PVGV);
3617 (void)SvOK_off(dstr);
3618 /* FIXME - why are we doing this, then turning it off and on again
3620 isGV_with_GP_on(dstr);
3622 GvSTASH(dstr) = GvSTASH(sstr);
3624 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3625 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3626 SvFAKE_on(dstr); /* can coerce to non-glob */
3629 if(GvGP(MUTABLE_GV(sstr))) {
3630 /* If source has method cache entry, clear it */
3632 SvREFCNT_dec(GvCV(sstr));
3636 /* If source has a real method, then a method is
3638 else if(GvCV((const GV *)sstr)) {
3643 /* If dest already had a real method, that's a change as well */
3644 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3648 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3651 gp_free(MUTABLE_GV(dstr));
3652 isGV_with_GP_off(dstr);
3653 (void)SvOK_off(dstr);
3654 isGV_with_GP_on(dstr);
3655 GvINTRO_off(dstr); /* one-shot flag */
3656 GvGP(dstr) = gp_ref(GvGP(sstr));
3657 if (SvTAINTED(sstr))
3659 if (GvIMPORTED(dstr) != GVf_IMPORTED
3660 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3662 GvIMPORTED_on(dstr);
3665 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3666 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3671 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3673 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3675 const int intro = GvINTRO(dstr);
3678 const U32 stype = SvTYPE(sref);
3679 bool mro_changes = FALSE;
3681 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3684 GvINTRO_off(dstr); /* one-shot flag */
3685 GvLINE(dstr) = CopLINE(PL_curcop);
3686 GvEGV(dstr) = MUTABLE_GV(dstr);
3691 location = (SV **) &GvCV(dstr);
3692 import_flag = GVf_IMPORTED_CV;
3695 location = (SV **) &GvHV(dstr);
3696 import_flag = GVf_IMPORTED_HV;
3699 location = (SV **) &GvAV(dstr);
3700 if (strEQ(GvNAME((GV*)dstr), "ISA"))
3702 import_flag = GVf_IMPORTED_AV;
3705 location = (SV **) &GvIOp(dstr);
3708 location = (SV **) &GvFORM(dstr);
3711 location = &GvSV(dstr);
3712 import_flag = GVf_IMPORTED_SV;
3715 if (stype == SVt_PVCV) {
3716 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3717 if (GvCVGEN(dstr)) {
3718 SvREFCNT_dec(GvCV(dstr));
3720 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3723 SAVEGENERICSV(*location);
3727 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3728 CV* const cv = MUTABLE_CV(*location);
3730 if (!GvCVGEN((const GV *)dstr) &&
3731 (CvROOT(cv) || CvXSUB(cv)))
3733 /* Redefining a sub - warning is mandatory if
3734 it was a const and its value changed. */
3735 if (CvCONST(cv) && CvCONST((const CV *)sref)
3737 == cv_const_sv((const CV *)sref)) {
3739 /* They are 2 constant subroutines generated from
3740 the same constant. This probably means that
3741 they are really the "same" proxy subroutine
3742 instantiated in 2 places. Most likely this is
3743 when a constant is exported twice. Don't warn.
3746 else if (ckWARN(WARN_REDEFINE)
3748 && (!CvCONST((const CV *)sref)
3749 || sv_cmp(cv_const_sv(cv),
3750 cv_const_sv((const CV *)
3752 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3755 ? "Constant subroutine %s::%s redefined"
3756 : "Subroutine %s::%s redefined"),
3757 HvNAME_get(GvSTASH((const GV *)dstr)),
3758 GvENAME(MUTABLE_GV(dstr)));
3762 cv_ckproto_len(cv, (const GV *)dstr,
3763 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3764 SvPOK(sref) ? SvCUR(sref) : 0);
3766 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3767 GvASSUMECV_on(dstr);
3768 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3771 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3772 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3773 GvFLAGS(dstr) |= import_flag;
3778 if (SvTAINTED(sstr))
3780 if (mro_changes) mro_isa_changed_in(GvSTASH(dstr));
3785 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3788 register U32 sflags;
3790 register svtype stype;
3792 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3797 if (SvIS_FREED(dstr)) {
3798 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3799 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3801 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3803 sstr = &PL_sv_undef;
3804 if (SvIS_FREED(sstr)) {
3805 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3806 (void*)sstr, (void*)dstr);
3808 stype = SvTYPE(sstr);
3809 dtype = SvTYPE(dstr);
3811 (void)SvAMAGIC_off(dstr);
3814 /* need to nuke the magic */
3818 /* There's a lot of redundancy below but we're going for speed here */
3823 if (dtype != SVt_PVGV) {
3824 (void)SvOK_off(dstr);
3832 sv_upgrade(dstr, SVt_IV);
3836 sv_upgrade(dstr, SVt_PVIV);
3839 goto end_of_first_switch;
3841 (void)SvIOK_only(dstr);
3842 SvIV_set(dstr, SvIVX(sstr));
3845 /* SvTAINTED can only be true if the SV has taint magic, which in
3846 turn means that the SV type is PVMG (or greater). This is the
3847 case statement for SVt_IV, so this cannot be true (whatever gcov
3849 assert(!SvTAINTED(sstr));
3854 if (dtype < SVt_PV && dtype != SVt_IV)
3855 sv_upgrade(dstr, SVt_IV);
3863 sv_upgrade(dstr, SVt_NV);
3867 sv_upgrade(dstr, SVt_PVNV);
3870 goto end_of_first_switch;
3872 SvNV_set(dstr, SvNVX(sstr));
3873 (void)SvNOK_only(dstr);
3874 /* SvTAINTED can only be true if the SV has taint magic, which in
3875 turn means that the SV type is PVMG (or greater). This is the
3876 case statement for SVt_NV, so this cannot be true (whatever gcov
3878 assert(!SvTAINTED(sstr));
3884 #ifdef PERL_OLD_COPY_ON_WRITE
3885 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3886 if (dtype < SVt_PVIV)
3887 sv_upgrade(dstr, SVt_PVIV);
3894 sv_upgrade(dstr, SVt_PV);
3897 if (dtype < SVt_PVIV)
3898 sv_upgrade(dstr, SVt_PVIV);
3901 if (dtype < SVt_PVNV)
3902 sv_upgrade(dstr, SVt_PVNV);
3906 const char * const type = sv_reftype(sstr,0);
3908 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3910 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3915 if (dtype < SVt_REGEXP)
3916 sv_upgrade(dstr, SVt_REGEXP);
3919 /* case SVt_BIND: */
3922 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3923 glob_assign_glob(dstr, sstr, dtype);
3926 /* SvVALID means that this PVGV is playing at being an FBM. */
3930 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3932 if (SvTYPE(sstr) != stype) {
3933 stype = SvTYPE(sstr);
3934 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3935 glob_assign_glob(dstr, sstr, dtype);
3940 if (stype == SVt_PVLV)
3941 SvUPGRADE(dstr, SVt_PVNV);
3943 SvUPGRADE(dstr, (svtype)stype);
3945 end_of_first_switch:
3947 /* dstr may have been upgraded. */
3948 dtype = SvTYPE(dstr);
3949 sflags = SvFLAGS(sstr);
3951 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3952 /* Assigning to a subroutine sets the prototype. */
3955 const char *const ptr = SvPV_const(sstr, len);
3957 SvGROW(dstr, len + 1);
3958 Copy(ptr, SvPVX(dstr), len + 1, char);
3959 SvCUR_set(dstr, len);
3961 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3965 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3966 const char * const type = sv_reftype(dstr,0);
3968 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3970 Perl_croak(aTHX_ "Cannot copy to %s", type);
3971 } else if (sflags & SVf_ROK) {
3972 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3973 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3976 if (GvIMPORTED(dstr) != GVf_IMPORTED
3977 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3979 GvIMPORTED_on(dstr);
3984 glob_assign_glob(dstr, sstr, dtype);
3988 if (dtype >= SVt_PV) {
3989 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3990 glob_assign_ref(dstr, sstr);
3993 if (SvPVX_const(dstr)) {
3999 (void)SvOK_off(dstr);
4000 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4001 SvFLAGS(dstr) |= sflags & SVf_ROK;
4002 assert(!(sflags & SVp_NOK));
4003 assert(!(sflags & SVp_IOK));
4004 assert(!(sflags & SVf_NOK));
4005 assert(!(sflags & SVf_IOK));
4007 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4008 if (!(sflags & SVf_OK)) {
4009 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4010 "Undefined value assigned to typeglob");
4013 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4014 if (dstr != (const SV *)gv) {
4016 gp_free(MUTABLE_GV(dstr));
4017 GvGP(dstr) = gp_ref(GvGP(gv));
4021 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4022 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4024 else if (sflags & SVp_POK) {
4028 * Check to see if we can just swipe the string. If so, it's a
4029 * possible small lose on short strings, but a big win on long ones.
4030 * It might even be a win on short strings if SvPVX_const(dstr)
4031 * has to be allocated and SvPVX_const(sstr) has to be freed.
4032 * Likewise if we can set up COW rather than doing an actual copy, we
4033 * drop to the else clause, as the swipe code and the COW setup code
4034 * have much in common.
4037 /* Whichever path we take through the next code, we want this true,
4038 and doing it now facilitates the COW check. */
4039 (void)SvPOK_only(dstr);
4042 /* If we're already COW then this clause is not true, and if COW
4043 is allowed then we drop down to the else and make dest COW
4044 with us. If caller hasn't said that we're allowed to COW
4045 shared hash keys then we don't do the COW setup, even if the
4046 source scalar is a shared hash key scalar. */
4047 (((flags & SV_COW_SHARED_HASH_KEYS)
4048 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4049 : 1 /* If making a COW copy is forbidden then the behaviour we
4050 desire is as if the source SV isn't actually already
4051 COW, even if it is. So we act as if the source flags
4052 are not COW, rather than actually testing them. */
4054 #ifndef PERL_OLD_COPY_ON_WRITE
4055 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4056 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4057 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4058 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4059 but in turn, it's somewhat dead code, never expected to go
4060 live, but more kept as a placeholder on how to do it better
4061 in a newer implementation. */
4062 /* If we are COW and dstr is a suitable target then we drop down
4063 into the else and make dest a COW of us. */
4064 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4069 (sflags & SVs_TEMP) && /* slated for free anyway? */
4070 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4071 (!(flags & SV_NOSTEAL)) &&
4072 /* and we're allowed to steal temps */
4073 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4074 SvLEN(sstr) && /* and really is a string */
4075 /* and won't be needed again, potentially */
4076 !(PL_op && PL_op->op_type == OP_AASSIGN))
4077 #ifdef PERL_OLD_COPY_ON_WRITE
4078 && ((flags & SV_COW_SHARED_HASH_KEYS)
4079 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4080 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4081 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4085 /* Failed the swipe test, and it's not a shared hash key either.
4086 Have to copy the string. */
4087 STRLEN len = SvCUR(sstr);
4088 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4089 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4090 SvCUR_set(dstr, len);
4091 *SvEND(dstr) = '\0';
4093 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4095 /* Either it's a shared hash key, or it's suitable for
4096 copy-on-write or we can swipe the string. */
4098 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4102 #ifdef PERL_OLD_COPY_ON_WRITE
4104 if ((sflags & (SVf_FAKE | SVf_READONLY))
4105 != (SVf_FAKE | SVf_READONLY)) {
4106 SvREADONLY_on(sstr);
4108 /* Make the source SV into a loop of 1.
4109 (about to become 2) */
4110 SV_COW_NEXT_SV_SET(sstr, sstr);
4114 /* Initial code is common. */
4115 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4120 /* making another shared SV. */
4121 STRLEN cur = SvCUR(sstr);
4122 STRLEN len = SvLEN(sstr);
4123 #ifdef PERL_OLD_COPY_ON_WRITE
4125 assert (SvTYPE(dstr) >= SVt_PVIV);
4126 /* SvIsCOW_normal */
4127 /* splice us in between source and next-after-source. */
4128 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4129 SV_COW_NEXT_SV_SET(sstr, dstr);
4130 SvPV_set(dstr, SvPVX_mutable(sstr));
4134 /* SvIsCOW_shared_hash */
4135 DEBUG_C(PerlIO_printf(Perl_debug_log,
4136 "Copy on write: Sharing hash\n"));
4138 assert (SvTYPE(dstr) >= SVt_PV);
4140 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4142 SvLEN_set(dstr, len);
4143 SvCUR_set(dstr, cur);
4144 SvREADONLY_on(dstr);
4148 { /* Passes the swipe test. */
4149 SvPV_set(dstr, SvPVX_mutable(sstr));
4150 SvLEN_set(dstr, SvLEN(sstr));
4151 SvCUR_set(dstr, SvCUR(sstr));
4154 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4155 SvPV_set(sstr, NULL);
4161 if (sflags & SVp_NOK) {
4162 SvNV_set(dstr, SvNVX(sstr));
4164 if (sflags & SVp_IOK) {
4165 SvIV_set(dstr, SvIVX(sstr));
4166 /* Must do this otherwise some other overloaded use of 0x80000000
4167 gets confused. I guess SVpbm_VALID */
4168 if (sflags & SVf_IVisUV)
4171 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4173 const MAGIC * const smg = SvVSTRING_mg(sstr);
4175 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4176 smg->mg_ptr, smg->mg_len);
4177 SvRMAGICAL_on(dstr);
4181 else if (sflags & (SVp_IOK|SVp_NOK)) {
4182 (void)SvOK_off(dstr);
4183 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4184 if (sflags & SVp_IOK) {
4185 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4186 SvIV_set(dstr, SvIVX(sstr));
4188 if (sflags & SVp_NOK) {
4189 SvNV_set(dstr, SvNVX(sstr));
4193 if (isGV_with_GP(sstr)) {
4194 /* This stringification rule for globs is spread in 3 places.
4195 This feels bad. FIXME. */
4196 const U32 wasfake = sflags & SVf_FAKE;
4198 /* FAKE globs can get coerced, so need to turn this off
4199 temporarily if it is on. */
4201 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4202 SvFLAGS(sstr) |= wasfake;
4205 (void)SvOK_off(dstr);
4207 if (SvTAINTED(sstr))
4212 =for apidoc sv_setsv_mg
4214 Like C<sv_setsv>, but also handles 'set' magic.
4220 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4222 PERL_ARGS_ASSERT_SV_SETSV_MG;
4224 sv_setsv(dstr,sstr);
4228 #ifdef PERL_OLD_COPY_ON_WRITE
4230 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4232 STRLEN cur = SvCUR(sstr);
4233 STRLEN len = SvLEN(sstr);
4234 register char *new_pv;
4236 PERL_ARGS_ASSERT_SV_SETSV_COW;
4239 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4240 (void*)sstr, (void*)dstr);
4247 if (SvTHINKFIRST(dstr))
4248 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4249 else if (SvPVX_const(dstr))
4250 Safefree(SvPVX_const(dstr));
4254 SvUPGRADE(dstr, SVt_PVIV);
4256 assert (SvPOK(sstr));
4257 assert (SvPOKp(sstr));
4258 assert (!SvIOK(sstr));
4259 assert (!SvIOKp(sstr));
4260 assert (!SvNOK(sstr));
4261 assert (!SvNOKp(sstr));
4263 if (SvIsCOW(sstr)) {
4265 if (SvLEN(sstr) == 0) {
4266 /* source is a COW shared hash key. */
4267 DEBUG_C(PerlIO_printf(Perl_debug_log,
4268 "Fast copy on write: Sharing hash\n"));
4269 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4272 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4274 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4275 SvUPGRADE(sstr, SVt_PVIV);
4276 SvREADONLY_on(sstr);
4278 DEBUG_C(PerlIO_printf(Perl_debug_log,
4279 "Fast copy on write: Converting sstr to COW\n"));
4280 SV_COW_NEXT_SV_SET(dstr, sstr);
4282 SV_COW_NEXT_SV_SET(sstr, dstr);
4283 new_pv = SvPVX_mutable(sstr);
4286 SvPV_set(dstr, new_pv);
4287 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4290 SvLEN_set(dstr, len);
4291 SvCUR_set(dstr, cur);
4300 =for apidoc sv_setpvn
4302 Copies a string into an SV. The C<len> parameter indicates the number of
4303 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4304 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4310 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4313 register char *dptr;
4315 PERL_ARGS_ASSERT_SV_SETPVN;
4317 SV_CHECK_THINKFIRST_COW_DROP(sv);
4323 /* len is STRLEN which is unsigned, need to copy to signed */
4326 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4328 SvUPGRADE(sv, SVt_PV);
4330 dptr = SvGROW(sv, len + 1);
4331 Move(ptr,dptr,len,char);
4334 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4339 =for apidoc sv_setpvn_mg
4341 Like C<sv_setpvn>, but also handles 'set' magic.
4347 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4349 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4351 sv_setpvn(sv,ptr,len);
4356 =for apidoc sv_setpv
4358 Copies a string into an SV. The string must be null-terminated. Does not
4359 handle 'set' magic. See C<sv_setpv_mg>.
4365 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4368 register STRLEN len;
4370 PERL_ARGS_ASSERT_SV_SETPV;
4372 SV_CHECK_THINKFIRST_COW_DROP(sv);
4378 SvUPGRADE(sv, SVt_PV);
4380 SvGROW(sv, len + 1);
4381 Move(ptr,SvPVX(sv),len+1,char);
4383 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4388 =for apidoc sv_setpv_mg
4390 Like C<sv_setpv>, but also handles 'set' magic.
4396 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4398 PERL_ARGS_ASSERT_SV_SETPV_MG;
4405 =for apidoc sv_usepvn_flags
4407 Tells an SV to use C<ptr> to find its string value. Normally the
4408 string is stored inside the SV but sv_usepvn allows the SV to use an
4409 outside string. The C<ptr> should point to memory that was allocated
4410 by C<malloc>. The string length, C<len>, must be supplied. By default
4411 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4412 so that pointer should not be freed or used by the programmer after
4413 giving it to sv_usepvn, and neither should any pointers from "behind"
4414 that pointer (e.g. ptr + 1) be used.
4416 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4417 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4418 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4419 C<len>, and already meets the requirements for storing in C<SvPVX>)
4425 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4430 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4432 SV_CHECK_THINKFIRST_COW_DROP(sv);
4433 SvUPGRADE(sv, SVt_PV);
4436 if (flags & SV_SMAGIC)
4440 if (SvPVX_const(sv))
4444 if (flags & SV_HAS_TRAILING_NUL)
4445 assert(ptr[len] == '\0');
4448 allocate = (flags & SV_HAS_TRAILING_NUL)
4450 #ifdef Perl_safesysmalloc_size
4453 PERL_STRLEN_ROUNDUP(len + 1);
4455 if (flags & SV_HAS_TRAILING_NUL) {
4456 /* It's long enough - do nothing.
4457 Specfically Perl_newCONSTSUB is relying on this. */
4460 /* Force a move to shake out bugs in callers. */
4461 char *new_ptr = (char*)safemalloc(allocate);
4462 Copy(ptr, new_ptr, len, char);
4463 PoisonFree(ptr,len,char);
4467 ptr = (char*) saferealloc (ptr, allocate);
4470 #ifdef Perl_safesysmalloc_size
4471 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4473 SvLEN_set(sv, allocate);
4477 if (!(flags & SV_HAS_TRAILING_NUL)) {
4480 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4482 if (flags & SV_SMAGIC)
4486 #ifdef PERL_OLD_COPY_ON_WRITE
4487 /* Need to do this *after* making the SV normal, as we need the buffer
4488 pointer to remain valid until after we've copied it. If we let go too early,
4489 another thread could invalidate it by unsharing last of the same hash key
4490 (which it can do by means other than releasing copy-on-write Svs)
4491 or by changing the other copy-on-write SVs in the loop. */
4493 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4495 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4497 { /* this SV was SvIsCOW_normal(sv) */
4498 /* we need to find the SV pointing to us. */
4499 SV *current = SV_COW_NEXT_SV(after);
4501 if (current == sv) {
4502 /* The SV we point to points back to us (there were only two of us
4504 Hence other SV is no longer copy on write either. */
4506 SvREADONLY_off(after);
4508 /* We need to follow the pointers around the loop. */
4510 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4513 /* don't loop forever if the structure is bust, and we have
4514 a pointer into a closed loop. */
4515 assert (current != after);
4516 assert (SvPVX_const(current) == pvx);
4518 /* Make the SV before us point to the SV after us. */
4519 SV_COW_NEXT_SV_SET(current, after);
4525 =for apidoc sv_force_normal_flags
4527 Undo various types of fakery on an SV: if the PV is a shared string, make
4528 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4529 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4530 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4531 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4532 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4533 set to some other value.) In addition, the C<flags> parameter gets passed to
4534 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4535 with flags set to 0.
4541 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4545 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4547 #ifdef PERL_OLD_COPY_ON_WRITE
4548 if (SvREADONLY(sv)) {
4550 const char * const pvx = SvPVX_const(sv);
4551 const STRLEN len = SvLEN(sv);
4552 const STRLEN cur = SvCUR(sv);
4553 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4554 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4555 we'll fail an assertion. */
4556 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4559 PerlIO_printf(Perl_debug_log,
4560 "Copy on write: Force normal %ld\n",
4566 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4569 if (flags & SV_COW_DROP_PV) {
4570 /* OK, so we don't need to copy our buffer. */
4573 SvGROW(sv, cur + 1);
4574 Move(pvx,SvPVX(sv),cur,char);
4579 sv_release_COW(sv, pvx, next);
4581 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4587 else if (IN_PERL_RUNTIME)
4588 Perl_croak(aTHX_ "%s", PL_no_modify);
4591 if (SvREADONLY(sv)) {
4593 const char * const pvx = SvPVX_const(sv);
4594 const STRLEN len = SvCUR(sv);
4599 SvGROW(sv, len + 1);
4600 Move(pvx,SvPVX(sv),len,char);
4602 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4604 else if (IN_PERL_RUNTIME)
4605 Perl_croak(aTHX_ "%s", PL_no_modify);
4609 sv_unref_flags(sv, flags);
4610 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4617 Efficient removal of characters from the beginning of the string buffer.
4618 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4619 the string buffer. The C<ptr> becomes the first character of the adjusted
4620 string. Uses the "OOK hack".
4621 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4622 refer to the same chunk of data.
4628 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4634 const U8 *real_start;
4638 PERL_ARGS_ASSERT_SV_CHOP;
4640 if (!ptr || !SvPOKp(sv))
4642 delta = ptr - SvPVX_const(sv);
4644 /* Nothing to do. */
4647 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4648 nothing uses the value of ptr any more. */
4649 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4650 if (ptr <= SvPVX_const(sv))
4651 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4652 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4653 SV_CHECK_THINKFIRST(sv);
4654 if (delta > max_delta)
4655 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4656 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4657 SvPVX_const(sv) + max_delta);
4660 if (!SvLEN(sv)) { /* make copy of shared string */
4661 const char *pvx = SvPVX_const(sv);
4662 const STRLEN len = SvCUR(sv);
4663 SvGROW(sv, len + 1);
4664 Move(pvx,SvPVX(sv),len,char);
4667 SvFLAGS(sv) |= SVf_OOK;
4670 SvOOK_offset(sv, old_delta);
4672 SvLEN_set(sv, SvLEN(sv) - delta);
4673 SvCUR_set(sv, SvCUR(sv) - delta);
4674 SvPV_set(sv, SvPVX(sv) + delta);
4676 p = (U8 *)SvPVX_const(sv);
4681 real_start = p - delta;
4685 if (delta < 0x100) {
4689 p -= sizeof(STRLEN);
4690 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4694 /* Fill the preceding buffer with sentinals to verify that no-one is
4696 while (p > real_start) {
4704 =for apidoc sv_catpvn
4706 Concatenates the string onto the end of the string which is in the SV. The
4707 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4708 status set, then the bytes appended should be valid UTF-8.
4709 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4711 =for apidoc sv_catpvn_flags
4713 Concatenates the string onto the end of the string which is in the SV. The
4714 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4715 status set, then the bytes appended should be valid UTF-8.
4716 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4717 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4718 in terms of this function.
4724 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4728 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4730 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4732 SvGROW(dsv, dlen + slen + 1);
4734 sstr = SvPVX_const(dsv);
4735 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4736 SvCUR_set(dsv, SvCUR(dsv) + slen);
4738 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4740 if (flags & SV_SMAGIC)
4745 =for apidoc sv_catsv
4747 Concatenates the string from SV C<ssv> onto the end of the string in
4748 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4749 not 'set' magic. See C<sv_catsv_mg>.
4751 =for apidoc sv_catsv_flags
4753 Concatenates the string from SV C<ssv> onto the end of the string in
4754 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4755 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4756 and C<sv_catsv_nomg> are implemented in terms of this function.
4761 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4765 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4769 const char *spv = SvPV_const(ssv, slen);
4771 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4772 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4773 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4774 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4775 dsv->sv_flags doesn't have that bit set.
4776 Andy Dougherty 12 Oct 2001
4778 const I32 sutf8 = DO_UTF8(ssv);
4781 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4783 dutf8 = DO_UTF8(dsv);
4785 if (dutf8 != sutf8) {
4787 /* Not modifying source SV, so taking a temporary copy. */
4788 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4790 sv_utf8_upgrade(csv);
4791 spv = SvPV_const(csv, slen);
4794 /* Leave enough space for the cat that's about to happen */
4795 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4797 sv_catpvn_nomg(dsv, spv, slen);
4800 if (flags & SV_SMAGIC)
4805 =for apidoc sv_catpv
4807 Concatenates the string onto the end of the string which is in the SV.
4808 If the SV has the UTF-8 status set, then the bytes appended should be
4809 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4814 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4817 register STRLEN len;
4821 PERL_ARGS_ASSERT_SV_CATPV;
4825 junk = SvPV_force(sv, tlen);
4827 SvGROW(sv, tlen + len + 1);
4829 ptr = SvPVX_const(sv);
4830 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4831 SvCUR_set(sv, SvCUR(sv) + len);
4832 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4837 =for apidoc sv_catpv_mg
4839 Like C<sv_catpv>, but also handles 'set' magic.
4845 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4847 PERL_ARGS_ASSERT_SV_CATPV_MG;
4856 Creates a new SV. A non-zero C<len> parameter indicates the number of
4857 bytes of preallocated string space the SV should have. An extra byte for a
4858 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4859 space is allocated.) The reference count for the new SV is set to 1.
4861 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4862 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4863 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4864 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4865 modules supporting older perls.
4871 Perl_newSV(pTHX_ const STRLEN len)
4878 sv_upgrade(sv, SVt_PV);
4879 SvGROW(sv, len + 1);
4884 =for apidoc sv_magicext
4886 Adds magic to an SV, upgrading it if necessary. Applies the
4887 supplied vtable and returns a pointer to the magic added.
4889 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4890 In particular, you can add magic to SvREADONLY SVs, and add more than
4891 one instance of the same 'how'.
4893 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4894 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4895 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4896 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4898 (This is now used as a subroutine by C<sv_magic>.)
4903 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4904 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4909 PERL_ARGS_ASSERT_SV_MAGICEXT;
4911 SvUPGRADE(sv, SVt_PVMG);
4912 Newxz(mg, 1, MAGIC);
4913 mg->mg_moremagic = SvMAGIC(sv);
4914 SvMAGIC_set(sv, mg);
4916 /* Sometimes a magic contains a reference loop, where the sv and
4917 object refer to each other. To prevent a reference loop that
4918 would prevent such objects being freed, we look for such loops
4919 and if we find one we avoid incrementing the object refcount.
4921 Note we cannot do this to avoid self-tie loops as intervening RV must
4922 have its REFCNT incremented to keep it in existence.
4925 if (!obj || obj == sv ||
4926 how == PERL_MAGIC_arylen ||
4927 how == PERL_MAGIC_symtab ||
4928 (SvTYPE(obj) == SVt_PVGV &&
4929 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4930 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4931 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4936 mg->mg_obj = SvREFCNT_inc_simple(obj);
4937 mg->mg_flags |= MGf_REFCOUNTED;
4940 /* Normal self-ties simply pass a null object, and instead of
4941 using mg_obj directly, use the SvTIED_obj macro to produce a
4942 new RV as needed. For glob "self-ties", we are tieing the PVIO
4943 with an RV obj pointing to the glob containing the PVIO. In
4944 this case, to avoid a reference loop, we need to weaken the
4948 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4949 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4955 mg->mg_len = namlen;
4958 mg->mg_ptr = savepvn(name, namlen);
4959 else if (namlen == HEf_SVKEY) {
4960 /* Yes, this is casting away const. This is only for the case of
4961 HEf_SVKEY. I think we need to document this abberation of the
4962 constness of the API, rather than making name non-const, as
4963 that change propagating outwards a long way. */
4964 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4966 mg->mg_ptr = (char *) name;
4968 mg->mg_virtual = (MGVTBL *) vtable;
4972 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4977 =for apidoc sv_magic
4979 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4980 then adds a new magic item of type C<how> to the head of the magic list.
4982 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4983 handling of the C<name> and C<namlen> arguments.
4985 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4986 to add more than one instance of the same 'how'.
4992 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4993 const char *const name, const I32 namlen)
4996 const MGVTBL *vtable;
4999 PERL_ARGS_ASSERT_SV_MAGIC;
5001 #ifdef PERL_OLD_COPY_ON_WRITE
5003 sv_force_normal_flags(sv, 0);
5005 if (SvREADONLY(sv)) {
5007 /* its okay to attach magic to shared strings; the subsequent
5008 * upgrade to PVMG will unshare the string */
5009 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5012 && how != PERL_MAGIC_regex_global
5013 && how != PERL_MAGIC_bm
5014 && how != PERL_MAGIC_fm
5015 && how != PERL_MAGIC_sv
5016 && how != PERL_MAGIC_backref
5019 Perl_croak(aTHX_ "%s", PL_no_modify);
5022 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5023 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5024 /* sv_magic() refuses to add a magic of the same 'how' as an
5027 if (how == PERL_MAGIC_taint) {
5029 /* Any scalar which already had taint magic on which someone
5030 (erroneously?) did SvIOK_on() or similar will now be
5031 incorrectly sporting public "OK" flags. */
5032 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5040 vtable = &PL_vtbl_sv;
5042 case PERL_MAGIC_overload:
5043 vtable = &PL_vtbl_amagic;
5045 case PERL_MAGIC_overload_elem:
5046 vtable = &PL_vtbl_amagicelem;
5048 case PERL_MAGIC_overload_table:
5049 vtable = &PL_vtbl_ovrld;
5052 vtable = &PL_vtbl_bm;
5054 case PERL_MAGIC_regdata:
5055 vtable = &PL_vtbl_regdata;
5057 case PERL_MAGIC_regdatum:
5058 vtable = &PL_vtbl_regdatum;
5060 case PERL_MAGIC_env:
5061 vtable = &PL_vtbl_env;
5064 vtable = &PL_vtbl_fm;
5066 case PERL_MAGIC_envelem:
5067 vtable = &PL_vtbl_envelem;
5069 case PERL_MAGIC_regex_global:
5070 vtable = &PL_vtbl_mglob;
5072 case PERL_MAGIC_isa:
5073 vtable = &PL_vtbl_isa;
5075 case PERL_MAGIC_isaelem:
5076 vtable = &PL_vtbl_isaelem;
5078 case PERL_MAGIC_nkeys:
5079 vtable = &PL_vtbl_nkeys;
5081 case PERL_MAGIC_dbfile:
5084 case PERL_MAGIC_dbline:
5085 vtable = &PL_vtbl_dbline;
5087 #ifdef USE_LOCALE_COLLATE
5088 case PERL_MAGIC_collxfrm:
5089 vtable = &PL_vtbl_collxfrm;
5091 #endif /* USE_LOCALE_COLLATE */
5092 case PERL_MAGIC_tied:
5093 vtable = &PL_vtbl_pack;
5095 case PERL_MAGIC_tiedelem:
5096 case PERL_MAGIC_tiedscalar:
5097 vtable = &PL_vtbl_packelem;
5100 vtable = &PL_vtbl_regexp;
5102 case PERL_MAGIC_sig:
5103 vtable = &PL_vtbl_sig;
5105 case PERL_MAGIC_sigelem:
5106 vtable = &PL_vtbl_sigelem;
5108 case PERL_MAGIC_taint:
5109 vtable = &PL_vtbl_taint;
5111 case PERL_MAGIC_uvar:
5112 vtable = &PL_vtbl_uvar;
5114 case PERL_MAGIC_vec:
5115 vtable = &PL_vtbl_vec;
5117 case PERL_MAGIC_arylen_p:
5118 case PERL_MAGIC_rhash:
5119 case PERL_MAGIC_symtab:
5120 case PERL_MAGIC_vstring:
5123 case PERL_MAGIC_utf8:
5124 vtable = &PL_vtbl_utf8;
5126 case PERL_MAGIC_substr:
5127 vtable = &PL_vtbl_substr;
5129 case PERL_MAGIC_defelem:
5130 vtable = &PL_vtbl_defelem;
5132 case PERL_MAGIC_arylen:
5133 vtable = &PL_vtbl_arylen;
5135 case PERL_MAGIC_pos:
5136 vtable = &PL_vtbl_pos;
5138 case PERL_MAGIC_backref:
5139 vtable = &PL_vtbl_backref;
5141 case PERL_MAGIC_hintselem:
5142 vtable = &PL_vtbl_hintselem;
5144 case PERL_MAGIC_hints:
5145 vtable = &PL_vtbl_hints;
5147 case PERL_MAGIC_ext:
5148 /* Reserved for use by extensions not perl internals. */
5149 /* Useful for attaching extension internal data to perl vars. */
5150 /* Note that multiple extensions may clash if magical scalars */
5151 /* etc holding private data from one are passed to another. */
5155 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5158 /* Rest of work is done else where */
5159 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5162 case PERL_MAGIC_taint:
5165 case PERL_MAGIC_ext:
5166 case PERL_MAGIC_dbfile:
5173 =for apidoc sv_unmagic
5175 Removes all magic of type C<type> from an SV.
5181 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5186 PERL_ARGS_ASSERT_SV_UNMAGIC;
5188 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5190 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5191 for (mg = *mgp; mg; mg = *mgp) {
5192 if (mg->mg_type == type) {
5193 const MGVTBL* const vtbl = mg->mg_virtual;
5194 *mgp = mg->mg_moremagic;
5195 if (vtbl && vtbl->svt_free)
5196 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5197 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5199 Safefree(mg->mg_ptr);
5200 else if (mg->mg_len == HEf_SVKEY)
5201 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5202 else if (mg->mg_type == PERL_MAGIC_utf8)
5203 Safefree(mg->mg_ptr);
5205 if (mg->mg_flags & MGf_REFCOUNTED)
5206 SvREFCNT_dec(mg->mg_obj);
5210 mgp = &mg->mg_moremagic;
5214 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5215 SvMAGIC_set(sv, NULL);
5222 =for apidoc sv_rvweaken
5224 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5225 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5226 push a back-reference to this RV onto the array of backreferences
5227 associated with that magic. If the RV is magical, set magic will be
5228 called after the RV is cleared.
5234 Perl_sv_rvweaken(pTHX_ SV *const sv)
5238 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5240 if (!SvOK(sv)) /* let undefs pass */
5243 Perl_croak(aTHX_ "Can't weaken a nonreference");
5244 else if (SvWEAKREF(sv)) {
5245 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5249 Perl_sv_add_backref(aTHX_ tsv, sv);
5255 /* Give tsv backref magic if it hasn't already got it, then push a
5256 * back-reference to sv onto the array associated with the backref magic.
5259 /* A discussion about the backreferences array and its refcount:
5261 * The AV holding the backreferences is pointed to either as the mg_obj of
5262 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5263 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5264 * have the standard magic instead.) The array is created with a refcount
5265 * of 2. This means that if during global destruction the array gets
5266 * picked on first to have its refcount decremented by the random zapper,
5267 * it won't actually be freed, meaning it's still theere for when its
5268 * parent gets freed.
5269 * When the parent SV is freed, in the case of magic, the magic is freed,
5270 * Perl_magic_killbackrefs is called which decrements one refcount, then
5271 * mg_obj is freed which kills the second count.
5272 * In the vase of a HV being freed, one ref is removed by
5273 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5278 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5283 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5285 if (SvTYPE(tsv) == SVt_PVHV) {
5286 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5290 /* There is no AV in the offical place - try a fixup. */
5291 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5294 /* Aha. They've got it stowed in magic. Bring it back. */
5295 av = MUTABLE_AV(mg->mg_obj);
5296 /* Stop mg_free decreasing the refernce count. */
5298 /* Stop mg_free even calling the destructor, given that
5299 there's no AV to free up. */
5301 sv_unmagic(tsv, PERL_MAGIC_backref);
5305 SvREFCNT_inc_simple_void(av); /* see discussion above */
5310 const MAGIC *const mg
5311 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5313 av = MUTABLE_AV(mg->mg_obj);
5317 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5318 /* av now has a refcnt of 2; see discussion above */
5321 if (AvFILLp(av) >= AvMAX(av)) {
5322 av_extend(av, AvFILLp(av)+1);
5324 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5327 /* delete a back-reference to ourselves from the backref magic associated
5328 * with the SV we point to.
5332 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5339 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5341 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5342 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5343 /* We mustn't attempt to "fix up" the hash here by moving the
5344 backreference array back to the hv_aux structure, as that is stored
5345 in the main HvARRAY(), and hfreentries assumes that no-one
5346 reallocates HvARRAY() while it is running. */
5349 const MAGIC *const mg
5350 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5352 av = MUTABLE_AV(mg->mg_obj);
5356 Perl_croak(aTHX_ "panic: del_backref");
5358 assert(!SvIS_FREED(av));
5361 /* We shouldn't be in here more than once, but for paranoia reasons lets
5363 for (i = AvFILLp(av); i >= 0; i--) {
5365 const SSize_t fill = AvFILLp(av);
5367 /* We weren't the last entry.
5368 An unordered list has this property that you can take the
5369 last element off the end to fill the hole, and it's still
5370 an unordered list :-)
5375 AvFILLp(av) = fill - 1;
5381 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5383 SV **svp = AvARRAY(av);
5385 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5386 PERL_UNUSED_ARG(sv);
5388 assert(!svp || !SvIS_FREED(av));
5390 SV *const *const last = svp + AvFILLp(av);
5392 while (svp <= last) {
5394 SV *const referrer = *svp;
5395 if (SvWEAKREF(referrer)) {
5396 /* XXX Should we check that it hasn't changed? */
5397 SvRV_set(referrer, 0);
5399 SvWEAKREF_off(referrer);
5400 SvSETMAGIC(referrer);
5401 } else if (SvTYPE(referrer) == SVt_PVGV ||
5402 SvTYPE(referrer) == SVt_PVLV) {
5403 /* You lookin' at me? */
5404 assert(GvSTASH(referrer));
5405 assert(GvSTASH(referrer) == (const HV *)sv);
5406 GvSTASH(referrer) = 0;
5409 "panic: magic_killbackrefs (flags=%"UVxf")",
5410 (UV)SvFLAGS(referrer));
5418 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5423 =for apidoc sv_insert
5425 Inserts a string at the specified offset/length within the SV. Similar to
5426 the Perl substr() function. Handles get magic.
5428 =for apidoc sv_insert_flags
5430 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5436 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5441 register char *midend;
5442 register char *bigend;
5446 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5449 Perl_croak(aTHX_ "Can't modify non-existent substring");
5450 SvPV_force_flags(bigstr, curlen, flags);
5451 (void)SvPOK_only_UTF8(bigstr);
5452 if (offset + len > curlen) {
5453 SvGROW(bigstr, offset+len+1);
5454 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5455 SvCUR_set(bigstr, offset+len);
5459 i = littlelen - len;
5460 if (i > 0) { /* string might grow */
5461 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5462 mid = big + offset + len;
5463 midend = bigend = big + SvCUR(bigstr);
5466 while (midend > mid) /* shove everything down */
5467 *--bigend = *--midend;
5468 Move(little,big+offset,littlelen,char);
5469 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5474 Move(little,SvPVX(bigstr)+offset,len,char);
5479 big = SvPVX(bigstr);
5482 bigend = big + SvCUR(bigstr);
5484 if (midend > bigend)
5485 Perl_croak(aTHX_ "panic: sv_insert");
5487 if (mid - big > bigend - midend) { /* faster to shorten from end */
5489 Move(little, mid, littlelen,char);
5492 i = bigend - midend;
5494 Move(midend, mid, i,char);
5498 SvCUR_set(bigstr, mid - big);
5500 else if ((i = mid - big)) { /* faster from front */
5501 midend -= littlelen;
5503 Move(big, midend - i, i, char);
5504 sv_chop(bigstr,midend-i);
5506 Move(little, mid, littlelen,char);
5508 else if (littlelen) {
5509 midend -= littlelen;
5510 sv_chop(bigstr,midend);
5511 Move(little,midend,littlelen,char);
5514 sv_chop(bigstr,midend);
5520 =for apidoc sv_replace
5522 Make the first argument a copy of the second, then delete the original.
5523 The target SV physically takes over ownership of the body of the source SV
5524 and inherits its flags; however, the target keeps any magic it owns,
5525 and any magic in the source is discarded.
5526 Note that this is a rather specialist SV copying operation; most of the
5527 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5533 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5536 const U32 refcnt = SvREFCNT(sv);
5538 PERL_ARGS_ASSERT_SV_REPLACE;
5540 SV_CHECK_THINKFIRST_COW_DROP(sv);
5541 if (SvREFCNT(nsv) != 1) {
5542 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5543 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5545 if (SvMAGICAL(sv)) {
5549 sv_upgrade(nsv, SVt_PVMG);
5550 SvMAGIC_set(nsv, SvMAGIC(sv));
5551 SvFLAGS(nsv) |= SvMAGICAL(sv);
5553 SvMAGIC_set(sv, NULL);
5557 assert(!SvREFCNT(sv));
5558 #ifdef DEBUG_LEAKING_SCALARS
5559 sv->sv_flags = nsv->sv_flags;
5560 sv->sv_any = nsv->sv_any;
5561 sv->sv_refcnt = nsv->sv_refcnt;
5562 sv->sv_u = nsv->sv_u;
5564 StructCopy(nsv,sv,SV);
5566 if(SvTYPE(sv) == SVt_IV) {
5568 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5572 #ifdef PERL_OLD_COPY_ON_WRITE
5573 if (SvIsCOW_normal(nsv)) {
5574 /* We need to follow the pointers around the loop to make the
5575 previous SV point to sv, rather than nsv. */
5578 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5581 assert(SvPVX_const(current) == SvPVX_const(nsv));
5583 /* Make the SV before us point to the SV after us. */
5585 PerlIO_printf(Perl_debug_log, "previous is\n");
5587 PerlIO_printf(Perl_debug_log,
5588 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5589 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5591 SV_COW_NEXT_SV_SET(current, sv);
5594 SvREFCNT(sv) = refcnt;
5595 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5601 =for apidoc sv_clear
5603 Clear an SV: call any destructors, free up any memory used by the body,
5604 and free the body itself. The SV's head is I<not> freed, although
5605 its type is set to all 1's so that it won't inadvertently be assumed
5606 to be live during global destruction etc.
5607 This function should only be called when REFCNT is zero. Most of the time
5608 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5615 Perl_sv_clear(pTHX_ register SV *const sv)
5618 const U32 type = SvTYPE(sv);
5619 const struct body_details *const sv_type_details
5620 = bodies_by_type + type;
5623 PERL_ARGS_ASSERT_SV_CLEAR;
5624 assert(SvREFCNT(sv) == 0);
5625 assert(SvTYPE(sv) != SVTYPEMASK);
5627 if (type <= SVt_IV) {
5628 /* See the comment in sv.h about the collusion between this early
5629 return and the overloading of the NULL and IV slots in the size
5632 SV * const target = SvRV(sv);
5634 sv_del_backref(target, sv);
5636 SvREFCNT_dec(target);
5638 SvFLAGS(sv) &= SVf_BREAK;
5639 SvFLAGS(sv) |= SVTYPEMASK;
5644 if (PL_defstash && /* Still have a symbol table? */
5651 stash = SvSTASH(sv);
5652 destructor = StashHANDLER(stash,DESTROY);
5654 /* A constant subroutine can have no side effects, so
5655 don't bother calling it. */
5656 && !CvCONST(destructor)
5657 /* Don't bother calling an empty destructor */
5658 && (CvISXSUB(destructor)
5659 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5661 SV* const tmpref = newRV(sv);
5662 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5664 PUSHSTACKi(PERLSI_DESTROY);
5669 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5675 if(SvREFCNT(tmpref) < 2) {
5676 /* tmpref is not kept alive! */
5678 SvRV_set(tmpref, NULL);
5681 SvREFCNT_dec(tmpref);
5683 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5687 if (PL_in_clean_objs)
5688 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5690 /* DESTROY gave object new lease on life */
5696 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5697 SvOBJECT_off(sv); /* Curse the object. */
5698 if (type != SVt_PVIO)
5699 --PL_sv_objcount; /* XXX Might want something more general */
5702 if (type >= SVt_PVMG) {
5703 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5704 SvREFCNT_dec(SvOURSTASH(sv));
5705 } else if (SvMAGIC(sv))
5707 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5708 SvREFCNT_dec(SvSTASH(sv));
5711 /* case SVt_BIND: */
5714 IoIFP(sv) != PerlIO_stdin() &&
5715 IoIFP(sv) != PerlIO_stdout() &&
5716 IoIFP(sv) != PerlIO_stderr())
5718 io_close(MUTABLE_IO(sv), FALSE);
5720 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5721 PerlDir_close(IoDIRP(sv));
5722 IoDIRP(sv) = (DIR*)NULL;
5723 Safefree(IoTOP_NAME(sv));
5724 Safefree(IoFMT_NAME(sv));
5725 Safefree(IoBOTTOM_NAME(sv));
5728 /* FIXME for plugins */
5729 pregfree2((REGEXP*) sv);
5733 cv_undef(MUTABLE_CV(sv));
5736 if (PL_last_swash_hv == (const HV *)sv) {
5737 PL_last_swash_hv = NULL;
5739 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5740 hv_undef(MUTABLE_HV(sv));
5743 if (PL_comppad == MUTABLE_AV(sv)) {
5747 av_undef(MUTABLE_AV(sv));
5750 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5751 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5752 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5753 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5755 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5756 SvREFCNT_dec(LvTARG(sv));
5758 if (isGV_with_GP(sv)) {
5759 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5760 && HvNAME_get(stash))
5761 mro_method_changed_in(stash);
5762 gp_free(MUTABLE_GV(sv));
5764 unshare_hek(GvNAME_HEK(sv));
5765 /* If we're in a stash, we don't own a reference to it. However it does
5766 have a back reference to us, which needs to be cleared. */
5767 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5768 sv_del_backref(MUTABLE_SV(stash), sv);
5770 /* FIXME. There are probably more unreferenced pointers to SVs in the
5771 interpreter struct that we should check and tidy in a similar
5773 if ((const GV *)sv == PL_last_in_gv)
5774 PL_last_in_gv = NULL;
5780 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5783 SvOOK_offset(sv, offset);
5784 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5785 /* Don't even bother with turning off the OOK flag. */
5788 SV * const target = SvRV(sv);
5790 sv_del_backref(target, sv);
5792 SvREFCNT_dec(target);
5794 #ifdef PERL_OLD_COPY_ON_WRITE
5795 else if (SvPVX_const(sv)) {
5798 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5802 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5804 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5808 } else if (SvLEN(sv)) {
5809 Safefree(SvPVX_const(sv));
5813 else if (SvPVX_const(sv) && SvLEN(sv))
5814 Safefree(SvPVX_mutable(sv));
5815 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5816 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5825 SvFLAGS(sv) &= SVf_BREAK;
5826 SvFLAGS(sv) |= SVTYPEMASK;
5828 if (sv_type_details->arena) {
5829 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5830 &PL_body_roots[type]);
5832 else if (sv_type_details->body_size) {
5833 my_safefree(SvANY(sv));
5838 =for apidoc sv_newref
5840 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5847 Perl_sv_newref(pTHX_ SV *const sv)
5849 PERL_UNUSED_CONTEXT;
5858 Decrement an SV's reference count, and if it drops to zero, call
5859 C<sv_clear> to invoke destructors and free up any memory used by
5860 the body; finally, deallocate the SV's head itself.
5861 Normally called via a wrapper macro C<SvREFCNT_dec>.
5867 Perl_sv_free(pTHX_ SV *const sv)
5872 if (SvREFCNT(sv) == 0) {
5873 if (SvFLAGS(sv) & SVf_BREAK)
5874 /* this SV's refcnt has been artificially decremented to
5875 * trigger cleanup */
5877 if (PL_in_clean_all) /* All is fair */
5879 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5880 /* make sure SvREFCNT(sv)==0 happens very seldom */
5881 SvREFCNT(sv) = (~(U32)0)/2;
5884 if (ckWARN_d(WARN_INTERNAL)) {
5885 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5886 Perl_dump_sv_child(aTHX_ sv);
5888 #ifdef DEBUG_LEAKING_SCALARS
5891 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5892 if (PL_warnhook == PERL_WARNHOOK_FATAL
5893 || ckDEAD(packWARN(WARN_INTERNAL))) {
5894 /* Don't let Perl_warner cause us to escape our fate: */
5898 /* This may not return: */
5899 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5900 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5901 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5904 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5909 if (--(SvREFCNT(sv)) > 0)
5911 Perl_sv_free2(aTHX_ sv);
5915 Perl_sv_free2(pTHX_ SV *const sv)
5919 PERL_ARGS_ASSERT_SV_FREE2;
5923 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5924 "Attempt to free temp prematurely: SV 0x%"UVxf
5925 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5929 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5930 /* make sure SvREFCNT(sv)==0 happens very seldom */
5931 SvREFCNT(sv) = (~(U32)0)/2;
5942 Returns the length of the string in the SV. Handles magic and type
5943 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5949 Perl_sv_len(pTHX_ register SV *const sv)
5957 len = mg_length(sv);
5959 (void)SvPV_const(sv, len);
5964 =for apidoc sv_len_utf8
5966 Returns the number of characters in the string in an SV, counting wide
5967 UTF-8 bytes as a single character. Handles magic and type coercion.
5973 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5974 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5975 * (Note that the mg_len is not the length of the mg_ptr field.
5976 * This allows the cache to store the character length of the string without
5977 * needing to malloc() extra storage to attach to the mg_ptr.)
5982 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5988 return mg_length(sv);
5992 const U8 *s = (U8*)SvPV_const(sv, len);
5996 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5998 if (mg && mg->mg_len != -1) {
6000 if (PL_utf8cache < 0) {
6001 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6003 /* Need to turn the assertions off otherwise we may
6004 recurse infinitely while printing error messages.
6006 SAVEI8(PL_utf8cache);
6008 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6009 " real %"UVuf" for %"SVf,
6010 (UV) ulen, (UV) real, SVfARG(sv));
6015 ulen = Perl_utf8_length(aTHX_ s, s + len);
6016 if (!SvREADONLY(sv)) {
6018 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6019 &PL_vtbl_utf8, 0, 0);
6027 return Perl_utf8_length(aTHX_ s, s + len);
6031 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6034 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6037 const U8 *s = start;
6039 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6041 while (s < send && uoffset--)
6044 /* This is the existing behaviour. Possibly it should be a croak, as
6045 it's actually a bounds error */
6051 /* Given the length of the string in both bytes and UTF-8 characters, decide
6052 whether to walk forwards or backwards to find the byte corresponding to
6053 the passed in UTF-8 offset. */
6055 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6056 const STRLEN uoffset, const STRLEN uend)
6058 STRLEN backw = uend - uoffset;
6060 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6062 if (uoffset < 2 * backw) {
6063 /* The assumption is that going forwards is twice the speed of going
6064 forward (that's where the 2 * backw comes from).
6065 (The real figure of course depends on the UTF-8 data.) */
6066 return sv_pos_u2b_forwards(start, send, uoffset);
6071 while (UTF8_IS_CONTINUATION(*send))
6074 return send - start;
6077 /* For the string representation of the given scalar, find the byte
6078 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6079 give another position in the string, *before* the sought offset, which
6080 (which is always true, as 0, 0 is a valid pair of positions), which should
6081 help reduce the amount of linear searching.
6082 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6083 will be used to reduce the amount of linear searching. The cache will be
6084 created if necessary, and the found value offered to it for update. */
6086 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6087 const U8 *const send, const STRLEN uoffset,
6088 STRLEN uoffset0, STRLEN boffset0)
6090 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6093 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6095 assert (uoffset >= uoffset0);
6097 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6098 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6099 if ((*mgp)->mg_ptr) {
6100 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6101 if (cache[0] == uoffset) {
6102 /* An exact match. */
6105 if (cache[2] == uoffset) {
6106 /* An exact match. */
6110 if (cache[0] < uoffset) {
6111 /* The cache already knows part of the way. */
6112 if (cache[0] > uoffset0) {
6113 /* The cache knows more than the passed in pair */
6114 uoffset0 = cache[0];
6115 boffset0 = cache[1];
6117 if ((*mgp)->mg_len != -1) {
6118 /* And we know the end too. */
6120 + sv_pos_u2b_midway(start + boffset0, send,
6122 (*mgp)->mg_len - uoffset0);
6125 + sv_pos_u2b_forwards(start + boffset0,
6126 send, uoffset - uoffset0);
6129 else if (cache[2] < uoffset) {
6130 /* We're between the two cache entries. */
6131 if (cache[2] > uoffset0) {
6132 /* and the cache knows more than the passed in pair */
6133 uoffset0 = cache[2];
6134 boffset0 = cache[3];
6138 + sv_pos_u2b_midway(start + boffset0,
6141 cache[0] - uoffset0);
6144 + sv_pos_u2b_midway(start + boffset0,
6147 cache[2] - uoffset0);
6151 else if ((*mgp)->mg_len != -1) {
6152 /* If we can take advantage of a passed in offset, do so. */
6153 /* In fact, offset0 is either 0, or less than offset, so don't
6154 need to worry about the other possibility. */
6156 + sv_pos_u2b_midway(start + boffset0, send,
6158 (*mgp)->mg_len - uoffset0);
6163 if (!found || PL_utf8cache < 0) {
6164 const STRLEN real_boffset
6165 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6166 send, uoffset - uoffset0);
6168 if (found && PL_utf8cache < 0) {
6169 if (real_boffset != boffset) {
6170 /* Need to turn the assertions off otherwise we may recurse
6171 infinitely while printing error messages. */
6172 SAVEI8(PL_utf8cache);
6174 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6175 " real %"UVuf" for %"SVf,
6176 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6179 boffset = real_boffset;
6183 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6189 =for apidoc sv_pos_u2b
6191 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6192 the start of the string, to a count of the equivalent number of bytes; if
6193 lenp is non-zero, it does the same to lenp, but this time starting from
6194 the offset, rather than from the start of the string. Handles magic and
6201 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6202 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6203 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6208 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6213 PERL_ARGS_ASSERT_SV_POS_U2B;
6218 start = (U8*)SvPV_const(sv, len);
6220 STRLEN uoffset = (STRLEN) *offsetp;
6221 const U8 * const send = start + len;
6223 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6226 *offsetp = (I32) boffset;
6229 /* Convert the relative offset to absolute. */
6230 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6231 const STRLEN boffset2
6232 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6233 uoffset, boffset) - boffset;
6247 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6248 byte length pairing. The (byte) length of the total SV is passed in too,
6249 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6250 may not have updated SvCUR, so we can't rely on reading it directly.
6252 The proffered utf8/byte length pairing isn't used if the cache already has
6253 two pairs, and swapping either for the proffered pair would increase the
6254 RMS of the intervals between known byte offsets.
6256 The cache itself consists of 4 STRLEN values
6257 0: larger UTF-8 offset
6258 1: corresponding byte offset
6259 2: smaller UTF-8 offset
6260 3: corresponding byte offset
6262 Unused cache pairs have the value 0, 0.
6263 Keeping the cache "backwards" means that the invariant of
6264 cache[0] >= cache[2] is maintained even with empty slots, which means that
6265 the code that uses it doesn't need to worry if only 1 entry has actually
6266 been set to non-zero. It also makes the "position beyond the end of the
6267 cache" logic much simpler, as the first slot is always the one to start
6271 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6272 const STRLEN utf8, const STRLEN blen)
6276 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6282 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6284 (*mgp)->mg_len = -1;
6288 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6289 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6290 (*mgp)->mg_ptr = (char *) cache;
6294 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6295 /* SvPOKp() because it's possible that sv has string overloading, and
6296 therefore is a reference, hence SvPVX() is actually a pointer.
6297 This cures the (very real) symptoms of RT 69422, but I'm not actually
6298 sure whether we should even be caching the results of UTF-8
6299 operations on overloading, given that nothing stops overloading
6300 returning a different value every time it's called. */
6301 const U8 *start = (const U8 *) SvPVX_const(sv);
6302 const STRLEN realutf8 = utf8_length(start, start + byte);
6304 if (realutf8 != utf8) {
6305 /* Need to turn the assertions off otherwise we may recurse
6306 infinitely while printing error messages. */
6307 SAVEI8(PL_utf8cache);
6309 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6310 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6314 /* Cache is held with the later position first, to simplify the code
6315 that deals with unbounded ends. */
6317 ASSERT_UTF8_CACHE(cache);
6318 if (cache[1] == 0) {
6319 /* Cache is totally empty */
6322 } else if (cache[3] == 0) {
6323 if (byte > cache[1]) {
6324 /* New one is larger, so goes first. */
6325 cache[2] = cache[0];
6326 cache[3] = cache[1];
6334 #define THREEWAY_SQUARE(a,b,c,d) \
6335 ((float)((d) - (c))) * ((float)((d) - (c))) \
6336 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6337 + ((float)((b) - (a))) * ((float)((b) - (a)))
6339 /* Cache has 2 slots in use, and we know three potential pairs.
6340 Keep the two that give the lowest RMS distance. Do the
6341 calcualation in bytes simply because we always know the byte
6342 length. squareroot has the same ordering as the positive value,
6343 so don't bother with the actual square root. */
6344 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6345 if (byte > cache[1]) {
6346 /* New position is after the existing pair of pairs. */
6347 const float keep_earlier
6348 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6349 const float keep_later
6350 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6352 if (keep_later < keep_earlier) {
6353 if (keep_later < existing) {
6354 cache[2] = cache[0];
6355 cache[3] = cache[1];
6361 if (keep_earlier < existing) {
6367 else if (byte > cache[3]) {
6368 /* New position is between the existing pair of pairs. */
6369 const float keep_earlier
6370 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6371 const float keep_later
6372 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6374 if (keep_later < keep_earlier) {
6375 if (keep_later < existing) {
6381 if (keep_earlier < existing) {
6388 /* New position is before the existing pair of pairs. */
6389 const float keep_earlier
6390 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6391 const float keep_later
6392 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6394 if (keep_later < keep_earlier) {
6395 if (keep_later < existing) {
6401 if (keep_earlier < existing) {
6402 cache[0] = cache[2];
6403 cache[1] = cache[3];
6410 ASSERT_UTF8_CACHE(cache);
6413 /* We already know all of the way, now we may be able to walk back. The same
6414 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6415 backward is half the speed of walking forward. */
6417 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6418 const U8 *end, STRLEN endu)
6420 const STRLEN forw = target - s;
6421 STRLEN backw = end - target;
6423 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6425 if (forw < 2 * backw) {
6426 return utf8_length(s, target);
6429 while (end > target) {
6431 while (UTF8_IS_CONTINUATION(*end)) {
6440 =for apidoc sv_pos_b2u
6442 Converts the value pointed to by offsetp from a count of bytes from the
6443 start of the string, to a count of the equivalent number of UTF-8 chars.
6444 Handles magic and type coercion.
6450 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6451 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6456 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6459 const STRLEN byte = *offsetp;
6460 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6466 PERL_ARGS_ASSERT_SV_POS_B2U;
6471 s = (const U8*)SvPV_const(sv, blen);
6474 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6478 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6479 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6481 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6482 if (cache[1] == byte) {
6483 /* An exact match. */
6484 *offsetp = cache[0];
6487 if (cache[3] == byte) {
6488 /* An exact match. */
6489 *offsetp = cache[2];
6493 if (cache[1] < byte) {
6494 /* We already know part of the way. */
6495 if (mg->mg_len != -1) {
6496 /* Actually, we know the end too. */
6498 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6499 s + blen, mg->mg_len - cache[0]);
6501 len = cache[0] + utf8_length(s + cache[1], send);
6504 else if (cache[3] < byte) {
6505 /* We're between the two cached pairs, so we do the calculation
6506 offset by the byte/utf-8 positions for the earlier pair,
6507 then add the utf-8 characters from the string start to
6509 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6510 s + cache[1], cache[0] - cache[2])
6514 else { /* cache[3] > byte */
6515 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6519 ASSERT_UTF8_CACHE(cache);
6521 } else if (mg->mg_len != -1) {
6522 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6526 if (!found || PL_utf8cache < 0) {
6527 const STRLEN real_len = utf8_length(s, send);
6529 if (found && PL_utf8cache < 0) {
6530 if (len != real_len) {
6531 /* Need to turn the assertions off otherwise we may recurse
6532 infinitely while printing error messages. */
6533 SAVEI8(PL_utf8cache);
6535 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6536 " real %"UVuf" for %"SVf,
6537 (UV) len, (UV) real_len, SVfARG(sv));
6545 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6551 Returns a boolean indicating whether the strings in the two SVs are
6552 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6553 coerce its args to strings if necessary.
6559 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6568 SV* svrecode = NULL;
6575 /* if pv1 and pv2 are the same, second SvPV_const call may
6576 * invalidate pv1, so we may need to make a copy */
6577 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6578 pv1 = SvPV_const(sv1, cur1);
6579 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6581 pv1 = SvPV_const(sv1, cur1);
6589 pv2 = SvPV_const(sv2, cur2);
6591 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6592 /* Differing utf8ness.
6593 * Do not UTF8size the comparands as a side-effect. */
6596 svrecode = newSVpvn(pv2, cur2);
6597 sv_recode_to_utf8(svrecode, PL_encoding);
6598 pv2 = SvPV_const(svrecode, cur2);
6601 svrecode = newSVpvn(pv1, cur1);
6602 sv_recode_to_utf8(svrecode, PL_encoding);
6603 pv1 = SvPV_const(svrecode, cur1);
6605 /* Now both are in UTF-8. */
6607 SvREFCNT_dec(svrecode);
6612 bool is_utf8 = TRUE;
6615 /* sv1 is the UTF-8 one,
6616 * if is equal it must be downgrade-able */
6617 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6623 /* sv2 is the UTF-8 one,
6624 * if is equal it must be downgrade-able */
6625 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6631 /* Downgrade not possible - cannot be eq */
6639 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6641 SvREFCNT_dec(svrecode);
6651 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6652 string in C<sv1> is less than, equal to, or greater than the string in
6653 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6654 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6660 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6664 const char *pv1, *pv2;
6667 SV *svrecode = NULL;
6674 pv1 = SvPV_const(sv1, cur1);
6681 pv2 = SvPV_const(sv2, cur2);
6683 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6684 /* Differing utf8ness.
6685 * Do not UTF8size the comparands as a side-effect. */
6688 svrecode = newSVpvn(pv2, cur2);
6689 sv_recode_to_utf8(svrecode, PL_encoding);
6690 pv2 = SvPV_const(svrecode, cur2);
6693 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6698 svrecode = newSVpvn(pv1, cur1);
6699 sv_recode_to_utf8(svrecode, PL_encoding);
6700 pv1 = SvPV_const(svrecode, cur1);
6703 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6709 cmp = cur2 ? -1 : 0;
6713 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6716 cmp = retval < 0 ? -1 : 1;
6717 } else if (cur1 == cur2) {
6720 cmp = cur1 < cur2 ? -1 : 1;
6724 SvREFCNT_dec(svrecode);
6732 =for apidoc sv_cmp_locale
6734 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6735 'use bytes' aware, handles get magic, and will coerce its args to strings
6736 if necessary. See also C<sv_cmp>.
6742 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6745 #ifdef USE_LOCALE_COLLATE
6751 if (PL_collation_standard)
6755 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6757 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6759 if (!pv1 || !len1) {
6770 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6773 return retval < 0 ? -1 : 1;
6776 * When the result of collation is equality, that doesn't mean
6777 * that there are no differences -- some locales exclude some
6778 * characters from consideration. So to avoid false equalities,
6779 * we use the raw string as a tiebreaker.
6785 #endif /* USE_LOCALE_COLLATE */
6787 return sv_cmp(sv1, sv2);
6791 #ifdef USE_LOCALE_COLLATE
6794 =for apidoc sv_collxfrm
6796 Add Collate Transform magic to an SV if it doesn't already have it.
6798 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6799 scalar data of the variable, but transformed to such a format that a normal
6800 memory comparison can be used to compare the data according to the locale
6807 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6812 PERL_ARGS_ASSERT_SV_COLLXFRM;
6814 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6815 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6821 Safefree(mg->mg_ptr);
6822 s = SvPV_const(sv, len);
6823 if ((xf = mem_collxfrm(s, len, &xlen))) {
6825 #ifdef PERL_OLD_COPY_ON_WRITE
6827 sv_force_normal_flags(sv, 0);
6829 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6843 if (mg && mg->mg_ptr) {
6845 return mg->mg_ptr + sizeof(PL_collation_ix);
6853 #endif /* USE_LOCALE_COLLATE */
6858 Get a line from the filehandle and store it into the SV, optionally
6859 appending to the currently-stored string.
6865 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6870 register STDCHAR rslast;
6871 register STDCHAR *bp;
6876 PERL_ARGS_ASSERT_SV_GETS;
6878 if (SvTHINKFIRST(sv))
6879 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6880 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6882 However, perlbench says it's slower, because the existing swipe code
6883 is faster than copy on write.
6884 Swings and roundabouts. */
6885 SvUPGRADE(sv, SVt_PV);
6890 if (PerlIO_isutf8(fp)) {
6892 sv_utf8_upgrade_nomg(sv);
6893 sv_pos_u2b(sv,&append,0);
6895 } else if (SvUTF8(sv)) {
6896 SV * const tsv = newSV(0);
6897 sv_gets(tsv, fp, 0);
6898 sv_utf8_upgrade_nomg(tsv);
6899 SvCUR_set(sv,append);
6902 goto return_string_or_null;
6907 if (PerlIO_isutf8(fp))
6910 if (IN_PERL_COMPILETIME) {
6911 /* we always read code in line mode */
6915 else if (RsSNARF(PL_rs)) {
6916 /* If it is a regular disk file use size from stat() as estimate
6917 of amount we are going to read -- may result in mallocing
6918 more memory than we really need if the layers below reduce
6919 the size we read (e.g. CRLF or a gzip layer).
6922 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6923 const Off_t offset = PerlIO_tell(fp);
6924 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6925 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6931 else if (RsRECORD(PL_rs)) {
6939 /* Grab the size of the record we're getting */
6940 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6941 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6944 /* VMS wants read instead of fread, because fread doesn't respect */
6945 /* RMS record boundaries. This is not necessarily a good thing to be */
6946 /* doing, but we've got no other real choice - except avoid stdio
6947 as implementation - perhaps write a :vms layer ?
6949 fd = PerlIO_fileno(fp);
6950 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6951 bytesread = PerlIO_read(fp, buffer, recsize);
6954 bytesread = PerlLIO_read(fd, buffer, recsize);
6957 bytesread = PerlIO_read(fp, buffer, recsize);
6961 SvCUR_set(sv, bytesread + append);
6962 buffer[bytesread] = '\0';
6963 goto return_string_or_null;
6965 else if (RsPARA(PL_rs)) {
6971 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6972 if (PerlIO_isutf8(fp)) {
6973 rsptr = SvPVutf8(PL_rs, rslen);
6976 if (SvUTF8(PL_rs)) {
6977 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6978 Perl_croak(aTHX_ "Wide character in $/");
6981 rsptr = SvPV_const(PL_rs, rslen);
6985 rslast = rslen ? rsptr[rslen - 1] : '\0';
6987 if (rspara) { /* have to do this both before and after */
6988 do { /* to make sure file boundaries work right */
6991 i = PerlIO_getc(fp);
6995 PerlIO_ungetc(fp,i);
7001 /* See if we know enough about I/O mechanism to cheat it ! */
7003 /* This used to be #ifdef test - it is made run-time test for ease
7004 of abstracting out stdio interface. One call should be cheap
7005 enough here - and may even be a macro allowing compile
7009 if (PerlIO_fast_gets(fp)) {
7012 * We're going to steal some values from the stdio struct
7013 * and put EVERYTHING in the innermost loop into registers.
7015 register STDCHAR *ptr;
7019 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7020 /* An ungetc()d char is handled separately from the regular
7021 * buffer, so we getc() it back out and stuff it in the buffer.
7023 i = PerlIO_getc(fp);
7024 if (i == EOF) return 0;
7025 *(--((*fp)->_ptr)) = (unsigned char) i;
7029 /* Here is some breathtakingly efficient cheating */
7031 cnt = PerlIO_get_cnt(fp); /* get count into register */
7032 /* make sure we have the room */
7033 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7034 /* Not room for all of it
7035 if we are looking for a separator and room for some
7037 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7038 /* just process what we have room for */
7039 shortbuffered = cnt - SvLEN(sv) + append + 1;
7040 cnt -= shortbuffered;
7044 /* remember that cnt can be negative */
7045 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7050 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7051 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7052 DEBUG_P(PerlIO_printf(Perl_debug_log,
7053 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7054 DEBUG_P(PerlIO_printf(Perl_debug_log,
7055 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7056 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7057 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7062 while (cnt > 0) { /* this | eat */
7064 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7065 goto thats_all_folks; /* screams | sed :-) */
7069 Copy(ptr, bp, cnt, char); /* this | eat */
7070 bp += cnt; /* screams | dust */
7071 ptr += cnt; /* louder | sed :-) */
7076 if (shortbuffered) { /* oh well, must extend */
7077 cnt = shortbuffered;
7079 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7081 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7082 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7086 DEBUG_P(PerlIO_printf(Perl_debug_log,
7087 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7088 PTR2UV(ptr),(long)cnt));
7089 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7091 DEBUG_P(PerlIO_printf(Perl_debug_log,
7092 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7093 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7094 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7096 /* This used to call 'filbuf' in stdio form, but as that behaves like
7097 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7098 another abstraction. */
7099 i = PerlIO_getc(fp); /* get more characters */
7101 DEBUG_P(PerlIO_printf(Perl_debug_log,
7102 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7103 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7104 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7106 cnt = PerlIO_get_cnt(fp);
7107 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7108 DEBUG_P(PerlIO_printf(Perl_debug_log,
7109 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7111 if (i == EOF) /* all done for ever? */
7112 goto thats_really_all_folks;
7114 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7116 SvGROW(sv, bpx + cnt + 2);
7117 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7119 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7121 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7122 goto thats_all_folks;
7126 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7127 memNE((char*)bp - rslen, rsptr, rslen))
7128 goto screamer; /* go back to the fray */
7129 thats_really_all_folks:
7131 cnt += shortbuffered;
7132 DEBUG_P(PerlIO_printf(Perl_debug_log,
7133 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7134 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7135 DEBUG_P(PerlIO_printf(Perl_debug_log,
7136 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7137 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7138 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7140 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7141 DEBUG_P(PerlIO_printf(Perl_debug_log,
7142 "Screamer: done, len=%ld, string=|%.*s|\n",
7143 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7147 /*The big, slow, and stupid way. */
7148 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7149 STDCHAR *buf = NULL;
7150 Newx(buf, 8192, STDCHAR);
7158 register const STDCHAR * const bpe = buf + sizeof(buf);
7160 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7161 ; /* keep reading */
7165 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7166 /* Accomodate broken VAXC compiler, which applies U8 cast to
7167 * both args of ?: operator, causing EOF to change into 255
7170 i = (U8)buf[cnt - 1];
7176 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7178 sv_catpvn(sv, (char *) buf, cnt);
7180 sv_setpvn(sv, (char *) buf, cnt);
7182 if (i != EOF && /* joy */
7184 SvCUR(sv) < rslen ||
7185 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7189 * If we're reading from a TTY and we get a short read,
7190 * indicating that the user hit his EOF character, we need
7191 * to notice it now, because if we try to read from the TTY
7192 * again, the EOF condition will disappear.
7194 * The comparison of cnt to sizeof(buf) is an optimization
7195 * that prevents unnecessary calls to feof().
7199 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7203 #ifdef USE_HEAP_INSTEAD_OF_STACK
7208 if (rspara) { /* have to do this both before and after */
7209 while (i != EOF) { /* to make sure file boundaries work right */
7210 i = PerlIO_getc(fp);
7212 PerlIO_ungetc(fp,i);
7218 return_string_or_null:
7219 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7225 Auto-increment of the value in the SV, doing string to numeric conversion
7226 if necessary. Handles 'get' magic.
7232 Perl_sv_inc(pTHX_ register SV *const sv)
7241 if (SvTHINKFIRST(sv)) {
7243 sv_force_normal_flags(sv, 0);
7244 if (SvREADONLY(sv)) {
7245 if (IN_PERL_RUNTIME)
7246 Perl_croak(aTHX_ "%s", PL_no_modify);
7250 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7252 i = PTR2IV(SvRV(sv));
7257 flags = SvFLAGS(sv);
7258 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7259 /* It's (privately or publicly) a float, but not tested as an
7260 integer, so test it to see. */
7262 flags = SvFLAGS(sv);
7264 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7265 /* It's publicly an integer, or privately an integer-not-float */
7266 #ifdef PERL_PRESERVE_IVUV
7270 if (SvUVX(sv) == UV_MAX)
7271 sv_setnv(sv, UV_MAX_P1);
7273 (void)SvIOK_only_UV(sv);
7274 SvUV_set(sv, SvUVX(sv) + 1);
7276 if (SvIVX(sv) == IV_MAX)
7277 sv_setuv(sv, (UV)IV_MAX + 1);
7279 (void)SvIOK_only(sv);
7280 SvIV_set(sv, SvIVX(sv) + 1);
7285 if (flags & SVp_NOK) {
7286 const NV was = SvNVX(sv);
7287 if (NV_OVERFLOWS_INTEGERS_AT &&
7288 was >= NV_OVERFLOWS_INTEGERS_AT) {
7289 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7290 "Lost precision when incrementing %" NVff " by 1",
7293 (void)SvNOK_only(sv);
7294 SvNV_set(sv, was + 1.0);
7298 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7299 if ((flags & SVTYPEMASK) < SVt_PVIV)
7300 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7301 (void)SvIOK_only(sv);
7306 while (isALPHA(*d)) d++;
7307 while (isDIGIT(*d)) d++;
7308 if (d < SvEND(sv)) {
7309 #ifdef PERL_PRESERVE_IVUV
7310 /* Got to punt this as an integer if needs be, but we don't issue
7311 warnings. Probably ought to make the sv_iv_please() that does
7312 the conversion if possible, and silently. */
7313 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7314 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7315 /* Need to try really hard to see if it's an integer.
7316 9.22337203685478e+18 is an integer.
7317 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7318 so $a="9.22337203685478e+18"; $a+0; $a++
7319 needs to be the same as $a="9.22337203685478e+18"; $a++
7326 /* sv_2iv *should* have made this an NV */
7327 if (flags & SVp_NOK) {
7328 (void)SvNOK_only(sv);
7329 SvNV_set(sv, SvNVX(sv) + 1.0);
7332 /* I don't think we can get here. Maybe I should assert this
7333 And if we do get here I suspect that sv_setnv will croak. NWC
7335 #if defined(USE_LONG_DOUBLE)
7336 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",
7337 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7339 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7340 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7343 #endif /* PERL_PRESERVE_IVUV */
7344 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7348 while (d >= SvPVX_const(sv)) {
7356 /* MKS: The original code here died if letters weren't consecutive.
7357 * at least it didn't have to worry about non-C locales. The
7358 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7359 * arranged in order (although not consecutively) and that only
7360 * [A-Za-z] are accepted by isALPHA in the C locale.
7362 if (*d != 'z' && *d != 'Z') {
7363 do { ++*d; } while (!isALPHA(*d));
7366 *(d--) -= 'z' - 'a';
7371 *(d--) -= 'z' - 'a' + 1;
7375 /* oh,oh, the number grew */
7376 SvGROW(sv, SvCUR(sv) + 2);
7377 SvCUR_set(sv, SvCUR(sv) + 1);
7378 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7389 Auto-decrement of the value in the SV, doing string to numeric conversion
7390 if necessary. Handles 'get' magic.
7396 Perl_sv_dec(pTHX_ register SV *const sv)
7404 if (SvTHINKFIRST(sv)) {
7406 sv_force_normal_flags(sv, 0);
7407 if (SvREADONLY(sv)) {
7408 if (IN_PERL_RUNTIME)
7409 Perl_croak(aTHX_ "%s", PL_no_modify);
7413 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7415 i = PTR2IV(SvRV(sv));
7420 /* Unlike sv_inc we don't have to worry about string-never-numbers
7421 and keeping them magic. But we mustn't warn on punting */
7422 flags = SvFLAGS(sv);
7423 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7424 /* It's publicly an integer, or privately an integer-not-float */
7425 #ifdef PERL_PRESERVE_IVUV
7429 if (SvUVX(sv) == 0) {
7430 (void)SvIOK_only(sv);
7434 (void)SvIOK_only_UV(sv);
7435 SvUV_set(sv, SvUVX(sv) - 1);
7438 if (SvIVX(sv) == IV_MIN) {
7439 sv_setnv(sv, (NV)IV_MIN);
7443 (void)SvIOK_only(sv);
7444 SvIV_set(sv, SvIVX(sv) - 1);
7449 if (flags & SVp_NOK) {
7452 const NV was = SvNVX(sv);
7453 if (NV_OVERFLOWS_INTEGERS_AT &&
7454 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7455 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7456 "Lost precision when decrementing %" NVff " by 1",
7459 (void)SvNOK_only(sv);
7460 SvNV_set(sv, was - 1.0);
7464 if (!(flags & SVp_POK)) {
7465 if ((flags & SVTYPEMASK) < SVt_PVIV)
7466 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7468 (void)SvIOK_only(sv);
7471 #ifdef PERL_PRESERVE_IVUV
7473 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7474 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7475 /* Need to try really hard to see if it's an integer.
7476 9.22337203685478e+18 is an integer.
7477 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7478 so $a="9.22337203685478e+18"; $a+0; $a--
7479 needs to be the same as $a="9.22337203685478e+18"; $a--
7486 /* sv_2iv *should* have made this an NV */
7487 if (flags & SVp_NOK) {
7488 (void)SvNOK_only(sv);
7489 SvNV_set(sv, SvNVX(sv) - 1.0);
7492 /* I don't think we can get here. Maybe I should assert this
7493 And if we do get here I suspect that sv_setnv will croak. NWC
7495 #if defined(USE_LONG_DOUBLE)
7496 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",
7497 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7499 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7500 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7504 #endif /* PERL_PRESERVE_IVUV */
7505 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7508 /* this define is used to eliminate a chunk of duplicated but shared logic
7509 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7510 * used anywhere but here - yves
7512 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7515 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7519 =for apidoc sv_mortalcopy
7521 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7522 The new SV is marked as mortal. It will be destroyed "soon", either by an
7523 explicit call to FREETMPS, or by an implicit call at places such as
7524 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7529 /* Make a string that will exist for the duration of the expression
7530 * evaluation. Actually, it may have to last longer than that, but
7531 * hopefully we won't free it until it has been assigned to a
7532 * permanent location. */
7535 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7541 sv_setsv(sv,oldstr);
7542 PUSH_EXTEND_MORTAL__SV_C(sv);
7548 =for apidoc sv_newmortal
7550 Creates a new null SV which is mortal. The reference count of the SV is
7551 set to 1. It will be destroyed "soon", either by an explicit call to
7552 FREETMPS, or by an implicit call at places such as statement boundaries.
7553 See also C<sv_mortalcopy> and C<sv_2mortal>.
7559 Perl_sv_newmortal(pTHX)
7565 SvFLAGS(sv) = SVs_TEMP;
7566 PUSH_EXTEND_MORTAL__SV_C(sv);
7572 =for apidoc newSVpvn_flags
7574 Creates a new SV and copies a string into it. The reference count for the
7575 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7576 string. You are responsible for ensuring that the source string is at least
7577 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7578 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7579 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7580 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7581 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7583 #define newSVpvn_utf8(s, len, u) \
7584 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7590 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7595 /* All the flags we don't support must be zero.
7596 And we're new code so I'm going to assert this from the start. */
7597 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7599 sv_setpvn(sv,s,len);
7601 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7602 * and do what it does outselves here.
7603 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7604 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7605 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7606 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7609 SvFLAGS(sv) |= flags;
7611 if(flags & SVs_TEMP){
7612 PUSH_EXTEND_MORTAL__SV_C(sv);
7619 =for apidoc sv_2mortal
7621 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7622 by an explicit call to FREETMPS, or by an implicit call at places such as
7623 statement boundaries. SvTEMP() is turned on which means that the SV's
7624 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7625 and C<sv_mortalcopy>.
7631 Perl_sv_2mortal(pTHX_ register SV *const sv)
7636 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7638 PUSH_EXTEND_MORTAL__SV_C(sv);
7646 Creates a new SV and copies a string into it. The reference count for the
7647 SV is set to 1. If C<len> is zero, Perl will compute the length using
7648 strlen(). For efficiency, consider using C<newSVpvn> instead.
7654 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7660 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7665 =for apidoc newSVpvn
7667 Creates a new SV and copies a string into it. The reference count for the
7668 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7669 string. You are responsible for ensuring that the source string is at least
7670 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7676 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7682 sv_setpvn(sv,s,len);
7687 =for apidoc newSVhek
7689 Creates a new SV from the hash key structure. It will generate scalars that
7690 point to the shared string table where possible. Returns a new (undefined)
7691 SV if the hek is NULL.
7697 Perl_newSVhek(pTHX_ const HEK *const hek)
7707 if (HEK_LEN(hek) == HEf_SVKEY) {
7708 return newSVsv(*(SV**)HEK_KEY(hek));
7710 const int flags = HEK_FLAGS(hek);
7711 if (flags & HVhek_WASUTF8) {
7713 Andreas would like keys he put in as utf8 to come back as utf8
7715 STRLEN utf8_len = HEK_LEN(hek);
7716 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7717 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7720 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7722 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7723 /* We don't have a pointer to the hv, so we have to replicate the
7724 flag into every HEK. This hv is using custom a hasing
7725 algorithm. Hence we can't return a shared string scalar, as
7726 that would contain the (wrong) hash value, and might get passed
7727 into an hv routine with a regular hash.
7728 Similarly, a hash that isn't using shared hash keys has to have
7729 the flag in every key so that we know not to try to call
7730 share_hek_kek on it. */
7732 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7737 /* This will be overwhelminly the most common case. */
7739 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7740 more efficient than sharepvn(). */
7744 sv_upgrade(sv, SVt_PV);
7745 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7746 SvCUR_set(sv, HEK_LEN(hek));
7759 =for apidoc newSVpvn_share
7761 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7762 table. If the string does not already exist in the table, it is created
7763 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7764 value is used; otherwise the hash is computed. The string's hash can be later
7765 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7766 that as the string table is used for shared hash keys these strings will have
7767 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7773 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7777 bool is_utf8 = FALSE;
7778 const char *const orig_src = src;
7781 STRLEN tmplen = -len;
7783 /* See the note in hv.c:hv_fetch() --jhi */
7784 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7788 PERL_HASH(hash, src, len);
7790 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7791 changes here, update it there too. */
7792 sv_upgrade(sv, SVt_PV);
7793 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7801 if (src != orig_src)
7807 #if defined(PERL_IMPLICIT_CONTEXT)
7809 /* pTHX_ magic can't cope with varargs, so this is a no-context
7810 * version of the main function, (which may itself be aliased to us).
7811 * Don't access this version directly.
7815 Perl_newSVpvf_nocontext(const char *const pat, ...)
7821 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7823 va_start(args, pat);
7824 sv = vnewSVpvf(pat, &args);
7831 =for apidoc newSVpvf
7833 Creates a new SV and initializes it with the string formatted like
7840 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7845 PERL_ARGS_ASSERT_NEWSVPVF;
7847 va_start(args, pat);
7848 sv = vnewSVpvf(pat, &args);
7853 /* backend for newSVpvf() and newSVpvf_nocontext() */
7856 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7861 PERL_ARGS_ASSERT_VNEWSVPVF;
7864 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7871 Creates a new SV and copies a floating point value into it.
7872 The reference count for the SV is set to 1.
7878 Perl_newSVnv(pTHX_ const NV n)
7891 Creates a new SV and copies an integer into it. The reference count for the
7898 Perl_newSViv(pTHX_ const IV i)
7911 Creates a new SV and copies an unsigned integer into it.
7912 The reference count for the SV is set to 1.
7918 Perl_newSVuv(pTHX_ const UV u)
7929 =for apidoc newSV_type
7931 Creates a new SV, of the type specified. The reference count for the new SV
7938 Perl_newSV_type(pTHX_ const svtype type)
7943 sv_upgrade(sv, type);
7948 =for apidoc newRV_noinc
7950 Creates an RV wrapper for an SV. The reference count for the original
7951 SV is B<not> incremented.
7957 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7960 register SV *sv = newSV_type(SVt_IV);
7962 PERL_ARGS_ASSERT_NEWRV_NOINC;
7965 SvRV_set(sv, tmpRef);
7970 /* newRV_inc is the official function name to use now.
7971 * newRV_inc is in fact #defined to newRV in sv.h
7975 Perl_newRV(pTHX_ SV *const sv)
7979 PERL_ARGS_ASSERT_NEWRV;
7981 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7987 Creates a new SV which is an exact duplicate of the original SV.
7994 Perl_newSVsv(pTHX_ register SV *const old)
8001 if (SvTYPE(old) == SVTYPEMASK) {
8002 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8006 /* SV_GMAGIC is the default for sv_setv()
8007 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8008 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8009 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8014 =for apidoc sv_reset
8016 Underlying implementation for the C<reset> Perl function.
8017 Note that the perl-level function is vaguely deprecated.
8023 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8026 char todo[PERL_UCHAR_MAX+1];
8028 PERL_ARGS_ASSERT_SV_RESET;
8033 if (!*s) { /* reset ?? searches */
8034 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8036 const U32 count = mg->mg_len / sizeof(PMOP**);
8037 PMOP **pmp = (PMOP**) mg->mg_ptr;
8038 PMOP *const *const end = pmp + count;
8042 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8044 (*pmp)->op_pmflags &= ~PMf_USED;
8052 /* reset variables */
8054 if (!HvARRAY(stash))
8057 Zero(todo, 256, char);
8060 I32 i = (unsigned char)*s;
8064 max = (unsigned char)*s++;
8065 for ( ; i <= max; i++) {
8068 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8070 for (entry = HvARRAY(stash)[i];
8072 entry = HeNEXT(entry))
8077 if (!todo[(U8)*HeKEY(entry)])
8079 gv = MUTABLE_GV(HeVAL(entry));
8082 if (SvTHINKFIRST(sv)) {
8083 if (!SvREADONLY(sv) && SvROK(sv))
8085 /* XXX Is this continue a bug? Why should THINKFIRST
8086 exempt us from resetting arrays and hashes? */
8090 if (SvTYPE(sv) >= SVt_PV) {
8092 if (SvPVX_const(sv) != NULL)
8100 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8102 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8105 # if defined(USE_ENVIRON_ARRAY)
8108 # endif /* USE_ENVIRON_ARRAY */
8119 Using various gambits, try to get an IO from an SV: the IO slot if its a
8120 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8121 named after the PV if we're a string.
8127 Perl_sv_2io(pTHX_ SV *const sv)
8132 PERL_ARGS_ASSERT_SV_2IO;
8134 switch (SvTYPE(sv)) {
8136 io = MUTABLE_IO(sv);
8139 if (isGV_with_GP(sv)) {
8140 gv = MUTABLE_GV(sv);
8143 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8149 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8151 return sv_2io(SvRV(sv));
8152 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8158 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8167 Using various gambits, try to get a CV from an SV; in addition, try if
8168 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8169 The flags in C<lref> are passed to gv_fetchsv.
8175 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8181 PERL_ARGS_ASSERT_SV_2CV;
8188 switch (SvTYPE(sv)) {
8192 return MUTABLE_CV(sv);
8199 if (isGV_with_GP(sv)) {
8200 gv = MUTABLE_GV(sv);
8209 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8211 tryAMAGICunDEREF(to_cv);
8214 if (SvTYPE(sv) == SVt_PVCV) {
8215 cv = MUTABLE_CV(sv);
8220 else if(isGV_with_GP(sv))
8221 gv = MUTABLE_GV(sv);
8223 Perl_croak(aTHX_ "Not a subroutine reference");
8225 else if (isGV_with_GP(sv)) {
8227 gv = MUTABLE_GV(sv);
8230 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8236 /* Some flags to gv_fetchsv mean don't really create the GV */
8237 if (!isGV_with_GP(gv)) {
8243 if (lref && !GvCVu(gv)) {
8247 gv_efullname3(tmpsv, gv, NULL);
8248 /* XXX this is probably not what they think they're getting.
8249 * It has the same effect as "sub name;", i.e. just a forward
8251 newSUB(start_subparse(FALSE, 0),
8252 newSVOP(OP_CONST, 0, tmpsv),
8256 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8257 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8266 Returns true if the SV has a true value by Perl's rules.
8267 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8268 instead use an in-line version.
8274 Perl_sv_true(pTHX_ register SV *const sv)
8279 register const XPV* const tXpv = (XPV*)SvANY(sv);
8281 (tXpv->xpv_cur > 1 ||
8282 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8289 return SvIVX(sv) != 0;
8292 return SvNVX(sv) != 0.0;
8294 return sv_2bool(sv);
8300 =for apidoc sv_pvn_force
8302 Get a sensible string out of the SV somehow.
8303 A private implementation of the C<SvPV_force> macro for compilers which
8304 can't cope with complex macro expressions. Always use the macro instead.
8306 =for apidoc sv_pvn_force_flags
8308 Get a sensible string out of the SV somehow.
8309 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8310 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8311 implemented in terms of this function.
8312 You normally want to use the various wrapper macros instead: see
8313 C<SvPV_force> and C<SvPV_force_nomg>
8319 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8323 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8325 if (SvTHINKFIRST(sv) && !SvROK(sv))
8326 sv_force_normal_flags(sv, 0);
8336 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8337 const char * const ref = sv_reftype(sv,0);
8339 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8340 ref, OP_NAME(PL_op));
8342 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8344 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8345 || isGV_with_GP(sv))
8346 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8348 s = sv_2pv_flags(sv, &len, flags);
8352 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8355 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8356 SvGROW(sv, len + 1);
8357 Move(s,SvPVX(sv),len,char);
8359 SvPVX(sv)[len] = '\0';
8362 SvPOK_on(sv); /* validate pointer */
8364 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8365 PTR2UV(sv),SvPVX_const(sv)));
8368 return SvPVX_mutable(sv);
8372 =for apidoc sv_pvbyten_force
8374 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8380 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8382 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8384 sv_pvn_force(sv,lp);
8385 sv_utf8_downgrade(sv,0);
8391 =for apidoc sv_pvutf8n_force
8393 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8399 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8401 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8403 sv_pvn_force(sv,lp);
8404 sv_utf8_upgrade(sv);
8410 =for apidoc sv_reftype
8412 Returns a string describing what the SV is a reference to.
8418 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8420 PERL_ARGS_ASSERT_SV_REFTYPE;
8422 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8423 inside return suggests a const propagation bug in g++. */
8424 if (ob && SvOBJECT(sv)) {
8425 char * const name = HvNAME_get(SvSTASH(sv));
8426 return name ? name : (char *) "__ANON__";
8429 switch (SvTYPE(sv)) {
8444 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8445 /* tied lvalues should appear to be
8446 * scalars for backwards compatitbility */
8447 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8448 ? "SCALAR" : "LVALUE");
8449 case SVt_PVAV: return "ARRAY";
8450 case SVt_PVHV: return "HASH";
8451 case SVt_PVCV: return "CODE";
8452 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8453 ? "GLOB" : "SCALAR");
8454 case SVt_PVFM: return "FORMAT";
8455 case SVt_PVIO: return "IO";
8456 case SVt_BIND: return "BIND";
8457 case SVt_REGEXP: return "REGEXP";
8458 default: return "UNKNOWN";
8464 =for apidoc sv_isobject
8466 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8467 object. If the SV is not an RV, or if the object is not blessed, then this
8474 Perl_sv_isobject(pTHX_ SV *sv)
8490 Returns a boolean indicating whether the SV is blessed into the specified
8491 class. This does not check for subtypes; use C<sv_derived_from> to verify
8492 an inheritance relationship.
8498 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8502 PERL_ARGS_ASSERT_SV_ISA;
8512 hvname = HvNAME_get(SvSTASH(sv));
8516 return strEQ(hvname, name);
8522 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8523 it will be upgraded to one. If C<classname> is non-null then the new SV will
8524 be blessed in the specified package. The new SV is returned and its
8525 reference count is 1.
8531 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8536 PERL_ARGS_ASSERT_NEWSVRV;
8540 SV_CHECK_THINKFIRST_COW_DROP(rv);
8541 (void)SvAMAGIC_off(rv);
8543 if (SvTYPE(rv) >= SVt_PVMG) {
8544 const U32 refcnt = SvREFCNT(rv);
8548 SvREFCNT(rv) = refcnt;
8550 sv_upgrade(rv, SVt_IV);
8551 } else if (SvROK(rv)) {
8552 SvREFCNT_dec(SvRV(rv));
8554 prepare_SV_for_RV(rv);
8562 HV* const stash = gv_stashpv(classname, GV_ADD);
8563 (void)sv_bless(rv, stash);
8569 =for apidoc sv_setref_pv
8571 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8572 argument will be upgraded to an RV. That RV will be modified to point to
8573 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8574 into the SV. The C<classname> argument indicates the package for the
8575 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8576 will have a reference count of 1, and the RV will be returned.
8578 Do not use with other Perl types such as HV, AV, SV, CV, because those
8579 objects will become corrupted by the pointer copy process.
8581 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8587 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8591 PERL_ARGS_ASSERT_SV_SETREF_PV;
8594 sv_setsv(rv, &PL_sv_undef);
8598 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8603 =for apidoc sv_setref_iv
8605 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8606 argument will be upgraded to an RV. That RV will be modified to point to
8607 the new SV. The C<classname> argument indicates the package for the
8608 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8609 will have a reference count of 1, and the RV will be returned.
8615 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8617 PERL_ARGS_ASSERT_SV_SETREF_IV;
8619 sv_setiv(newSVrv(rv,classname), iv);
8624 =for apidoc sv_setref_uv
8626 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8627 argument will be upgraded to an RV. That RV will be modified to point to
8628 the new SV. The C<classname> argument indicates the package for the
8629 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8630 will have a reference count of 1, and the RV will be returned.
8636 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8638 PERL_ARGS_ASSERT_SV_SETREF_UV;
8640 sv_setuv(newSVrv(rv,classname), uv);
8645 =for apidoc sv_setref_nv
8647 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8648 argument will be upgraded to an RV. That RV will be modified to point to
8649 the new SV. The C<classname> argument indicates the package for the
8650 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8651 will have a reference count of 1, and the RV will be returned.
8657 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8659 PERL_ARGS_ASSERT_SV_SETREF_NV;
8661 sv_setnv(newSVrv(rv,classname), nv);
8666 =for apidoc sv_setref_pvn
8668 Copies a string into a new SV, optionally blessing the SV. The length of the
8669 string must be specified with C<n>. The C<rv> argument will be upgraded to
8670 an RV. That RV will be modified to point to the new SV. The C<classname>
8671 argument indicates the package for the blessing. Set C<classname> to
8672 C<NULL> to avoid the blessing. The new SV will have a reference count
8673 of 1, and the RV will be returned.
8675 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8681 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8682 const char *const pv, const STRLEN n)
8684 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8686 sv_setpvn(newSVrv(rv,classname), pv, n);
8691 =for apidoc sv_bless
8693 Blesses an SV into a specified package. The SV must be an RV. The package
8694 must be designated by its stash (see C<gv_stashpv()>). The reference count
8695 of the SV is unaffected.
8701 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8706 PERL_ARGS_ASSERT_SV_BLESS;
8709 Perl_croak(aTHX_ "Can't bless non-reference value");
8711 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8712 if (SvIsCOW(tmpRef))
8713 sv_force_normal_flags(tmpRef, 0);
8714 if (SvREADONLY(tmpRef))
8715 Perl_croak(aTHX_ "%s", PL_no_modify);
8716 if (SvOBJECT(tmpRef)) {
8717 if (SvTYPE(tmpRef) != SVt_PVIO)
8719 SvREFCNT_dec(SvSTASH(tmpRef));
8722 SvOBJECT_on(tmpRef);
8723 if (SvTYPE(tmpRef) != SVt_PVIO)
8725 SvUPGRADE(tmpRef, SVt_PVMG);
8726 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8731 (void)SvAMAGIC_off(sv);
8733 if(SvSMAGICAL(tmpRef))
8734 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8742 /* Downgrades a PVGV to a PVMG.
8746 S_sv_unglob(pTHX_ SV *const sv)
8751 SV * const temp = sv_newmortal();
8753 PERL_ARGS_ASSERT_SV_UNGLOB;
8755 assert(SvTYPE(sv) == SVt_PVGV);
8757 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8760 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8761 && HvNAME_get(stash))
8762 mro_method_changed_in(stash);
8763 gp_free(MUTABLE_GV(sv));
8766 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8770 if (GvNAME_HEK(sv)) {
8771 unshare_hek(GvNAME_HEK(sv));
8773 isGV_with_GP_off(sv);
8775 /* need to keep SvANY(sv) in the right arena */
8776 xpvmg = new_XPVMG();
8777 StructCopy(SvANY(sv), xpvmg, XPVMG);
8778 del_XPVGV(SvANY(sv));
8781 SvFLAGS(sv) &= ~SVTYPEMASK;
8782 SvFLAGS(sv) |= SVt_PVMG;
8784 /* Intentionally not calling any local SET magic, as this isn't so much a
8785 set operation as merely an internal storage change. */
8786 sv_setsv_flags(sv, temp, 0);
8790 =for apidoc sv_unref_flags
8792 Unsets the RV status of the SV, and decrements the reference count of
8793 whatever was being referenced by the RV. This can almost be thought of
8794 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8795 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8796 (otherwise the decrementing is conditional on the reference count being
8797 different from one or the reference being a readonly SV).
8804 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8806 SV* const target = SvRV(ref);
8808 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8810 if (SvWEAKREF(ref)) {
8811 sv_del_backref(target, ref);
8813 SvRV_set(ref, NULL);
8816 SvRV_set(ref, NULL);
8818 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8819 assigned to as BEGIN {$a = \"Foo"} will fail. */
8820 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8821 SvREFCNT_dec(target);
8822 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8823 sv_2mortal(target); /* Schedule for freeing later */
8827 =for apidoc sv_untaint
8829 Untaint an SV. Use C<SvTAINTED_off> instead.
8834 Perl_sv_untaint(pTHX_ SV *const sv)
8836 PERL_ARGS_ASSERT_SV_UNTAINT;
8838 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8839 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8846 =for apidoc sv_tainted
8848 Test an SV for taintedness. Use C<SvTAINTED> instead.
8853 Perl_sv_tainted(pTHX_ SV *const sv)
8855 PERL_ARGS_ASSERT_SV_TAINTED;
8857 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8858 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8859 if (mg && (mg->mg_len & 1) )
8866 =for apidoc sv_setpviv
8868 Copies an integer into the given SV, also updating its string value.
8869 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8875 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8877 char buf[TYPE_CHARS(UV)];
8879 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8881 PERL_ARGS_ASSERT_SV_SETPVIV;
8883 sv_setpvn(sv, ptr, ebuf - ptr);
8887 =for apidoc sv_setpviv_mg
8889 Like C<sv_setpviv>, but also handles 'set' magic.
8895 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8897 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8903 #if defined(PERL_IMPLICIT_CONTEXT)
8905 /* pTHX_ magic can't cope with varargs, so this is a no-context
8906 * version of the main function, (which may itself be aliased to us).
8907 * Don't access this version directly.
8911 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8916 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8918 va_start(args, pat);
8919 sv_vsetpvf(sv, pat, &args);
8923 /* pTHX_ magic can't cope with varargs, so this is a no-context
8924 * version of the main function, (which may itself be aliased to us).
8925 * Don't access this version directly.
8929 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8934 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8936 va_start(args, pat);
8937 sv_vsetpvf_mg(sv, pat, &args);
8943 =for apidoc sv_setpvf
8945 Works like C<sv_catpvf> but copies the text into the SV instead of
8946 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8952 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8956 PERL_ARGS_ASSERT_SV_SETPVF;
8958 va_start(args, pat);
8959 sv_vsetpvf(sv, pat, &args);
8964 =for apidoc sv_vsetpvf
8966 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8967 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8969 Usually used via its frontend C<sv_setpvf>.
8975 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8977 PERL_ARGS_ASSERT_SV_VSETPVF;
8979 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8983 =for apidoc sv_setpvf_mg
8985 Like C<sv_setpvf>, but also handles 'set' magic.
8991 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8995 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8997 va_start(args, pat);
8998 sv_vsetpvf_mg(sv, pat, &args);
9003 =for apidoc sv_vsetpvf_mg
9005 Like C<sv_vsetpvf>, but also handles 'set' magic.
9007 Usually used via its frontend C<sv_setpvf_mg>.
9013 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9015 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9017 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9021 #if defined(PERL_IMPLICIT_CONTEXT)
9023 /* pTHX_ magic can't cope with varargs, so this is a no-context
9024 * version of the main function, (which may itself be aliased to us).
9025 * Don't access this version directly.
9029 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9034 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9036 va_start(args, pat);
9037 sv_vcatpvf(sv, pat, &args);
9041 /* pTHX_ magic can't cope with varargs, so this is a no-context
9042 * version of the main function, (which may itself be aliased to us).
9043 * Don't access this version directly.
9047 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9052 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9054 va_start(args, pat);
9055 sv_vcatpvf_mg(sv, pat, &args);
9061 =for apidoc sv_catpvf
9063 Processes its arguments like C<sprintf> and appends the formatted
9064 output to an SV. If the appended data contains "wide" characters
9065 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9066 and characters >255 formatted with %c), the original SV might get
9067 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9068 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9069 valid UTF-8; if the original SV was bytes, the pattern should be too.
9074 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9078 PERL_ARGS_ASSERT_SV_CATPVF;
9080 va_start(args, pat);
9081 sv_vcatpvf(sv, pat, &args);
9086 =for apidoc sv_vcatpvf
9088 Processes its arguments like C<vsprintf> and appends the formatted output
9089 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9091 Usually used via its frontend C<sv_catpvf>.
9097 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9099 PERL_ARGS_ASSERT_SV_VCATPVF;
9101 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9105 =for apidoc sv_catpvf_mg
9107 Like C<sv_catpvf>, but also handles 'set' magic.
9113 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9117 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9119 va_start(args, pat);
9120 sv_vcatpvf_mg(sv, pat, &args);
9125 =for apidoc sv_vcatpvf_mg
9127 Like C<sv_vcatpvf>, but also handles 'set' magic.
9129 Usually used via its frontend C<sv_catpvf_mg>.
9135 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9137 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9139 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9144 =for apidoc sv_vsetpvfn
9146 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9149 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9155 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9156 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9158 PERL_ARGS_ASSERT_SV_VSETPVFN;
9161 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9165 S_expect_number(pTHX_ char **const pattern)
9170 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9172 switch (**pattern) {
9173 case '1': case '2': case '3':
9174 case '4': case '5': case '6':
9175 case '7': case '8': case '9':
9176 var = *(*pattern)++ - '0';
9177 while (isDIGIT(**pattern)) {
9178 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9180 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9188 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9190 const int neg = nv < 0;
9193 PERL_ARGS_ASSERT_F0CONVERT;
9201 if (uv & 1 && uv == nv)
9202 uv--; /* Round to even */
9204 const unsigned dig = uv % 10;
9217 =for apidoc sv_vcatpvfn
9219 Processes its arguments like C<vsprintf> and appends the formatted output
9220 to an SV. Uses an array of SVs if the C style variable argument list is
9221 missing (NULL). When running with taint checks enabled, indicates via
9222 C<maybe_tainted> if results are untrustworthy (often due to the use of
9225 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9231 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9232 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9233 vec_utf8 = DO_UTF8(vecsv);
9235 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9238 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9239 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9247 static const char nullstr[] = "(null)";
9249 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9250 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9252 /* Times 4: a decimal digit takes more than 3 binary digits.
9253 * NV_DIG: mantissa takes than many decimal digits.
9254 * Plus 32: Playing safe. */
9255 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9256 /* large enough for "%#.#f" --chip */
9257 /* what about long double NVs? --jhi */
9259 PERL_ARGS_ASSERT_SV_VCATPVFN;
9260 PERL_UNUSED_ARG(maybe_tainted);
9262 /* no matter what, this is a string now */
9263 (void)SvPV_force(sv, origlen);
9265 /* special-case "", "%s", and "%-p" (SVf - see below) */
9268 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9270 const char * const s = va_arg(*args, char*);
9271 sv_catpv(sv, s ? s : nullstr);
9273 else if (svix < svmax) {
9274 sv_catsv(sv, *svargs);
9278 if (args && patlen == 3 && pat[0] == '%' &&
9279 pat[1] == '-' && pat[2] == 'p') {
9280 argsv = MUTABLE_SV(va_arg(*args, void*));
9281 sv_catsv(sv, argsv);
9285 #ifndef USE_LONG_DOUBLE
9286 /* special-case "%.<number>[gf]" */
9287 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9288 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9289 unsigned digits = 0;
9293 while (*pp >= '0' && *pp <= '9')
9294 digits = 10 * digits + (*pp++ - '0');
9295 if (pp - pat == (int)patlen - 1) {
9303 /* Add check for digits != 0 because it seems that some
9304 gconverts are buggy in this case, and we don't yet have
9305 a Configure test for this. */
9306 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9307 /* 0, point, slack */
9308 Gconvert(nv, (int)digits, 0, ebuf);
9310 if (*ebuf) /* May return an empty string for digits==0 */
9313 } else if (!digits) {
9316 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9317 sv_catpvn(sv, p, l);
9323 #endif /* !USE_LONG_DOUBLE */
9325 if (!args && svix < svmax && DO_UTF8(*svargs))
9328 patend = (char*)pat + patlen;
9329 for (p = (char*)pat; p < patend; p = q) {
9332 bool vectorize = FALSE;
9333 bool vectorarg = FALSE;
9334 bool vec_utf8 = FALSE;
9340 bool has_precis = FALSE;
9342 const I32 osvix = svix;
9343 bool is_utf8 = FALSE; /* is this item utf8? */
9344 #ifdef HAS_LDBL_SPRINTF_BUG
9345 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9346 with sfio - Allen <allens@cpan.org> */
9347 bool fix_ldbl_sprintf_bug = FALSE;
9351 U8 utf8buf[UTF8_MAXBYTES+1];
9352 STRLEN esignlen = 0;
9354 const char *eptr = NULL;
9355 const char *fmtstart;
9358 const U8 *vecstr = NULL;
9365 /* we need a long double target in case HAS_LONG_DOUBLE but
9368 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9376 const char *dotstr = ".";
9377 STRLEN dotstrlen = 1;
9378 I32 efix = 0; /* explicit format parameter index */
9379 I32 ewix = 0; /* explicit width index */
9380 I32 epix = 0; /* explicit precision index */
9381 I32 evix = 0; /* explicit vector index */
9382 bool asterisk = FALSE;
9384 /* echo everything up to the next format specification */
9385 for (q = p; q < patend && *q != '%'; ++q) ;
9387 if (has_utf8 && !pat_utf8)
9388 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9390 sv_catpvn(sv, p, q - p);
9399 We allow format specification elements in this order:
9400 \d+\$ explicit format parameter index
9402 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9403 0 flag (as above): repeated to allow "v02"
9404 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9405 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9407 [%bcdefginopsuxDFOUX] format (mandatory)
9412 As of perl5.9.3, printf format checking is on by default.
9413 Internally, perl uses %p formats to provide an escape to
9414 some extended formatting. This block deals with those
9415 extensions: if it does not match, (char*)q is reset and
9416 the normal format processing code is used.
9418 Currently defined extensions are:
9419 %p include pointer address (standard)
9420 %-p (SVf) include an SV (previously %_)
9421 %-<num>p include an SV with precision <num>
9422 %<num>p reserved for future extensions
9424 Robin Barker 2005-07-14
9426 %1p (VDf) removed. RMB 2007-10-19
9433 n = expect_number(&q);
9440 argsv = MUTABLE_SV(va_arg(*args, void*));
9441 eptr = SvPV_const(argsv, elen);
9447 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9448 "internal %%<num>p might conflict with future printf extensions");
9454 if ( (width = expect_number(&q)) ) {
9469 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9498 if ( (ewix = expect_number(&q)) )
9507 if ((vectorarg = asterisk)) {
9520 width = expect_number(&q);
9526 vecsv = va_arg(*args, SV*);
9528 vecsv = (evix > 0 && evix <= svmax)
9529 ? svargs[evix-1] : &PL_sv_undef;
9531 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9533 dotstr = SvPV_const(vecsv, dotstrlen);
9534 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9535 bad with tied or overloaded values that return UTF8. */
9538 else if (has_utf8) {
9539 vecsv = sv_mortalcopy(vecsv);
9540 sv_utf8_upgrade(vecsv);
9541 dotstr = SvPV_const(vecsv, dotstrlen);
9548 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9549 vecsv = svargs[efix ? efix-1 : svix++];
9550 vecstr = (U8*)SvPV_const(vecsv,veclen);
9551 vec_utf8 = DO_UTF8(vecsv);
9553 /* if this is a version object, we need to convert
9554 * back into v-string notation and then let the
9555 * vectorize happen normally
9557 if (sv_derived_from(vecsv, "version")) {
9558 char *version = savesvpv(vecsv);
9559 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9560 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9561 "vector argument not supported with alpha versions");
9564 vecsv = sv_newmortal();
9565 scan_vstring(version, version + veclen, vecsv);
9566 vecstr = (U8*)SvPV_const(vecsv, veclen);
9567 vec_utf8 = DO_UTF8(vecsv);
9579 i = va_arg(*args, int);
9581 i = (ewix ? ewix <= svmax : svix < svmax) ?
9582 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9584 width = (i < 0) ? -i : i;
9594 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9596 /* XXX: todo, support specified precision parameter */
9600 i = va_arg(*args, int);
9602 i = (ewix ? ewix <= svmax : svix < svmax)
9603 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9605 has_precis = !(i < 0);
9610 precis = precis * 10 + (*q++ - '0');
9619 case 'I': /* Ix, I32x, and I64x */
9621 if (q[1] == '6' && q[2] == '4') {
9627 if (q[1] == '3' && q[2] == '2') {
9637 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9648 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9649 if (*(q + 1) == 'l') { /* lld, llf */
9675 if (!vectorize && !args) {
9677 const I32 i = efix-1;
9678 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9680 argsv = (svix >= 0 && svix < svmax)
9681 ? svargs[svix++] : &PL_sv_undef;
9692 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9694 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9696 eptr = (char*)utf8buf;
9697 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9711 eptr = va_arg(*args, char*);
9713 elen = strlen(eptr);
9715 eptr = (char *)nullstr;
9716 elen = sizeof nullstr - 1;
9720 eptr = SvPV_const(argsv, elen);
9721 if (DO_UTF8(argsv)) {
9722 STRLEN old_precis = precis;
9723 if (has_precis && precis < elen) {
9724 STRLEN ulen = sv_len_utf8(argsv);
9725 I32 p = precis > ulen ? ulen : precis;
9726 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9729 if (width) { /* fudge width (can't fudge elen) */
9730 if (has_precis && precis < elen)
9731 width += precis - old_precis;
9733 width += elen - sv_len_utf8(argsv);
9740 if (has_precis && precis < elen)
9747 if (alt || vectorize)
9749 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9770 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9779 esignbuf[esignlen++] = plus;
9783 case 'h': iv = (short)va_arg(*args, int); break;
9784 case 'l': iv = va_arg(*args, long); break;
9785 case 'V': iv = va_arg(*args, IV); break;
9786 default: iv = va_arg(*args, int); break;
9789 iv = va_arg(*args, Quad_t); break;
9796 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9798 case 'h': iv = (short)tiv; break;
9799 case 'l': iv = (long)tiv; break;
9801 default: iv = tiv; break;
9804 iv = (Quad_t)tiv; break;
9810 if ( !vectorize ) /* we already set uv above */
9815 esignbuf[esignlen++] = plus;
9819 esignbuf[esignlen++] = '-';
9863 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9874 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9875 case 'l': uv = va_arg(*args, unsigned long); break;
9876 case 'V': uv = va_arg(*args, UV); break;
9877 default: uv = va_arg(*args, unsigned); break;
9880 uv = va_arg(*args, Uquad_t); break;
9887 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9889 case 'h': uv = (unsigned short)tuv; break;
9890 case 'l': uv = (unsigned long)tuv; break;
9892 default: uv = tuv; break;
9895 uv = (Uquad_t)tuv; break;
9904 char *ptr = ebuf + sizeof ebuf;
9905 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9911 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9917 esignbuf[esignlen++] = '0';
9918 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9926 if (alt && *ptr != '0')
9935 esignbuf[esignlen++] = '0';
9936 esignbuf[esignlen++] = c;
9939 default: /* it had better be ten or less */
9943 } while (uv /= base);
9946 elen = (ebuf + sizeof ebuf) - ptr;
9950 zeros = precis - elen;
9951 else if (precis == 0 && elen == 1 && *eptr == '0'
9952 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9955 /* a precision nullifies the 0 flag. */
9962 /* FLOATING POINT */
9965 c = 'f'; /* maybe %F isn't supported here */
9973 /* This is evil, but floating point is even more evil */
9975 /* for SV-style calling, we can only get NV
9976 for C-style calling, we assume %f is double;
9977 for simplicity we allow any of %Lf, %llf, %qf for long double
9981 #if defined(USE_LONG_DOUBLE)
9985 /* [perl #20339] - we should accept and ignore %lf rather than die */
9989 #if defined(USE_LONG_DOUBLE)
9990 intsize = args ? 0 : 'q';
9994 #if defined(HAS_LONG_DOUBLE)
10003 /* now we need (long double) if intsize == 'q', else (double) */
10005 #if LONG_DOUBLESIZE > DOUBLESIZE
10007 va_arg(*args, long double) :
10008 va_arg(*args, double)
10010 va_arg(*args, double)
10015 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10016 else. frexp() has some unspecified behaviour for those three */
10017 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10019 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10020 will cast our (long double) to (double) */
10021 (void)Perl_frexp(nv, &i);
10022 if (i == PERL_INT_MIN)
10023 Perl_die(aTHX_ "panic: frexp");
10025 need = BIT_DIGITS(i);
10027 need += has_precis ? precis : 6; /* known default */
10032 #ifdef HAS_LDBL_SPRINTF_BUG
10033 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10034 with sfio - Allen <allens@cpan.org> */
10037 # define MY_DBL_MAX DBL_MAX
10038 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10039 # if DOUBLESIZE >= 8
10040 # define MY_DBL_MAX 1.7976931348623157E+308L
10042 # define MY_DBL_MAX 3.40282347E+38L
10046 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10047 # define MY_DBL_MAX_BUG 1L
10049 # define MY_DBL_MAX_BUG MY_DBL_MAX
10053 # define MY_DBL_MIN DBL_MIN
10054 # else /* XXX guessing! -Allen */
10055 # if DOUBLESIZE >= 8
10056 # define MY_DBL_MIN 2.2250738585072014E-308L
10058 # define MY_DBL_MIN 1.17549435E-38L
10062 if ((intsize == 'q') && (c == 'f') &&
10063 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10064 (need < DBL_DIG)) {
10065 /* it's going to be short enough that
10066 * long double precision is not needed */
10068 if ((nv <= 0L) && (nv >= -0L))
10069 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10071 /* would use Perl_fp_class as a double-check but not
10072 * functional on IRIX - see perl.h comments */
10074 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10075 /* It's within the range that a double can represent */
10076 #if defined(DBL_MAX) && !defined(DBL_MIN)
10077 if ((nv >= ((long double)1/DBL_MAX)) ||
10078 (nv <= (-(long double)1/DBL_MAX)))
10080 fix_ldbl_sprintf_bug = TRUE;
10083 if (fix_ldbl_sprintf_bug == TRUE) {
10093 # undef MY_DBL_MAX_BUG
10096 #endif /* HAS_LDBL_SPRINTF_BUG */
10098 need += 20; /* fudge factor */
10099 if (PL_efloatsize < need) {
10100 Safefree(PL_efloatbuf);
10101 PL_efloatsize = need + 20; /* more fudge */
10102 Newx(PL_efloatbuf, PL_efloatsize, char);
10103 PL_efloatbuf[0] = '\0';
10106 if ( !(width || left || plus || alt) && fill != '0'
10107 && has_precis && intsize != 'q' ) { /* Shortcuts */
10108 /* See earlier comment about buggy Gconvert when digits,
10110 if ( c == 'g' && precis) {
10111 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10112 /* May return an empty string for digits==0 */
10113 if (*PL_efloatbuf) {
10114 elen = strlen(PL_efloatbuf);
10115 goto float_converted;
10117 } else if ( c == 'f' && !precis) {
10118 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10123 char *ptr = ebuf + sizeof ebuf;
10126 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10127 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10128 if (intsize == 'q') {
10129 /* Copy the one or more characters in a long double
10130 * format before the 'base' ([efgEFG]) character to
10131 * the format string. */
10132 static char const prifldbl[] = PERL_PRIfldbl;
10133 char const *p = prifldbl + sizeof(prifldbl) - 3;
10134 while (p >= prifldbl) { *--ptr = *p--; }
10139 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10144 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10156 /* No taint. Otherwise we are in the strange situation
10157 * where printf() taints but print($float) doesn't.
10159 #if defined(HAS_LONG_DOUBLE)
10160 elen = ((intsize == 'q')
10161 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10162 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10164 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10168 eptr = PL_efloatbuf;
10176 i = SvCUR(sv) - origlen;
10179 case 'h': *(va_arg(*args, short*)) = i; break;
10180 default: *(va_arg(*args, int*)) = i; break;
10181 case 'l': *(va_arg(*args, long*)) = i; break;
10182 case 'V': *(va_arg(*args, IV*)) = i; break;
10185 *(va_arg(*args, Quad_t*)) = i; break;
10192 sv_setuv_mg(argsv, (UV)i);
10193 continue; /* not "break" */
10200 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10201 && ckWARN(WARN_PRINTF))
10203 SV * const msg = sv_newmortal();
10204 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10205 (PL_op->op_type == OP_PRTF) ? "" : "s");
10206 if (fmtstart < patend) {
10207 const char * const fmtend = q < patend ? q : patend;
10209 sv_catpvs(msg, "\"%");
10210 for (f = fmtstart; f < fmtend; f++) {
10212 sv_catpvn(msg, f, 1);
10214 Perl_sv_catpvf(aTHX_ msg,
10215 "\\%03"UVof, (UV)*f & 0xFF);
10218 sv_catpvs(msg, "\"");
10220 sv_catpvs(msg, "end of string");
10222 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10225 /* output mangled stuff ... */
10231 /* ... right here, because formatting flags should not apply */
10232 SvGROW(sv, SvCUR(sv) + elen + 1);
10234 Copy(eptr, p, elen, char);
10237 SvCUR_set(sv, p - SvPVX_const(sv));
10239 continue; /* not "break" */
10242 if (is_utf8 != has_utf8) {
10245 sv_utf8_upgrade(sv);
10248 const STRLEN old_elen = elen;
10249 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10250 sv_utf8_upgrade(nsv);
10251 eptr = SvPVX_const(nsv);
10254 if (width) { /* fudge width (can't fudge elen) */
10255 width += elen - old_elen;
10261 have = esignlen + zeros + elen;
10263 Perl_croak_nocontext("%s", PL_memory_wrap);
10265 need = (have > width ? have : width);
10268 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10269 Perl_croak_nocontext("%s", PL_memory_wrap);
10270 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10272 if (esignlen && fill == '0') {
10274 for (i = 0; i < (int)esignlen; i++)
10275 *p++ = esignbuf[i];
10277 if (gap && !left) {
10278 memset(p, fill, gap);
10281 if (esignlen && fill != '0') {
10283 for (i = 0; i < (int)esignlen; i++)
10284 *p++ = esignbuf[i];
10288 for (i = zeros; i; i--)
10292 Copy(eptr, p, elen, char);
10296 memset(p, ' ', gap);
10301 Copy(dotstr, p, dotstrlen, char);
10305 vectorize = FALSE; /* done iterating over vecstr */
10312 SvCUR_set(sv, p - SvPVX_const(sv));
10320 /* =========================================================================
10322 =head1 Cloning an interpreter
10324 All the macros and functions in this section are for the private use of
10325 the main function, perl_clone().
10327 The foo_dup() functions make an exact copy of an existing foo thingy.
10328 During the course of a cloning, a hash table is used to map old addresses
10329 to new addresses. The table is created and manipulated with the
10330 ptr_table_* functions.
10334 * =========================================================================*/
10337 #if defined(USE_ITHREADS)
10339 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10340 #ifndef GpREFCNT_inc
10341 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10345 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10346 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10347 If this changes, please unmerge ss_dup.
10348 Likewise, sv_dup_inc_multiple() relies on this fact. */
10349 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10350 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10351 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10352 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10353 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10354 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10355 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10356 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10357 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10358 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10359 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10360 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10361 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10362 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10364 /* clone a parser */
10367 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10371 PERL_ARGS_ASSERT_PARSER_DUP;
10376 /* look for it in the table first */
10377 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10381 /* create anew and remember what it is */
10382 Newxz(parser, 1, yy_parser);
10383 ptr_table_store(PL_ptr_table, proto, parser);
10385 parser->yyerrstatus = 0;
10386 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10388 /* XXX these not yet duped */
10389 parser->old_parser = NULL;
10390 parser->stack = NULL;
10392 parser->stack_size = 0;
10393 /* XXX parser->stack->state = 0; */
10395 /* XXX eventually, just Copy() most of the parser struct ? */
10397 parser->lex_brackets = proto->lex_brackets;
10398 parser->lex_casemods = proto->lex_casemods;
10399 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10400 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10401 parser->lex_casestack = savepvn(proto->lex_casestack,
10402 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10403 parser->lex_defer = proto->lex_defer;
10404 parser->lex_dojoin = proto->lex_dojoin;
10405 parser->lex_expect = proto->lex_expect;
10406 parser->lex_formbrack = proto->lex_formbrack;
10407 parser->lex_inpat = proto->lex_inpat;
10408 parser->lex_inwhat = proto->lex_inwhat;
10409 parser->lex_op = proto->lex_op;
10410 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10411 parser->lex_starts = proto->lex_starts;
10412 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10413 parser->multi_close = proto->multi_close;
10414 parser->multi_open = proto->multi_open;
10415 parser->multi_start = proto->multi_start;
10416 parser->multi_end = proto->multi_end;
10417 parser->pending_ident = proto->pending_ident;
10418 parser->preambled = proto->preambled;
10419 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10420 parser->linestr = sv_dup_inc(proto->linestr, param);
10421 parser->expect = proto->expect;
10422 parser->copline = proto->copline;
10423 parser->last_lop_op = proto->last_lop_op;
10424 parser->lex_state = proto->lex_state;
10425 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10426 /* rsfp_filters entries have fake IoDIRP() */
10427 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10428 parser->in_my = proto->in_my;
10429 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10430 parser->error_count = proto->error_count;
10433 parser->linestr = sv_dup_inc(proto->linestr, param);
10436 char * const ols = SvPVX(proto->linestr);
10437 char * const ls = SvPVX(parser->linestr);
10439 parser->bufptr = ls + (proto->bufptr >= ols ?
10440 proto->bufptr - ols : 0);
10441 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10442 proto->oldbufptr - ols : 0);
10443 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10444 proto->oldoldbufptr - ols : 0);
10445 parser->linestart = ls + (proto->linestart >= ols ?
10446 proto->linestart - ols : 0);
10447 parser->last_uni = ls + (proto->last_uni >= ols ?
10448 proto->last_uni - ols : 0);
10449 parser->last_lop = ls + (proto->last_lop >= ols ?
10450 proto->last_lop - ols : 0);
10452 parser->bufend = ls + SvCUR(parser->linestr);
10455 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10459 parser->endwhite = proto->endwhite;
10460 parser->faketokens = proto->faketokens;
10461 parser->lasttoke = proto->lasttoke;
10462 parser->nextwhite = proto->nextwhite;
10463 parser->realtokenstart = proto->realtokenstart;
10464 parser->skipwhite = proto->skipwhite;
10465 parser->thisclose = proto->thisclose;
10466 parser->thismad = proto->thismad;
10467 parser->thisopen = proto->thisopen;
10468 parser->thisstuff = proto->thisstuff;
10469 parser->thistoken = proto->thistoken;
10470 parser->thiswhite = proto->thiswhite;
10472 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10473 parser->curforce = proto->curforce;
10475 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10476 Copy(proto->nexttype, parser->nexttype, 5, I32);
10477 parser->nexttoke = proto->nexttoke;
10480 /* XXX should clone saved_curcop here, but we aren't passed
10481 * proto_perl; so do it in perl_clone_using instead */
10487 /* duplicate a file handle */
10490 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10494 PERL_ARGS_ASSERT_FP_DUP;
10495 PERL_UNUSED_ARG(type);
10498 return (PerlIO*)NULL;
10500 /* look for it in the table first */
10501 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10505 /* create anew and remember what it is */
10506 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10507 ptr_table_store(PL_ptr_table, fp, ret);
10511 /* duplicate a directory handle */
10514 Perl_dirp_dup(pTHX_ DIR *const dp)
10516 PERL_UNUSED_CONTEXT;
10523 /* duplicate a typeglob */
10526 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10530 PERL_ARGS_ASSERT_GP_DUP;
10534 /* look for it in the table first */
10535 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10539 /* create anew and remember what it is */
10541 ptr_table_store(PL_ptr_table, gp, ret);
10544 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10545 on Newxz() to do this for us. */
10546 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10547 ret->gp_io = io_dup_inc(gp->gp_io, param);
10548 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10549 ret->gp_av = av_dup_inc(gp->gp_av, param);
10550 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10551 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10552 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10553 ret->gp_cvgen = gp->gp_cvgen;
10554 ret->gp_line = gp->gp_line;
10555 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10559 /* duplicate a chain of magic */
10562 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10564 MAGIC *mgret = NULL;
10565 MAGIC **mgprev_p = &mgret;
10567 PERL_ARGS_ASSERT_MG_DUP;
10569 for (; mg; mg = mg->mg_moremagic) {
10571 Newx(nmg, 1, MAGIC);
10573 mgprev_p = &(nmg->mg_moremagic);
10575 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10576 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10577 from the original commit adding Perl_mg_dup() - revision 4538.
10578 Similarly there is the annotation "XXX random ptr?" next to the
10579 assignment to nmg->mg_ptr. */
10582 /* FIXME for plugins
10583 if (nmg->mg_type == PERL_MAGIC_qr) {
10584 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10588 if(nmg->mg_type == PERL_MAGIC_backref) {
10589 /* The backref AV has its reference count deliberately bumped by
10592 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10595 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10596 ? sv_dup_inc(nmg->mg_obj, param)
10597 : sv_dup(nmg->mg_obj, param);
10600 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10601 if (nmg->mg_len > 0) {
10602 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10603 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10604 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10606 AMT * const namtp = (AMT*)nmg->mg_ptr;
10607 sv_dup_inc_multiple((SV**)(namtp->table),
10608 (SV**)(namtp->table), NofAMmeth, param);
10611 else if (nmg->mg_len == HEf_SVKEY)
10612 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10614 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10615 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10621 #endif /* USE_ITHREADS */
10623 /* create a new pointer-mapping table */
10626 Perl_ptr_table_new(pTHX)
10629 PERL_UNUSED_CONTEXT;
10631 Newx(tbl, 1, PTR_TBL_t);
10632 tbl->tbl_max = 511;
10633 tbl->tbl_items = 0;
10634 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10638 #define PTR_TABLE_HASH(ptr) \
10639 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10642 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10643 following define) and at call to new_body_inline made below in
10644 Perl_ptr_table_store()
10647 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10649 /* map an existing pointer using a table */
10651 STATIC PTR_TBL_ENT_t *
10652 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10654 PTR_TBL_ENT_t *tblent;
10655 const UV hash = PTR_TABLE_HASH(sv);
10657 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10659 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10660 for (; tblent; tblent = tblent->next) {
10661 if (tblent->oldval == sv)
10668 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10670 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10672 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10673 PERL_UNUSED_CONTEXT;
10675 return tblent ? tblent->newval : NULL;
10678 /* add a new entry to a pointer-mapping table */
10681 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10683 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10685 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10686 PERL_UNUSED_CONTEXT;
10689 tblent->newval = newsv;
10691 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10693 new_body_inline(tblent, PTE_SVSLOT);
10695 tblent->oldval = oldsv;
10696 tblent->newval = newsv;
10697 tblent->next = tbl->tbl_ary[entry];
10698 tbl->tbl_ary[entry] = tblent;
10700 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10701 ptr_table_split(tbl);
10705 /* double the hash bucket size of an existing ptr table */
10708 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10710 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10711 const UV oldsize = tbl->tbl_max + 1;
10712 UV newsize = oldsize * 2;
10715 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10716 PERL_UNUSED_CONTEXT;
10718 Renew(ary, newsize, PTR_TBL_ENT_t*);
10719 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10720 tbl->tbl_max = --newsize;
10721 tbl->tbl_ary = ary;
10722 for (i=0; i < oldsize; i++, ary++) {
10723 PTR_TBL_ENT_t **curentp, **entp, *ent;
10726 curentp = ary + oldsize;
10727 for (entp = ary, ent = *ary; ent; ent = *entp) {
10728 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10730 ent->next = *curentp;
10740 /* remove all the entries from a ptr table */
10743 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10745 if (tbl && tbl->tbl_items) {
10746 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10747 UV riter = tbl->tbl_max;
10750 PTR_TBL_ENT_t *entry = array[riter];
10753 PTR_TBL_ENT_t * const oentry = entry;
10754 entry = entry->next;
10759 tbl->tbl_items = 0;
10763 /* clear and free a ptr table */
10766 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10771 ptr_table_clear(tbl);
10772 Safefree(tbl->tbl_ary);
10776 #if defined(USE_ITHREADS)
10779 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10781 PERL_ARGS_ASSERT_RVPV_DUP;
10784 SvRV_set(dstr, SvWEAKREF(sstr)
10785 ? sv_dup(SvRV_const(sstr), param)
10786 : sv_dup_inc(SvRV_const(sstr), param));
10789 else if (SvPVX_const(sstr)) {
10790 /* Has something there */
10792 /* Normal PV - clone whole allocated space */
10793 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10794 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10795 /* Not that normal - actually sstr is copy on write.
10796 But we are a true, independant SV, so: */
10797 SvREADONLY_off(dstr);
10802 /* Special case - not normally malloced for some reason */
10803 if (isGV_with_GP(sstr)) {
10804 /* Don't need to do anything here. */
10806 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10807 /* A "shared" PV - clone it as "shared" PV */
10809 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10813 /* Some other special case - random pointer */
10814 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10819 /* Copy the NULL */
10820 SvPV_set(dstr, NULL);
10824 /* duplicate a list of SVs. source and dest may point to the same memory. */
10826 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10827 SSize_t items, CLONE_PARAMS *const param)
10829 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10831 while (items-- > 0) {
10832 *dest++ = sv_dup_inc(*source++, param);
10838 /* duplicate an SV of any type (including AV, HV etc) */
10841 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10846 PERL_ARGS_ASSERT_SV_DUP;
10850 if (SvTYPE(sstr) == SVTYPEMASK) {
10851 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10856 /* look for it in the table first */
10857 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10861 if(param->flags & CLONEf_JOIN_IN) {
10862 /** We are joining here so we don't want do clone
10863 something that is bad **/
10864 if (SvTYPE(sstr) == SVt_PVHV) {
10865 const HEK * const hvname = HvNAME_HEK(sstr);
10867 /** don't clone stashes if they already exist **/
10868 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10872 /* create anew and remember what it is */
10875 #ifdef DEBUG_LEAKING_SCALARS
10876 dstr->sv_debug_optype = sstr->sv_debug_optype;
10877 dstr->sv_debug_line = sstr->sv_debug_line;
10878 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10879 dstr->sv_debug_cloned = 1;
10880 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10883 ptr_table_store(PL_ptr_table, sstr, dstr);
10886 SvFLAGS(dstr) = SvFLAGS(sstr);
10887 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10888 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10891 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10892 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10893 (void*)PL_watch_pvx, SvPVX_const(sstr));
10896 /* don't clone objects whose class has asked us not to */
10897 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10902 switch (SvTYPE(sstr)) {
10904 SvANY(dstr) = NULL;
10907 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10909 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10911 SvIV_set(dstr, SvIVX(sstr));
10915 SvANY(dstr) = new_XNV();
10916 SvNV_set(dstr, SvNVX(sstr));
10918 /* case SVt_BIND: */
10921 /* These are all the types that need complex bodies allocating. */
10923 const svtype sv_type = SvTYPE(sstr);
10924 const struct body_details *const sv_type_details
10925 = bodies_by_type + sv_type;
10929 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10944 assert(sv_type_details->body_size);
10945 if (sv_type_details->arena) {
10946 new_body_inline(new_body, sv_type);
10948 = (void*)((char*)new_body - sv_type_details->offset);
10950 new_body = new_NOARENA(sv_type_details);
10954 SvANY(dstr) = new_body;
10957 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10958 ((char*)SvANY(dstr)) + sv_type_details->offset,
10959 sv_type_details->copy, char);
10961 Copy(((char*)SvANY(sstr)),
10962 ((char*)SvANY(dstr)),
10963 sv_type_details->body_size + sv_type_details->offset, char);
10966 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10967 && !isGV_with_GP(dstr))
10968 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10970 /* The Copy above means that all the source (unduplicated) pointers
10971 are now in the destination. We can check the flags and the
10972 pointers in either, but it's possible that there's less cache
10973 missing by always going for the destination.
10974 FIXME - instrument and check that assumption */
10975 if (sv_type >= SVt_PVMG) {
10976 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10977 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10978 } else if (SvMAGIC(dstr))
10979 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10981 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10984 /* The cast silences a GCC warning about unhandled types. */
10985 switch ((int)sv_type) {
10995 /* FIXME for plugins */
10996 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10999 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11000 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11001 LvTARG(dstr) = dstr;
11002 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11003 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11005 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11007 if(isGV_with_GP(sstr)) {
11008 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11009 /* Don't call sv_add_backref here as it's going to be
11010 created as part of the magic cloning of the symbol
11012 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11013 at the point of this comment. */
11014 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11015 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11016 (void)GpREFCNT_inc(GvGP(dstr));
11018 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11021 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11022 if (IoOFP(dstr) == IoIFP(sstr))
11023 IoOFP(dstr) = IoIFP(dstr);
11025 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11026 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11027 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11028 /* I have no idea why fake dirp (rsfps)
11029 should be treated differently but otherwise
11030 we end up with leaks -- sky*/
11031 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11032 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11033 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11035 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11036 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11037 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11038 if (IoDIRP(dstr)) {
11039 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11042 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11045 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11046 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11047 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11050 /* avoid cloning an empty array */
11051 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11052 SV **dst_ary, **src_ary;
11053 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11055 src_ary = AvARRAY((const AV *)sstr);
11056 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11057 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11058 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11059 AvALLOC((const AV *)dstr) = dst_ary;
11060 if (AvREAL((const AV *)sstr)) {
11061 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11065 while (items-- > 0)
11066 *dst_ary++ = sv_dup(*src_ary++, param);
11068 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11069 while (items-- > 0) {
11070 *dst_ary++ = &PL_sv_undef;
11074 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11075 AvALLOC((const AV *)dstr) = (SV**)NULL;
11076 AvMAX( (const AV *)dstr) = -1;
11077 AvFILLp((const AV *)dstr) = -1;
11081 if (HvARRAY((const HV *)sstr)) {
11083 const bool sharekeys = !!HvSHAREKEYS(sstr);
11084 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11085 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11087 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11088 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11090 HvARRAY(dstr) = (HE**)darray;
11091 while (i <= sxhv->xhv_max) {
11092 const HE * const source = HvARRAY(sstr)[i];
11093 HvARRAY(dstr)[i] = source
11094 ? he_dup(source, sharekeys, param) : 0;
11099 const struct xpvhv_aux * const saux = HvAUX(sstr);
11100 struct xpvhv_aux * const daux = HvAUX(dstr);
11101 /* This flag isn't copied. */
11102 /* SvOOK_on(hv) attacks the IV flags. */
11103 SvFLAGS(dstr) |= SVf_OOK;
11105 hvname = saux->xhv_name;
11106 daux->xhv_name = hek_dup(hvname, param);
11108 daux->xhv_riter = saux->xhv_riter;
11109 daux->xhv_eiter = saux->xhv_eiter
11110 ? he_dup(saux->xhv_eiter,
11111 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11112 /* backref array needs refcnt=2; see sv_add_backref */
11113 daux->xhv_backreferences =
11114 saux->xhv_backreferences
11115 ? MUTABLE_AV(SvREFCNT_inc(
11116 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11119 daux->xhv_mro_meta = saux->xhv_mro_meta
11120 ? mro_meta_dup(saux->xhv_mro_meta, param)
11123 /* Record stashes for possible cloning in Perl_clone(). */
11125 av_push(param->stashes, dstr);
11129 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11132 if (!(param->flags & CLONEf_COPY_STACKS)) {
11136 /* NOTE: not refcounted */
11137 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11139 if (!CvISXSUB(dstr))
11140 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11142 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11143 CvXSUBANY(dstr).any_ptr =
11144 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11146 /* don't dup if copying back - CvGV isn't refcounted, so the
11147 * duped GV may never be freed. A bit of a hack! DAPM */
11148 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11149 NULL : gv_dup(CvGV(dstr), param) ;
11150 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11152 CvWEAKOUTSIDE(sstr)
11153 ? cv_dup( CvOUTSIDE(dstr), param)
11154 : cv_dup_inc(CvOUTSIDE(dstr), param);
11155 if (!CvISXSUB(dstr))
11156 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11162 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11168 /* duplicate a context */
11171 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11173 PERL_CONTEXT *ncxs;
11175 PERL_ARGS_ASSERT_CX_DUP;
11178 return (PERL_CONTEXT*)NULL;
11180 /* look for it in the table first */
11181 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11185 /* create anew and remember what it is */
11186 Newx(ncxs, max + 1, PERL_CONTEXT);
11187 ptr_table_store(PL_ptr_table, cxs, ncxs);
11188 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11191 PERL_CONTEXT * const ncx = &ncxs[ix];
11192 if (CxTYPE(ncx) == CXt_SUBST) {
11193 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11196 switch (CxTYPE(ncx)) {
11198 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11199 ? cv_dup_inc(ncx->blk_sub.cv, param)
11200 : cv_dup(ncx->blk_sub.cv,param));
11201 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11202 ? av_dup_inc(ncx->blk_sub.argarray,
11205 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11207 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11208 ncx->blk_sub.oldcomppad);
11211 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11213 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11215 case CXt_LOOP_LAZYSV:
11216 ncx->blk_loop.state_u.lazysv.end
11217 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11218 /* We are taking advantage of av_dup_inc and sv_dup_inc
11219 actually being the same function, and order equivalance of
11221 We can assert the later [but only at run time :-(] */
11222 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11223 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11225 ncx->blk_loop.state_u.ary.ary
11226 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11227 case CXt_LOOP_LAZYIV:
11228 case CXt_LOOP_PLAIN:
11229 if (CxPADLOOP(ncx)) {
11230 ncx->blk_loop.oldcomppad
11231 = (PAD*)ptr_table_fetch(PL_ptr_table,
11232 ncx->blk_loop.oldcomppad);
11234 ncx->blk_loop.oldcomppad
11235 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11240 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11241 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11242 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11255 /* duplicate a stack info structure */
11258 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11262 PERL_ARGS_ASSERT_SI_DUP;
11265 return (PERL_SI*)NULL;
11267 /* look for it in the table first */
11268 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11272 /* create anew and remember what it is */
11273 Newxz(nsi, 1, PERL_SI);
11274 ptr_table_store(PL_ptr_table, si, nsi);
11276 nsi->si_stack = av_dup_inc(si->si_stack, param);
11277 nsi->si_cxix = si->si_cxix;
11278 nsi->si_cxmax = si->si_cxmax;
11279 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11280 nsi->si_type = si->si_type;
11281 nsi->si_prev = si_dup(si->si_prev, param);
11282 nsi->si_next = si_dup(si->si_next, param);
11283 nsi->si_markoff = si->si_markoff;
11288 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11289 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11290 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11291 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11292 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11293 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11294 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11295 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11296 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11297 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11298 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11299 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11300 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11301 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11304 #define pv_dup_inc(p) SAVEPV(p)
11305 #define pv_dup(p) SAVEPV(p)
11306 #define svp_dup_inc(p,pp) any_dup(p,pp)
11308 /* map any object to the new equivent - either something in the
11309 * ptr table, or something in the interpreter structure
11313 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11317 PERL_ARGS_ASSERT_ANY_DUP;
11320 return (void*)NULL;
11322 /* look for it in the table first */
11323 ret = ptr_table_fetch(PL_ptr_table, v);
11327 /* see if it is part of the interpreter structure */
11328 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11329 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11337 /* duplicate the save stack */
11340 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11343 ANY * const ss = proto_perl->Isavestack;
11344 const I32 max = proto_perl->Isavestack_max;
11345 I32 ix = proto_perl->Isavestack_ix;
11358 void (*dptr) (void*);
11359 void (*dxptr) (pTHX_ void*);
11361 PERL_ARGS_ASSERT_SS_DUP;
11363 Newxz(nss, max, ANY);
11366 const I32 type = POPINT(ss,ix);
11367 TOPINT(nss,ix) = type;
11369 case SAVEt_HELEM: /* hash element */
11370 sv = (const SV *)POPPTR(ss,ix);
11371 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11373 case SAVEt_ITEM: /* normal string */
11374 case SAVEt_SV: /* scalar reference */
11375 sv = (const SV *)POPPTR(ss,ix);
11376 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11379 case SAVEt_MORTALIZESV:
11380 sv = (const SV *)POPPTR(ss,ix);
11381 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11383 case SAVEt_SHARED_PVREF: /* char* in shared space */
11384 c = (char*)POPPTR(ss,ix);
11385 TOPPTR(nss,ix) = savesharedpv(c);
11386 ptr = POPPTR(ss,ix);
11387 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11389 case SAVEt_GENERIC_SVREF: /* generic sv */
11390 case SAVEt_SVREF: /* scalar reference */
11391 sv = (const SV *)POPPTR(ss,ix);
11392 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11393 ptr = POPPTR(ss,ix);
11394 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11396 case SAVEt_HV: /* hash reference */
11397 case SAVEt_AV: /* array reference */
11398 sv = (const SV *) POPPTR(ss,ix);
11399 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11401 case SAVEt_COMPPAD:
11403 sv = (const SV *) POPPTR(ss,ix);
11404 TOPPTR(nss,ix) = sv_dup(sv, param);
11406 case SAVEt_INT: /* int reference */
11407 ptr = POPPTR(ss,ix);
11408 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11409 intval = (int)POPINT(ss,ix);
11410 TOPINT(nss,ix) = intval;
11412 case SAVEt_LONG: /* long reference */
11413 ptr = POPPTR(ss,ix);
11414 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11416 case SAVEt_CLEARSV:
11417 longval = (long)POPLONG(ss,ix);
11418 TOPLONG(nss,ix) = longval;
11420 case SAVEt_I32: /* I32 reference */
11421 case SAVEt_I16: /* I16 reference */
11422 case SAVEt_I8: /* I8 reference */
11423 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11424 ptr = POPPTR(ss,ix);
11425 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11427 TOPINT(nss,ix) = i;
11429 case SAVEt_IV: /* IV reference */
11430 ptr = POPPTR(ss,ix);
11431 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11433 TOPIV(nss,ix) = iv;
11435 case SAVEt_HPTR: /* HV* reference */
11436 case SAVEt_APTR: /* AV* reference */
11437 case SAVEt_SPTR: /* SV* reference */
11438 ptr = POPPTR(ss,ix);
11439 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11440 sv = (const SV *)POPPTR(ss,ix);
11441 TOPPTR(nss,ix) = sv_dup(sv, param);
11443 case SAVEt_VPTR: /* random* reference */
11444 ptr = POPPTR(ss,ix);
11445 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11446 ptr = POPPTR(ss,ix);
11447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11449 case SAVEt_GENERIC_PVREF: /* generic char* */
11450 case SAVEt_PPTR: /* char* reference */
11451 ptr = POPPTR(ss,ix);
11452 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11453 c = (char*)POPPTR(ss,ix);
11454 TOPPTR(nss,ix) = pv_dup(c);
11456 case SAVEt_GP: /* scalar reference */
11457 gp = (GP*)POPPTR(ss,ix);
11458 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11459 (void)GpREFCNT_inc(gp);
11460 gv = (const GV *)POPPTR(ss,ix);
11461 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11464 ptr = POPPTR(ss,ix);
11465 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11466 /* these are assumed to be refcounted properly */
11468 switch (((OP*)ptr)->op_type) {
11470 case OP_LEAVESUBLV:
11474 case OP_LEAVEWRITE:
11475 TOPPTR(nss,ix) = ptr;
11478 (void) OpREFCNT_inc(o);
11482 TOPPTR(nss,ix) = NULL;
11487 TOPPTR(nss,ix) = NULL;
11490 hv = (const HV *)POPPTR(ss,ix);
11491 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11493 TOPINT(nss,ix) = i;
11496 c = (char*)POPPTR(ss,ix);
11497 TOPPTR(nss,ix) = pv_dup_inc(c);
11499 case SAVEt_STACK_POS: /* Position on Perl stack */
11501 TOPINT(nss,ix) = i;
11503 case SAVEt_DESTRUCTOR:
11504 ptr = POPPTR(ss,ix);
11505 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11506 dptr = POPDPTR(ss,ix);
11507 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11508 any_dup(FPTR2DPTR(void *, dptr),
11511 case SAVEt_DESTRUCTOR_X:
11512 ptr = POPPTR(ss,ix);
11513 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11514 dxptr = POPDXPTR(ss,ix);
11515 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11516 any_dup(FPTR2DPTR(void *, dxptr),
11519 case SAVEt_REGCONTEXT:
11522 TOPINT(nss,ix) = i;
11525 case SAVEt_AELEM: /* array element */
11526 sv = (const SV *)POPPTR(ss,ix);
11527 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11529 TOPINT(nss,ix) = i;
11530 av = (const AV *)POPPTR(ss,ix);
11531 TOPPTR(nss,ix) = av_dup_inc(av, param);
11534 ptr = POPPTR(ss,ix);
11535 TOPPTR(nss,ix) = ptr;
11538 ptr = POPPTR(ss,ix);
11541 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11542 HINTS_REFCNT_UNLOCK;
11544 TOPPTR(nss,ix) = ptr;
11546 TOPINT(nss,ix) = i;
11547 if (i & HINT_LOCALIZE_HH) {
11548 hv = (const HV *)POPPTR(ss,ix);
11549 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11552 case SAVEt_PADSV_AND_MORTALIZE:
11553 longval = (long)POPLONG(ss,ix);
11554 TOPLONG(nss,ix) = longval;
11555 ptr = POPPTR(ss,ix);
11556 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11557 sv = (const SV *)POPPTR(ss,ix);
11558 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11561 ptr = POPPTR(ss,ix);
11562 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11563 longval = (long)POPBOOL(ss,ix);
11564 TOPBOOL(nss,ix) = (bool)longval;
11566 case SAVEt_SET_SVFLAGS:
11568 TOPINT(nss,ix) = i;
11570 TOPINT(nss,ix) = i;
11571 sv = (const SV *)POPPTR(ss,ix);
11572 TOPPTR(nss,ix) = sv_dup(sv, param);
11574 case SAVEt_RE_STATE:
11576 const struct re_save_state *const old_state
11577 = (struct re_save_state *)
11578 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11579 struct re_save_state *const new_state
11580 = (struct re_save_state *)
11581 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11583 Copy(old_state, new_state, 1, struct re_save_state);
11584 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11586 new_state->re_state_bostr
11587 = pv_dup(old_state->re_state_bostr);
11588 new_state->re_state_reginput
11589 = pv_dup(old_state->re_state_reginput);
11590 new_state->re_state_regeol
11591 = pv_dup(old_state->re_state_regeol);
11592 new_state->re_state_regoffs
11593 = (regexp_paren_pair*)
11594 any_dup(old_state->re_state_regoffs, proto_perl);
11595 new_state->re_state_reglastparen
11596 = (U32*) any_dup(old_state->re_state_reglastparen,
11598 new_state->re_state_reglastcloseparen
11599 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11601 /* XXX This just has to be broken. The old save_re_context
11602 code did SAVEGENERICPV(PL_reg_start_tmp);
11603 PL_reg_start_tmp is char **.
11604 Look above to what the dup code does for
11605 SAVEt_GENERIC_PVREF
11606 It can never have worked.
11607 So this is merely a faithful copy of the exiting bug: */
11608 new_state->re_state_reg_start_tmp
11609 = (char **) pv_dup((char *)
11610 old_state->re_state_reg_start_tmp);
11611 /* I assume that it only ever "worked" because no-one called
11612 (pseudo)fork while the regexp engine had re-entered itself.
11614 #ifdef PERL_OLD_COPY_ON_WRITE
11615 new_state->re_state_nrs
11616 = sv_dup(old_state->re_state_nrs, param);
11618 new_state->re_state_reg_magic
11619 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11621 new_state->re_state_reg_oldcurpm
11622 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11624 new_state->re_state_reg_curpm
11625 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11627 new_state->re_state_reg_oldsaved
11628 = pv_dup(old_state->re_state_reg_oldsaved);
11629 new_state->re_state_reg_poscache
11630 = pv_dup(old_state->re_state_reg_poscache);
11631 new_state->re_state_reg_starttry
11632 = pv_dup(old_state->re_state_reg_starttry);
11635 case SAVEt_COMPILE_WARNINGS:
11636 ptr = POPPTR(ss,ix);
11637 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11640 ptr = POPPTR(ss,ix);
11641 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11645 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11653 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11654 * flag to the result. This is done for each stash before cloning starts,
11655 * so we know which stashes want their objects cloned */
11658 do_mark_cloneable_stash(pTHX_ SV *const sv)
11660 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11662 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11663 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11664 if (cloner && GvCV(cloner)) {
11671 mXPUSHs(newSVhek(hvname));
11673 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11680 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11688 =for apidoc perl_clone
11690 Create and return a new interpreter by cloning the current one.
11692 perl_clone takes these flags as parameters:
11694 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11695 without it we only clone the data and zero the stacks,
11696 with it we copy the stacks and the new perl interpreter is
11697 ready to run at the exact same point as the previous one.
11698 The pseudo-fork code uses COPY_STACKS while the
11699 threads->create doesn't.
11701 CLONEf_KEEP_PTR_TABLE
11702 perl_clone keeps a ptr_table with the pointer of the old
11703 variable as a key and the new variable as a value,
11704 this allows it to check if something has been cloned and not
11705 clone it again but rather just use the value and increase the
11706 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11707 the ptr_table using the function
11708 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11709 reason to keep it around is if you want to dup some of your own
11710 variable who are outside the graph perl scans, example of this
11711 code is in threads.xs create
11714 This is a win32 thing, it is ignored on unix, it tells perls
11715 win32host code (which is c++) to clone itself, this is needed on
11716 win32 if you want to run two threads at the same time,
11717 if you just want to do some stuff in a separate perl interpreter
11718 and then throw it away and return to the original one,
11719 you don't need to do anything.
11724 /* XXX the above needs expanding by someone who actually understands it ! */
11725 EXTERN_C PerlInterpreter *
11726 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11729 perl_clone(PerlInterpreter *proto_perl, UV flags)
11732 #ifdef PERL_IMPLICIT_SYS
11734 PERL_ARGS_ASSERT_PERL_CLONE;
11736 /* perlhost.h so we need to call into it
11737 to clone the host, CPerlHost should have a c interface, sky */
11739 if (flags & CLONEf_CLONE_HOST) {
11740 return perl_clone_host(proto_perl,flags);
11742 return perl_clone_using(proto_perl, flags,
11744 proto_perl->IMemShared,
11745 proto_perl->IMemParse,
11747 proto_perl->IStdIO,
11751 proto_perl->IProc);
11755 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11756 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11757 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11758 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11759 struct IPerlDir* ipD, struct IPerlSock* ipS,
11760 struct IPerlProc* ipP)
11762 /* XXX many of the string copies here can be optimized if they're
11763 * constants; they need to be allocated as common memory and just
11764 * their pointers copied. */
11767 CLONE_PARAMS clone_params;
11768 CLONE_PARAMS* const param = &clone_params;
11770 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11772 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11774 /* for each stash, determine whether its objects should be cloned */
11775 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11776 PERL_SET_THX(my_perl);
11779 PoisonNew(my_perl, 1, PerlInterpreter);
11784 PL_scopestack_name = 0;
11786 PL_savestack_ix = 0;
11787 PL_savestack_max = -1;
11788 PL_sig_pending = 0;
11790 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11791 # else /* !DEBUGGING */
11792 Zero(my_perl, 1, PerlInterpreter);
11793 # endif /* DEBUGGING */
11795 /* host pointers */
11797 PL_MemShared = ipMS;
11798 PL_MemParse = ipMP;
11805 #else /* !PERL_IMPLICIT_SYS */
11807 CLONE_PARAMS clone_params;
11808 CLONE_PARAMS* param = &clone_params;
11809 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11811 PERL_ARGS_ASSERT_PERL_CLONE;
11813 /* for each stash, determine whether its objects should be cloned */
11814 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11815 PERL_SET_THX(my_perl);
11818 PoisonNew(my_perl, 1, PerlInterpreter);
11823 PL_scopestack_name = 0;
11825 PL_savestack_ix = 0;
11826 PL_savestack_max = -1;
11827 PL_sig_pending = 0;
11829 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11830 # else /* !DEBUGGING */
11831 Zero(my_perl, 1, PerlInterpreter);
11832 # endif /* DEBUGGING */
11833 #endif /* PERL_IMPLICIT_SYS */
11834 param->flags = flags;
11835 param->proto_perl = proto_perl;
11837 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11839 PL_body_arenas = NULL;
11840 Zero(&PL_body_roots, 1, PL_body_roots);
11842 PL_nice_chunk = NULL;
11843 PL_nice_chunk_size = 0;
11845 PL_sv_objcount = 0;
11847 PL_sv_arenaroot = NULL;
11849 PL_debug = proto_perl->Idebug;
11851 PL_hash_seed = proto_perl->Ihash_seed;
11852 PL_rehash_seed = proto_perl->Irehash_seed;
11854 #ifdef USE_REENTRANT_API
11855 /* XXX: things like -Dm will segfault here in perlio, but doing
11856 * PERL_SET_CONTEXT(proto_perl);
11857 * breaks too many other things
11859 Perl_reentrant_init(aTHX);
11862 /* create SV map for pointer relocation */
11863 PL_ptr_table = ptr_table_new();
11865 /* initialize these special pointers as early as possible */
11866 SvANY(&PL_sv_undef) = NULL;
11867 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11868 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11869 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11871 SvANY(&PL_sv_no) = new_XPVNV();
11872 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11873 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11874 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11875 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11876 SvCUR_set(&PL_sv_no, 0);
11877 SvLEN_set(&PL_sv_no, 1);
11878 SvIV_set(&PL_sv_no, 0);
11879 SvNV_set(&PL_sv_no, 0);
11880 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11882 SvANY(&PL_sv_yes) = new_XPVNV();
11883 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11884 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11885 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11886 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11887 SvCUR_set(&PL_sv_yes, 1);
11888 SvLEN_set(&PL_sv_yes, 2);
11889 SvIV_set(&PL_sv_yes, 1);
11890 SvNV_set(&PL_sv_yes, 1);
11891 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11893 /* create (a non-shared!) shared string table */
11894 PL_strtab = newHV();
11895 HvSHAREKEYS_off(PL_strtab);
11896 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11897 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11899 PL_compiling = proto_perl->Icompiling;
11901 /* These two PVs will be free'd special way so must set them same way op.c does */
11902 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11903 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11905 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11906 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11908 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11909 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11910 if (PL_compiling.cop_hints_hash) {
11912 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11913 HINTS_REFCNT_UNLOCK;
11915 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11916 #ifdef PERL_DEBUG_READONLY_OPS
11921 /* pseudo environmental stuff */
11922 PL_origargc = proto_perl->Iorigargc;
11923 PL_origargv = proto_perl->Iorigargv;
11925 param->stashes = newAV(); /* Setup array of objects to call clone on */
11927 /* Set tainting stuff before PerlIO_debug can possibly get called */
11928 PL_tainting = proto_perl->Itainting;
11929 PL_taint_warn = proto_perl->Itaint_warn;
11931 #ifdef PERLIO_LAYERS
11932 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11933 PerlIO_clone(aTHX_ proto_perl, param);
11936 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11937 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11938 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11939 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11940 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11941 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11944 PL_minus_c = proto_perl->Iminus_c;
11945 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11946 PL_localpatches = proto_perl->Ilocalpatches;
11947 PL_splitstr = proto_perl->Isplitstr;
11948 PL_minus_n = proto_perl->Iminus_n;
11949 PL_minus_p = proto_perl->Iminus_p;
11950 PL_minus_l = proto_perl->Iminus_l;
11951 PL_minus_a = proto_perl->Iminus_a;
11952 PL_minus_E = proto_perl->Iminus_E;
11953 PL_minus_F = proto_perl->Iminus_F;
11954 PL_doswitches = proto_perl->Idoswitches;
11955 PL_dowarn = proto_perl->Idowarn;
11956 PL_doextract = proto_perl->Idoextract;
11957 PL_sawampersand = proto_perl->Isawampersand;
11958 PL_unsafe = proto_perl->Iunsafe;
11959 PL_inplace = SAVEPV(proto_perl->Iinplace);
11960 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11961 PL_perldb = proto_perl->Iperldb;
11962 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11963 PL_exit_flags = proto_perl->Iexit_flags;
11965 /* magical thingies */
11966 /* XXX time(&PL_basetime) when asked for? */
11967 PL_basetime = proto_perl->Ibasetime;
11968 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11970 PL_maxsysfd = proto_perl->Imaxsysfd;
11971 PL_statusvalue = proto_perl->Istatusvalue;
11973 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11975 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11977 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11979 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11980 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11981 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11984 /* RE engine related */
11985 Zero(&PL_reg_state, 1, struct re_save_state);
11986 PL_reginterp_cnt = 0;
11987 PL_regmatch_slab = NULL;
11989 /* Clone the regex array */
11990 /* ORANGE FIXME for plugins, probably in the SV dup code.
11991 newSViv(PTR2IV(CALLREGDUPE(
11992 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11994 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11995 PL_regex_pad = AvARRAY(PL_regex_padav);
11997 /* shortcuts to various I/O objects */
11998 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11999 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12000 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12001 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12002 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12003 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12004 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12006 /* shortcuts to regexp stuff */
12007 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12009 /* shortcuts to misc objects */
12010 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12012 /* shortcuts to debugging objects */
12013 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12014 PL_DBline = gv_dup(proto_perl->IDBline, param);
12015 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12016 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12017 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12018 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12019 PL_dbargs = av_dup(proto_perl->Idbargs, param);
12021 /* symbol tables */
12022 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12023 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12024 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12025 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12026 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12028 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12029 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12030 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12031 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12032 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12033 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12034 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12035 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12037 PL_sub_generation = proto_perl->Isub_generation;
12038 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12040 /* funky return mechanisms */
12041 PL_forkprocess = proto_perl->Iforkprocess;
12043 /* subprocess state */
12044 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12046 /* internal state */
12047 PL_maxo = proto_perl->Imaxo;
12048 if (proto_perl->Iop_mask)
12049 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12052 /* PL_asserting = proto_perl->Iasserting; */
12054 /* current interpreter roots */
12055 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12057 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12059 PL_main_start = proto_perl->Imain_start;
12060 PL_eval_root = proto_perl->Ieval_root;
12061 PL_eval_start = proto_perl->Ieval_start;
12063 /* runtime control stuff */
12064 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12066 PL_filemode = proto_perl->Ifilemode;
12067 PL_lastfd = proto_perl->Ilastfd;
12068 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12071 PL_gensym = proto_perl->Igensym;
12072 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12073 PL_laststatval = proto_perl->Ilaststatval;
12074 PL_laststype = proto_perl->Ilaststype;
12077 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12079 /* interpreter atexit processing */
12080 PL_exitlistlen = proto_perl->Iexitlistlen;
12081 if (PL_exitlistlen) {
12082 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12083 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12086 PL_exitlist = (PerlExitListEntry*)NULL;
12088 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12089 if (PL_my_cxt_size) {
12090 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12091 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12092 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12093 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12094 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12098 PL_my_cxt_list = (void**)NULL;
12099 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12100 PL_my_cxt_keys = (const char**)NULL;
12103 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12104 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12105 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12107 PL_profiledata = NULL;
12109 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12111 PAD_CLONE_VARS(proto_perl, param);
12113 #ifdef HAVE_INTERP_INTERN
12114 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12117 /* more statics moved here */
12118 PL_generation = proto_perl->Igeneration;
12119 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12121 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12122 PL_in_clean_all = proto_perl->Iin_clean_all;
12124 PL_uid = proto_perl->Iuid;
12125 PL_euid = proto_perl->Ieuid;
12126 PL_gid = proto_perl->Igid;
12127 PL_egid = proto_perl->Iegid;
12128 PL_nomemok = proto_perl->Inomemok;
12129 PL_an = proto_perl->Ian;
12130 PL_evalseq = proto_perl->Ievalseq;
12131 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12132 PL_origalen = proto_perl->Iorigalen;
12133 #ifdef PERL_USES_PL_PIDSTATUS
12134 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12136 PL_osname = SAVEPV(proto_perl->Iosname);
12137 PL_sighandlerp = proto_perl->Isighandlerp;
12139 PL_runops = proto_perl->Irunops;
12141 PL_parser = parser_dup(proto_perl->Iparser, param);
12143 /* XXX this only works if the saved cop has already been cloned */
12144 if (proto_perl->Iparser) {
12145 PL_parser->saved_curcop = (COP*)any_dup(
12146 proto_perl->Iparser->saved_curcop,
12150 PL_subline = proto_perl->Isubline;
12151 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12154 PL_cryptseen = proto_perl->Icryptseen;
12157 PL_hints = proto_perl->Ihints;
12159 PL_amagic_generation = proto_perl->Iamagic_generation;
12161 #ifdef USE_LOCALE_COLLATE
12162 PL_collation_ix = proto_perl->Icollation_ix;
12163 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12164 PL_collation_standard = proto_perl->Icollation_standard;
12165 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12166 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12167 #endif /* USE_LOCALE_COLLATE */
12169 #ifdef USE_LOCALE_NUMERIC
12170 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12171 PL_numeric_standard = proto_perl->Inumeric_standard;
12172 PL_numeric_local = proto_perl->Inumeric_local;
12173 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12174 #endif /* !USE_LOCALE_NUMERIC */
12176 /* utf8 character classes */
12177 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12178 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12179 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12180 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12181 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12182 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12183 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12184 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12185 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12186 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12187 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12188 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12189 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12190 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12191 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12192 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12193 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12194 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12195 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12197 /* Did the locale setup indicate UTF-8? */
12198 PL_utf8locale = proto_perl->Iutf8locale;
12199 /* Unicode features (see perlrun/-C) */
12200 PL_unicode = proto_perl->Iunicode;
12202 /* Pre-5.8 signals control */
12203 PL_signals = proto_perl->Isignals;
12205 /* times() ticks per second */
12206 PL_clocktick = proto_perl->Iclocktick;
12208 /* Recursion stopper for PerlIO_find_layer */
12209 PL_in_load_module = proto_perl->Iin_load_module;
12211 /* sort() routine */
12212 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12214 /* Not really needed/useful since the reenrant_retint is "volatile",
12215 * but do it for consistency's sake. */
12216 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12218 /* Hooks to shared SVs and locks. */
12219 PL_sharehook = proto_perl->Isharehook;
12220 PL_lockhook = proto_perl->Ilockhook;
12221 PL_unlockhook = proto_perl->Iunlockhook;
12222 PL_threadhook = proto_perl->Ithreadhook;
12223 PL_destroyhook = proto_perl->Idestroyhook;
12225 #ifdef THREADS_HAVE_PIDS
12226 PL_ppid = proto_perl->Ippid;
12230 PL_last_swash_hv = NULL; /* reinits on demand */
12231 PL_last_swash_klen = 0;
12232 PL_last_swash_key[0]= '\0';
12233 PL_last_swash_tmps = (U8*)NULL;
12234 PL_last_swash_slen = 0;
12236 PL_glob_index = proto_perl->Iglob_index;
12237 PL_srand_called = proto_perl->Isrand_called;
12239 if (proto_perl->Ipsig_pend) {
12240 Newxz(PL_psig_pend, SIG_SIZE, int);
12243 PL_psig_pend = (int*)NULL;
12246 if (proto_perl->Ipsig_name) {
12247 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12248 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12250 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12253 PL_psig_ptr = (SV**)NULL;
12254 PL_psig_name = (SV**)NULL;
12257 /* intrpvar.h stuff */
12259 if (flags & CLONEf_COPY_STACKS) {
12260 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12261 PL_tmps_ix = proto_perl->Itmps_ix;
12262 PL_tmps_max = proto_perl->Itmps_max;
12263 PL_tmps_floor = proto_perl->Itmps_floor;
12264 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12265 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12266 PL_tmps_ix+1, param);
12268 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12269 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12270 Newxz(PL_markstack, i, I32);
12271 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12272 - proto_perl->Imarkstack);
12273 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12274 - proto_perl->Imarkstack);
12275 Copy(proto_perl->Imarkstack, PL_markstack,
12276 PL_markstack_ptr - PL_markstack + 1, I32);
12278 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12279 * NOTE: unlike the others! */
12280 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12281 PL_scopestack_max = proto_perl->Iscopestack_max;
12282 Newxz(PL_scopestack, PL_scopestack_max, I32);
12283 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12286 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12287 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12289 /* NOTE: si_dup() looks at PL_markstack */
12290 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12292 /* PL_curstack = PL_curstackinfo->si_stack; */
12293 PL_curstack = av_dup(proto_perl->Icurstack, param);
12294 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12296 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12297 PL_stack_base = AvARRAY(PL_curstack);
12298 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12299 - proto_perl->Istack_base);
12300 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12302 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12303 * NOTE: unlike the others! */
12304 PL_savestack_ix = proto_perl->Isavestack_ix;
12305 PL_savestack_max = proto_perl->Isavestack_max;
12306 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12307 PL_savestack = ss_dup(proto_perl, param);
12311 ENTER; /* perl_destruct() wants to LEAVE; */
12313 /* although we're not duplicating the tmps stack, we should still
12314 * add entries for any SVs on the tmps stack that got cloned by a
12315 * non-refcount means (eg a temp in @_); otherwise they will be
12318 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12319 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12320 proto_perl->Itmps_stack[i]));
12321 if (nsv && !SvREFCNT(nsv)) {
12322 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12327 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12328 PL_top_env = &PL_start_env;
12330 PL_op = proto_perl->Iop;
12333 PL_Xpv = (XPV*)NULL;
12334 my_perl->Ina = proto_perl->Ina;
12336 PL_statbuf = proto_perl->Istatbuf;
12337 PL_statcache = proto_perl->Istatcache;
12338 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12339 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12341 PL_timesbuf = proto_perl->Itimesbuf;
12344 PL_tainted = proto_perl->Itainted;
12345 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12346 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12347 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12348 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12349 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12350 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12351 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12352 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12354 PL_restartop = proto_perl->Irestartop;
12355 PL_in_eval = proto_perl->Iin_eval;
12356 PL_delaymagic = proto_perl->Idelaymagic;
12357 PL_dirty = proto_perl->Idirty;
12358 PL_localizing = proto_perl->Ilocalizing;
12360 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12361 PL_hv_fetch_ent_mh = NULL;
12362 PL_modcount = proto_perl->Imodcount;
12363 PL_lastgotoprobe = NULL;
12364 PL_dumpindent = proto_perl->Idumpindent;
12366 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12367 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12368 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12369 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12370 PL_efloatbuf = NULL; /* reinits on demand */
12371 PL_efloatsize = 0; /* reinits on demand */
12375 PL_screamfirst = NULL;
12376 PL_screamnext = NULL;
12377 PL_maxscream = -1; /* reinits on demand */
12378 PL_lastscream = NULL;
12381 PL_regdummy = proto_perl->Iregdummy;
12382 PL_colorset = 0; /* reinits PL_colors[] */
12383 /*PL_colors[6] = {0,0,0,0,0,0};*/
12387 /* Pluggable optimizer */
12388 PL_peepp = proto_perl->Ipeepp;
12389 /* op_free() hook */
12390 PL_opfreehook = proto_perl->Iopfreehook;
12392 PL_stashcache = newHV();
12394 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12395 proto_perl->Iwatchaddr);
12396 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12397 if (PL_debug && PL_watchaddr) {
12398 PerlIO_printf(Perl_debug_log,
12399 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12400 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12401 PTR2UV(PL_watchok));
12404 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12406 /* Call the ->CLONE method, if it exists, for each of the stashes
12407 identified by sv_dup() above.
12409 while(av_len(param->stashes) != -1) {
12410 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12411 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12412 if (cloner && GvCV(cloner)) {
12417 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12419 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12425 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12426 ptr_table_free(PL_ptr_table);
12427 PL_ptr_table = NULL;
12431 SvREFCNT_dec(param->stashes);
12433 /* orphaned? eg threads->new inside BEGIN or use */
12434 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12435 SvREFCNT_inc_simple_void(PL_compcv);
12436 SAVEFREESV(PL_compcv);
12442 #endif /* USE_ITHREADS */
12445 =head1 Unicode Support
12447 =for apidoc sv_recode_to_utf8
12449 The encoding is assumed to be an Encode object, on entry the PV
12450 of the sv is assumed to be octets in that encoding, and the sv
12451 will be converted into Unicode (and UTF-8).
12453 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12454 is not a reference, nothing is done to the sv. If the encoding is not
12455 an C<Encode::XS> Encoding object, bad things will happen.
12456 (See F<lib/encoding.pm> and L<Encode>).
12458 The PV of the sv is returned.
12463 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12467 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12469 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12483 Passing sv_yes is wrong - it needs to be or'ed set of constants
12484 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12485 remove converted chars from source.
12487 Both will default the value - let them.
12489 XPUSHs(&PL_sv_yes);
12492 call_method("decode", G_SCALAR);
12496 s = SvPV_const(uni, len);
12497 if (s != SvPVX_const(sv)) {
12498 SvGROW(sv, len + 1);
12499 Move(s, SvPVX(sv), len + 1, char);
12500 SvCUR_set(sv, len);
12507 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12511 =for apidoc sv_cat_decode
12513 The encoding is assumed to be an Encode object, the PV of the ssv is
12514 assumed to be octets in that encoding and decoding the input starts
12515 from the position which (PV + *offset) pointed to. The dsv will be
12516 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12517 when the string tstr appears in decoding output or the input ends on
12518 the PV of the ssv. The value which the offset points will be modified
12519 to the last input position on the ssv.
12521 Returns TRUE if the terminator was found, else returns FALSE.
12526 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12527 SV *ssv, int *offset, char *tstr, int tlen)
12532 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12534 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12545 offsv = newSViv(*offset);
12547 mXPUSHp(tstr, tlen);
12549 call_method("cat_decode", G_SCALAR);
12551 ret = SvTRUE(TOPs);
12552 *offset = SvIV(offsv);
12558 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12563 /* ---------------------------------------------------------------------
12565 * support functions for report_uninit()
12568 /* the maxiumum size of array or hash where we will scan looking
12569 * for the undefined element that triggered the warning */
12571 #define FUV_MAX_SEARCH_SIZE 1000
12573 /* Look for an entry in the hash whose value has the same SV as val;
12574 * If so, return a mortal copy of the key. */
12577 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12580 register HE **array;
12583 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12585 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12586 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12589 array = HvARRAY(hv);
12591 for (i=HvMAX(hv); i>0; i--) {
12592 register HE *entry;
12593 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12594 if (HeVAL(entry) != val)
12596 if ( HeVAL(entry) == &PL_sv_undef ||
12597 HeVAL(entry) == &PL_sv_placeholder)
12601 if (HeKLEN(entry) == HEf_SVKEY)
12602 return sv_mortalcopy(HeKEY_sv(entry));
12603 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12609 /* Look for an entry in the array whose value has the same SV as val;
12610 * If so, return the index, otherwise return -1. */
12613 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12617 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12619 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12620 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12623 if (val != &PL_sv_undef) {
12624 SV ** const svp = AvARRAY(av);
12627 for (i=AvFILLp(av); i>=0; i--)
12634 /* S_varname(): return the name of a variable, optionally with a subscript.
12635 * If gv is non-zero, use the name of that global, along with gvtype (one
12636 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12637 * targ. Depending on the value of the subscript_type flag, return:
12640 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12641 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12642 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12643 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12646 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12647 const SV *const keyname, I32 aindex, int subscript_type)
12650 SV * const name = sv_newmortal();
12653 buffer[0] = gvtype;
12656 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12658 gv_fullname4(name, gv, buffer, 0);
12660 if ((unsigned int)SvPVX(name)[1] <= 26) {
12662 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12664 /* Swap the 1 unprintable control character for the 2 byte pretty
12665 version - ie substr($name, 1, 1) = $buffer; */
12666 sv_insert(name, 1, 1, buffer, 2);
12670 CV * const cv = find_runcv(NULL);
12674 if (!cv || !CvPADLIST(cv))
12676 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12677 sv = *av_fetch(av, targ, FALSE);
12678 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12681 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12682 SV * const sv = newSV(0);
12683 *SvPVX(name) = '$';
12684 Perl_sv_catpvf(aTHX_ name, "{%s}",
12685 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12688 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12689 *SvPVX(name) = '$';
12690 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12692 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12693 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12694 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12702 =for apidoc find_uninit_var
12704 Find the name of the undefined variable (if any) that caused the operator o
12705 to issue a "Use of uninitialized value" warning.
12706 If match is true, only return a name if it's value matches uninit_sv.
12707 So roughly speaking, if a unary operator (such as OP_COS) generates a
12708 warning, then following the direct child of the op may yield an
12709 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12710 other hand, with OP_ADD there are two branches to follow, so we only print
12711 the variable name if we get an exact match.
12713 The name is returned as a mortal SV.
12715 Assumes that PL_op is the op that originally triggered the error, and that
12716 PL_comppad/PL_curpad points to the currently executing pad.
12722 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12728 const OP *o, *o2, *kid;
12730 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12731 uninit_sv == &PL_sv_placeholder)))
12734 switch (obase->op_type) {
12741 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12742 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12745 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12747 if (pad) { /* @lex, %lex */
12748 sv = PAD_SVl(obase->op_targ);
12752 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12753 /* @global, %global */
12754 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12757 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12759 else /* @{expr}, %{expr} */
12760 return find_uninit_var(cUNOPx(obase)->op_first,
12764 /* attempt to find a match within the aggregate */
12766 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12768 subscript_type = FUV_SUBSCRIPT_HASH;
12771 index = find_array_subscript((const AV *)sv, uninit_sv);
12773 subscript_type = FUV_SUBSCRIPT_ARRAY;
12776 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12779 return varname(gv, hash ? '%' : '@', obase->op_targ,
12780 keysv, index, subscript_type);
12784 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12786 return varname(NULL, '$', obase->op_targ,
12787 NULL, 0, FUV_SUBSCRIPT_NONE);
12790 gv = cGVOPx_gv(obase);
12791 if (!gv || (match && GvSV(gv) != uninit_sv))
12793 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12796 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12799 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12800 if (!av || SvRMAGICAL(av))
12802 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12803 if (!svp || *svp != uninit_sv)
12806 return varname(NULL, '$', obase->op_targ,
12807 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12810 gv = cGVOPx_gv(obase);
12815 AV *const av = GvAV(gv);
12816 if (!av || SvRMAGICAL(av))
12818 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12819 if (!svp || *svp != uninit_sv)
12822 return varname(gv, '$', 0,
12823 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12828 o = cUNOPx(obase)->op_first;
12829 if (!o || o->op_type != OP_NULL ||
12830 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12832 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12836 if (PL_op == obase)
12837 /* $a[uninit_expr] or $h{uninit_expr} */
12838 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12841 o = cBINOPx(obase)->op_first;
12842 kid = cBINOPx(obase)->op_last;
12844 /* get the av or hv, and optionally the gv */
12846 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12847 sv = PAD_SV(o->op_targ);
12849 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12850 && cUNOPo->op_first->op_type == OP_GV)
12852 gv = cGVOPx_gv(cUNOPo->op_first);
12856 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12861 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12862 /* index is constant */
12866 if (obase->op_type == OP_HELEM) {
12867 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12868 if (!he || HeVAL(he) != uninit_sv)
12872 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12873 if (!svp || *svp != uninit_sv)
12877 if (obase->op_type == OP_HELEM)
12878 return varname(gv, '%', o->op_targ,
12879 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12881 return varname(gv, '@', o->op_targ, NULL,
12882 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12885 /* index is an expression;
12886 * attempt to find a match within the aggregate */
12887 if (obase->op_type == OP_HELEM) {
12888 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12890 return varname(gv, '%', o->op_targ,
12891 keysv, 0, FUV_SUBSCRIPT_HASH);
12895 = find_array_subscript((const AV *)sv, uninit_sv);
12897 return varname(gv, '@', o->op_targ,
12898 NULL, index, FUV_SUBSCRIPT_ARRAY);
12903 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12905 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12910 /* only examine RHS */
12911 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12914 o = cUNOPx(obase)->op_first;
12915 if (o->op_type == OP_PUSHMARK)
12918 if (!o->op_sibling) {
12919 /* one-arg version of open is highly magical */
12921 if (o->op_type == OP_GV) { /* open FOO; */
12923 if (match && GvSV(gv) != uninit_sv)
12925 return varname(gv, '$', 0,
12926 NULL, 0, FUV_SUBSCRIPT_NONE);
12928 /* other possibilities not handled are:
12929 * open $x; or open my $x; should return '${*$x}'
12930 * open expr; should return '$'.expr ideally
12936 /* ops where $_ may be an implicit arg */
12940 if ( !(obase->op_flags & OPf_STACKED)) {
12941 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12942 ? PAD_SVl(obase->op_targ)
12945 sv = sv_newmortal();
12946 sv_setpvs(sv, "$_");
12955 match = 1; /* print etc can return undef on defined args */
12956 /* skip filehandle as it can't produce 'undef' warning */
12957 o = cUNOPx(obase)->op_first;
12958 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12959 o = o->op_sibling->op_sibling;
12963 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12965 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12967 /* the following ops are capable of returning PL_sv_undef even for
12968 * defined arg(s) */
12987 case OP_GETPEERNAME:
13035 case OP_SMARTMATCH:
13044 /* XXX tmp hack: these two may call an XS sub, and currently
13045 XS subs don't have a SUB entry on the context stack, so CV and
13046 pad determination goes wrong, and BAD things happen. So, just
13047 don't try to determine the value under those circumstances.
13048 Need a better fix at dome point. DAPM 11/2007 */
13054 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13055 if (gv && GvSV(gv) == uninit_sv)
13056 return newSVpvs_flags("$.", SVs_TEMP);
13061 /* def-ness of rval pos() is independent of the def-ness of its arg */
13062 if ( !(obase->op_flags & OPf_MOD))
13067 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13068 return newSVpvs_flags("${$/}", SVs_TEMP);
13073 if (!(obase->op_flags & OPf_KIDS))
13075 o = cUNOPx(obase)->op_first;
13081 /* if all except one arg are constant, or have no side-effects,
13082 * or are optimized away, then it's unambiguous */
13084 for (kid=o; kid; kid = kid->op_sibling) {
13086 const OPCODE type = kid->op_type;
13087 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13088 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13089 || (type == OP_PUSHMARK)
13093 if (o2) { /* more than one found */
13100 return find_uninit_var(o2, uninit_sv, match);
13102 /* scan all args */
13104 sv = find_uninit_var(o, uninit_sv, 1);
13116 =for apidoc report_uninit
13118 Print appropriate "Use of uninitialized variable" warning
13124 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13128 SV* varname = NULL;
13130 varname = find_uninit_var(PL_op, uninit_sv,0);
13132 sv_insert(varname, 0, 0, " ", 1);
13134 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13135 varname ? SvPV_nolen_const(varname) : "",
13136 " in ", OP_DESC(PL_op));
13139 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13145 * c-indentation-style: bsd
13146 * c-basic-offset: 4
13147 * indent-tabs-mode: t
13150 * ex: set ts=8 sts=4 sw=4 noet: