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("FileHandle::", 0, SVt_PVHV);
1437 /* Clear the stashcache because a new IO could overrule a package
1439 hv_clear(PL_stashcache);
1441 /* unless exists($main::{FileHandle}) and
1442 defined(%main::FileHandle::) */
1443 if (!(iogv && GvHV(iogv) && HvARRAY(GvHV(iogv))))
1444 iogv = gv_fetchpvs("IO::Handle::", GV_ADD, SVt_PVHV);
1445 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1446 IoPAGE_LEN(sv) = 60;
1448 if (old_type < SVt_PV) {
1449 /* referant will be NULL unless the old type was SVt_IV emulating
1451 sv->sv_u.svu_rv = referant;
1455 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1456 (unsigned long)new_type);
1459 if (old_type_details->arena) {
1460 /* If there was an old body, then we need to free it.
1461 Note that there is an assumption that all bodies of types that
1462 can be upgraded came from arenas. Only the more complex non-
1463 upgradable types are allowed to be directly malloc()ed. */
1465 my_safefree(old_body);
1467 del_body((void*)((char*)old_body + old_type_details->offset),
1468 &PL_body_roots[old_type]);
1474 =for apidoc sv_backoff
1476 Remove any string offset. You should normally use the C<SvOOK_off> macro
1483 Perl_sv_backoff(pTHX_ register SV *const sv)
1486 const char * const s = SvPVX_const(sv);
1488 PERL_ARGS_ASSERT_SV_BACKOFF;
1489 PERL_UNUSED_CONTEXT;
1492 assert(SvTYPE(sv) != SVt_PVHV);
1493 assert(SvTYPE(sv) != SVt_PVAV);
1495 SvOOK_offset(sv, delta);
1497 SvLEN_set(sv, SvLEN(sv) + delta);
1498 SvPV_set(sv, SvPVX(sv) - delta);
1499 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1500 SvFLAGS(sv) &= ~SVf_OOK;
1507 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1508 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1509 Use the C<SvGROW> wrapper instead.
1515 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1519 PERL_ARGS_ASSERT_SV_GROW;
1521 if (PL_madskills && newlen >= 0x100000) {
1522 PerlIO_printf(Perl_debug_log,
1523 "Allocation too large: %"UVxf"\n", (UV)newlen);
1525 #ifdef HAS_64K_LIMIT
1526 if (newlen >= 0x10000) {
1527 PerlIO_printf(Perl_debug_log,
1528 "Allocation too large: %"UVxf"\n", (UV)newlen);
1531 #endif /* HAS_64K_LIMIT */
1534 if (SvTYPE(sv) < SVt_PV) {
1535 sv_upgrade(sv, SVt_PV);
1536 s = SvPVX_mutable(sv);
1538 else if (SvOOK(sv)) { /* pv is offset? */
1540 s = SvPVX_mutable(sv);
1541 if (newlen > SvLEN(sv))
1542 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1543 #ifdef HAS_64K_LIMIT
1544 if (newlen >= 0x10000)
1549 s = SvPVX_mutable(sv);
1551 if (newlen > SvLEN(sv)) { /* need more room? */
1552 #ifndef Perl_safesysmalloc_size
1553 newlen = PERL_STRLEN_ROUNDUP(newlen);
1555 if (SvLEN(sv) && s) {
1556 s = (char*)saferealloc(s, newlen);
1559 s = (char*)safemalloc(newlen);
1560 if (SvPVX_const(sv) && SvCUR(sv)) {
1561 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1565 #ifdef Perl_safesysmalloc_size
1566 /* Do this here, do it once, do it right, and then we will never get
1567 called back into sv_grow() unless there really is some growing
1569 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1571 SvLEN_set(sv, newlen);
1578 =for apidoc sv_setiv
1580 Copies an integer into the given SV, upgrading first if necessary.
1581 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1587 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1591 PERL_ARGS_ASSERT_SV_SETIV;
1593 SV_CHECK_THINKFIRST_COW_DROP(sv);
1594 switch (SvTYPE(sv)) {
1597 sv_upgrade(sv, SVt_IV);
1600 sv_upgrade(sv, SVt_PVIV);
1604 if (!isGV_with_GP(sv))
1611 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1615 (void)SvIOK_only(sv); /* validate number */
1621 =for apidoc sv_setiv_mg
1623 Like C<sv_setiv>, but also handles 'set' magic.
1629 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1631 PERL_ARGS_ASSERT_SV_SETIV_MG;
1638 =for apidoc sv_setuv
1640 Copies an unsigned integer into the given SV, upgrading first if necessary.
1641 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1647 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1649 PERL_ARGS_ASSERT_SV_SETUV;
1651 /* With these two if statements:
1652 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1655 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1657 If you wish to remove them, please benchmark to see what the effect is
1659 if (u <= (UV)IV_MAX) {
1660 sv_setiv(sv, (IV)u);
1669 =for apidoc sv_setuv_mg
1671 Like C<sv_setuv>, but also handles 'set' magic.
1677 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1679 PERL_ARGS_ASSERT_SV_SETUV_MG;
1686 =for apidoc sv_setnv
1688 Copies a double into the given SV, upgrading first if necessary.
1689 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1695 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1699 PERL_ARGS_ASSERT_SV_SETNV;
1701 SV_CHECK_THINKFIRST_COW_DROP(sv);
1702 switch (SvTYPE(sv)) {
1705 sv_upgrade(sv, SVt_NV);
1709 sv_upgrade(sv, SVt_PVNV);
1713 if (!isGV_with_GP(sv))
1720 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1725 (void)SvNOK_only(sv); /* validate number */
1730 =for apidoc sv_setnv_mg
1732 Like C<sv_setnv>, but also handles 'set' magic.
1738 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1740 PERL_ARGS_ASSERT_SV_SETNV_MG;
1746 /* Print an "isn't numeric" warning, using a cleaned-up,
1747 * printable version of the offending string
1751 S_not_a_number(pTHX_ SV *const sv)
1758 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1761 dsv = newSVpvs_flags("", SVs_TEMP);
1762 pv = sv_uni_display(dsv, sv, 10, 0);
1765 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1766 /* each *s can expand to 4 chars + "...\0",
1767 i.e. need room for 8 chars */
1769 const char *s = SvPVX_const(sv);
1770 const char * const end = s + SvCUR(sv);
1771 for ( ; s < end && d < limit; s++ ) {
1773 if (ch & 128 && !isPRINT_LC(ch)) {
1782 else if (ch == '\r') {
1786 else if (ch == '\f') {
1790 else if (ch == '\\') {
1794 else if (ch == '\0') {
1798 else if (isPRINT_LC(ch))
1815 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1816 "Argument \"%s\" isn't numeric in %s", pv,
1819 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1820 "Argument \"%s\" isn't numeric", pv);
1824 =for apidoc looks_like_number
1826 Test if the content of an SV looks like a number (or is a number).
1827 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1828 non-numeric warning), even if your atof() doesn't grok them.
1834 Perl_looks_like_number(pTHX_ SV *const sv)
1836 register const char *sbegin;
1839 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1842 sbegin = SvPVX_const(sv);
1845 else if (SvPOKp(sv))
1846 sbegin = SvPV_const(sv, len);
1848 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1849 return grok_number(sbegin, len, NULL);
1853 S_glob_2number(pTHX_ GV * const gv)
1855 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1856 SV *const buffer = sv_newmortal();
1858 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1860 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1863 gv_efullname3(buffer, gv, "*");
1864 SvFLAGS(gv) |= wasfake;
1866 /* We know that all GVs stringify to something that is not-a-number,
1867 so no need to test that. */
1868 if (ckWARN(WARN_NUMERIC))
1869 not_a_number(buffer);
1870 /* We just want something true to return, so that S_sv_2iuv_common
1871 can tail call us and return true. */
1875 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1876 until proven guilty, assume that things are not that bad... */
1881 As 64 bit platforms often have an NV that doesn't preserve all bits of
1882 an IV (an assumption perl has been based on to date) it becomes necessary
1883 to remove the assumption that the NV always carries enough precision to
1884 recreate the IV whenever needed, and that the NV is the canonical form.
1885 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1886 precision as a side effect of conversion (which would lead to insanity
1887 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1888 1) to distinguish between IV/UV/NV slots that have cached a valid
1889 conversion where precision was lost and IV/UV/NV slots that have a
1890 valid conversion which has lost no precision
1891 2) to ensure that if a numeric conversion to one form is requested that
1892 would lose precision, the precise conversion (or differently
1893 imprecise conversion) is also performed and cached, to prevent
1894 requests for different numeric formats on the same SV causing
1895 lossy conversion chains. (lossless conversion chains are perfectly
1900 SvIOKp is true if the IV slot contains a valid value
1901 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1902 SvNOKp is true if the NV slot contains a valid value
1903 SvNOK is true only if the NV value is accurate
1906 while converting from PV to NV, check to see if converting that NV to an
1907 IV(or UV) would lose accuracy over a direct conversion from PV to
1908 IV(or UV). If it would, cache both conversions, return NV, but mark
1909 SV as IOK NOKp (ie not NOK).
1911 While converting from PV to IV, check to see if converting that IV to an
1912 NV would lose accuracy over a direct conversion from PV to NV. If it
1913 would, cache both conversions, flag similarly.
1915 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1916 correctly because if IV & NV were set NV *always* overruled.
1917 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1918 changes - now IV and NV together means that the two are interchangeable:
1919 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1921 The benefit of this is that operations such as pp_add know that if
1922 SvIOK is true for both left and right operands, then integer addition
1923 can be used instead of floating point (for cases where the result won't
1924 overflow). Before, floating point was always used, which could lead to
1925 loss of precision compared with integer addition.
1927 * making IV and NV equal status should make maths accurate on 64 bit
1929 * may speed up maths somewhat if pp_add and friends start to use
1930 integers when possible instead of fp. (Hopefully the overhead in
1931 looking for SvIOK and checking for overflow will not outweigh the
1932 fp to integer speedup)
1933 * will slow down integer operations (callers of SvIV) on "inaccurate"
1934 values, as the change from SvIOK to SvIOKp will cause a call into
1935 sv_2iv each time rather than a macro access direct to the IV slot
1936 * should speed up number->string conversion on integers as IV is
1937 favoured when IV and NV are equally accurate
1939 ####################################################################
1940 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1941 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1942 On the other hand, SvUOK is true iff UV.
1943 ####################################################################
1945 Your mileage will vary depending your CPU's relative fp to integer
1949 #ifndef NV_PRESERVES_UV
1950 # define IS_NUMBER_UNDERFLOW_IV 1
1951 # define IS_NUMBER_UNDERFLOW_UV 2
1952 # define IS_NUMBER_IV_AND_UV 2
1953 # define IS_NUMBER_OVERFLOW_IV 4
1954 # define IS_NUMBER_OVERFLOW_UV 5
1956 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1958 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1960 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1968 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1970 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));
1971 if (SvNVX(sv) < (NV)IV_MIN) {
1972 (void)SvIOKp_on(sv);
1974 SvIV_set(sv, IV_MIN);
1975 return IS_NUMBER_UNDERFLOW_IV;
1977 if (SvNVX(sv) > (NV)UV_MAX) {
1978 (void)SvIOKp_on(sv);
1981 SvUV_set(sv, UV_MAX);
1982 return IS_NUMBER_OVERFLOW_UV;
1984 (void)SvIOKp_on(sv);
1986 /* Can't use strtol etc to convert this string. (See truth table in
1988 if (SvNVX(sv) <= (UV)IV_MAX) {
1989 SvIV_set(sv, I_V(SvNVX(sv)));
1990 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1991 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1993 /* Integer is imprecise. NOK, IOKp */
1995 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1998 SvUV_set(sv, U_V(SvNVX(sv)));
1999 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2000 if (SvUVX(sv) == UV_MAX) {
2001 /* As we know that NVs don't preserve UVs, UV_MAX cannot
2002 possibly be preserved by NV. Hence, it must be overflow.
2004 return IS_NUMBER_OVERFLOW_UV;
2006 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2008 /* Integer is imprecise. NOK, IOKp */
2010 return IS_NUMBER_OVERFLOW_IV;
2012 #endif /* !NV_PRESERVES_UV*/
2015 S_sv_2iuv_common(pTHX_ SV *const sv)
2019 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2022 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2023 * without also getting a cached IV/UV from it at the same time
2024 * (ie PV->NV conversion should detect loss of accuracy and cache
2025 * IV or UV at same time to avoid this. */
2026 /* IV-over-UV optimisation - choose to cache IV if possible */
2028 if (SvTYPE(sv) == SVt_NV)
2029 sv_upgrade(sv, SVt_PVNV);
2031 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2032 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2033 certainly cast into the IV range at IV_MAX, whereas the correct
2034 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2036 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2037 if (Perl_isnan(SvNVX(sv))) {
2043 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2044 SvIV_set(sv, I_V(SvNVX(sv)));
2045 if (SvNVX(sv) == (NV) SvIVX(sv)
2046 #ifndef NV_PRESERVES_UV
2047 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2048 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2049 /* Don't flag it as "accurately an integer" if the number
2050 came from a (by definition imprecise) NV operation, and
2051 we're outside the range of NV integer precision */
2055 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2057 /* scalar has trailing garbage, eg "42a" */
2059 DEBUG_c(PerlIO_printf(Perl_debug_log,
2060 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2066 /* IV not precise. No need to convert from PV, as NV
2067 conversion would already have cached IV if it detected
2068 that PV->IV would be better than PV->NV->IV
2069 flags already correct - don't set public IOK. */
2070 DEBUG_c(PerlIO_printf(Perl_debug_log,
2071 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2076 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2077 but the cast (NV)IV_MIN rounds to a the value less (more
2078 negative) than IV_MIN which happens to be equal to SvNVX ??
2079 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2080 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2081 (NV)UVX == NVX are both true, but the values differ. :-(
2082 Hopefully for 2s complement IV_MIN is something like
2083 0x8000000000000000 which will be exact. NWC */
2086 SvUV_set(sv, U_V(SvNVX(sv)));
2088 (SvNVX(sv) == (NV) SvUVX(sv))
2089 #ifndef NV_PRESERVES_UV
2090 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2091 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2092 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2093 /* Don't flag it as "accurately an integer" if the number
2094 came from a (by definition imprecise) NV operation, and
2095 we're outside the range of NV integer precision */
2101 DEBUG_c(PerlIO_printf(Perl_debug_log,
2102 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2108 else if (SvPOKp(sv) && SvLEN(sv)) {
2110 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2111 /* We want to avoid a possible problem when we cache an IV/ a UV which
2112 may be later translated to an NV, and the resulting NV is not
2113 the same as the direct translation of the initial string
2114 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2115 be careful to ensure that the value with the .456 is around if the
2116 NV value is requested in the future).
2118 This means that if we cache such an IV/a UV, we need to cache the
2119 NV as well. Moreover, we trade speed for space, and do not
2120 cache the NV if we are sure it's not needed.
2123 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2124 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2125 == IS_NUMBER_IN_UV) {
2126 /* It's definitely an integer, only upgrade to PVIV */
2127 if (SvTYPE(sv) < SVt_PVIV)
2128 sv_upgrade(sv, SVt_PVIV);
2130 } else if (SvTYPE(sv) < SVt_PVNV)
2131 sv_upgrade(sv, SVt_PVNV);
2133 /* If NVs preserve UVs then we only use the UV value if we know that
2134 we aren't going to call atof() below. If NVs don't preserve UVs
2135 then the value returned may have more precision than atof() will
2136 return, even though value isn't perfectly accurate. */
2137 if ((numtype & (IS_NUMBER_IN_UV
2138 #ifdef NV_PRESERVES_UV
2141 )) == IS_NUMBER_IN_UV) {
2142 /* This won't turn off the public IOK flag if it was set above */
2143 (void)SvIOKp_on(sv);
2145 if (!(numtype & IS_NUMBER_NEG)) {
2147 if (value <= (UV)IV_MAX) {
2148 SvIV_set(sv, (IV)value);
2150 /* it didn't overflow, and it was positive. */
2151 SvUV_set(sv, value);
2155 /* 2s complement assumption */
2156 if (value <= (UV)IV_MIN) {
2157 SvIV_set(sv, -(IV)value);
2159 /* Too negative for an IV. This is a double upgrade, but
2160 I'm assuming it will be rare. */
2161 if (SvTYPE(sv) < SVt_PVNV)
2162 sv_upgrade(sv, SVt_PVNV);
2166 SvNV_set(sv, -(NV)value);
2167 SvIV_set(sv, IV_MIN);
2171 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2172 will be in the previous block to set the IV slot, and the next
2173 block to set the NV slot. So no else here. */
2175 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2176 != IS_NUMBER_IN_UV) {
2177 /* It wasn't an (integer that doesn't overflow the UV). */
2178 SvNV_set(sv, Atof(SvPVX_const(sv)));
2180 if (! numtype && ckWARN(WARN_NUMERIC))
2183 #if defined(USE_LONG_DOUBLE)
2184 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2185 PTR2UV(sv), SvNVX(sv)));
2187 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2188 PTR2UV(sv), SvNVX(sv)));
2191 #ifdef NV_PRESERVES_UV
2192 (void)SvIOKp_on(sv);
2194 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2195 SvIV_set(sv, I_V(SvNVX(sv)));
2196 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2199 NOOP; /* Integer is imprecise. NOK, IOKp */
2201 /* UV will not work better than IV */
2203 if (SvNVX(sv) > (NV)UV_MAX) {
2205 /* Integer is inaccurate. NOK, IOKp, is UV */
2206 SvUV_set(sv, UV_MAX);
2208 SvUV_set(sv, U_V(SvNVX(sv)));
2209 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2210 NV preservse UV so can do correct comparison. */
2211 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2214 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2219 #else /* NV_PRESERVES_UV */
2220 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2221 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2222 /* The IV/UV slot will have been set from value returned by
2223 grok_number above. The NV slot has just been set using
2226 assert (SvIOKp(sv));
2228 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2229 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2230 /* Small enough to preserve all bits. */
2231 (void)SvIOKp_on(sv);
2233 SvIV_set(sv, I_V(SvNVX(sv)));
2234 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2236 /* Assumption: first non-preserved integer is < IV_MAX,
2237 this NV is in the preserved range, therefore: */
2238 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2240 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);
2244 0 0 already failed to read UV.
2245 0 1 already failed to read UV.
2246 1 0 you won't get here in this case. IV/UV
2247 slot set, public IOK, Atof() unneeded.
2248 1 1 already read UV.
2249 so there's no point in sv_2iuv_non_preserve() attempting
2250 to use atol, strtol, strtoul etc. */
2252 sv_2iuv_non_preserve (sv, numtype);
2254 sv_2iuv_non_preserve (sv);
2258 #endif /* NV_PRESERVES_UV */
2259 /* It might be more code efficient to go through the entire logic above
2260 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2261 gets complex and potentially buggy, so more programmer efficient
2262 to do it this way, by turning off the public flags: */
2264 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2268 if (isGV_with_GP(sv))
2269 return glob_2number(MUTABLE_GV(sv));
2271 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2272 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2275 if (SvTYPE(sv) < SVt_IV)
2276 /* Typically the caller expects that sv_any is not NULL now. */
2277 sv_upgrade(sv, SVt_IV);
2278 /* Return 0 from the caller. */
2285 =for apidoc sv_2iv_flags
2287 Return the integer value of an SV, doing any necessary string
2288 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2289 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2295 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2300 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2301 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2302 cache IVs just in case. In practice it seems that they never
2303 actually anywhere accessible by user Perl code, let alone get used
2304 in anything other than a string context. */
2305 if (flags & SV_GMAGIC)
2310 return I_V(SvNVX(sv));
2312 if (SvPOKp(sv) && SvLEN(sv)) {
2315 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2317 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2318 == IS_NUMBER_IN_UV) {
2319 /* It's definitely an integer */
2320 if (numtype & IS_NUMBER_NEG) {
2321 if (value < (UV)IV_MIN)
2324 if (value < (UV)IV_MAX)
2329 if (ckWARN(WARN_NUMERIC))
2332 return I_V(Atof(SvPVX_const(sv)));
2337 assert(SvTYPE(sv) >= SVt_PVMG);
2338 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2339 } else if (SvTHINKFIRST(sv)) {
2343 SV * const tmpstr=AMG_CALLun(sv,numer);
2344 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2345 return SvIV(tmpstr);
2348 return PTR2IV(SvRV(sv));
2351 sv_force_normal_flags(sv, 0);
2353 if (SvREADONLY(sv) && !SvOK(sv)) {
2354 if (ckWARN(WARN_UNINITIALIZED))
2360 if (S_sv_2iuv_common(aTHX_ sv))
2363 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2364 PTR2UV(sv),SvIVX(sv)));
2365 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2369 =for apidoc sv_2uv_flags
2371 Return the unsigned integer value of an SV, doing any necessary string
2372 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2373 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2379 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2384 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2385 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2386 cache IVs just in case. */
2387 if (flags & SV_GMAGIC)
2392 return U_V(SvNVX(sv));
2393 if (SvPOKp(sv) && SvLEN(sv)) {
2396 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2398 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2399 == IS_NUMBER_IN_UV) {
2400 /* It's definitely an integer */
2401 if (!(numtype & IS_NUMBER_NEG))
2405 if (ckWARN(WARN_NUMERIC))
2408 return U_V(Atof(SvPVX_const(sv)));
2413 assert(SvTYPE(sv) >= SVt_PVMG);
2414 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2415 } else if (SvTHINKFIRST(sv)) {
2419 SV *const tmpstr = AMG_CALLun(sv,numer);
2420 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2421 return SvUV(tmpstr);
2424 return PTR2UV(SvRV(sv));
2427 sv_force_normal_flags(sv, 0);
2429 if (SvREADONLY(sv) && !SvOK(sv)) {
2430 if (ckWARN(WARN_UNINITIALIZED))
2436 if (S_sv_2iuv_common(aTHX_ sv))
2440 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2441 PTR2UV(sv),SvUVX(sv)));
2442 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2448 Return the num value of an SV, doing any necessary string or integer
2449 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2456 Perl_sv_2nv(pTHX_ register SV *const sv)
2461 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2462 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2463 cache IVs just in case. */
2467 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2468 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2469 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2471 return Atof(SvPVX_const(sv));
2475 return (NV)SvUVX(sv);
2477 return (NV)SvIVX(sv);
2482 assert(SvTYPE(sv) >= SVt_PVMG);
2483 /* This falls through to the report_uninit near the end of the
2485 } else if (SvTHINKFIRST(sv)) {
2489 SV *const tmpstr = AMG_CALLun(sv,numer);
2490 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2491 return SvNV(tmpstr);
2494 return PTR2NV(SvRV(sv));
2497 sv_force_normal_flags(sv, 0);
2499 if (SvREADONLY(sv) && !SvOK(sv)) {
2500 if (ckWARN(WARN_UNINITIALIZED))
2505 if (SvTYPE(sv) < SVt_NV) {
2506 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2507 sv_upgrade(sv, SVt_NV);
2508 #ifdef USE_LONG_DOUBLE
2510 STORE_NUMERIC_LOCAL_SET_STANDARD();
2511 PerlIO_printf(Perl_debug_log,
2512 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2513 PTR2UV(sv), SvNVX(sv));
2514 RESTORE_NUMERIC_LOCAL();
2518 STORE_NUMERIC_LOCAL_SET_STANDARD();
2519 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2520 PTR2UV(sv), SvNVX(sv));
2521 RESTORE_NUMERIC_LOCAL();
2525 else if (SvTYPE(sv) < SVt_PVNV)
2526 sv_upgrade(sv, SVt_PVNV);
2531 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2532 #ifdef NV_PRESERVES_UV
2538 /* Only set the public NV OK flag if this NV preserves the IV */
2539 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2541 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2542 : (SvIVX(sv) == I_V(SvNVX(sv))))
2548 else if (SvPOKp(sv) && SvLEN(sv)) {
2550 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2551 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2553 #ifdef NV_PRESERVES_UV
2554 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2555 == IS_NUMBER_IN_UV) {
2556 /* It's definitely an integer */
2557 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2559 SvNV_set(sv, Atof(SvPVX_const(sv)));
2565 SvNV_set(sv, Atof(SvPVX_const(sv)));
2566 /* Only set the public NV OK flag if this NV preserves the value in
2567 the PV at least as well as an IV/UV would.
2568 Not sure how to do this 100% reliably. */
2569 /* if that shift count is out of range then Configure's test is
2570 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2572 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2573 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2574 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2575 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2576 /* Can't use strtol etc to convert this string, so don't try.
2577 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2580 /* value has been set. It may not be precise. */
2581 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2582 /* 2s complement assumption for (UV)IV_MIN */
2583 SvNOK_on(sv); /* Integer is too negative. */
2588 if (numtype & IS_NUMBER_NEG) {
2589 SvIV_set(sv, -(IV)value);
2590 } else if (value <= (UV)IV_MAX) {
2591 SvIV_set(sv, (IV)value);
2593 SvUV_set(sv, value);
2597 if (numtype & IS_NUMBER_NOT_INT) {
2598 /* I believe that even if the original PV had decimals,
2599 they are lost beyond the limit of the FP precision.
2600 However, neither is canonical, so both only get p
2601 flags. NWC, 2000/11/25 */
2602 /* Both already have p flags, so do nothing */
2604 const NV nv = SvNVX(sv);
2605 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2606 if (SvIVX(sv) == I_V(nv)) {
2609 /* It had no "." so it must be integer. */
2613 /* between IV_MAX and NV(UV_MAX).
2614 Could be slightly > UV_MAX */
2616 if (numtype & IS_NUMBER_NOT_INT) {
2617 /* UV and NV both imprecise. */
2619 const UV nv_as_uv = U_V(nv);
2621 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2630 /* It might be more code efficient to go through the entire logic above
2631 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2632 gets complex and potentially buggy, so more programmer efficient
2633 to do it this way, by turning off the public flags: */
2635 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2636 #endif /* NV_PRESERVES_UV */
2639 if (isGV_with_GP(sv)) {
2640 glob_2number(MUTABLE_GV(sv));
2644 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2646 assert (SvTYPE(sv) >= SVt_NV);
2647 /* Typically the caller expects that sv_any is not NULL now. */
2648 /* XXX Ilya implies that this is a bug in callers that assume this
2649 and ideally should be fixed. */
2652 #if defined(USE_LONG_DOUBLE)
2654 STORE_NUMERIC_LOCAL_SET_STANDARD();
2655 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2656 PTR2UV(sv), SvNVX(sv));
2657 RESTORE_NUMERIC_LOCAL();
2661 STORE_NUMERIC_LOCAL_SET_STANDARD();
2662 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2663 PTR2UV(sv), SvNVX(sv));
2664 RESTORE_NUMERIC_LOCAL();
2673 Return an SV with the numeric value of the source SV, doing any necessary
2674 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2675 access this function.
2681 Perl_sv_2num(pTHX_ register SV *const sv)
2683 PERL_ARGS_ASSERT_SV_2NUM;
2688 SV * const tmpsv = AMG_CALLun(sv,numer);
2689 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2690 return sv_2num(tmpsv);
2692 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2695 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2696 * UV as a string towards the end of buf, and return pointers to start and
2699 * We assume that buf is at least TYPE_CHARS(UV) long.
2703 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2705 char *ptr = buf + TYPE_CHARS(UV);
2706 char * const ebuf = ptr;
2709 PERL_ARGS_ASSERT_UIV_2BUF;
2721 *--ptr = '0' + (char)(uv % 10);
2730 =for apidoc sv_2pv_flags
2732 Returns a pointer to the string value of an SV, and sets *lp to its length.
2733 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2735 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2736 usually end up here too.
2742 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2752 if (SvGMAGICAL(sv)) {
2753 if (flags & SV_GMAGIC)
2758 if (flags & SV_MUTABLE_RETURN)
2759 return SvPVX_mutable(sv);
2760 if (flags & SV_CONST_RETURN)
2761 return (char *)SvPVX_const(sv);
2764 if (SvIOKp(sv) || SvNOKp(sv)) {
2765 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2770 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2771 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2773 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2780 #ifdef FIXNEGATIVEZERO
2781 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2787 SvUPGRADE(sv, SVt_PV);
2790 s = SvGROW_mutable(sv, len + 1);
2793 return (char*)memcpy(s, tbuf, len + 1);
2799 assert(SvTYPE(sv) >= SVt_PVMG);
2800 /* This falls through to the report_uninit near the end of the
2802 } else if (SvTHINKFIRST(sv)) {
2806 SV *const tmpstr = AMG_CALLun(sv,string);
2807 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2809 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2813 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2814 if (flags & SV_CONST_RETURN) {
2815 pv = (char *) SvPVX_const(tmpstr);
2817 pv = (flags & SV_MUTABLE_RETURN)
2818 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2821 *lp = SvCUR(tmpstr);
2823 pv = sv_2pv_flags(tmpstr, lp, flags);
2836 SV *const referent = SvRV(sv);
2840 retval = buffer = savepvn("NULLREF", len);
2841 } else if (SvTYPE(referent) == SVt_REGEXP) {
2842 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2847 /* If the regex is UTF-8 we want the containing scalar to
2848 have an UTF-8 flag too */
2854 if ((seen_evals = RX_SEEN_EVALS(re)))
2855 PL_reginterp_cnt += seen_evals;
2858 *lp = RX_WRAPLEN(re);
2860 return RX_WRAPPED(re);
2862 const char *const typestr = sv_reftype(referent, 0);
2863 const STRLEN typelen = strlen(typestr);
2864 UV addr = PTR2UV(referent);
2865 const char *stashname = NULL;
2866 STRLEN stashnamelen = 0; /* hush, gcc */
2867 const char *buffer_end;
2869 if (SvOBJECT(referent)) {
2870 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2873 stashname = HEK_KEY(name);
2874 stashnamelen = HEK_LEN(name);
2876 if (HEK_UTF8(name)) {
2882 stashname = "__ANON__";
2885 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2886 + 2 * sizeof(UV) + 2 /* )\0 */;
2888 len = typelen + 3 /* (0x */
2889 + 2 * sizeof(UV) + 2 /* )\0 */;
2892 Newx(buffer, len, char);
2893 buffer_end = retval = buffer + len;
2895 /* Working backwards */
2899 *--retval = PL_hexdigit[addr & 15];
2900 } while (addr >>= 4);
2906 memcpy(retval, typestr, typelen);
2910 retval -= stashnamelen;
2911 memcpy(retval, stashname, stashnamelen);
2913 /* retval may not neccesarily have reached the start of the
2915 assert (retval >= buffer);
2917 len = buffer_end - retval - 1; /* -1 for that \0 */
2925 if (SvREADONLY(sv) && !SvOK(sv)) {
2928 if (flags & SV_UNDEF_RETURNS_NULL)
2930 if (ckWARN(WARN_UNINITIALIZED))
2935 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2936 /* I'm assuming that if both IV and NV are equally valid then
2937 converting the IV is going to be more efficient */
2938 const U32 isUIOK = SvIsUV(sv);
2939 char buf[TYPE_CHARS(UV)];
2943 if (SvTYPE(sv) < SVt_PVIV)
2944 sv_upgrade(sv, SVt_PVIV);
2945 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2947 /* inlined from sv_setpvn */
2948 s = SvGROW_mutable(sv, len + 1);
2949 Move(ptr, s, len, char);
2953 else if (SvNOKp(sv)) {
2955 if (SvTYPE(sv) < SVt_PVNV)
2956 sv_upgrade(sv, SVt_PVNV);
2957 /* The +20 is pure guesswork. Configure test needed. --jhi */
2958 s = SvGROW_mutable(sv, NV_DIG + 20);
2959 /* some Xenix systems wipe out errno here */
2961 if (SvNVX(sv) == 0.0)
2962 my_strlcpy(s, "0", SvLEN(sv));
2966 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2969 #ifdef FIXNEGATIVEZERO
2970 if (*s == '-' && s[1] == '0' && !s[2]) {
2982 if (isGV_with_GP(sv)) {
2983 GV *const gv = MUTABLE_GV(sv);
2984 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2985 SV *const buffer = sv_newmortal();
2987 /* FAKE globs can get coerced, so need to turn this off temporarily
2990 gv_efullname3(buffer, gv, "*");
2991 SvFLAGS(gv) |= wasfake;
2993 assert(SvPOK(buffer));
2995 *lp = SvCUR(buffer);
2997 return SvPVX(buffer);
3002 if (flags & SV_UNDEF_RETURNS_NULL)
3004 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3006 if (SvTYPE(sv) < SVt_PV)
3007 /* Typically the caller expects that sv_any is not NULL now. */
3008 sv_upgrade(sv, SVt_PV);
3012 const STRLEN len = s - SvPVX_const(sv);
3018 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3019 PTR2UV(sv),SvPVX_const(sv)));
3020 if (flags & SV_CONST_RETURN)
3021 return (char *)SvPVX_const(sv);
3022 if (flags & SV_MUTABLE_RETURN)
3023 return SvPVX_mutable(sv);
3028 =for apidoc sv_copypv
3030 Copies a stringified representation of the source SV into the
3031 destination SV. Automatically performs any necessary mg_get and
3032 coercion of numeric values into strings. Guaranteed to preserve
3033 UTF8 flag even from overloaded objects. Similar in nature to
3034 sv_2pv[_flags] but operates directly on an SV instead of just the
3035 string. Mostly uses sv_2pv_flags to do its work, except when that
3036 would lose the UTF-8'ness of the PV.
3042 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3045 const char * const s = SvPV_const(ssv,len);
3047 PERL_ARGS_ASSERT_SV_COPYPV;
3049 sv_setpvn(dsv,s,len);
3057 =for apidoc sv_2pvbyte
3059 Return a pointer to the byte-encoded representation of the SV, and set *lp
3060 to its length. May cause the SV to be downgraded from UTF-8 as a
3063 Usually accessed via the C<SvPVbyte> macro.
3069 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3071 PERL_ARGS_ASSERT_SV_2PVBYTE;
3073 sv_utf8_downgrade(sv,0);
3074 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3078 =for apidoc sv_2pvutf8
3080 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3081 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3083 Usually accessed via the C<SvPVutf8> macro.
3089 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3091 PERL_ARGS_ASSERT_SV_2PVUTF8;
3093 sv_utf8_upgrade(sv);
3094 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3099 =for apidoc sv_2bool
3101 This function is only called on magical items, and is only used by
3102 sv_true() or its macro equivalent.
3108 Perl_sv_2bool(pTHX_ register SV *const sv)
3112 PERL_ARGS_ASSERT_SV_2BOOL;
3120 SV * const tmpsv = AMG_CALLun(sv,bool_);
3121 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3122 return (bool)SvTRUE(tmpsv);
3124 return SvRV(sv) != 0;
3127 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3129 (*sv->sv_u.svu_pv > '0' ||
3130 Xpvtmp->xpv_cur > 1 ||
3131 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3138 return SvIVX(sv) != 0;
3141 return SvNVX(sv) != 0.0;
3143 if (isGV_with_GP(sv))
3153 =for apidoc sv_utf8_upgrade
3155 Converts the PV of an SV to its UTF-8-encoded form.
3156 Forces the SV to string form if it is not already.
3157 Will C<mg_get> on C<sv> if appropriate.
3158 Always sets the SvUTF8 flag to avoid future validity checks even
3159 if the whole string is the same in UTF-8 as not.
3160 Returns the number of bytes in the converted string
3162 This is not as a general purpose byte encoding to Unicode interface:
3163 use the Encode extension for that.
3165 =for apidoc sv_utf8_upgrade_nomg
3167 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3169 =for apidoc sv_utf8_upgrade_flags
3171 Converts the PV of an SV to its UTF-8-encoded form.
3172 Forces the SV to string form if it is not already.
3173 Always sets the SvUTF8 flag to avoid future validity checks even
3174 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3175 will C<mg_get> on C<sv> if appropriate, else not.
3176 Returns the number of bytes in the converted string
3177 C<sv_utf8_upgrade> and
3178 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3180 This is not as a general purpose byte encoding to Unicode interface:
3181 use the Encode extension for that.
3185 The grow version is currently not externally documented. It adds a parameter,
3186 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3187 have free after it upon return. This allows the caller to reserve extra space
3188 that it intends to fill, to avoid extra grows.
3190 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3191 which can be used to tell this function to not first check to see if there are
3192 any characters that are different in UTF-8 (variant characters) which would
3193 force it to allocate a new string to sv, but to assume there are. Typically
3194 this flag is used by a routine that has already parsed the string to find that
3195 there are such characters, and passes this information on so that the work
3196 doesn't have to be repeated.
3198 (One might think that the calling routine could pass in the position of the
3199 first such variant, so it wouldn't have to be found again. But that is not the
3200 case, because typically when the caller is likely to use this flag, it won't be
3201 calling this routine unless it finds something that won't fit into a byte.
3202 Otherwise it tries to not upgrade and just use bytes. But some things that
3203 do fit into a byte are variants in utf8, and the caller may not have been
3204 keeping track of these.)
3206 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3207 isn't guaranteed due to having other routines do the work in some input cases,
3208 or if the input is already flagged as being in utf8.
3210 The speed of this could perhaps be improved for many cases if someone wanted to
3211 write a fast function that counts the number of variant characters in a string,
3212 especially if it could return the position of the first one.
3217 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3221 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3223 if (sv == &PL_sv_undef)
3227 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3228 (void) sv_2pv_flags(sv,&len, flags);
3230 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3234 (void) SvPV_force(sv,len);
3239 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3244 sv_force_normal_flags(sv, 0);
3247 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3248 sv_recode_to_utf8(sv, PL_encoding);
3249 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3253 if (SvCUR(sv) > 0) { /* Assume Latin-1/EBCDIC */
3254 /* This function could be much more efficient if we
3255 * had a FLAG in SVs to signal if there are any variant
3256 * chars in the PV. Given that there isn't such a flag
3257 * make the loop as fast as possible (although there are certainly ways
3258 * to speed this up, eg. through vectorization) */
3259 U8 * s = (U8 *) SvPVX_const(sv);
3260 U8 * e = (U8 *) SvEND(sv);
3262 STRLEN two_byte_count = 0;
3264 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3266 /* See if really will need to convert to utf8. We mustn't rely on our
3267 * incoming SV being well formed and having a trailing '\0', as certain
3268 * code in pp_formline can send us partially built SVs. */
3272 if (NATIVE_IS_INVARIANT(ch)) continue;
3274 t--; /* t already incremented; re-point to first variant */
3279 /* utf8 conversion not needed because all are invariants. Mark as
3280 * UTF-8 even if no variant - saves scanning loop */
3286 /* Here, the string should be converted to utf8, either because of an
3287 * input flag (two_byte_count = 0), or because a character that
3288 * requires 2 bytes was found (two_byte_count = 1). t points either to
3289 * the beginning of the string (if we didn't examine anything), or to
3290 * the first variant. In either case, everything from s to t - 1 will
3291 * occupy only 1 byte each on output.
3293 * There are two main ways to convert. One is to create a new string
3294 * and go through the input starting from the beginning, appending each
3295 * converted value onto the new string as we go along. It's probably
3296 * best to allocate enough space in the string for the worst possible
3297 * case rather than possibly running out of space and having to
3298 * reallocate and then copy what we've done so far. Since everything
3299 * from s to t - 1 is invariant, the destination can be initialized
3300 * with these using a fast memory copy
3302 * The other way is to figure out exactly how big the string should be
3303 * by parsing the entire input. Then you don't have to make it big
3304 * enough to handle the worst possible case, and more importantly, if
3305 * the string you already have is large enough, you don't have to
3306 * allocate a new string, you can copy the last character in the input
3307 * string to the final position(s) that will be occupied by the
3308 * converted string and go backwards, stopping at t, since everything
3309 * before that is invariant.
3311 * There are advantages and disadvantages to each method.
3313 * In the first method, we can allocate a new string, do the memory
3314 * copy from the s to t - 1, and then proceed through the rest of the
3315 * string byte-by-byte.
3317 * In the second method, we proceed through the rest of the input
3318 * string just calculating how big the converted string will be. Then
3319 * there are two cases:
3320 * 1) if the string has enough extra space to handle the converted
3321 * value. We go backwards through the string, converting until we
3322 * get to the position we are at now, and then stop. If this
3323 * position is far enough along in the string, this method is
3324 * faster than the other method. If the memory copy were the same
3325 * speed as the byte-by-byte loop, that position would be about
3326 * half-way, as at the half-way mark, parsing to the end and back
3327 * is one complete string's parse, the same amount as starting
3328 * over and going all the way through. Actually, it would be
3329 * somewhat less than half-way, as it's faster to just count bytes
3330 * than to also copy, and we don't have the overhead of allocating
3331 * a new string, changing the scalar to use it, and freeing the
3332 * existing one. But if the memory copy is fast, the break-even
3333 * point is somewhere after half way. The counting loop could be
3334 * sped up by vectorization, etc, to move the break-even point
3335 * further towards the beginning.
3336 * 2) if the string doesn't have enough space to handle the converted
3337 * value. A new string will have to be allocated, and one might
3338 * as well, given that, start from the beginning doing the first
3339 * method. We've spent extra time parsing the string and in
3340 * exchange all we've gotten is that we know precisely how big to
3341 * make the new one. Perl is more optimized for time than space,
3342 * so this case is a loser.
3343 * So what I've decided to do is not use the 2nd method unless it is
3344 * guaranteed that a new string won't have to be allocated, assuming
3345 * the worst case. I also decided not to put any more conditions on it
3346 * than this, for now. It seems likely that, since the worst case is
3347 * twice as big as the unknown portion of the string (plus 1), we won't
3348 * be guaranteed enough space, causing us to go to the first method,
3349 * unless the string is short, or the first variant character is near
3350 * the end of it. In either of these cases, it seems best to use the
3351 * 2nd method. The only circumstance I can think of where this would
3352 * be really slower is if the string had once had much more data in it
3353 * than it does now, but there is still a substantial amount in it */
3356 STRLEN invariant_head = t - s;
3357 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3358 if (SvLEN(sv) < size) {
3360 /* Here, have decided to allocate a new string */
3365 Newx(dst, size, U8);
3367 /* If no known invariants at the beginning of the input string,
3368 * set so starts from there. Otherwise, can use memory copy to
3369 * get up to where we are now, and then start from here */
3371 if (invariant_head <= 0) {
3374 Copy(s, dst, invariant_head, char);
3375 d = dst + invariant_head;
3379 const UV uv = NATIVE8_TO_UNI(*t++);
3380 if (UNI_IS_INVARIANT(uv))
3381 *d++ = (U8)UNI_TO_NATIVE(uv);
3383 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3384 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3388 SvPV_free(sv); /* No longer using pre-existing string */
3389 SvPV_set(sv, (char*)dst);
3390 SvCUR_set(sv, d - dst);
3391 SvLEN_set(sv, size);
3394 /* Here, have decided to get the exact size of the string.
3395 * Currently this happens only when we know that there is
3396 * guaranteed enough space to fit the converted string, so
3397 * don't have to worry about growing. If two_byte_count is 0,
3398 * then t points to the first byte of the string which hasn't
3399 * been examined yet. Otherwise two_byte_count is 1, and t
3400 * points to the first byte in the string that will expand to
3401 * two. Depending on this, start examining at t or 1 after t.
3404 U8 *d = t + two_byte_count;
3407 /* Count up the remaining bytes that expand to two */
3410 const U8 chr = *d++;
3411 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3414 /* The string will expand by just the number of bytes that
3415 * occupy two positions. But we are one afterwards because of
3416 * the increment just above. This is the place to put the
3417 * trailing NUL, and to set the length before we decrement */
3419 d += two_byte_count;
3420 SvCUR_set(sv, d - s);
3424 /* Having decremented d, it points to the position to put the
3425 * very last byte of the expanded string. Go backwards through
3426 * the string, copying and expanding as we go, stopping when we
3427 * get to the part that is invariant the rest of the way down */
3431 const U8 ch = NATIVE8_TO_UNI(*e--);
3432 if (UNI_IS_INVARIANT(ch)) {
3433 *d-- = UNI_TO_NATIVE(ch);
3435 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3436 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3443 /* Mark as UTF-8 even if no variant - saves scanning loop */
3449 =for apidoc sv_utf8_downgrade
3451 Attempts to convert the PV of an SV from characters to bytes.
3452 If the PV contains a character that cannot fit
3453 in a byte, this conversion will fail;
3454 in this case, either returns false or, if C<fail_ok> is not
3457 This is not as a general purpose Unicode to byte encoding interface:
3458 use the Encode extension for that.
3464 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3468 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3470 if (SvPOKp(sv) && SvUTF8(sv)) {
3476 sv_force_normal_flags(sv, 0);
3478 s = (U8 *) SvPV(sv, len);
3479 if (!utf8_to_bytes(s, &len)) {
3484 Perl_croak(aTHX_ "Wide character in %s",
3487 Perl_croak(aTHX_ "Wide character");
3498 =for apidoc sv_utf8_encode
3500 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3501 flag off so that it looks like octets again.
3507 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3509 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3512 sv_force_normal_flags(sv, 0);
3514 if (SvREADONLY(sv)) {
3515 Perl_croak(aTHX_ "%s", PL_no_modify);
3517 (void) sv_utf8_upgrade(sv);
3522 =for apidoc sv_utf8_decode
3524 If the PV of the SV is an octet sequence in UTF-8
3525 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3526 so that it looks like a character. If the PV contains only single-byte
3527 characters, the C<SvUTF8> flag stays being off.
3528 Scans PV for validity and returns false if the PV is invalid UTF-8.
3534 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3536 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3542 /* The octets may have got themselves encoded - get them back as
3545 if (!sv_utf8_downgrade(sv, TRUE))
3548 /* it is actually just a matter of turning the utf8 flag on, but
3549 * we want to make sure everything inside is valid utf8 first.
3551 c = (const U8 *) SvPVX_const(sv);
3552 if (!is_utf8_string(c, SvCUR(sv)+1))
3554 e = (const U8 *) SvEND(sv);
3557 if (!UTF8_IS_INVARIANT(ch)) {
3567 =for apidoc sv_setsv
3569 Copies the contents of the source SV C<ssv> into the destination SV
3570 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3571 function if the source SV needs to be reused. Does not handle 'set' magic.
3572 Loosely speaking, it performs a copy-by-value, obliterating any previous
3573 content of the destination.
3575 You probably want to use one of the assortment of wrappers, such as
3576 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3577 C<SvSetMagicSV_nosteal>.
3579 =for apidoc sv_setsv_flags
3581 Copies the contents of the source SV C<ssv> into the destination SV
3582 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3583 function if the source SV needs to be reused. Does not handle 'set' magic.
3584 Loosely speaking, it performs a copy-by-value, obliterating any previous
3585 content of the destination.
3586 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3587 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3588 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3589 and C<sv_setsv_nomg> are implemented in terms of this function.
3591 You probably want to use one of the assortment of wrappers, such as
3592 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3593 C<SvSetMagicSV_nosteal>.
3595 This is the primary function for copying scalars, and most other
3596 copy-ish functions and macros use this underneath.
3602 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3604 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3606 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3608 if (dtype != SVt_PVGV) {
3609 const char * const name = GvNAME(sstr);
3610 const STRLEN len = GvNAMELEN(sstr);
3612 if (dtype >= SVt_PV) {
3618 SvUPGRADE(dstr, SVt_PVGV);
3619 (void)SvOK_off(dstr);
3620 /* FIXME - why are we doing this, then turning it off and on again
3622 isGV_with_GP_on(dstr);
3624 GvSTASH(dstr) = GvSTASH(sstr);
3626 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3627 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3628 SvFAKE_on(dstr); /* can coerce to non-glob */
3631 if(GvGP(MUTABLE_GV(sstr))) {
3632 /* If source has method cache entry, clear it */
3634 SvREFCNT_dec(GvCV(sstr));
3638 /* If source has a real method, then a method is
3640 else if(GvCV((const GV *)sstr)) {
3645 /* If dest already had a real method, that's a change as well */
3646 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3650 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3653 gp_free(MUTABLE_GV(dstr));
3654 isGV_with_GP_off(dstr);
3655 (void)SvOK_off(dstr);
3656 isGV_with_GP_on(dstr);
3657 GvINTRO_off(dstr); /* one-shot flag */
3658 GvGP(dstr) = gp_ref(GvGP(sstr));
3659 if (SvTAINTED(sstr))
3661 if (GvIMPORTED(dstr) != GVf_IMPORTED
3662 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3664 GvIMPORTED_on(dstr);
3667 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3668 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3673 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3675 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3677 const int intro = GvINTRO(dstr);
3680 const U32 stype = SvTYPE(sref);
3681 bool mro_changes = FALSE;
3683 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3686 GvINTRO_off(dstr); /* one-shot flag */
3687 GvLINE(dstr) = CopLINE(PL_curcop);
3688 GvEGV(dstr) = MUTABLE_GV(dstr);
3693 location = (SV **) &GvCV(dstr);
3694 import_flag = GVf_IMPORTED_CV;
3697 location = (SV **) &GvHV(dstr);
3698 import_flag = GVf_IMPORTED_HV;
3701 location = (SV **) &GvAV(dstr);
3702 if (strEQ(GvNAME((GV*)dstr), "ISA"))
3704 import_flag = GVf_IMPORTED_AV;
3707 location = (SV **) &GvIOp(dstr);
3710 location = (SV **) &GvFORM(dstr);
3713 location = &GvSV(dstr);
3714 import_flag = GVf_IMPORTED_SV;
3717 if (stype == SVt_PVCV) {
3718 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3719 if (GvCVGEN(dstr)) {
3720 SvREFCNT_dec(GvCV(dstr));
3722 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3725 SAVEGENERICSV(*location);
3729 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3730 CV* const cv = MUTABLE_CV(*location);
3732 if (!GvCVGEN((const GV *)dstr) &&
3733 (CvROOT(cv) || CvXSUB(cv)))
3735 /* Redefining a sub - warning is mandatory if
3736 it was a const and its value changed. */
3737 if (CvCONST(cv) && CvCONST((const CV *)sref)
3739 == cv_const_sv((const CV *)sref)) {
3741 /* They are 2 constant subroutines generated from
3742 the same constant. This probably means that
3743 they are really the "same" proxy subroutine
3744 instantiated in 2 places. Most likely this is
3745 when a constant is exported twice. Don't warn.
3748 else if (ckWARN(WARN_REDEFINE)
3750 && (!CvCONST((const CV *)sref)
3751 || sv_cmp(cv_const_sv(cv),
3752 cv_const_sv((const CV *)
3754 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3757 ? "Constant subroutine %s::%s redefined"
3758 : "Subroutine %s::%s redefined"),
3759 HvNAME_get(GvSTASH((const GV *)dstr)),
3760 GvENAME(MUTABLE_GV(dstr)));
3764 cv_ckproto_len(cv, (const GV *)dstr,
3765 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3766 SvPOK(sref) ? SvCUR(sref) : 0);
3768 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3769 GvASSUMECV_on(dstr);
3770 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3773 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3774 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3775 GvFLAGS(dstr) |= import_flag;
3780 if (SvTAINTED(sstr))
3782 if (mro_changes) mro_isa_changed_in(GvSTASH(dstr));
3787 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3790 register U32 sflags;
3792 register svtype stype;
3794 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3799 if (SvIS_FREED(dstr)) {
3800 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3801 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3803 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3805 sstr = &PL_sv_undef;
3806 if (SvIS_FREED(sstr)) {
3807 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3808 (void*)sstr, (void*)dstr);
3810 stype = SvTYPE(sstr);
3811 dtype = SvTYPE(dstr);
3813 (void)SvAMAGIC_off(dstr);
3816 /* need to nuke the magic */
3820 /* There's a lot of redundancy below but we're going for speed here */
3825 if (dtype != SVt_PVGV) {
3826 (void)SvOK_off(dstr);
3834 sv_upgrade(dstr, SVt_IV);
3838 sv_upgrade(dstr, SVt_PVIV);
3841 goto end_of_first_switch;
3843 (void)SvIOK_only(dstr);
3844 SvIV_set(dstr, SvIVX(sstr));
3847 /* SvTAINTED can only be true if the SV has taint magic, which in
3848 turn means that the SV type is PVMG (or greater). This is the
3849 case statement for SVt_IV, so this cannot be true (whatever gcov
3851 assert(!SvTAINTED(sstr));
3856 if (dtype < SVt_PV && dtype != SVt_IV)
3857 sv_upgrade(dstr, SVt_IV);
3865 sv_upgrade(dstr, SVt_NV);
3869 sv_upgrade(dstr, SVt_PVNV);
3872 goto end_of_first_switch;
3874 SvNV_set(dstr, SvNVX(sstr));
3875 (void)SvNOK_only(dstr);
3876 /* SvTAINTED can only be true if the SV has taint magic, which in
3877 turn means that the SV type is PVMG (or greater). This is the
3878 case statement for SVt_NV, so this cannot be true (whatever gcov
3880 assert(!SvTAINTED(sstr));
3886 #ifdef PERL_OLD_COPY_ON_WRITE
3887 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3888 if (dtype < SVt_PVIV)
3889 sv_upgrade(dstr, SVt_PVIV);
3896 sv_upgrade(dstr, SVt_PV);
3899 if (dtype < SVt_PVIV)
3900 sv_upgrade(dstr, SVt_PVIV);
3903 if (dtype < SVt_PVNV)
3904 sv_upgrade(dstr, SVt_PVNV);
3908 const char * const type = sv_reftype(sstr,0);
3910 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3912 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3917 if (dtype < SVt_REGEXP)
3918 sv_upgrade(dstr, SVt_REGEXP);
3921 /* case SVt_BIND: */
3924 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3925 glob_assign_glob(dstr, sstr, dtype);
3928 /* SvVALID means that this PVGV is playing at being an FBM. */
3932 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3934 if (SvTYPE(sstr) != stype) {
3935 stype = SvTYPE(sstr);
3936 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3937 glob_assign_glob(dstr, sstr, dtype);
3942 if (stype == SVt_PVLV)
3943 SvUPGRADE(dstr, SVt_PVNV);
3945 SvUPGRADE(dstr, (svtype)stype);
3947 end_of_first_switch:
3949 /* dstr may have been upgraded. */
3950 dtype = SvTYPE(dstr);
3951 sflags = SvFLAGS(sstr);
3953 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3954 /* Assigning to a subroutine sets the prototype. */
3957 const char *const ptr = SvPV_const(sstr, len);
3959 SvGROW(dstr, len + 1);
3960 Copy(ptr, SvPVX(dstr), len + 1, char);
3961 SvCUR_set(dstr, len);
3963 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3967 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3968 const char * const type = sv_reftype(dstr,0);
3970 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3972 Perl_croak(aTHX_ "Cannot copy to %s", type);
3973 } else if (sflags & SVf_ROK) {
3974 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3975 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3978 if (GvIMPORTED(dstr) != GVf_IMPORTED
3979 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3981 GvIMPORTED_on(dstr);
3986 glob_assign_glob(dstr, sstr, dtype);
3990 if (dtype >= SVt_PV) {
3991 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3992 glob_assign_ref(dstr, sstr);
3995 if (SvPVX_const(dstr)) {
4001 (void)SvOK_off(dstr);
4002 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4003 SvFLAGS(dstr) |= sflags & SVf_ROK;
4004 assert(!(sflags & SVp_NOK));
4005 assert(!(sflags & SVp_IOK));
4006 assert(!(sflags & SVf_NOK));
4007 assert(!(sflags & SVf_IOK));
4009 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4010 if (!(sflags & SVf_OK)) {
4011 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4012 "Undefined value assigned to typeglob");
4015 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4016 if (dstr != (const SV *)gv) {
4018 gp_free(MUTABLE_GV(dstr));
4019 GvGP(dstr) = gp_ref(GvGP(gv));
4023 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4024 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4026 else if (sflags & SVp_POK) {
4030 * Check to see if we can just swipe the string. If so, it's a
4031 * possible small lose on short strings, but a big win on long ones.
4032 * It might even be a win on short strings if SvPVX_const(dstr)
4033 * has to be allocated and SvPVX_const(sstr) has to be freed.
4034 * Likewise if we can set up COW rather than doing an actual copy, we
4035 * drop to the else clause, as the swipe code and the COW setup code
4036 * have much in common.
4039 /* Whichever path we take through the next code, we want this true,
4040 and doing it now facilitates the COW check. */
4041 (void)SvPOK_only(dstr);
4044 /* If we're already COW then this clause is not true, and if COW
4045 is allowed then we drop down to the else and make dest COW
4046 with us. If caller hasn't said that we're allowed to COW
4047 shared hash keys then we don't do the COW setup, even if the
4048 source scalar is a shared hash key scalar. */
4049 (((flags & SV_COW_SHARED_HASH_KEYS)
4050 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4051 : 1 /* If making a COW copy is forbidden then the behaviour we
4052 desire is as if the source SV isn't actually already
4053 COW, even if it is. So we act as if the source flags
4054 are not COW, rather than actually testing them. */
4056 #ifndef PERL_OLD_COPY_ON_WRITE
4057 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4058 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4059 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4060 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4061 but in turn, it's somewhat dead code, never expected to go
4062 live, but more kept as a placeholder on how to do it better
4063 in a newer implementation. */
4064 /* If we are COW and dstr is a suitable target then we drop down
4065 into the else and make dest a COW of us. */
4066 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4071 (sflags & SVs_TEMP) && /* slated for free anyway? */
4072 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4073 (!(flags & SV_NOSTEAL)) &&
4074 /* and we're allowed to steal temps */
4075 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4076 SvLEN(sstr) && /* and really is a string */
4077 /* and won't be needed again, potentially */
4078 !(PL_op && PL_op->op_type == OP_AASSIGN))
4079 #ifdef PERL_OLD_COPY_ON_WRITE
4080 && ((flags & SV_COW_SHARED_HASH_KEYS)
4081 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4082 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4083 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4087 /* Failed the swipe test, and it's not a shared hash key either.
4088 Have to copy the string. */
4089 STRLEN len = SvCUR(sstr);
4090 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4091 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4092 SvCUR_set(dstr, len);
4093 *SvEND(dstr) = '\0';
4095 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4097 /* Either it's a shared hash key, or it's suitable for
4098 copy-on-write or we can swipe the string. */
4100 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4104 #ifdef PERL_OLD_COPY_ON_WRITE
4106 if ((sflags & (SVf_FAKE | SVf_READONLY))
4107 != (SVf_FAKE | SVf_READONLY)) {
4108 SvREADONLY_on(sstr);
4110 /* Make the source SV into a loop of 1.
4111 (about to become 2) */
4112 SV_COW_NEXT_SV_SET(sstr, sstr);
4116 /* Initial code is common. */
4117 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4122 /* making another shared SV. */
4123 STRLEN cur = SvCUR(sstr);
4124 STRLEN len = SvLEN(sstr);
4125 #ifdef PERL_OLD_COPY_ON_WRITE
4127 assert (SvTYPE(dstr) >= SVt_PVIV);
4128 /* SvIsCOW_normal */
4129 /* splice us in between source and next-after-source. */
4130 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4131 SV_COW_NEXT_SV_SET(sstr, dstr);
4132 SvPV_set(dstr, SvPVX_mutable(sstr));
4136 /* SvIsCOW_shared_hash */
4137 DEBUG_C(PerlIO_printf(Perl_debug_log,
4138 "Copy on write: Sharing hash\n"));
4140 assert (SvTYPE(dstr) >= SVt_PV);
4142 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4144 SvLEN_set(dstr, len);
4145 SvCUR_set(dstr, cur);
4146 SvREADONLY_on(dstr);
4150 { /* Passes the swipe test. */
4151 SvPV_set(dstr, SvPVX_mutable(sstr));
4152 SvLEN_set(dstr, SvLEN(sstr));
4153 SvCUR_set(dstr, SvCUR(sstr));
4156 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4157 SvPV_set(sstr, NULL);
4163 if (sflags & SVp_NOK) {
4164 SvNV_set(dstr, SvNVX(sstr));
4166 if (sflags & SVp_IOK) {
4167 SvIV_set(dstr, SvIVX(sstr));
4168 /* Must do this otherwise some other overloaded use of 0x80000000
4169 gets confused. I guess SVpbm_VALID */
4170 if (sflags & SVf_IVisUV)
4173 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4175 const MAGIC * const smg = SvVSTRING_mg(sstr);
4177 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4178 smg->mg_ptr, smg->mg_len);
4179 SvRMAGICAL_on(dstr);
4183 else if (sflags & (SVp_IOK|SVp_NOK)) {
4184 (void)SvOK_off(dstr);
4185 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4186 if (sflags & SVp_IOK) {
4187 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4188 SvIV_set(dstr, SvIVX(sstr));
4190 if (sflags & SVp_NOK) {
4191 SvNV_set(dstr, SvNVX(sstr));
4195 if (isGV_with_GP(sstr)) {
4196 /* This stringification rule for globs is spread in 3 places.
4197 This feels bad. FIXME. */
4198 const U32 wasfake = sflags & SVf_FAKE;
4200 /* FAKE globs can get coerced, so need to turn this off
4201 temporarily if it is on. */
4203 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4204 SvFLAGS(sstr) |= wasfake;
4207 (void)SvOK_off(dstr);
4209 if (SvTAINTED(sstr))
4214 =for apidoc sv_setsv_mg
4216 Like C<sv_setsv>, but also handles 'set' magic.
4222 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4224 PERL_ARGS_ASSERT_SV_SETSV_MG;
4226 sv_setsv(dstr,sstr);
4230 #ifdef PERL_OLD_COPY_ON_WRITE
4232 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4234 STRLEN cur = SvCUR(sstr);
4235 STRLEN len = SvLEN(sstr);
4236 register char *new_pv;
4238 PERL_ARGS_ASSERT_SV_SETSV_COW;
4241 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4242 (void*)sstr, (void*)dstr);
4249 if (SvTHINKFIRST(dstr))
4250 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4251 else if (SvPVX_const(dstr))
4252 Safefree(SvPVX_const(dstr));
4256 SvUPGRADE(dstr, SVt_PVIV);
4258 assert (SvPOK(sstr));
4259 assert (SvPOKp(sstr));
4260 assert (!SvIOK(sstr));
4261 assert (!SvIOKp(sstr));
4262 assert (!SvNOK(sstr));
4263 assert (!SvNOKp(sstr));
4265 if (SvIsCOW(sstr)) {
4267 if (SvLEN(sstr) == 0) {
4268 /* source is a COW shared hash key. */
4269 DEBUG_C(PerlIO_printf(Perl_debug_log,
4270 "Fast copy on write: Sharing hash\n"));
4271 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4274 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4276 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4277 SvUPGRADE(sstr, SVt_PVIV);
4278 SvREADONLY_on(sstr);
4280 DEBUG_C(PerlIO_printf(Perl_debug_log,
4281 "Fast copy on write: Converting sstr to COW\n"));
4282 SV_COW_NEXT_SV_SET(dstr, sstr);
4284 SV_COW_NEXT_SV_SET(sstr, dstr);
4285 new_pv = SvPVX_mutable(sstr);
4288 SvPV_set(dstr, new_pv);
4289 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4292 SvLEN_set(dstr, len);
4293 SvCUR_set(dstr, cur);
4302 =for apidoc sv_setpvn
4304 Copies a string into an SV. The C<len> parameter indicates the number of
4305 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4306 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4312 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4315 register char *dptr;
4317 PERL_ARGS_ASSERT_SV_SETPVN;
4319 SV_CHECK_THINKFIRST_COW_DROP(sv);
4325 /* len is STRLEN which is unsigned, need to copy to signed */
4328 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4330 SvUPGRADE(sv, SVt_PV);
4332 dptr = SvGROW(sv, len + 1);
4333 Move(ptr,dptr,len,char);
4336 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4341 =for apidoc sv_setpvn_mg
4343 Like C<sv_setpvn>, but also handles 'set' magic.
4349 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4351 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4353 sv_setpvn(sv,ptr,len);
4358 =for apidoc sv_setpv
4360 Copies a string into an SV. The string must be null-terminated. Does not
4361 handle 'set' magic. See C<sv_setpv_mg>.
4367 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4370 register STRLEN len;
4372 PERL_ARGS_ASSERT_SV_SETPV;
4374 SV_CHECK_THINKFIRST_COW_DROP(sv);
4380 SvUPGRADE(sv, SVt_PV);
4382 SvGROW(sv, len + 1);
4383 Move(ptr,SvPVX(sv),len+1,char);
4385 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4390 =for apidoc sv_setpv_mg
4392 Like C<sv_setpv>, but also handles 'set' magic.
4398 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4400 PERL_ARGS_ASSERT_SV_SETPV_MG;
4407 =for apidoc sv_usepvn_flags
4409 Tells an SV to use C<ptr> to find its string value. Normally the
4410 string is stored inside the SV but sv_usepvn allows the SV to use an
4411 outside string. The C<ptr> should point to memory that was allocated
4412 by C<malloc>. The string length, C<len>, must be supplied. By default
4413 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4414 so that pointer should not be freed or used by the programmer after
4415 giving it to sv_usepvn, and neither should any pointers from "behind"
4416 that pointer (e.g. ptr + 1) be used.
4418 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4419 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4420 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4421 C<len>, and already meets the requirements for storing in C<SvPVX>)
4427 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4432 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4434 SV_CHECK_THINKFIRST_COW_DROP(sv);
4435 SvUPGRADE(sv, SVt_PV);
4438 if (flags & SV_SMAGIC)
4442 if (SvPVX_const(sv))
4446 if (flags & SV_HAS_TRAILING_NUL)
4447 assert(ptr[len] == '\0');
4450 allocate = (flags & SV_HAS_TRAILING_NUL)
4452 #ifdef Perl_safesysmalloc_size
4455 PERL_STRLEN_ROUNDUP(len + 1);
4457 if (flags & SV_HAS_TRAILING_NUL) {
4458 /* It's long enough - do nothing.
4459 Specfically Perl_newCONSTSUB is relying on this. */
4462 /* Force a move to shake out bugs in callers. */
4463 char *new_ptr = (char*)safemalloc(allocate);
4464 Copy(ptr, new_ptr, len, char);
4465 PoisonFree(ptr,len,char);
4469 ptr = (char*) saferealloc (ptr, allocate);
4472 #ifdef Perl_safesysmalloc_size
4473 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4475 SvLEN_set(sv, allocate);
4479 if (!(flags & SV_HAS_TRAILING_NUL)) {
4482 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4484 if (flags & SV_SMAGIC)
4488 #ifdef PERL_OLD_COPY_ON_WRITE
4489 /* Need to do this *after* making the SV normal, as we need the buffer
4490 pointer to remain valid until after we've copied it. If we let go too early,
4491 another thread could invalidate it by unsharing last of the same hash key
4492 (which it can do by means other than releasing copy-on-write Svs)
4493 or by changing the other copy-on-write SVs in the loop. */
4495 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4497 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4499 { /* this SV was SvIsCOW_normal(sv) */
4500 /* we need to find the SV pointing to us. */
4501 SV *current = SV_COW_NEXT_SV(after);
4503 if (current == sv) {
4504 /* The SV we point to points back to us (there were only two of us
4506 Hence other SV is no longer copy on write either. */
4508 SvREADONLY_off(after);
4510 /* We need to follow the pointers around the loop. */
4512 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4515 /* don't loop forever if the structure is bust, and we have
4516 a pointer into a closed loop. */
4517 assert (current != after);
4518 assert (SvPVX_const(current) == pvx);
4520 /* Make the SV before us point to the SV after us. */
4521 SV_COW_NEXT_SV_SET(current, after);
4527 =for apidoc sv_force_normal_flags
4529 Undo various types of fakery on an SV: if the PV is a shared string, make
4530 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4531 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4532 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4533 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4534 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4535 set to some other value.) In addition, the C<flags> parameter gets passed to
4536 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4537 with flags set to 0.
4543 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4547 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4549 #ifdef PERL_OLD_COPY_ON_WRITE
4550 if (SvREADONLY(sv)) {
4552 const char * const pvx = SvPVX_const(sv);
4553 const STRLEN len = SvLEN(sv);
4554 const STRLEN cur = SvCUR(sv);
4555 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4556 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4557 we'll fail an assertion. */
4558 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4561 PerlIO_printf(Perl_debug_log,
4562 "Copy on write: Force normal %ld\n",
4568 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4571 if (flags & SV_COW_DROP_PV) {
4572 /* OK, so we don't need to copy our buffer. */
4575 SvGROW(sv, cur + 1);
4576 Move(pvx,SvPVX(sv),cur,char);
4581 sv_release_COW(sv, pvx, next);
4583 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4589 else if (IN_PERL_RUNTIME)
4590 Perl_croak(aTHX_ "%s", PL_no_modify);
4593 if (SvREADONLY(sv)) {
4595 const char * const pvx = SvPVX_const(sv);
4596 const STRLEN len = SvCUR(sv);
4601 SvGROW(sv, len + 1);
4602 Move(pvx,SvPVX(sv),len,char);
4604 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4606 else if (IN_PERL_RUNTIME)
4607 Perl_croak(aTHX_ "%s", PL_no_modify);
4611 sv_unref_flags(sv, flags);
4612 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4619 Efficient removal of characters from the beginning of the string buffer.
4620 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4621 the string buffer. The C<ptr> becomes the first character of the adjusted
4622 string. Uses the "OOK hack".
4623 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4624 refer to the same chunk of data.
4630 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4636 const U8 *real_start;
4640 PERL_ARGS_ASSERT_SV_CHOP;
4642 if (!ptr || !SvPOKp(sv))
4644 delta = ptr - SvPVX_const(sv);
4646 /* Nothing to do. */
4649 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4650 nothing uses the value of ptr any more. */
4651 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4652 if (ptr <= SvPVX_const(sv))
4653 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4654 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4655 SV_CHECK_THINKFIRST(sv);
4656 if (delta > max_delta)
4657 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4658 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4659 SvPVX_const(sv) + max_delta);
4662 if (!SvLEN(sv)) { /* make copy of shared string */
4663 const char *pvx = SvPVX_const(sv);
4664 const STRLEN len = SvCUR(sv);
4665 SvGROW(sv, len + 1);
4666 Move(pvx,SvPVX(sv),len,char);
4669 SvFLAGS(sv) |= SVf_OOK;
4672 SvOOK_offset(sv, old_delta);
4674 SvLEN_set(sv, SvLEN(sv) - delta);
4675 SvCUR_set(sv, SvCUR(sv) - delta);
4676 SvPV_set(sv, SvPVX(sv) + delta);
4678 p = (U8 *)SvPVX_const(sv);
4683 real_start = p - delta;
4687 if (delta < 0x100) {
4691 p -= sizeof(STRLEN);
4692 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4696 /* Fill the preceding buffer with sentinals to verify that no-one is
4698 while (p > real_start) {
4706 =for apidoc sv_catpvn
4708 Concatenates the string onto the end of the string which is in the SV. The
4709 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4710 status set, then the bytes appended should be valid UTF-8.
4711 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4713 =for apidoc sv_catpvn_flags
4715 Concatenates the string onto the end of the string which is in the SV. The
4716 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4717 status set, then the bytes appended should be valid UTF-8.
4718 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4719 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4720 in terms of this function.
4726 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4730 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4732 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4734 SvGROW(dsv, dlen + slen + 1);
4736 sstr = SvPVX_const(dsv);
4737 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4738 SvCUR_set(dsv, SvCUR(dsv) + slen);
4740 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4742 if (flags & SV_SMAGIC)
4747 =for apidoc sv_catsv
4749 Concatenates the string from SV C<ssv> onto the end of the string in
4750 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4751 not 'set' magic. See C<sv_catsv_mg>.
4753 =for apidoc sv_catsv_flags
4755 Concatenates the string from SV C<ssv> onto the end of the string in
4756 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4757 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4758 and C<sv_catsv_nomg> are implemented in terms of this function.
4763 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4767 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4771 const char *spv = SvPV_const(ssv, slen);
4773 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4774 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4775 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4776 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4777 dsv->sv_flags doesn't have that bit set.
4778 Andy Dougherty 12 Oct 2001
4780 const I32 sutf8 = DO_UTF8(ssv);
4783 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4785 dutf8 = DO_UTF8(dsv);
4787 if (dutf8 != sutf8) {
4789 /* Not modifying source SV, so taking a temporary copy. */
4790 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4792 sv_utf8_upgrade(csv);
4793 spv = SvPV_const(csv, slen);
4796 /* Leave enough space for the cat that's about to happen */
4797 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4799 sv_catpvn_nomg(dsv, spv, slen);
4802 if (flags & SV_SMAGIC)
4807 =for apidoc sv_catpv
4809 Concatenates the string onto the end of the string which is in the SV.
4810 If the SV has the UTF-8 status set, then the bytes appended should be
4811 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4816 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4819 register STRLEN len;
4823 PERL_ARGS_ASSERT_SV_CATPV;
4827 junk = SvPV_force(sv, tlen);
4829 SvGROW(sv, tlen + len + 1);
4831 ptr = SvPVX_const(sv);
4832 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4833 SvCUR_set(sv, SvCUR(sv) + len);
4834 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4839 =for apidoc sv_catpv_mg
4841 Like C<sv_catpv>, but also handles 'set' magic.
4847 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4849 PERL_ARGS_ASSERT_SV_CATPV_MG;
4858 Creates a new SV. A non-zero C<len> parameter indicates the number of
4859 bytes of preallocated string space the SV should have. An extra byte for a
4860 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4861 space is allocated.) The reference count for the new SV is set to 1.
4863 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4864 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4865 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4866 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4867 modules supporting older perls.
4873 Perl_newSV(pTHX_ const STRLEN len)
4880 sv_upgrade(sv, SVt_PV);
4881 SvGROW(sv, len + 1);
4886 =for apidoc sv_magicext
4888 Adds magic to an SV, upgrading it if necessary. Applies the
4889 supplied vtable and returns a pointer to the magic added.
4891 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4892 In particular, you can add magic to SvREADONLY SVs, and add more than
4893 one instance of the same 'how'.
4895 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4896 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4897 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4898 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4900 (This is now used as a subroutine by C<sv_magic>.)
4905 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4906 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4911 PERL_ARGS_ASSERT_SV_MAGICEXT;
4913 SvUPGRADE(sv, SVt_PVMG);
4914 Newxz(mg, 1, MAGIC);
4915 mg->mg_moremagic = SvMAGIC(sv);
4916 SvMAGIC_set(sv, mg);
4918 /* Sometimes a magic contains a reference loop, where the sv and
4919 object refer to each other. To prevent a reference loop that
4920 would prevent such objects being freed, we look for such loops
4921 and if we find one we avoid incrementing the object refcount.
4923 Note we cannot do this to avoid self-tie loops as intervening RV must
4924 have its REFCNT incremented to keep it in existence.
4927 if (!obj || obj == sv ||
4928 how == PERL_MAGIC_arylen ||
4929 how == PERL_MAGIC_symtab ||
4930 (SvTYPE(obj) == SVt_PVGV &&
4931 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4932 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4933 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4938 mg->mg_obj = SvREFCNT_inc_simple(obj);
4939 mg->mg_flags |= MGf_REFCOUNTED;
4942 /* Normal self-ties simply pass a null object, and instead of
4943 using mg_obj directly, use the SvTIED_obj macro to produce a
4944 new RV as needed. For glob "self-ties", we are tieing the PVIO
4945 with an RV obj pointing to the glob containing the PVIO. In
4946 this case, to avoid a reference loop, we need to weaken the
4950 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4951 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4957 mg->mg_len = namlen;
4960 mg->mg_ptr = savepvn(name, namlen);
4961 else if (namlen == HEf_SVKEY) {
4962 /* Yes, this is casting away const. This is only for the case of
4963 HEf_SVKEY. I think we need to document this abberation of the
4964 constness of the API, rather than making name non-const, as
4965 that change propagating outwards a long way. */
4966 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4968 mg->mg_ptr = (char *) name;
4970 mg->mg_virtual = (MGVTBL *) vtable;
4974 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4979 =for apidoc sv_magic
4981 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4982 then adds a new magic item of type C<how> to the head of the magic list.
4984 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4985 handling of the C<name> and C<namlen> arguments.
4987 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4988 to add more than one instance of the same 'how'.
4994 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4995 const char *const name, const I32 namlen)
4998 const MGVTBL *vtable;
5001 PERL_ARGS_ASSERT_SV_MAGIC;
5003 #ifdef PERL_OLD_COPY_ON_WRITE
5005 sv_force_normal_flags(sv, 0);
5007 if (SvREADONLY(sv)) {
5009 /* its okay to attach magic to shared strings; the subsequent
5010 * upgrade to PVMG will unshare the string */
5011 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5014 && how != PERL_MAGIC_regex_global
5015 && how != PERL_MAGIC_bm
5016 && how != PERL_MAGIC_fm
5017 && how != PERL_MAGIC_sv
5018 && how != PERL_MAGIC_backref
5021 Perl_croak(aTHX_ "%s", PL_no_modify);
5024 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5025 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5026 /* sv_magic() refuses to add a magic of the same 'how' as an
5029 if (how == PERL_MAGIC_taint) {
5031 /* Any scalar which already had taint magic on which someone
5032 (erroneously?) did SvIOK_on() or similar will now be
5033 incorrectly sporting public "OK" flags. */
5034 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5042 vtable = &PL_vtbl_sv;
5044 case PERL_MAGIC_overload:
5045 vtable = &PL_vtbl_amagic;
5047 case PERL_MAGIC_overload_elem:
5048 vtable = &PL_vtbl_amagicelem;
5050 case PERL_MAGIC_overload_table:
5051 vtable = &PL_vtbl_ovrld;
5054 vtable = &PL_vtbl_bm;
5056 case PERL_MAGIC_regdata:
5057 vtable = &PL_vtbl_regdata;
5059 case PERL_MAGIC_regdatum:
5060 vtable = &PL_vtbl_regdatum;
5062 case PERL_MAGIC_env:
5063 vtable = &PL_vtbl_env;
5066 vtable = &PL_vtbl_fm;
5068 case PERL_MAGIC_envelem:
5069 vtable = &PL_vtbl_envelem;
5071 case PERL_MAGIC_regex_global:
5072 vtable = &PL_vtbl_mglob;
5074 case PERL_MAGIC_isa:
5075 vtable = &PL_vtbl_isa;
5077 case PERL_MAGIC_isaelem:
5078 vtable = &PL_vtbl_isaelem;
5080 case PERL_MAGIC_nkeys:
5081 vtable = &PL_vtbl_nkeys;
5083 case PERL_MAGIC_dbfile:
5086 case PERL_MAGIC_dbline:
5087 vtable = &PL_vtbl_dbline;
5089 #ifdef USE_LOCALE_COLLATE
5090 case PERL_MAGIC_collxfrm:
5091 vtable = &PL_vtbl_collxfrm;
5093 #endif /* USE_LOCALE_COLLATE */
5094 case PERL_MAGIC_tied:
5095 vtable = &PL_vtbl_pack;
5097 case PERL_MAGIC_tiedelem:
5098 case PERL_MAGIC_tiedscalar:
5099 vtable = &PL_vtbl_packelem;
5102 vtable = &PL_vtbl_regexp;
5104 case PERL_MAGIC_sig:
5105 vtable = &PL_vtbl_sig;
5107 case PERL_MAGIC_sigelem:
5108 vtable = &PL_vtbl_sigelem;
5110 case PERL_MAGIC_taint:
5111 vtable = &PL_vtbl_taint;
5113 case PERL_MAGIC_uvar:
5114 vtable = &PL_vtbl_uvar;
5116 case PERL_MAGIC_vec:
5117 vtable = &PL_vtbl_vec;
5119 case PERL_MAGIC_arylen_p:
5120 case PERL_MAGIC_rhash:
5121 case PERL_MAGIC_symtab:
5122 case PERL_MAGIC_vstring:
5125 case PERL_MAGIC_utf8:
5126 vtable = &PL_vtbl_utf8;
5128 case PERL_MAGIC_substr:
5129 vtable = &PL_vtbl_substr;
5131 case PERL_MAGIC_defelem:
5132 vtable = &PL_vtbl_defelem;
5134 case PERL_MAGIC_arylen:
5135 vtable = &PL_vtbl_arylen;
5137 case PERL_MAGIC_pos:
5138 vtable = &PL_vtbl_pos;
5140 case PERL_MAGIC_backref:
5141 vtable = &PL_vtbl_backref;
5143 case PERL_MAGIC_hintselem:
5144 vtable = &PL_vtbl_hintselem;
5146 case PERL_MAGIC_hints:
5147 vtable = &PL_vtbl_hints;
5149 case PERL_MAGIC_ext:
5150 /* Reserved for use by extensions not perl internals. */
5151 /* Useful for attaching extension internal data to perl vars. */
5152 /* Note that multiple extensions may clash if magical scalars */
5153 /* etc holding private data from one are passed to another. */
5157 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5160 /* Rest of work is done else where */
5161 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5164 case PERL_MAGIC_taint:
5167 case PERL_MAGIC_ext:
5168 case PERL_MAGIC_dbfile:
5175 =for apidoc sv_unmagic
5177 Removes all magic of type C<type> from an SV.
5183 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5188 PERL_ARGS_ASSERT_SV_UNMAGIC;
5190 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5192 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5193 for (mg = *mgp; mg; mg = *mgp) {
5194 if (mg->mg_type == type) {
5195 const MGVTBL* const vtbl = mg->mg_virtual;
5196 *mgp = mg->mg_moremagic;
5197 if (vtbl && vtbl->svt_free)
5198 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5199 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5201 Safefree(mg->mg_ptr);
5202 else if (mg->mg_len == HEf_SVKEY)
5203 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5204 else if (mg->mg_type == PERL_MAGIC_utf8)
5205 Safefree(mg->mg_ptr);
5207 if (mg->mg_flags & MGf_REFCOUNTED)
5208 SvREFCNT_dec(mg->mg_obj);
5212 mgp = &mg->mg_moremagic;
5216 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5217 SvMAGIC_set(sv, NULL);
5224 =for apidoc sv_rvweaken
5226 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5227 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5228 push a back-reference to this RV onto the array of backreferences
5229 associated with that magic. If the RV is magical, set magic will be
5230 called after the RV is cleared.
5236 Perl_sv_rvweaken(pTHX_ SV *const sv)
5240 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5242 if (!SvOK(sv)) /* let undefs pass */
5245 Perl_croak(aTHX_ "Can't weaken a nonreference");
5246 else if (SvWEAKREF(sv)) {
5247 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5251 Perl_sv_add_backref(aTHX_ tsv, sv);
5257 /* Give tsv backref magic if it hasn't already got it, then push a
5258 * back-reference to sv onto the array associated with the backref magic.
5261 /* A discussion about the backreferences array and its refcount:
5263 * The AV holding the backreferences is pointed to either as the mg_obj of
5264 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5265 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5266 * have the standard magic instead.) The array is created with a refcount
5267 * of 2. This means that if during global destruction the array gets
5268 * picked on first to have its refcount decremented by the random zapper,
5269 * it won't actually be freed, meaning it's still theere for when its
5270 * parent gets freed.
5271 * When the parent SV is freed, in the case of magic, the magic is freed,
5272 * Perl_magic_killbackrefs is called which decrements one refcount, then
5273 * mg_obj is freed which kills the second count.
5274 * In the vase of a HV being freed, one ref is removed by
5275 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5280 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5285 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5287 if (SvTYPE(tsv) == SVt_PVHV) {
5288 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5292 /* There is no AV in the offical place - try a fixup. */
5293 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5296 /* Aha. They've got it stowed in magic. Bring it back. */
5297 av = MUTABLE_AV(mg->mg_obj);
5298 /* Stop mg_free decreasing the refernce count. */
5300 /* Stop mg_free even calling the destructor, given that
5301 there's no AV to free up. */
5303 sv_unmagic(tsv, PERL_MAGIC_backref);
5307 SvREFCNT_inc_simple_void(av); /* see discussion above */
5312 const MAGIC *const mg
5313 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5315 av = MUTABLE_AV(mg->mg_obj);
5319 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5320 /* av now has a refcnt of 2; see discussion above */
5323 if (AvFILLp(av) >= AvMAX(av)) {
5324 av_extend(av, AvFILLp(av)+1);
5326 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5329 /* delete a back-reference to ourselves from the backref magic associated
5330 * with the SV we point to.
5334 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5341 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5343 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5344 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5345 /* We mustn't attempt to "fix up" the hash here by moving the
5346 backreference array back to the hv_aux structure, as that is stored
5347 in the main HvARRAY(), and hfreentries assumes that no-one
5348 reallocates HvARRAY() while it is running. */
5351 const MAGIC *const mg
5352 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5354 av = MUTABLE_AV(mg->mg_obj);
5358 Perl_croak(aTHX_ "panic: del_backref");
5360 assert(!SvIS_FREED(av));
5363 /* We shouldn't be in here more than once, but for paranoia reasons lets
5365 for (i = AvFILLp(av); i >= 0; i--) {
5367 const SSize_t fill = AvFILLp(av);
5369 /* We weren't the last entry.
5370 An unordered list has this property that you can take the
5371 last element off the end to fill the hole, and it's still
5372 an unordered list :-)
5377 AvFILLp(av) = fill - 1;
5383 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5385 SV **svp = AvARRAY(av);
5387 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5388 PERL_UNUSED_ARG(sv);
5390 assert(!svp || !SvIS_FREED(av));
5392 SV *const *const last = svp + AvFILLp(av);
5394 while (svp <= last) {
5396 SV *const referrer = *svp;
5397 if (SvWEAKREF(referrer)) {
5398 /* XXX Should we check that it hasn't changed? */
5399 SvRV_set(referrer, 0);
5401 SvWEAKREF_off(referrer);
5402 SvSETMAGIC(referrer);
5403 } else if (SvTYPE(referrer) == SVt_PVGV ||
5404 SvTYPE(referrer) == SVt_PVLV) {
5405 /* You lookin' at me? */
5406 assert(GvSTASH(referrer));
5407 assert(GvSTASH(referrer) == (const HV *)sv);
5408 GvSTASH(referrer) = 0;
5411 "panic: magic_killbackrefs (flags=%"UVxf")",
5412 (UV)SvFLAGS(referrer));
5420 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5425 =for apidoc sv_insert
5427 Inserts a string at the specified offset/length within the SV. Similar to
5428 the Perl substr() function. Handles get magic.
5430 =for apidoc sv_insert_flags
5432 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5438 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5443 register char *midend;
5444 register char *bigend;
5448 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5451 Perl_croak(aTHX_ "Can't modify non-existent substring");
5452 SvPV_force_flags(bigstr, curlen, flags);
5453 (void)SvPOK_only_UTF8(bigstr);
5454 if (offset + len > curlen) {
5455 SvGROW(bigstr, offset+len+1);
5456 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5457 SvCUR_set(bigstr, offset+len);
5461 i = littlelen - len;
5462 if (i > 0) { /* string might grow */
5463 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5464 mid = big + offset + len;
5465 midend = bigend = big + SvCUR(bigstr);
5468 while (midend > mid) /* shove everything down */
5469 *--bigend = *--midend;
5470 Move(little,big+offset,littlelen,char);
5471 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5476 Move(little,SvPVX(bigstr)+offset,len,char);
5481 big = SvPVX(bigstr);
5484 bigend = big + SvCUR(bigstr);
5486 if (midend > bigend)
5487 Perl_croak(aTHX_ "panic: sv_insert");
5489 if (mid - big > bigend - midend) { /* faster to shorten from end */
5491 Move(little, mid, littlelen,char);
5494 i = bigend - midend;
5496 Move(midend, mid, i,char);
5500 SvCUR_set(bigstr, mid - big);
5502 else if ((i = mid - big)) { /* faster from front */
5503 midend -= littlelen;
5505 Move(big, midend - i, i, char);
5506 sv_chop(bigstr,midend-i);
5508 Move(little, mid, littlelen,char);
5510 else if (littlelen) {
5511 midend -= littlelen;
5512 sv_chop(bigstr,midend);
5513 Move(little,midend,littlelen,char);
5516 sv_chop(bigstr,midend);
5522 =for apidoc sv_replace
5524 Make the first argument a copy of the second, then delete the original.
5525 The target SV physically takes over ownership of the body of the source SV
5526 and inherits its flags; however, the target keeps any magic it owns,
5527 and any magic in the source is discarded.
5528 Note that this is a rather specialist SV copying operation; most of the
5529 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5535 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5538 const U32 refcnt = SvREFCNT(sv);
5540 PERL_ARGS_ASSERT_SV_REPLACE;
5542 SV_CHECK_THINKFIRST_COW_DROP(sv);
5543 if (SvREFCNT(nsv) != 1) {
5544 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5545 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5547 if (SvMAGICAL(sv)) {
5551 sv_upgrade(nsv, SVt_PVMG);
5552 SvMAGIC_set(nsv, SvMAGIC(sv));
5553 SvFLAGS(nsv) |= SvMAGICAL(sv);
5555 SvMAGIC_set(sv, NULL);
5559 assert(!SvREFCNT(sv));
5560 #ifdef DEBUG_LEAKING_SCALARS
5561 sv->sv_flags = nsv->sv_flags;
5562 sv->sv_any = nsv->sv_any;
5563 sv->sv_refcnt = nsv->sv_refcnt;
5564 sv->sv_u = nsv->sv_u;
5566 StructCopy(nsv,sv,SV);
5568 if(SvTYPE(sv) == SVt_IV) {
5570 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5574 #ifdef PERL_OLD_COPY_ON_WRITE
5575 if (SvIsCOW_normal(nsv)) {
5576 /* We need to follow the pointers around the loop to make the
5577 previous SV point to sv, rather than nsv. */
5580 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5583 assert(SvPVX_const(current) == SvPVX_const(nsv));
5585 /* Make the SV before us point to the SV after us. */
5587 PerlIO_printf(Perl_debug_log, "previous is\n");
5589 PerlIO_printf(Perl_debug_log,
5590 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5591 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5593 SV_COW_NEXT_SV_SET(current, sv);
5596 SvREFCNT(sv) = refcnt;
5597 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5603 =for apidoc sv_clear
5605 Clear an SV: call any destructors, free up any memory used by the body,
5606 and free the body itself. The SV's head is I<not> freed, although
5607 its type is set to all 1's so that it won't inadvertently be assumed
5608 to be live during global destruction etc.
5609 This function should only be called when REFCNT is zero. Most of the time
5610 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5617 Perl_sv_clear(pTHX_ register SV *const sv)
5620 const U32 type = SvTYPE(sv);
5621 const struct body_details *const sv_type_details
5622 = bodies_by_type + type;
5625 PERL_ARGS_ASSERT_SV_CLEAR;
5626 assert(SvREFCNT(sv) == 0);
5627 assert(SvTYPE(sv) != SVTYPEMASK);
5629 if (type <= SVt_IV) {
5630 /* See the comment in sv.h about the collusion between this early
5631 return and the overloading of the NULL and IV slots in the size
5634 SV * const target = SvRV(sv);
5636 sv_del_backref(target, sv);
5638 SvREFCNT_dec(target);
5640 SvFLAGS(sv) &= SVf_BREAK;
5641 SvFLAGS(sv) |= SVTYPEMASK;
5646 if (PL_defstash && /* Still have a symbol table? */
5653 stash = SvSTASH(sv);
5654 destructor = StashHANDLER(stash,DESTROY);
5656 /* A constant subroutine can have no side effects, so
5657 don't bother calling it. */
5658 && !CvCONST(destructor)
5659 /* Don't bother calling an empty destructor */
5660 && (CvISXSUB(destructor)
5661 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5663 SV* const tmpref = newRV(sv);
5664 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5666 PUSHSTACKi(PERLSI_DESTROY);
5671 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5677 if(SvREFCNT(tmpref) < 2) {
5678 /* tmpref is not kept alive! */
5680 SvRV_set(tmpref, NULL);
5683 SvREFCNT_dec(tmpref);
5685 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5689 if (PL_in_clean_objs)
5690 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5692 /* DESTROY gave object new lease on life */
5698 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5699 SvOBJECT_off(sv); /* Curse the object. */
5700 if (type != SVt_PVIO)
5701 --PL_sv_objcount; /* XXX Might want something more general */
5704 if (type >= SVt_PVMG) {
5705 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5706 SvREFCNT_dec(SvOURSTASH(sv));
5707 } else if (SvMAGIC(sv))
5709 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5710 SvREFCNT_dec(SvSTASH(sv));
5713 /* case SVt_BIND: */
5716 IoIFP(sv) != PerlIO_stdin() &&
5717 IoIFP(sv) != PerlIO_stdout() &&
5718 IoIFP(sv) != PerlIO_stderr())
5720 io_close(MUTABLE_IO(sv), FALSE);
5722 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5723 PerlDir_close(IoDIRP(sv));
5724 IoDIRP(sv) = (DIR*)NULL;
5725 Safefree(IoTOP_NAME(sv));
5726 Safefree(IoFMT_NAME(sv));
5727 Safefree(IoBOTTOM_NAME(sv));
5730 /* FIXME for plugins */
5731 pregfree2((REGEXP*) sv);
5735 cv_undef(MUTABLE_CV(sv));
5738 if (PL_last_swash_hv == (const HV *)sv) {
5739 PL_last_swash_hv = NULL;
5741 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5742 hv_undef(MUTABLE_HV(sv));
5745 if (PL_comppad == MUTABLE_AV(sv)) {
5749 av_undef(MUTABLE_AV(sv));
5752 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5753 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5754 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5755 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5757 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5758 SvREFCNT_dec(LvTARG(sv));
5760 if (isGV_with_GP(sv)) {
5761 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5762 && HvNAME_get(stash))
5763 mro_method_changed_in(stash);
5764 gp_free(MUTABLE_GV(sv));
5766 unshare_hek(GvNAME_HEK(sv));
5767 /* If we're in a stash, we don't own a reference to it. However it does
5768 have a back reference to us, which needs to be cleared. */
5769 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5770 sv_del_backref(MUTABLE_SV(stash), sv);
5772 /* FIXME. There are probably more unreferenced pointers to SVs in the
5773 interpreter struct that we should check and tidy in a similar
5775 if ((const GV *)sv == PL_last_in_gv)
5776 PL_last_in_gv = NULL;
5782 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5785 SvOOK_offset(sv, offset);
5786 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5787 /* Don't even bother with turning off the OOK flag. */
5790 SV * const target = SvRV(sv);
5792 sv_del_backref(target, sv);
5794 SvREFCNT_dec(target);
5796 #ifdef PERL_OLD_COPY_ON_WRITE
5797 else if (SvPVX_const(sv)) {
5800 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5804 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5806 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5810 } else if (SvLEN(sv)) {
5811 Safefree(SvPVX_const(sv));
5815 else if (SvPVX_const(sv) && SvLEN(sv))
5816 Safefree(SvPVX_mutable(sv));
5817 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5818 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5827 SvFLAGS(sv) &= SVf_BREAK;
5828 SvFLAGS(sv) |= SVTYPEMASK;
5830 if (sv_type_details->arena) {
5831 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5832 &PL_body_roots[type]);
5834 else if (sv_type_details->body_size) {
5835 my_safefree(SvANY(sv));
5840 =for apidoc sv_newref
5842 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5849 Perl_sv_newref(pTHX_ SV *const sv)
5851 PERL_UNUSED_CONTEXT;
5860 Decrement an SV's reference count, and if it drops to zero, call
5861 C<sv_clear> to invoke destructors and free up any memory used by
5862 the body; finally, deallocate the SV's head itself.
5863 Normally called via a wrapper macro C<SvREFCNT_dec>.
5869 Perl_sv_free(pTHX_ SV *const sv)
5874 if (SvREFCNT(sv) == 0) {
5875 if (SvFLAGS(sv) & SVf_BREAK)
5876 /* this SV's refcnt has been artificially decremented to
5877 * trigger cleanup */
5879 if (PL_in_clean_all) /* All is fair */
5881 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5882 /* make sure SvREFCNT(sv)==0 happens very seldom */
5883 SvREFCNT(sv) = (~(U32)0)/2;
5886 if (ckWARN_d(WARN_INTERNAL)) {
5887 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5888 Perl_dump_sv_child(aTHX_ sv);
5890 #ifdef DEBUG_LEAKING_SCALARS
5893 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5894 if (PL_warnhook == PERL_WARNHOOK_FATAL
5895 || ckDEAD(packWARN(WARN_INTERNAL))) {
5896 /* Don't let Perl_warner cause us to escape our fate: */
5900 /* This may not return: */
5901 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5902 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5903 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5906 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5911 if (--(SvREFCNT(sv)) > 0)
5913 Perl_sv_free2(aTHX_ sv);
5917 Perl_sv_free2(pTHX_ SV *const sv)
5921 PERL_ARGS_ASSERT_SV_FREE2;
5925 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5926 "Attempt to free temp prematurely: SV 0x%"UVxf
5927 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5931 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5932 /* make sure SvREFCNT(sv)==0 happens very seldom */
5933 SvREFCNT(sv) = (~(U32)0)/2;
5944 Returns the length of the string in the SV. Handles magic and type
5945 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5951 Perl_sv_len(pTHX_ register SV *const sv)
5959 len = mg_length(sv);
5961 (void)SvPV_const(sv, len);
5966 =for apidoc sv_len_utf8
5968 Returns the number of characters in the string in an SV, counting wide
5969 UTF-8 bytes as a single character. Handles magic and type coercion.
5975 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5976 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5977 * (Note that the mg_len is not the length of the mg_ptr field.
5978 * This allows the cache to store the character length of the string without
5979 * needing to malloc() extra storage to attach to the mg_ptr.)
5984 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5990 return mg_length(sv);
5994 const U8 *s = (U8*)SvPV_const(sv, len);
5998 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6000 if (mg && mg->mg_len != -1) {
6002 if (PL_utf8cache < 0) {
6003 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6005 /* Need to turn the assertions off otherwise we may
6006 recurse infinitely while printing error messages.
6008 SAVEI8(PL_utf8cache);
6010 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6011 " real %"UVuf" for %"SVf,
6012 (UV) ulen, (UV) real, SVfARG(sv));
6017 ulen = Perl_utf8_length(aTHX_ s, s + len);
6018 if (!SvREADONLY(sv)) {
6020 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6021 &PL_vtbl_utf8, 0, 0);
6029 return Perl_utf8_length(aTHX_ s, s + len);
6033 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6036 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6039 const U8 *s = start;
6041 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6043 while (s < send && uoffset--)
6046 /* This is the existing behaviour. Possibly it should be a croak, as
6047 it's actually a bounds error */
6053 /* Given the length of the string in both bytes and UTF-8 characters, decide
6054 whether to walk forwards or backwards to find the byte corresponding to
6055 the passed in UTF-8 offset. */
6057 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6058 const STRLEN uoffset, const STRLEN uend)
6060 STRLEN backw = uend - uoffset;
6062 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6064 if (uoffset < 2 * backw) {
6065 /* The assumption is that going forwards is twice the speed of going
6066 forward (that's where the 2 * backw comes from).
6067 (The real figure of course depends on the UTF-8 data.) */
6068 return sv_pos_u2b_forwards(start, send, uoffset);
6073 while (UTF8_IS_CONTINUATION(*send))
6076 return send - start;
6079 /* For the string representation of the given scalar, find the byte
6080 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6081 give another position in the string, *before* the sought offset, which
6082 (which is always true, as 0, 0 is a valid pair of positions), which should
6083 help reduce the amount of linear searching.
6084 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6085 will be used to reduce the amount of linear searching. The cache will be
6086 created if necessary, and the found value offered to it for update. */
6088 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6089 const U8 *const send, const STRLEN uoffset,
6090 STRLEN uoffset0, STRLEN boffset0)
6092 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6095 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6097 assert (uoffset >= uoffset0);
6099 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6100 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6101 if ((*mgp)->mg_ptr) {
6102 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6103 if (cache[0] == uoffset) {
6104 /* An exact match. */
6107 if (cache[2] == uoffset) {
6108 /* An exact match. */
6112 if (cache[0] < uoffset) {
6113 /* The cache already knows part of the way. */
6114 if (cache[0] > uoffset0) {
6115 /* The cache knows more than the passed in pair */
6116 uoffset0 = cache[0];
6117 boffset0 = cache[1];
6119 if ((*mgp)->mg_len != -1) {
6120 /* And we know the end too. */
6122 + sv_pos_u2b_midway(start + boffset0, send,
6124 (*mgp)->mg_len - uoffset0);
6127 + sv_pos_u2b_forwards(start + boffset0,
6128 send, uoffset - uoffset0);
6131 else if (cache[2] < uoffset) {
6132 /* We're between the two cache entries. */
6133 if (cache[2] > uoffset0) {
6134 /* and the cache knows more than the passed in pair */
6135 uoffset0 = cache[2];
6136 boffset0 = cache[3];
6140 + sv_pos_u2b_midway(start + boffset0,
6143 cache[0] - uoffset0);
6146 + sv_pos_u2b_midway(start + boffset0,
6149 cache[2] - uoffset0);
6153 else if ((*mgp)->mg_len != -1) {
6154 /* If we can take advantage of a passed in offset, do so. */
6155 /* In fact, offset0 is either 0, or less than offset, so don't
6156 need to worry about the other possibility. */
6158 + sv_pos_u2b_midway(start + boffset0, send,
6160 (*mgp)->mg_len - uoffset0);
6165 if (!found || PL_utf8cache < 0) {
6166 const STRLEN real_boffset
6167 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6168 send, uoffset - uoffset0);
6170 if (found && PL_utf8cache < 0) {
6171 if (real_boffset != boffset) {
6172 /* Need to turn the assertions off otherwise we may recurse
6173 infinitely while printing error messages. */
6174 SAVEI8(PL_utf8cache);
6176 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6177 " real %"UVuf" for %"SVf,
6178 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6181 boffset = real_boffset;
6185 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6191 =for apidoc sv_pos_u2b
6193 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6194 the start of the string, to a count of the equivalent number of bytes; if
6195 lenp is non-zero, it does the same to lenp, but this time starting from
6196 the offset, rather than from the start of the string. Handles magic and
6203 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6204 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6205 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6210 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6215 PERL_ARGS_ASSERT_SV_POS_U2B;
6220 start = (U8*)SvPV_const(sv, len);
6222 STRLEN uoffset = (STRLEN) *offsetp;
6223 const U8 * const send = start + len;
6225 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6228 *offsetp = (I32) boffset;
6231 /* Convert the relative offset to absolute. */
6232 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6233 const STRLEN boffset2
6234 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6235 uoffset, boffset) - boffset;
6249 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6250 byte length pairing. The (byte) length of the total SV is passed in too,
6251 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6252 may not have updated SvCUR, so we can't rely on reading it directly.
6254 The proffered utf8/byte length pairing isn't used if the cache already has
6255 two pairs, and swapping either for the proffered pair would increase the
6256 RMS of the intervals between known byte offsets.
6258 The cache itself consists of 4 STRLEN values
6259 0: larger UTF-8 offset
6260 1: corresponding byte offset
6261 2: smaller UTF-8 offset
6262 3: corresponding byte offset
6264 Unused cache pairs have the value 0, 0.
6265 Keeping the cache "backwards" means that the invariant of
6266 cache[0] >= cache[2] is maintained even with empty slots, which means that
6267 the code that uses it doesn't need to worry if only 1 entry has actually
6268 been set to non-zero. It also makes the "position beyond the end of the
6269 cache" logic much simpler, as the first slot is always the one to start
6273 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6274 const STRLEN utf8, const STRLEN blen)
6278 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6284 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6286 (*mgp)->mg_len = -1;
6290 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6291 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6292 (*mgp)->mg_ptr = (char *) cache;
6296 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6297 /* SvPOKp() because it's possible that sv has string overloading, and
6298 therefore is a reference, hence SvPVX() is actually a pointer.
6299 This cures the (very real) symptoms of RT 69422, but I'm not actually
6300 sure whether we should even be caching the results of UTF-8
6301 operations on overloading, given that nothing stops overloading
6302 returning a different value every time it's called. */
6303 const U8 *start = (const U8 *) SvPVX_const(sv);
6304 const STRLEN realutf8 = utf8_length(start, start + byte);
6306 if (realutf8 != utf8) {
6307 /* Need to turn the assertions off otherwise we may recurse
6308 infinitely while printing error messages. */
6309 SAVEI8(PL_utf8cache);
6311 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6312 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6316 /* Cache is held with the later position first, to simplify the code
6317 that deals with unbounded ends. */
6319 ASSERT_UTF8_CACHE(cache);
6320 if (cache[1] == 0) {
6321 /* Cache is totally empty */
6324 } else if (cache[3] == 0) {
6325 if (byte > cache[1]) {
6326 /* New one is larger, so goes first. */
6327 cache[2] = cache[0];
6328 cache[3] = cache[1];
6336 #define THREEWAY_SQUARE(a,b,c,d) \
6337 ((float)((d) - (c))) * ((float)((d) - (c))) \
6338 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6339 + ((float)((b) - (a))) * ((float)((b) - (a)))
6341 /* Cache has 2 slots in use, and we know three potential pairs.
6342 Keep the two that give the lowest RMS distance. Do the
6343 calcualation in bytes simply because we always know the byte
6344 length. squareroot has the same ordering as the positive value,
6345 so don't bother with the actual square root. */
6346 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6347 if (byte > cache[1]) {
6348 /* New position is after the existing pair of pairs. */
6349 const float keep_earlier
6350 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6351 const float keep_later
6352 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6354 if (keep_later < keep_earlier) {
6355 if (keep_later < existing) {
6356 cache[2] = cache[0];
6357 cache[3] = cache[1];
6363 if (keep_earlier < existing) {
6369 else if (byte > cache[3]) {
6370 /* New position is between the existing pair of pairs. */
6371 const float keep_earlier
6372 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6373 const float keep_later
6374 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6376 if (keep_later < keep_earlier) {
6377 if (keep_later < existing) {
6383 if (keep_earlier < existing) {
6390 /* New position is before the existing pair of pairs. */
6391 const float keep_earlier
6392 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6393 const float keep_later
6394 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6396 if (keep_later < keep_earlier) {
6397 if (keep_later < existing) {
6403 if (keep_earlier < existing) {
6404 cache[0] = cache[2];
6405 cache[1] = cache[3];
6412 ASSERT_UTF8_CACHE(cache);
6415 /* We already know all of the way, now we may be able to walk back. The same
6416 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6417 backward is half the speed of walking forward. */
6419 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6420 const U8 *end, STRLEN endu)
6422 const STRLEN forw = target - s;
6423 STRLEN backw = end - target;
6425 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6427 if (forw < 2 * backw) {
6428 return utf8_length(s, target);
6431 while (end > target) {
6433 while (UTF8_IS_CONTINUATION(*end)) {
6442 =for apidoc sv_pos_b2u
6444 Converts the value pointed to by offsetp from a count of bytes from the
6445 start of the string, to a count of the equivalent number of UTF-8 chars.
6446 Handles magic and type coercion.
6452 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6453 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6458 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6461 const STRLEN byte = *offsetp;
6462 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6468 PERL_ARGS_ASSERT_SV_POS_B2U;
6473 s = (const U8*)SvPV_const(sv, blen);
6476 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6480 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6481 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6483 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6484 if (cache[1] == byte) {
6485 /* An exact match. */
6486 *offsetp = cache[0];
6489 if (cache[3] == byte) {
6490 /* An exact match. */
6491 *offsetp = cache[2];
6495 if (cache[1] < byte) {
6496 /* We already know part of the way. */
6497 if (mg->mg_len != -1) {
6498 /* Actually, we know the end too. */
6500 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6501 s + blen, mg->mg_len - cache[0]);
6503 len = cache[0] + utf8_length(s + cache[1], send);
6506 else if (cache[3] < byte) {
6507 /* We're between the two cached pairs, so we do the calculation
6508 offset by the byte/utf-8 positions for the earlier pair,
6509 then add the utf-8 characters from the string start to
6511 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6512 s + cache[1], cache[0] - cache[2])
6516 else { /* cache[3] > byte */
6517 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6521 ASSERT_UTF8_CACHE(cache);
6523 } else if (mg->mg_len != -1) {
6524 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6528 if (!found || PL_utf8cache < 0) {
6529 const STRLEN real_len = utf8_length(s, send);
6531 if (found && PL_utf8cache < 0) {
6532 if (len != real_len) {
6533 /* Need to turn the assertions off otherwise we may recurse
6534 infinitely while printing error messages. */
6535 SAVEI8(PL_utf8cache);
6537 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6538 " real %"UVuf" for %"SVf,
6539 (UV) len, (UV) real_len, SVfARG(sv));
6547 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6553 Returns a boolean indicating whether the strings in the two SVs are
6554 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6555 coerce its args to strings if necessary.
6561 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6570 SV* svrecode = NULL;
6577 /* if pv1 and pv2 are the same, second SvPV_const call may
6578 * invalidate pv1, so we may need to make a copy */
6579 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6580 pv1 = SvPV_const(sv1, cur1);
6581 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6583 pv1 = SvPV_const(sv1, cur1);
6591 pv2 = SvPV_const(sv2, cur2);
6593 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6594 /* Differing utf8ness.
6595 * Do not UTF8size the comparands as a side-effect. */
6598 svrecode = newSVpvn(pv2, cur2);
6599 sv_recode_to_utf8(svrecode, PL_encoding);
6600 pv2 = SvPV_const(svrecode, cur2);
6603 svrecode = newSVpvn(pv1, cur1);
6604 sv_recode_to_utf8(svrecode, PL_encoding);
6605 pv1 = SvPV_const(svrecode, cur1);
6607 /* Now both are in UTF-8. */
6609 SvREFCNT_dec(svrecode);
6614 bool is_utf8 = TRUE;
6617 /* sv1 is the UTF-8 one,
6618 * if is equal it must be downgrade-able */
6619 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6625 /* sv2 is the UTF-8 one,
6626 * if is equal it must be downgrade-able */
6627 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6633 /* Downgrade not possible - cannot be eq */
6641 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6643 SvREFCNT_dec(svrecode);
6653 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6654 string in C<sv1> is less than, equal to, or greater than the string in
6655 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6656 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6662 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6666 const char *pv1, *pv2;
6669 SV *svrecode = NULL;
6676 pv1 = SvPV_const(sv1, cur1);
6683 pv2 = SvPV_const(sv2, cur2);
6685 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6686 /* Differing utf8ness.
6687 * Do not UTF8size the comparands as a side-effect. */
6690 svrecode = newSVpvn(pv2, cur2);
6691 sv_recode_to_utf8(svrecode, PL_encoding);
6692 pv2 = SvPV_const(svrecode, cur2);
6695 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6700 svrecode = newSVpvn(pv1, cur1);
6701 sv_recode_to_utf8(svrecode, PL_encoding);
6702 pv1 = SvPV_const(svrecode, cur1);
6705 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6711 cmp = cur2 ? -1 : 0;
6715 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6718 cmp = retval < 0 ? -1 : 1;
6719 } else if (cur1 == cur2) {
6722 cmp = cur1 < cur2 ? -1 : 1;
6726 SvREFCNT_dec(svrecode);
6734 =for apidoc sv_cmp_locale
6736 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6737 'use bytes' aware, handles get magic, and will coerce its args to strings
6738 if necessary. See also C<sv_cmp>.
6744 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6747 #ifdef USE_LOCALE_COLLATE
6753 if (PL_collation_standard)
6757 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6759 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6761 if (!pv1 || !len1) {
6772 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6775 return retval < 0 ? -1 : 1;
6778 * When the result of collation is equality, that doesn't mean
6779 * that there are no differences -- some locales exclude some
6780 * characters from consideration. So to avoid false equalities,
6781 * we use the raw string as a tiebreaker.
6787 #endif /* USE_LOCALE_COLLATE */
6789 return sv_cmp(sv1, sv2);
6793 #ifdef USE_LOCALE_COLLATE
6796 =for apidoc sv_collxfrm
6798 Add Collate Transform magic to an SV if it doesn't already have it.
6800 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6801 scalar data of the variable, but transformed to such a format that a normal
6802 memory comparison can be used to compare the data according to the locale
6809 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6814 PERL_ARGS_ASSERT_SV_COLLXFRM;
6816 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6817 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6823 Safefree(mg->mg_ptr);
6824 s = SvPV_const(sv, len);
6825 if ((xf = mem_collxfrm(s, len, &xlen))) {
6827 #ifdef PERL_OLD_COPY_ON_WRITE
6829 sv_force_normal_flags(sv, 0);
6831 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6845 if (mg && mg->mg_ptr) {
6847 return mg->mg_ptr + sizeof(PL_collation_ix);
6855 #endif /* USE_LOCALE_COLLATE */
6860 Get a line from the filehandle and store it into the SV, optionally
6861 appending to the currently-stored string.
6867 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6872 register STDCHAR rslast;
6873 register STDCHAR *bp;
6878 PERL_ARGS_ASSERT_SV_GETS;
6880 if (SvTHINKFIRST(sv))
6881 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6882 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6884 However, perlbench says it's slower, because the existing swipe code
6885 is faster than copy on write.
6886 Swings and roundabouts. */
6887 SvUPGRADE(sv, SVt_PV);
6892 if (PerlIO_isutf8(fp)) {
6894 sv_utf8_upgrade_nomg(sv);
6895 sv_pos_u2b(sv,&append,0);
6897 } else if (SvUTF8(sv)) {
6898 SV * const tsv = newSV(0);
6899 sv_gets(tsv, fp, 0);
6900 sv_utf8_upgrade_nomg(tsv);
6901 SvCUR_set(sv,append);
6904 goto return_string_or_null;
6909 if (PerlIO_isutf8(fp))
6912 if (IN_PERL_COMPILETIME) {
6913 /* we always read code in line mode */
6917 else if (RsSNARF(PL_rs)) {
6918 /* If it is a regular disk file use size from stat() as estimate
6919 of amount we are going to read -- may result in mallocing
6920 more memory than we really need if the layers below reduce
6921 the size we read (e.g. CRLF or a gzip layer).
6924 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6925 const Off_t offset = PerlIO_tell(fp);
6926 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6927 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6933 else if (RsRECORD(PL_rs)) {
6941 /* Grab the size of the record we're getting */
6942 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6943 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6946 /* VMS wants read instead of fread, because fread doesn't respect */
6947 /* RMS record boundaries. This is not necessarily a good thing to be */
6948 /* doing, but we've got no other real choice - except avoid stdio
6949 as implementation - perhaps write a :vms layer ?
6951 fd = PerlIO_fileno(fp);
6952 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6953 bytesread = PerlIO_read(fp, buffer, recsize);
6956 bytesread = PerlLIO_read(fd, buffer, recsize);
6959 bytesread = PerlIO_read(fp, buffer, recsize);
6963 SvCUR_set(sv, bytesread + append);
6964 buffer[bytesread] = '\0';
6965 goto return_string_or_null;
6967 else if (RsPARA(PL_rs)) {
6973 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6974 if (PerlIO_isutf8(fp)) {
6975 rsptr = SvPVutf8(PL_rs, rslen);
6978 if (SvUTF8(PL_rs)) {
6979 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6980 Perl_croak(aTHX_ "Wide character in $/");
6983 rsptr = SvPV_const(PL_rs, rslen);
6987 rslast = rslen ? rsptr[rslen - 1] : '\0';
6989 if (rspara) { /* have to do this both before and after */
6990 do { /* to make sure file boundaries work right */
6993 i = PerlIO_getc(fp);
6997 PerlIO_ungetc(fp,i);
7003 /* See if we know enough about I/O mechanism to cheat it ! */
7005 /* This used to be #ifdef test - it is made run-time test for ease
7006 of abstracting out stdio interface. One call should be cheap
7007 enough here - and may even be a macro allowing compile
7011 if (PerlIO_fast_gets(fp)) {
7014 * We're going to steal some values from the stdio struct
7015 * and put EVERYTHING in the innermost loop into registers.
7017 register STDCHAR *ptr;
7021 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7022 /* An ungetc()d char is handled separately from the regular
7023 * buffer, so we getc() it back out and stuff it in the buffer.
7025 i = PerlIO_getc(fp);
7026 if (i == EOF) return 0;
7027 *(--((*fp)->_ptr)) = (unsigned char) i;
7031 /* Here is some breathtakingly efficient cheating */
7033 cnt = PerlIO_get_cnt(fp); /* get count into register */
7034 /* make sure we have the room */
7035 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7036 /* Not room for all of it
7037 if we are looking for a separator and room for some
7039 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7040 /* just process what we have room for */
7041 shortbuffered = cnt - SvLEN(sv) + append + 1;
7042 cnt -= shortbuffered;
7046 /* remember that cnt can be negative */
7047 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7052 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7053 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7054 DEBUG_P(PerlIO_printf(Perl_debug_log,
7055 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7056 DEBUG_P(PerlIO_printf(Perl_debug_log,
7057 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7058 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7059 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7064 while (cnt > 0) { /* this | eat */
7066 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7067 goto thats_all_folks; /* screams | sed :-) */
7071 Copy(ptr, bp, cnt, char); /* this | eat */
7072 bp += cnt; /* screams | dust */
7073 ptr += cnt; /* louder | sed :-) */
7078 if (shortbuffered) { /* oh well, must extend */
7079 cnt = shortbuffered;
7081 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7083 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7084 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7088 DEBUG_P(PerlIO_printf(Perl_debug_log,
7089 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7090 PTR2UV(ptr),(long)cnt));
7091 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7093 DEBUG_P(PerlIO_printf(Perl_debug_log,
7094 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7095 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7096 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7098 /* This used to call 'filbuf' in stdio form, but as that behaves like
7099 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7100 another abstraction. */
7101 i = PerlIO_getc(fp); /* get more characters */
7103 DEBUG_P(PerlIO_printf(Perl_debug_log,
7104 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7105 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7106 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7108 cnt = PerlIO_get_cnt(fp);
7109 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7110 DEBUG_P(PerlIO_printf(Perl_debug_log,
7111 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7113 if (i == EOF) /* all done for ever? */
7114 goto thats_really_all_folks;
7116 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7118 SvGROW(sv, bpx + cnt + 2);
7119 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7121 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7123 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7124 goto thats_all_folks;
7128 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7129 memNE((char*)bp - rslen, rsptr, rslen))
7130 goto screamer; /* go back to the fray */
7131 thats_really_all_folks:
7133 cnt += shortbuffered;
7134 DEBUG_P(PerlIO_printf(Perl_debug_log,
7135 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7136 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7137 DEBUG_P(PerlIO_printf(Perl_debug_log,
7138 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7139 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7140 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7142 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7143 DEBUG_P(PerlIO_printf(Perl_debug_log,
7144 "Screamer: done, len=%ld, string=|%.*s|\n",
7145 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7149 /*The big, slow, and stupid way. */
7150 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7151 STDCHAR *buf = NULL;
7152 Newx(buf, 8192, STDCHAR);
7160 register const STDCHAR * const bpe = buf + sizeof(buf);
7162 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7163 ; /* keep reading */
7167 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7168 /* Accomodate broken VAXC compiler, which applies U8 cast to
7169 * both args of ?: operator, causing EOF to change into 255
7172 i = (U8)buf[cnt - 1];
7178 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7180 sv_catpvn(sv, (char *) buf, cnt);
7182 sv_setpvn(sv, (char *) buf, cnt);
7184 if (i != EOF && /* joy */
7186 SvCUR(sv) < rslen ||
7187 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7191 * If we're reading from a TTY and we get a short read,
7192 * indicating that the user hit his EOF character, we need
7193 * to notice it now, because if we try to read from the TTY
7194 * again, the EOF condition will disappear.
7196 * The comparison of cnt to sizeof(buf) is an optimization
7197 * that prevents unnecessary calls to feof().
7201 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7205 #ifdef USE_HEAP_INSTEAD_OF_STACK
7210 if (rspara) { /* have to do this both before and after */
7211 while (i != EOF) { /* to make sure file boundaries work right */
7212 i = PerlIO_getc(fp);
7214 PerlIO_ungetc(fp,i);
7220 return_string_or_null:
7221 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7227 Auto-increment of the value in the SV, doing string to numeric conversion
7228 if necessary. Handles 'get' magic.
7234 Perl_sv_inc(pTHX_ register SV *const sv)
7243 if (SvTHINKFIRST(sv)) {
7245 sv_force_normal_flags(sv, 0);
7246 if (SvREADONLY(sv)) {
7247 if (IN_PERL_RUNTIME)
7248 Perl_croak(aTHX_ "%s", PL_no_modify);
7252 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7254 i = PTR2IV(SvRV(sv));
7259 flags = SvFLAGS(sv);
7260 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7261 /* It's (privately or publicly) a float, but not tested as an
7262 integer, so test it to see. */
7264 flags = SvFLAGS(sv);
7266 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7267 /* It's publicly an integer, or privately an integer-not-float */
7268 #ifdef PERL_PRESERVE_IVUV
7272 if (SvUVX(sv) == UV_MAX)
7273 sv_setnv(sv, UV_MAX_P1);
7275 (void)SvIOK_only_UV(sv);
7276 SvUV_set(sv, SvUVX(sv) + 1);
7278 if (SvIVX(sv) == IV_MAX)
7279 sv_setuv(sv, (UV)IV_MAX + 1);
7281 (void)SvIOK_only(sv);
7282 SvIV_set(sv, SvIVX(sv) + 1);
7287 if (flags & SVp_NOK) {
7288 const NV was = SvNVX(sv);
7289 if (NV_OVERFLOWS_INTEGERS_AT &&
7290 was >= NV_OVERFLOWS_INTEGERS_AT) {
7291 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7292 "Lost precision when incrementing %" NVff " by 1",
7295 (void)SvNOK_only(sv);
7296 SvNV_set(sv, was + 1.0);
7300 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7301 if ((flags & SVTYPEMASK) < SVt_PVIV)
7302 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7303 (void)SvIOK_only(sv);
7308 while (isALPHA(*d)) d++;
7309 while (isDIGIT(*d)) d++;
7310 if (d < SvEND(sv)) {
7311 #ifdef PERL_PRESERVE_IVUV
7312 /* Got to punt this as an integer if needs be, but we don't issue
7313 warnings. Probably ought to make the sv_iv_please() that does
7314 the conversion if possible, and silently. */
7315 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7316 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7317 /* Need to try really hard to see if it's an integer.
7318 9.22337203685478e+18 is an integer.
7319 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7320 so $a="9.22337203685478e+18"; $a+0; $a++
7321 needs to be the same as $a="9.22337203685478e+18"; $a++
7328 /* sv_2iv *should* have made this an NV */
7329 if (flags & SVp_NOK) {
7330 (void)SvNOK_only(sv);
7331 SvNV_set(sv, SvNVX(sv) + 1.0);
7334 /* I don't think we can get here. Maybe I should assert this
7335 And if we do get here I suspect that sv_setnv will croak. NWC
7337 #if defined(USE_LONG_DOUBLE)
7338 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",
7339 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7341 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7342 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7345 #endif /* PERL_PRESERVE_IVUV */
7346 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7350 while (d >= SvPVX_const(sv)) {
7358 /* MKS: The original code here died if letters weren't consecutive.
7359 * at least it didn't have to worry about non-C locales. The
7360 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7361 * arranged in order (although not consecutively) and that only
7362 * [A-Za-z] are accepted by isALPHA in the C locale.
7364 if (*d != 'z' && *d != 'Z') {
7365 do { ++*d; } while (!isALPHA(*d));
7368 *(d--) -= 'z' - 'a';
7373 *(d--) -= 'z' - 'a' + 1;
7377 /* oh,oh, the number grew */
7378 SvGROW(sv, SvCUR(sv) + 2);
7379 SvCUR_set(sv, SvCUR(sv) + 1);
7380 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7391 Auto-decrement of the value in the SV, doing string to numeric conversion
7392 if necessary. Handles 'get' magic.
7398 Perl_sv_dec(pTHX_ register SV *const sv)
7406 if (SvTHINKFIRST(sv)) {
7408 sv_force_normal_flags(sv, 0);
7409 if (SvREADONLY(sv)) {
7410 if (IN_PERL_RUNTIME)
7411 Perl_croak(aTHX_ "%s", PL_no_modify);
7415 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7417 i = PTR2IV(SvRV(sv));
7422 /* Unlike sv_inc we don't have to worry about string-never-numbers
7423 and keeping them magic. But we mustn't warn on punting */
7424 flags = SvFLAGS(sv);
7425 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7426 /* It's publicly an integer, or privately an integer-not-float */
7427 #ifdef PERL_PRESERVE_IVUV
7431 if (SvUVX(sv) == 0) {
7432 (void)SvIOK_only(sv);
7436 (void)SvIOK_only_UV(sv);
7437 SvUV_set(sv, SvUVX(sv) - 1);
7440 if (SvIVX(sv) == IV_MIN) {
7441 sv_setnv(sv, (NV)IV_MIN);
7445 (void)SvIOK_only(sv);
7446 SvIV_set(sv, SvIVX(sv) - 1);
7451 if (flags & SVp_NOK) {
7454 const NV was = SvNVX(sv);
7455 if (NV_OVERFLOWS_INTEGERS_AT &&
7456 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7457 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7458 "Lost precision when decrementing %" NVff " by 1",
7461 (void)SvNOK_only(sv);
7462 SvNV_set(sv, was - 1.0);
7466 if (!(flags & SVp_POK)) {
7467 if ((flags & SVTYPEMASK) < SVt_PVIV)
7468 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7470 (void)SvIOK_only(sv);
7473 #ifdef PERL_PRESERVE_IVUV
7475 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7476 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7477 /* Need to try really hard to see if it's an integer.
7478 9.22337203685478e+18 is an integer.
7479 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7480 so $a="9.22337203685478e+18"; $a+0; $a--
7481 needs to be the same as $a="9.22337203685478e+18"; $a--
7488 /* sv_2iv *should* have made this an NV */
7489 if (flags & SVp_NOK) {
7490 (void)SvNOK_only(sv);
7491 SvNV_set(sv, SvNVX(sv) - 1.0);
7494 /* I don't think we can get here. Maybe I should assert this
7495 And if we do get here I suspect that sv_setnv will croak. NWC
7497 #if defined(USE_LONG_DOUBLE)
7498 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",
7499 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7501 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7502 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7506 #endif /* PERL_PRESERVE_IVUV */
7507 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7510 /* this define is used to eliminate a chunk of duplicated but shared logic
7511 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7512 * used anywhere but here - yves
7514 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7517 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7521 =for apidoc sv_mortalcopy
7523 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7524 The new SV is marked as mortal. It will be destroyed "soon", either by an
7525 explicit call to FREETMPS, or by an implicit call at places such as
7526 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7531 /* Make a string that will exist for the duration of the expression
7532 * evaluation. Actually, it may have to last longer than that, but
7533 * hopefully we won't free it until it has been assigned to a
7534 * permanent location. */
7537 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7543 sv_setsv(sv,oldstr);
7544 PUSH_EXTEND_MORTAL__SV_C(sv);
7550 =for apidoc sv_newmortal
7552 Creates a new null SV which is mortal. The reference count of the SV is
7553 set to 1. It will be destroyed "soon", either by an explicit call to
7554 FREETMPS, or by an implicit call at places such as statement boundaries.
7555 See also C<sv_mortalcopy> and C<sv_2mortal>.
7561 Perl_sv_newmortal(pTHX)
7567 SvFLAGS(sv) = SVs_TEMP;
7568 PUSH_EXTEND_MORTAL__SV_C(sv);
7574 =for apidoc newSVpvn_flags
7576 Creates a new SV and copies a string into it. The reference count for the
7577 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7578 string. You are responsible for ensuring that the source string is at least
7579 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7580 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7581 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7582 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7583 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7585 #define newSVpvn_utf8(s, len, u) \
7586 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7592 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7597 /* All the flags we don't support must be zero.
7598 And we're new code so I'm going to assert this from the start. */
7599 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7601 sv_setpvn(sv,s,len);
7603 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7604 * and do what it does outselves here.
7605 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7606 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7607 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7608 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7611 SvFLAGS(sv) |= flags;
7613 if(flags & SVs_TEMP){
7614 PUSH_EXTEND_MORTAL__SV_C(sv);
7621 =for apidoc sv_2mortal
7623 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7624 by an explicit call to FREETMPS, or by an implicit call at places such as
7625 statement boundaries. SvTEMP() is turned on which means that the SV's
7626 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7627 and C<sv_mortalcopy>.
7633 Perl_sv_2mortal(pTHX_ register SV *const sv)
7638 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7640 PUSH_EXTEND_MORTAL__SV_C(sv);
7648 Creates a new SV and copies a string into it. The reference count for the
7649 SV is set to 1. If C<len> is zero, Perl will compute the length using
7650 strlen(). For efficiency, consider using C<newSVpvn> instead.
7656 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7662 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7667 =for apidoc newSVpvn
7669 Creates a new SV and copies a string into it. The reference count for the
7670 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7671 string. You are responsible for ensuring that the source string is at least
7672 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7678 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7684 sv_setpvn(sv,s,len);
7689 =for apidoc newSVhek
7691 Creates a new SV from the hash key structure. It will generate scalars that
7692 point to the shared string table where possible. Returns a new (undefined)
7693 SV if the hek is NULL.
7699 Perl_newSVhek(pTHX_ const HEK *const hek)
7709 if (HEK_LEN(hek) == HEf_SVKEY) {
7710 return newSVsv(*(SV**)HEK_KEY(hek));
7712 const int flags = HEK_FLAGS(hek);
7713 if (flags & HVhek_WASUTF8) {
7715 Andreas would like keys he put in as utf8 to come back as utf8
7717 STRLEN utf8_len = HEK_LEN(hek);
7718 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7719 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7722 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7724 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7725 /* We don't have a pointer to the hv, so we have to replicate the
7726 flag into every HEK. This hv is using custom a hasing
7727 algorithm. Hence we can't return a shared string scalar, as
7728 that would contain the (wrong) hash value, and might get passed
7729 into an hv routine with a regular hash.
7730 Similarly, a hash that isn't using shared hash keys has to have
7731 the flag in every key so that we know not to try to call
7732 share_hek_kek on it. */
7734 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7739 /* This will be overwhelminly the most common case. */
7741 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7742 more efficient than sharepvn(). */
7746 sv_upgrade(sv, SVt_PV);
7747 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7748 SvCUR_set(sv, HEK_LEN(hek));
7761 =for apidoc newSVpvn_share
7763 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7764 table. If the string does not already exist in the table, it is created
7765 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7766 value is used; otherwise the hash is computed. The string's hash can be later
7767 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7768 that as the string table is used for shared hash keys these strings will have
7769 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7775 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7779 bool is_utf8 = FALSE;
7780 const char *const orig_src = src;
7783 STRLEN tmplen = -len;
7785 /* See the note in hv.c:hv_fetch() --jhi */
7786 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7790 PERL_HASH(hash, src, len);
7792 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7793 changes here, update it there too. */
7794 sv_upgrade(sv, SVt_PV);
7795 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7803 if (src != orig_src)
7809 #if defined(PERL_IMPLICIT_CONTEXT)
7811 /* pTHX_ magic can't cope with varargs, so this is a no-context
7812 * version of the main function, (which may itself be aliased to us).
7813 * Don't access this version directly.
7817 Perl_newSVpvf_nocontext(const char *const pat, ...)
7823 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7825 va_start(args, pat);
7826 sv = vnewSVpvf(pat, &args);
7833 =for apidoc newSVpvf
7835 Creates a new SV and initializes it with the string formatted like
7842 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7847 PERL_ARGS_ASSERT_NEWSVPVF;
7849 va_start(args, pat);
7850 sv = vnewSVpvf(pat, &args);
7855 /* backend for newSVpvf() and newSVpvf_nocontext() */
7858 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7863 PERL_ARGS_ASSERT_VNEWSVPVF;
7866 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7873 Creates a new SV and copies a floating point value into it.
7874 The reference count for the SV is set to 1.
7880 Perl_newSVnv(pTHX_ const NV n)
7893 Creates a new SV and copies an integer into it. The reference count for the
7900 Perl_newSViv(pTHX_ const IV i)
7913 Creates a new SV and copies an unsigned integer into it.
7914 The reference count for the SV is set to 1.
7920 Perl_newSVuv(pTHX_ const UV u)
7931 =for apidoc newSV_type
7933 Creates a new SV, of the type specified. The reference count for the new SV
7940 Perl_newSV_type(pTHX_ const svtype type)
7945 sv_upgrade(sv, type);
7950 =for apidoc newRV_noinc
7952 Creates an RV wrapper for an SV. The reference count for the original
7953 SV is B<not> incremented.
7959 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7962 register SV *sv = newSV_type(SVt_IV);
7964 PERL_ARGS_ASSERT_NEWRV_NOINC;
7967 SvRV_set(sv, tmpRef);
7972 /* newRV_inc is the official function name to use now.
7973 * newRV_inc is in fact #defined to newRV in sv.h
7977 Perl_newRV(pTHX_ SV *const sv)
7981 PERL_ARGS_ASSERT_NEWRV;
7983 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7989 Creates a new SV which is an exact duplicate of the original SV.
7996 Perl_newSVsv(pTHX_ register SV *const old)
8003 if (SvTYPE(old) == SVTYPEMASK) {
8004 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8008 /* SV_GMAGIC is the default for sv_setv()
8009 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8010 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8011 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8016 =for apidoc sv_reset
8018 Underlying implementation for the C<reset> Perl function.
8019 Note that the perl-level function is vaguely deprecated.
8025 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8028 char todo[PERL_UCHAR_MAX+1];
8030 PERL_ARGS_ASSERT_SV_RESET;
8035 if (!*s) { /* reset ?? searches */
8036 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8038 const U32 count = mg->mg_len / sizeof(PMOP**);
8039 PMOP **pmp = (PMOP**) mg->mg_ptr;
8040 PMOP *const *const end = pmp + count;
8044 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8046 (*pmp)->op_pmflags &= ~PMf_USED;
8054 /* reset variables */
8056 if (!HvARRAY(stash))
8059 Zero(todo, 256, char);
8062 I32 i = (unsigned char)*s;
8066 max = (unsigned char)*s++;
8067 for ( ; i <= max; i++) {
8070 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8072 for (entry = HvARRAY(stash)[i];
8074 entry = HeNEXT(entry))
8079 if (!todo[(U8)*HeKEY(entry)])
8081 gv = MUTABLE_GV(HeVAL(entry));
8084 if (SvTHINKFIRST(sv)) {
8085 if (!SvREADONLY(sv) && SvROK(sv))
8087 /* XXX Is this continue a bug? Why should THINKFIRST
8088 exempt us from resetting arrays and hashes? */
8092 if (SvTYPE(sv) >= SVt_PV) {
8094 if (SvPVX_const(sv) != NULL)
8102 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8104 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8107 # if defined(USE_ENVIRON_ARRAY)
8110 # endif /* USE_ENVIRON_ARRAY */
8121 Using various gambits, try to get an IO from an SV: the IO slot if its a
8122 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8123 named after the PV if we're a string.
8129 Perl_sv_2io(pTHX_ SV *const sv)
8134 PERL_ARGS_ASSERT_SV_2IO;
8136 switch (SvTYPE(sv)) {
8138 io = MUTABLE_IO(sv);
8141 if (isGV_with_GP(sv)) {
8142 gv = MUTABLE_GV(sv);
8145 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8151 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8153 return sv_2io(SvRV(sv));
8154 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8160 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8169 Using various gambits, try to get a CV from an SV; in addition, try if
8170 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8171 The flags in C<lref> are passed to gv_fetchsv.
8177 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8183 PERL_ARGS_ASSERT_SV_2CV;
8190 switch (SvTYPE(sv)) {
8194 return MUTABLE_CV(sv);
8201 if (isGV_with_GP(sv)) {
8202 gv = MUTABLE_GV(sv);
8211 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8213 tryAMAGICunDEREF(to_cv);
8216 if (SvTYPE(sv) == SVt_PVCV) {
8217 cv = MUTABLE_CV(sv);
8222 else if(isGV_with_GP(sv))
8223 gv = MUTABLE_GV(sv);
8225 Perl_croak(aTHX_ "Not a subroutine reference");
8227 else if (isGV_with_GP(sv)) {
8229 gv = MUTABLE_GV(sv);
8232 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8238 /* Some flags to gv_fetchsv mean don't really create the GV */
8239 if (!isGV_with_GP(gv)) {
8245 if (lref && !GvCVu(gv)) {
8249 gv_efullname3(tmpsv, gv, NULL);
8250 /* XXX this is probably not what they think they're getting.
8251 * It has the same effect as "sub name;", i.e. just a forward
8253 newSUB(start_subparse(FALSE, 0),
8254 newSVOP(OP_CONST, 0, tmpsv),
8258 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8259 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8268 Returns true if the SV has a true value by Perl's rules.
8269 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8270 instead use an in-line version.
8276 Perl_sv_true(pTHX_ register SV *const sv)
8281 register const XPV* const tXpv = (XPV*)SvANY(sv);
8283 (tXpv->xpv_cur > 1 ||
8284 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8291 return SvIVX(sv) != 0;
8294 return SvNVX(sv) != 0.0;
8296 return sv_2bool(sv);
8302 =for apidoc sv_pvn_force
8304 Get a sensible string out of the SV somehow.
8305 A private implementation of the C<SvPV_force> macro for compilers which
8306 can't cope with complex macro expressions. Always use the macro instead.
8308 =for apidoc sv_pvn_force_flags
8310 Get a sensible string out of the SV somehow.
8311 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8312 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8313 implemented in terms of this function.
8314 You normally want to use the various wrapper macros instead: see
8315 C<SvPV_force> and C<SvPV_force_nomg>
8321 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8325 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8327 if (SvTHINKFIRST(sv) && !SvROK(sv))
8328 sv_force_normal_flags(sv, 0);
8338 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8339 const char * const ref = sv_reftype(sv,0);
8341 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8342 ref, OP_NAME(PL_op));
8344 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8346 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8347 || isGV_with_GP(sv))
8348 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8350 s = sv_2pv_flags(sv, &len, flags);
8354 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8357 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8358 SvGROW(sv, len + 1);
8359 Move(s,SvPVX(sv),len,char);
8361 SvPVX(sv)[len] = '\0';
8364 SvPOK_on(sv); /* validate pointer */
8366 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8367 PTR2UV(sv),SvPVX_const(sv)));
8370 return SvPVX_mutable(sv);
8374 =for apidoc sv_pvbyten_force
8376 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8382 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8384 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8386 sv_pvn_force(sv,lp);
8387 sv_utf8_downgrade(sv,0);
8393 =for apidoc sv_pvutf8n_force
8395 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8401 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8403 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8405 sv_pvn_force(sv,lp);
8406 sv_utf8_upgrade(sv);
8412 =for apidoc sv_reftype
8414 Returns a string describing what the SV is a reference to.
8420 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8422 PERL_ARGS_ASSERT_SV_REFTYPE;
8424 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8425 inside return suggests a const propagation bug in g++. */
8426 if (ob && SvOBJECT(sv)) {
8427 char * const name = HvNAME_get(SvSTASH(sv));
8428 return name ? name : (char *) "__ANON__";
8431 switch (SvTYPE(sv)) {
8446 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8447 /* tied lvalues should appear to be
8448 * scalars for backwards compatitbility */
8449 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8450 ? "SCALAR" : "LVALUE");
8451 case SVt_PVAV: return "ARRAY";
8452 case SVt_PVHV: return "HASH";
8453 case SVt_PVCV: return "CODE";
8454 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8455 ? "GLOB" : "SCALAR");
8456 case SVt_PVFM: return "FORMAT";
8457 case SVt_PVIO: return "IO";
8458 case SVt_BIND: return "BIND";
8459 case SVt_REGEXP: return "REGEXP";
8460 default: return "UNKNOWN";
8466 =for apidoc sv_isobject
8468 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8469 object. If the SV is not an RV, or if the object is not blessed, then this
8476 Perl_sv_isobject(pTHX_ SV *sv)
8492 Returns a boolean indicating whether the SV is blessed into the specified
8493 class. This does not check for subtypes; use C<sv_derived_from> to verify
8494 an inheritance relationship.
8500 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8504 PERL_ARGS_ASSERT_SV_ISA;
8514 hvname = HvNAME_get(SvSTASH(sv));
8518 return strEQ(hvname, name);
8524 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8525 it will be upgraded to one. If C<classname> is non-null then the new SV will
8526 be blessed in the specified package. The new SV is returned and its
8527 reference count is 1.
8533 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8538 PERL_ARGS_ASSERT_NEWSVRV;
8542 SV_CHECK_THINKFIRST_COW_DROP(rv);
8543 (void)SvAMAGIC_off(rv);
8545 if (SvTYPE(rv) >= SVt_PVMG) {
8546 const U32 refcnt = SvREFCNT(rv);
8550 SvREFCNT(rv) = refcnt;
8552 sv_upgrade(rv, SVt_IV);
8553 } else if (SvROK(rv)) {
8554 SvREFCNT_dec(SvRV(rv));
8556 prepare_SV_for_RV(rv);
8564 HV* const stash = gv_stashpv(classname, GV_ADD);
8565 (void)sv_bless(rv, stash);
8571 =for apidoc sv_setref_pv
8573 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8574 argument will be upgraded to an RV. That RV will be modified to point to
8575 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8576 into the SV. The C<classname> argument indicates the package for the
8577 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8578 will have a reference count of 1, and the RV will be returned.
8580 Do not use with other Perl types such as HV, AV, SV, CV, because those
8581 objects will become corrupted by the pointer copy process.
8583 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8589 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8593 PERL_ARGS_ASSERT_SV_SETREF_PV;
8596 sv_setsv(rv, &PL_sv_undef);
8600 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8605 =for apidoc sv_setref_iv
8607 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8608 argument will be upgraded to an RV. That RV will be modified to point to
8609 the new SV. The C<classname> argument indicates the package for the
8610 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8611 will have a reference count of 1, and the RV will be returned.
8617 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8619 PERL_ARGS_ASSERT_SV_SETREF_IV;
8621 sv_setiv(newSVrv(rv,classname), iv);
8626 =for apidoc sv_setref_uv
8628 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8629 argument will be upgraded to an RV. That RV will be modified to point to
8630 the new SV. The C<classname> argument indicates the package for the
8631 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8632 will have a reference count of 1, and the RV will be returned.
8638 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8640 PERL_ARGS_ASSERT_SV_SETREF_UV;
8642 sv_setuv(newSVrv(rv,classname), uv);
8647 =for apidoc sv_setref_nv
8649 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8650 argument will be upgraded to an RV. That RV will be modified to point to
8651 the new SV. The C<classname> argument indicates the package for the
8652 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8653 will have a reference count of 1, and the RV will be returned.
8659 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8661 PERL_ARGS_ASSERT_SV_SETREF_NV;
8663 sv_setnv(newSVrv(rv,classname), nv);
8668 =for apidoc sv_setref_pvn
8670 Copies a string into a new SV, optionally blessing the SV. The length of the
8671 string must be specified with C<n>. The C<rv> argument will be upgraded to
8672 an RV. That RV will be modified to point to the new SV. The C<classname>
8673 argument indicates the package for the blessing. Set C<classname> to
8674 C<NULL> to avoid the blessing. The new SV will have a reference count
8675 of 1, and the RV will be returned.
8677 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8683 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8684 const char *const pv, const STRLEN n)
8686 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8688 sv_setpvn(newSVrv(rv,classname), pv, n);
8693 =for apidoc sv_bless
8695 Blesses an SV into a specified package. The SV must be an RV. The package
8696 must be designated by its stash (see C<gv_stashpv()>). The reference count
8697 of the SV is unaffected.
8703 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8708 PERL_ARGS_ASSERT_SV_BLESS;
8711 Perl_croak(aTHX_ "Can't bless non-reference value");
8713 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8714 if (SvIsCOW(tmpRef))
8715 sv_force_normal_flags(tmpRef, 0);
8716 if (SvREADONLY(tmpRef))
8717 Perl_croak(aTHX_ "%s", PL_no_modify);
8718 if (SvOBJECT(tmpRef)) {
8719 if (SvTYPE(tmpRef) != SVt_PVIO)
8721 SvREFCNT_dec(SvSTASH(tmpRef));
8724 SvOBJECT_on(tmpRef);
8725 if (SvTYPE(tmpRef) != SVt_PVIO)
8727 SvUPGRADE(tmpRef, SVt_PVMG);
8728 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8733 (void)SvAMAGIC_off(sv);
8735 if(SvSMAGICAL(tmpRef))
8736 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8744 /* Downgrades a PVGV to a PVMG.
8748 S_sv_unglob(pTHX_ SV *const sv)
8753 SV * const temp = sv_newmortal();
8755 PERL_ARGS_ASSERT_SV_UNGLOB;
8757 assert(SvTYPE(sv) == SVt_PVGV);
8759 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8762 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8763 && HvNAME_get(stash))
8764 mro_method_changed_in(stash);
8765 gp_free(MUTABLE_GV(sv));
8768 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8772 if (GvNAME_HEK(sv)) {
8773 unshare_hek(GvNAME_HEK(sv));
8775 isGV_with_GP_off(sv);
8777 /* need to keep SvANY(sv) in the right arena */
8778 xpvmg = new_XPVMG();
8779 StructCopy(SvANY(sv), xpvmg, XPVMG);
8780 del_XPVGV(SvANY(sv));
8783 SvFLAGS(sv) &= ~SVTYPEMASK;
8784 SvFLAGS(sv) |= SVt_PVMG;
8786 /* Intentionally not calling any local SET magic, as this isn't so much a
8787 set operation as merely an internal storage change. */
8788 sv_setsv_flags(sv, temp, 0);
8792 =for apidoc sv_unref_flags
8794 Unsets the RV status of the SV, and decrements the reference count of
8795 whatever was being referenced by the RV. This can almost be thought of
8796 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8797 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8798 (otherwise the decrementing is conditional on the reference count being
8799 different from one or the reference being a readonly SV).
8806 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8808 SV* const target = SvRV(ref);
8810 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8812 if (SvWEAKREF(ref)) {
8813 sv_del_backref(target, ref);
8815 SvRV_set(ref, NULL);
8818 SvRV_set(ref, NULL);
8820 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8821 assigned to as BEGIN {$a = \"Foo"} will fail. */
8822 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8823 SvREFCNT_dec(target);
8824 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8825 sv_2mortal(target); /* Schedule for freeing later */
8829 =for apidoc sv_untaint
8831 Untaint an SV. Use C<SvTAINTED_off> instead.
8836 Perl_sv_untaint(pTHX_ SV *const sv)
8838 PERL_ARGS_ASSERT_SV_UNTAINT;
8840 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8841 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8848 =for apidoc sv_tainted
8850 Test an SV for taintedness. Use C<SvTAINTED> instead.
8855 Perl_sv_tainted(pTHX_ SV *const sv)
8857 PERL_ARGS_ASSERT_SV_TAINTED;
8859 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8860 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8861 if (mg && (mg->mg_len & 1) )
8868 =for apidoc sv_setpviv
8870 Copies an integer into the given SV, also updating its string value.
8871 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8877 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8879 char buf[TYPE_CHARS(UV)];
8881 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8883 PERL_ARGS_ASSERT_SV_SETPVIV;
8885 sv_setpvn(sv, ptr, ebuf - ptr);
8889 =for apidoc sv_setpviv_mg
8891 Like C<sv_setpviv>, but also handles 'set' magic.
8897 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8899 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8905 #if defined(PERL_IMPLICIT_CONTEXT)
8907 /* pTHX_ magic can't cope with varargs, so this is a no-context
8908 * version of the main function, (which may itself be aliased to us).
8909 * Don't access this version directly.
8913 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8918 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8920 va_start(args, pat);
8921 sv_vsetpvf(sv, pat, &args);
8925 /* pTHX_ magic can't cope with varargs, so this is a no-context
8926 * version of the main function, (which may itself be aliased to us).
8927 * Don't access this version directly.
8931 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8936 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8938 va_start(args, pat);
8939 sv_vsetpvf_mg(sv, pat, &args);
8945 =for apidoc sv_setpvf
8947 Works like C<sv_catpvf> but copies the text into the SV instead of
8948 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8954 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8958 PERL_ARGS_ASSERT_SV_SETPVF;
8960 va_start(args, pat);
8961 sv_vsetpvf(sv, pat, &args);
8966 =for apidoc sv_vsetpvf
8968 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8969 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8971 Usually used via its frontend C<sv_setpvf>.
8977 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8979 PERL_ARGS_ASSERT_SV_VSETPVF;
8981 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8985 =for apidoc sv_setpvf_mg
8987 Like C<sv_setpvf>, but also handles 'set' magic.
8993 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8997 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8999 va_start(args, pat);
9000 sv_vsetpvf_mg(sv, pat, &args);
9005 =for apidoc sv_vsetpvf_mg
9007 Like C<sv_vsetpvf>, but also handles 'set' magic.
9009 Usually used via its frontend C<sv_setpvf_mg>.
9015 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9017 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9019 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9023 #if defined(PERL_IMPLICIT_CONTEXT)
9025 /* pTHX_ magic can't cope with varargs, so this is a no-context
9026 * version of the main function, (which may itself be aliased to us).
9027 * Don't access this version directly.
9031 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9036 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9038 va_start(args, pat);
9039 sv_vcatpvf(sv, pat, &args);
9043 /* pTHX_ magic can't cope with varargs, so this is a no-context
9044 * version of the main function, (which may itself be aliased to us).
9045 * Don't access this version directly.
9049 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9054 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9056 va_start(args, pat);
9057 sv_vcatpvf_mg(sv, pat, &args);
9063 =for apidoc sv_catpvf
9065 Processes its arguments like C<sprintf> and appends the formatted
9066 output to an SV. If the appended data contains "wide" characters
9067 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9068 and characters >255 formatted with %c), the original SV might get
9069 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9070 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9071 valid UTF-8; if the original SV was bytes, the pattern should be too.
9076 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9080 PERL_ARGS_ASSERT_SV_CATPVF;
9082 va_start(args, pat);
9083 sv_vcatpvf(sv, pat, &args);
9088 =for apidoc sv_vcatpvf
9090 Processes its arguments like C<vsprintf> and appends the formatted output
9091 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9093 Usually used via its frontend C<sv_catpvf>.
9099 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9101 PERL_ARGS_ASSERT_SV_VCATPVF;
9103 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9107 =for apidoc sv_catpvf_mg
9109 Like C<sv_catpvf>, but also handles 'set' magic.
9115 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9119 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9121 va_start(args, pat);
9122 sv_vcatpvf_mg(sv, pat, &args);
9127 =for apidoc sv_vcatpvf_mg
9129 Like C<sv_vcatpvf>, but also handles 'set' magic.
9131 Usually used via its frontend C<sv_catpvf_mg>.
9137 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9139 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9141 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9146 =for apidoc sv_vsetpvfn
9148 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9151 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9157 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9158 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9160 PERL_ARGS_ASSERT_SV_VSETPVFN;
9163 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9167 S_expect_number(pTHX_ char **const pattern)
9172 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9174 switch (**pattern) {
9175 case '1': case '2': case '3':
9176 case '4': case '5': case '6':
9177 case '7': case '8': case '9':
9178 var = *(*pattern)++ - '0';
9179 while (isDIGIT(**pattern)) {
9180 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9182 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9190 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9192 const int neg = nv < 0;
9195 PERL_ARGS_ASSERT_F0CONVERT;
9203 if (uv & 1 && uv == nv)
9204 uv--; /* Round to even */
9206 const unsigned dig = uv % 10;
9219 =for apidoc sv_vcatpvfn
9221 Processes its arguments like C<vsprintf> and appends the formatted output
9222 to an SV. Uses an array of SVs if the C style variable argument list is
9223 missing (NULL). When running with taint checks enabled, indicates via
9224 C<maybe_tainted> if results are untrustworthy (often due to the use of
9227 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9233 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9234 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9235 vec_utf8 = DO_UTF8(vecsv);
9237 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9240 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9241 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9249 static const char nullstr[] = "(null)";
9251 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9252 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9254 /* Times 4: a decimal digit takes more than 3 binary digits.
9255 * NV_DIG: mantissa takes than many decimal digits.
9256 * Plus 32: Playing safe. */
9257 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9258 /* large enough for "%#.#f" --chip */
9259 /* what about long double NVs? --jhi */
9261 PERL_ARGS_ASSERT_SV_VCATPVFN;
9262 PERL_UNUSED_ARG(maybe_tainted);
9264 /* no matter what, this is a string now */
9265 (void)SvPV_force(sv, origlen);
9267 /* special-case "", "%s", and "%-p" (SVf - see below) */
9270 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9272 const char * const s = va_arg(*args, char*);
9273 sv_catpv(sv, s ? s : nullstr);
9275 else if (svix < svmax) {
9276 sv_catsv(sv, *svargs);
9280 if (args && patlen == 3 && pat[0] == '%' &&
9281 pat[1] == '-' && pat[2] == 'p') {
9282 argsv = MUTABLE_SV(va_arg(*args, void*));
9283 sv_catsv(sv, argsv);
9287 #ifndef USE_LONG_DOUBLE
9288 /* special-case "%.<number>[gf]" */
9289 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9290 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9291 unsigned digits = 0;
9295 while (*pp >= '0' && *pp <= '9')
9296 digits = 10 * digits + (*pp++ - '0');
9297 if (pp - pat == (int)patlen - 1) {
9305 /* Add check for digits != 0 because it seems that some
9306 gconverts are buggy in this case, and we don't yet have
9307 a Configure test for this. */
9308 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9309 /* 0, point, slack */
9310 Gconvert(nv, (int)digits, 0, ebuf);
9312 if (*ebuf) /* May return an empty string for digits==0 */
9315 } else if (!digits) {
9318 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9319 sv_catpvn(sv, p, l);
9325 #endif /* !USE_LONG_DOUBLE */
9327 if (!args && svix < svmax && DO_UTF8(*svargs))
9330 patend = (char*)pat + patlen;
9331 for (p = (char*)pat; p < patend; p = q) {
9334 bool vectorize = FALSE;
9335 bool vectorarg = FALSE;
9336 bool vec_utf8 = FALSE;
9342 bool has_precis = FALSE;
9344 const I32 osvix = svix;
9345 bool is_utf8 = FALSE; /* is this item utf8? */
9346 #ifdef HAS_LDBL_SPRINTF_BUG
9347 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9348 with sfio - Allen <allens@cpan.org> */
9349 bool fix_ldbl_sprintf_bug = FALSE;
9353 U8 utf8buf[UTF8_MAXBYTES+1];
9354 STRLEN esignlen = 0;
9356 const char *eptr = NULL;
9357 const char *fmtstart;
9360 const U8 *vecstr = NULL;
9367 /* we need a long double target in case HAS_LONG_DOUBLE but
9370 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9378 const char *dotstr = ".";
9379 STRLEN dotstrlen = 1;
9380 I32 efix = 0; /* explicit format parameter index */
9381 I32 ewix = 0; /* explicit width index */
9382 I32 epix = 0; /* explicit precision index */
9383 I32 evix = 0; /* explicit vector index */
9384 bool asterisk = FALSE;
9386 /* echo everything up to the next format specification */
9387 for (q = p; q < patend && *q != '%'; ++q) ;
9389 if (has_utf8 && !pat_utf8)
9390 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9392 sv_catpvn(sv, p, q - p);
9401 We allow format specification elements in this order:
9402 \d+\$ explicit format parameter index
9404 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9405 0 flag (as above): repeated to allow "v02"
9406 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9407 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9409 [%bcdefginopsuxDFOUX] format (mandatory)
9414 As of perl5.9.3, printf format checking is on by default.
9415 Internally, perl uses %p formats to provide an escape to
9416 some extended formatting. This block deals with those
9417 extensions: if it does not match, (char*)q is reset and
9418 the normal format processing code is used.
9420 Currently defined extensions are:
9421 %p include pointer address (standard)
9422 %-p (SVf) include an SV (previously %_)
9423 %-<num>p include an SV with precision <num>
9424 %<num>p reserved for future extensions
9426 Robin Barker 2005-07-14
9428 %1p (VDf) removed. RMB 2007-10-19
9435 n = expect_number(&q);
9442 argsv = MUTABLE_SV(va_arg(*args, void*));
9443 eptr = SvPV_const(argsv, elen);
9449 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9450 "internal %%<num>p might conflict with future printf extensions");
9456 if ( (width = expect_number(&q)) ) {
9471 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9500 if ( (ewix = expect_number(&q)) )
9509 if ((vectorarg = asterisk)) {
9522 width = expect_number(&q);
9528 vecsv = va_arg(*args, SV*);
9530 vecsv = (evix > 0 && evix <= svmax)
9531 ? svargs[evix-1] : &PL_sv_undef;
9533 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9535 dotstr = SvPV_const(vecsv, dotstrlen);
9536 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9537 bad with tied or overloaded values that return UTF8. */
9540 else if (has_utf8) {
9541 vecsv = sv_mortalcopy(vecsv);
9542 sv_utf8_upgrade(vecsv);
9543 dotstr = SvPV_const(vecsv, dotstrlen);
9550 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9551 vecsv = svargs[efix ? efix-1 : svix++];
9552 vecstr = (U8*)SvPV_const(vecsv,veclen);
9553 vec_utf8 = DO_UTF8(vecsv);
9555 /* if this is a version object, we need to convert
9556 * back into v-string notation and then let the
9557 * vectorize happen normally
9559 if (sv_derived_from(vecsv, "version")) {
9560 char *version = savesvpv(vecsv);
9561 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9562 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9563 "vector argument not supported with alpha versions");
9566 vecsv = sv_newmortal();
9567 scan_vstring(version, version + veclen, vecsv);
9568 vecstr = (U8*)SvPV_const(vecsv, veclen);
9569 vec_utf8 = DO_UTF8(vecsv);
9581 i = va_arg(*args, int);
9583 i = (ewix ? ewix <= svmax : svix < svmax) ?
9584 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9586 width = (i < 0) ? -i : i;
9596 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9598 /* XXX: todo, support specified precision parameter */
9602 i = va_arg(*args, int);
9604 i = (ewix ? ewix <= svmax : svix < svmax)
9605 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9607 has_precis = !(i < 0);
9612 precis = precis * 10 + (*q++ - '0');
9621 case 'I': /* Ix, I32x, and I64x */
9623 if (q[1] == '6' && q[2] == '4') {
9629 if (q[1] == '3' && q[2] == '2') {
9639 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9650 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9651 if (*(q + 1) == 'l') { /* lld, llf */
9677 if (!vectorize && !args) {
9679 const I32 i = efix-1;
9680 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9682 argsv = (svix >= 0 && svix < svmax)
9683 ? svargs[svix++] : &PL_sv_undef;
9694 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9696 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9698 eptr = (char*)utf8buf;
9699 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9713 eptr = va_arg(*args, char*);
9715 elen = strlen(eptr);
9717 eptr = (char *)nullstr;
9718 elen = sizeof nullstr - 1;
9722 eptr = SvPV_const(argsv, elen);
9723 if (DO_UTF8(argsv)) {
9724 STRLEN old_precis = precis;
9725 if (has_precis && precis < elen) {
9726 STRLEN ulen = sv_len_utf8(argsv);
9727 I32 p = precis > ulen ? ulen : precis;
9728 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9731 if (width) { /* fudge width (can't fudge elen) */
9732 if (has_precis && precis < elen)
9733 width += precis - old_precis;
9735 width += elen - sv_len_utf8(argsv);
9742 if (has_precis && precis < elen)
9749 if (alt || vectorize)
9751 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9772 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9781 esignbuf[esignlen++] = plus;
9785 case 'h': iv = (short)va_arg(*args, int); break;
9786 case 'l': iv = va_arg(*args, long); break;
9787 case 'V': iv = va_arg(*args, IV); break;
9788 default: iv = va_arg(*args, int); break;
9791 iv = va_arg(*args, Quad_t); break;
9798 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9800 case 'h': iv = (short)tiv; break;
9801 case 'l': iv = (long)tiv; break;
9803 default: iv = tiv; break;
9806 iv = (Quad_t)tiv; break;
9812 if ( !vectorize ) /* we already set uv above */
9817 esignbuf[esignlen++] = plus;
9821 esignbuf[esignlen++] = '-';
9865 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9876 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9877 case 'l': uv = va_arg(*args, unsigned long); break;
9878 case 'V': uv = va_arg(*args, UV); break;
9879 default: uv = va_arg(*args, unsigned); break;
9882 uv = va_arg(*args, Uquad_t); break;
9889 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9891 case 'h': uv = (unsigned short)tuv; break;
9892 case 'l': uv = (unsigned long)tuv; break;
9894 default: uv = tuv; break;
9897 uv = (Uquad_t)tuv; break;
9906 char *ptr = ebuf + sizeof ebuf;
9907 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9913 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9919 esignbuf[esignlen++] = '0';
9920 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9928 if (alt && *ptr != '0')
9937 esignbuf[esignlen++] = '0';
9938 esignbuf[esignlen++] = c;
9941 default: /* it had better be ten or less */
9945 } while (uv /= base);
9948 elen = (ebuf + sizeof ebuf) - ptr;
9952 zeros = precis - elen;
9953 else if (precis == 0 && elen == 1 && *eptr == '0'
9954 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9957 /* a precision nullifies the 0 flag. */
9964 /* FLOATING POINT */
9967 c = 'f'; /* maybe %F isn't supported here */
9975 /* This is evil, but floating point is even more evil */
9977 /* for SV-style calling, we can only get NV
9978 for C-style calling, we assume %f is double;
9979 for simplicity we allow any of %Lf, %llf, %qf for long double
9983 #if defined(USE_LONG_DOUBLE)
9987 /* [perl #20339] - we should accept and ignore %lf rather than die */
9991 #if defined(USE_LONG_DOUBLE)
9992 intsize = args ? 0 : 'q';
9996 #if defined(HAS_LONG_DOUBLE)
10005 /* now we need (long double) if intsize == 'q', else (double) */
10007 #if LONG_DOUBLESIZE > DOUBLESIZE
10009 va_arg(*args, long double) :
10010 va_arg(*args, double)
10012 va_arg(*args, double)
10017 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10018 else. frexp() has some unspecified behaviour for those three */
10019 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10021 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10022 will cast our (long double) to (double) */
10023 (void)Perl_frexp(nv, &i);
10024 if (i == PERL_INT_MIN)
10025 Perl_die(aTHX_ "panic: frexp");
10027 need = BIT_DIGITS(i);
10029 need += has_precis ? precis : 6; /* known default */
10034 #ifdef HAS_LDBL_SPRINTF_BUG
10035 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10036 with sfio - Allen <allens@cpan.org> */
10039 # define MY_DBL_MAX DBL_MAX
10040 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10041 # if DOUBLESIZE >= 8
10042 # define MY_DBL_MAX 1.7976931348623157E+308L
10044 # define MY_DBL_MAX 3.40282347E+38L
10048 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10049 # define MY_DBL_MAX_BUG 1L
10051 # define MY_DBL_MAX_BUG MY_DBL_MAX
10055 # define MY_DBL_MIN DBL_MIN
10056 # else /* XXX guessing! -Allen */
10057 # if DOUBLESIZE >= 8
10058 # define MY_DBL_MIN 2.2250738585072014E-308L
10060 # define MY_DBL_MIN 1.17549435E-38L
10064 if ((intsize == 'q') && (c == 'f') &&
10065 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10066 (need < DBL_DIG)) {
10067 /* it's going to be short enough that
10068 * long double precision is not needed */
10070 if ((nv <= 0L) && (nv >= -0L))
10071 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10073 /* would use Perl_fp_class as a double-check but not
10074 * functional on IRIX - see perl.h comments */
10076 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10077 /* It's within the range that a double can represent */
10078 #if defined(DBL_MAX) && !defined(DBL_MIN)
10079 if ((nv >= ((long double)1/DBL_MAX)) ||
10080 (nv <= (-(long double)1/DBL_MAX)))
10082 fix_ldbl_sprintf_bug = TRUE;
10085 if (fix_ldbl_sprintf_bug == TRUE) {
10095 # undef MY_DBL_MAX_BUG
10098 #endif /* HAS_LDBL_SPRINTF_BUG */
10100 need += 20; /* fudge factor */
10101 if (PL_efloatsize < need) {
10102 Safefree(PL_efloatbuf);
10103 PL_efloatsize = need + 20; /* more fudge */
10104 Newx(PL_efloatbuf, PL_efloatsize, char);
10105 PL_efloatbuf[0] = '\0';
10108 if ( !(width || left || plus || alt) && fill != '0'
10109 && has_precis && intsize != 'q' ) { /* Shortcuts */
10110 /* See earlier comment about buggy Gconvert when digits,
10112 if ( c == 'g' && precis) {
10113 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10114 /* May return an empty string for digits==0 */
10115 if (*PL_efloatbuf) {
10116 elen = strlen(PL_efloatbuf);
10117 goto float_converted;
10119 } else if ( c == 'f' && !precis) {
10120 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10125 char *ptr = ebuf + sizeof ebuf;
10128 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10129 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10130 if (intsize == 'q') {
10131 /* Copy the one or more characters in a long double
10132 * format before the 'base' ([efgEFG]) character to
10133 * the format string. */
10134 static char const prifldbl[] = PERL_PRIfldbl;
10135 char const *p = prifldbl + sizeof(prifldbl) - 3;
10136 while (p >= prifldbl) { *--ptr = *p--; }
10141 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10146 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10158 /* No taint. Otherwise we are in the strange situation
10159 * where printf() taints but print($float) doesn't.
10161 #if defined(HAS_LONG_DOUBLE)
10162 elen = ((intsize == 'q')
10163 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10164 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10166 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10170 eptr = PL_efloatbuf;
10178 i = SvCUR(sv) - origlen;
10181 case 'h': *(va_arg(*args, short*)) = i; break;
10182 default: *(va_arg(*args, int*)) = i; break;
10183 case 'l': *(va_arg(*args, long*)) = i; break;
10184 case 'V': *(va_arg(*args, IV*)) = i; break;
10187 *(va_arg(*args, Quad_t*)) = i; break;
10194 sv_setuv_mg(argsv, (UV)i);
10195 continue; /* not "break" */
10202 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10203 && ckWARN(WARN_PRINTF))
10205 SV * const msg = sv_newmortal();
10206 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10207 (PL_op->op_type == OP_PRTF) ? "" : "s");
10208 if (fmtstart < patend) {
10209 const char * const fmtend = q < patend ? q : patend;
10211 sv_catpvs(msg, "\"%");
10212 for (f = fmtstart; f < fmtend; f++) {
10214 sv_catpvn(msg, f, 1);
10216 Perl_sv_catpvf(aTHX_ msg,
10217 "\\%03"UVof, (UV)*f & 0xFF);
10220 sv_catpvs(msg, "\"");
10222 sv_catpvs(msg, "end of string");
10224 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10227 /* output mangled stuff ... */
10233 /* ... right here, because formatting flags should not apply */
10234 SvGROW(sv, SvCUR(sv) + elen + 1);
10236 Copy(eptr, p, elen, char);
10239 SvCUR_set(sv, p - SvPVX_const(sv));
10241 continue; /* not "break" */
10244 if (is_utf8 != has_utf8) {
10247 sv_utf8_upgrade(sv);
10250 const STRLEN old_elen = elen;
10251 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10252 sv_utf8_upgrade(nsv);
10253 eptr = SvPVX_const(nsv);
10256 if (width) { /* fudge width (can't fudge elen) */
10257 width += elen - old_elen;
10263 have = esignlen + zeros + elen;
10265 Perl_croak_nocontext("%s", PL_memory_wrap);
10267 need = (have > width ? have : width);
10270 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10271 Perl_croak_nocontext("%s", PL_memory_wrap);
10272 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10274 if (esignlen && fill == '0') {
10276 for (i = 0; i < (int)esignlen; i++)
10277 *p++ = esignbuf[i];
10279 if (gap && !left) {
10280 memset(p, fill, gap);
10283 if (esignlen && fill != '0') {
10285 for (i = 0; i < (int)esignlen; i++)
10286 *p++ = esignbuf[i];
10290 for (i = zeros; i; i--)
10294 Copy(eptr, p, elen, char);
10298 memset(p, ' ', gap);
10303 Copy(dotstr, p, dotstrlen, char);
10307 vectorize = FALSE; /* done iterating over vecstr */
10314 SvCUR_set(sv, p - SvPVX_const(sv));
10322 /* =========================================================================
10324 =head1 Cloning an interpreter
10326 All the macros and functions in this section are for the private use of
10327 the main function, perl_clone().
10329 The foo_dup() functions make an exact copy of an existing foo thingy.
10330 During the course of a cloning, a hash table is used to map old addresses
10331 to new addresses. The table is created and manipulated with the
10332 ptr_table_* functions.
10336 * =========================================================================*/
10339 #if defined(USE_ITHREADS)
10341 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10342 #ifndef GpREFCNT_inc
10343 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10347 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10348 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10349 If this changes, please unmerge ss_dup.
10350 Likewise, sv_dup_inc_multiple() relies on this fact. */
10351 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10352 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10353 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10354 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10355 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10356 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10357 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10358 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10359 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10360 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10361 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10362 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10363 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10364 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10366 /* clone a parser */
10369 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10373 PERL_ARGS_ASSERT_PARSER_DUP;
10378 /* look for it in the table first */
10379 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10383 /* create anew and remember what it is */
10384 Newxz(parser, 1, yy_parser);
10385 ptr_table_store(PL_ptr_table, proto, parser);
10387 parser->yyerrstatus = 0;
10388 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10390 /* XXX these not yet duped */
10391 parser->old_parser = NULL;
10392 parser->stack = NULL;
10394 parser->stack_size = 0;
10395 /* XXX parser->stack->state = 0; */
10397 /* XXX eventually, just Copy() most of the parser struct ? */
10399 parser->lex_brackets = proto->lex_brackets;
10400 parser->lex_casemods = proto->lex_casemods;
10401 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10402 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10403 parser->lex_casestack = savepvn(proto->lex_casestack,
10404 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10405 parser->lex_defer = proto->lex_defer;
10406 parser->lex_dojoin = proto->lex_dojoin;
10407 parser->lex_expect = proto->lex_expect;
10408 parser->lex_formbrack = proto->lex_formbrack;
10409 parser->lex_inpat = proto->lex_inpat;
10410 parser->lex_inwhat = proto->lex_inwhat;
10411 parser->lex_op = proto->lex_op;
10412 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10413 parser->lex_starts = proto->lex_starts;
10414 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10415 parser->multi_close = proto->multi_close;
10416 parser->multi_open = proto->multi_open;
10417 parser->multi_start = proto->multi_start;
10418 parser->multi_end = proto->multi_end;
10419 parser->pending_ident = proto->pending_ident;
10420 parser->preambled = proto->preambled;
10421 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10422 parser->linestr = sv_dup_inc(proto->linestr, param);
10423 parser->expect = proto->expect;
10424 parser->copline = proto->copline;
10425 parser->last_lop_op = proto->last_lop_op;
10426 parser->lex_state = proto->lex_state;
10427 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10428 /* rsfp_filters entries have fake IoDIRP() */
10429 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10430 parser->in_my = proto->in_my;
10431 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10432 parser->error_count = proto->error_count;
10435 parser->linestr = sv_dup_inc(proto->linestr, param);
10438 char * const ols = SvPVX(proto->linestr);
10439 char * const ls = SvPVX(parser->linestr);
10441 parser->bufptr = ls + (proto->bufptr >= ols ?
10442 proto->bufptr - ols : 0);
10443 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10444 proto->oldbufptr - ols : 0);
10445 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10446 proto->oldoldbufptr - ols : 0);
10447 parser->linestart = ls + (proto->linestart >= ols ?
10448 proto->linestart - ols : 0);
10449 parser->last_uni = ls + (proto->last_uni >= ols ?
10450 proto->last_uni - ols : 0);
10451 parser->last_lop = ls + (proto->last_lop >= ols ?
10452 proto->last_lop - ols : 0);
10454 parser->bufend = ls + SvCUR(parser->linestr);
10457 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10461 parser->endwhite = proto->endwhite;
10462 parser->faketokens = proto->faketokens;
10463 parser->lasttoke = proto->lasttoke;
10464 parser->nextwhite = proto->nextwhite;
10465 parser->realtokenstart = proto->realtokenstart;
10466 parser->skipwhite = proto->skipwhite;
10467 parser->thisclose = proto->thisclose;
10468 parser->thismad = proto->thismad;
10469 parser->thisopen = proto->thisopen;
10470 parser->thisstuff = proto->thisstuff;
10471 parser->thistoken = proto->thistoken;
10472 parser->thiswhite = proto->thiswhite;
10474 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10475 parser->curforce = proto->curforce;
10477 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10478 Copy(proto->nexttype, parser->nexttype, 5, I32);
10479 parser->nexttoke = proto->nexttoke;
10482 /* XXX should clone saved_curcop here, but we aren't passed
10483 * proto_perl; so do it in perl_clone_using instead */
10489 /* duplicate a file handle */
10492 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10496 PERL_ARGS_ASSERT_FP_DUP;
10497 PERL_UNUSED_ARG(type);
10500 return (PerlIO*)NULL;
10502 /* look for it in the table first */
10503 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10507 /* create anew and remember what it is */
10508 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10509 ptr_table_store(PL_ptr_table, fp, ret);
10513 /* duplicate a directory handle */
10516 Perl_dirp_dup(pTHX_ DIR *const dp)
10518 PERL_UNUSED_CONTEXT;
10525 /* duplicate a typeglob */
10528 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10532 PERL_ARGS_ASSERT_GP_DUP;
10536 /* look for it in the table first */
10537 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10541 /* create anew and remember what it is */
10543 ptr_table_store(PL_ptr_table, gp, ret);
10546 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10547 on Newxz() to do this for us. */
10548 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10549 ret->gp_io = io_dup_inc(gp->gp_io, param);
10550 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10551 ret->gp_av = av_dup_inc(gp->gp_av, param);
10552 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10553 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10554 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10555 ret->gp_cvgen = gp->gp_cvgen;
10556 ret->gp_line = gp->gp_line;
10557 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10561 /* duplicate a chain of magic */
10564 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10566 MAGIC *mgret = NULL;
10567 MAGIC **mgprev_p = &mgret;
10569 PERL_ARGS_ASSERT_MG_DUP;
10571 for (; mg; mg = mg->mg_moremagic) {
10573 Newx(nmg, 1, MAGIC);
10575 mgprev_p = &(nmg->mg_moremagic);
10577 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10578 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10579 from the original commit adding Perl_mg_dup() - revision 4538.
10580 Similarly there is the annotation "XXX random ptr?" next to the
10581 assignment to nmg->mg_ptr. */
10584 /* FIXME for plugins
10585 if (nmg->mg_type == PERL_MAGIC_qr) {
10586 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10590 if(nmg->mg_type == PERL_MAGIC_backref) {
10591 /* The backref AV has its reference count deliberately bumped by
10594 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10597 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10598 ? sv_dup_inc(nmg->mg_obj, param)
10599 : sv_dup(nmg->mg_obj, param);
10602 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10603 if (nmg->mg_len > 0) {
10604 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10605 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10606 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10608 AMT * const namtp = (AMT*)nmg->mg_ptr;
10609 sv_dup_inc_multiple((SV**)(namtp->table),
10610 (SV**)(namtp->table), NofAMmeth, param);
10613 else if (nmg->mg_len == HEf_SVKEY)
10614 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10616 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10617 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10623 #endif /* USE_ITHREADS */
10625 /* create a new pointer-mapping table */
10628 Perl_ptr_table_new(pTHX)
10631 PERL_UNUSED_CONTEXT;
10633 Newx(tbl, 1, PTR_TBL_t);
10634 tbl->tbl_max = 511;
10635 tbl->tbl_items = 0;
10636 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10640 #define PTR_TABLE_HASH(ptr) \
10641 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10644 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10645 following define) and at call to new_body_inline made below in
10646 Perl_ptr_table_store()
10649 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10651 /* map an existing pointer using a table */
10653 STATIC PTR_TBL_ENT_t *
10654 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10656 PTR_TBL_ENT_t *tblent;
10657 const UV hash = PTR_TABLE_HASH(sv);
10659 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10661 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10662 for (; tblent; tblent = tblent->next) {
10663 if (tblent->oldval == sv)
10670 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10672 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10674 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10675 PERL_UNUSED_CONTEXT;
10677 return tblent ? tblent->newval : NULL;
10680 /* add a new entry to a pointer-mapping table */
10683 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10685 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10687 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10688 PERL_UNUSED_CONTEXT;
10691 tblent->newval = newsv;
10693 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10695 new_body_inline(tblent, PTE_SVSLOT);
10697 tblent->oldval = oldsv;
10698 tblent->newval = newsv;
10699 tblent->next = tbl->tbl_ary[entry];
10700 tbl->tbl_ary[entry] = tblent;
10702 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10703 ptr_table_split(tbl);
10707 /* double the hash bucket size of an existing ptr table */
10710 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10712 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10713 const UV oldsize = tbl->tbl_max + 1;
10714 UV newsize = oldsize * 2;
10717 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10718 PERL_UNUSED_CONTEXT;
10720 Renew(ary, newsize, PTR_TBL_ENT_t*);
10721 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10722 tbl->tbl_max = --newsize;
10723 tbl->tbl_ary = ary;
10724 for (i=0; i < oldsize; i++, ary++) {
10725 PTR_TBL_ENT_t **curentp, **entp, *ent;
10728 curentp = ary + oldsize;
10729 for (entp = ary, ent = *ary; ent; ent = *entp) {
10730 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10732 ent->next = *curentp;
10742 /* remove all the entries from a ptr table */
10745 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10747 if (tbl && tbl->tbl_items) {
10748 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10749 UV riter = tbl->tbl_max;
10752 PTR_TBL_ENT_t *entry = array[riter];
10755 PTR_TBL_ENT_t * const oentry = entry;
10756 entry = entry->next;
10761 tbl->tbl_items = 0;
10765 /* clear and free a ptr table */
10768 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10773 ptr_table_clear(tbl);
10774 Safefree(tbl->tbl_ary);
10778 #if defined(USE_ITHREADS)
10781 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10783 PERL_ARGS_ASSERT_RVPV_DUP;
10786 SvRV_set(dstr, SvWEAKREF(sstr)
10787 ? sv_dup(SvRV_const(sstr), param)
10788 : sv_dup_inc(SvRV_const(sstr), param));
10791 else if (SvPVX_const(sstr)) {
10792 /* Has something there */
10794 /* Normal PV - clone whole allocated space */
10795 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10796 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10797 /* Not that normal - actually sstr is copy on write.
10798 But we are a true, independant SV, so: */
10799 SvREADONLY_off(dstr);
10804 /* Special case - not normally malloced for some reason */
10805 if (isGV_with_GP(sstr)) {
10806 /* Don't need to do anything here. */
10808 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10809 /* A "shared" PV - clone it as "shared" PV */
10811 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10815 /* Some other special case - random pointer */
10816 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10821 /* Copy the NULL */
10822 SvPV_set(dstr, NULL);
10826 /* duplicate a list of SVs. source and dest may point to the same memory. */
10828 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10829 SSize_t items, CLONE_PARAMS *const param)
10831 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10833 while (items-- > 0) {
10834 *dest++ = sv_dup_inc(*source++, param);
10840 /* duplicate an SV of any type (including AV, HV etc) */
10843 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10848 PERL_ARGS_ASSERT_SV_DUP;
10852 if (SvTYPE(sstr) == SVTYPEMASK) {
10853 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10858 /* look for it in the table first */
10859 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10863 if(param->flags & CLONEf_JOIN_IN) {
10864 /** We are joining here so we don't want do clone
10865 something that is bad **/
10866 if (SvTYPE(sstr) == SVt_PVHV) {
10867 const HEK * const hvname = HvNAME_HEK(sstr);
10869 /** don't clone stashes if they already exist **/
10870 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10874 /* create anew and remember what it is */
10877 #ifdef DEBUG_LEAKING_SCALARS
10878 dstr->sv_debug_optype = sstr->sv_debug_optype;
10879 dstr->sv_debug_line = sstr->sv_debug_line;
10880 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10881 dstr->sv_debug_cloned = 1;
10882 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10885 ptr_table_store(PL_ptr_table, sstr, dstr);
10888 SvFLAGS(dstr) = SvFLAGS(sstr);
10889 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10890 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10893 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10894 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10895 (void*)PL_watch_pvx, SvPVX_const(sstr));
10898 /* don't clone objects whose class has asked us not to */
10899 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10904 switch (SvTYPE(sstr)) {
10906 SvANY(dstr) = NULL;
10909 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10911 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10913 SvIV_set(dstr, SvIVX(sstr));
10917 SvANY(dstr) = new_XNV();
10918 SvNV_set(dstr, SvNVX(sstr));
10920 /* case SVt_BIND: */
10923 /* These are all the types that need complex bodies allocating. */
10925 const svtype sv_type = SvTYPE(sstr);
10926 const struct body_details *const sv_type_details
10927 = bodies_by_type + sv_type;
10931 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10946 assert(sv_type_details->body_size);
10947 if (sv_type_details->arena) {
10948 new_body_inline(new_body, sv_type);
10950 = (void*)((char*)new_body - sv_type_details->offset);
10952 new_body = new_NOARENA(sv_type_details);
10956 SvANY(dstr) = new_body;
10959 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10960 ((char*)SvANY(dstr)) + sv_type_details->offset,
10961 sv_type_details->copy, char);
10963 Copy(((char*)SvANY(sstr)),
10964 ((char*)SvANY(dstr)),
10965 sv_type_details->body_size + sv_type_details->offset, char);
10968 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10969 && !isGV_with_GP(dstr))
10970 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10972 /* The Copy above means that all the source (unduplicated) pointers
10973 are now in the destination. We can check the flags and the
10974 pointers in either, but it's possible that there's less cache
10975 missing by always going for the destination.
10976 FIXME - instrument and check that assumption */
10977 if (sv_type >= SVt_PVMG) {
10978 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10979 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10980 } else if (SvMAGIC(dstr))
10981 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10983 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10986 /* The cast silences a GCC warning about unhandled types. */
10987 switch ((int)sv_type) {
10997 /* FIXME for plugins */
10998 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11001 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11002 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11003 LvTARG(dstr) = dstr;
11004 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11005 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11007 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11009 if(isGV_with_GP(sstr)) {
11010 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11011 /* Don't call sv_add_backref here as it's going to be
11012 created as part of the magic cloning of the symbol
11014 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11015 at the point of this comment. */
11016 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11017 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11018 (void)GpREFCNT_inc(GvGP(dstr));
11020 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11023 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11024 if (IoOFP(dstr) == IoIFP(sstr))
11025 IoOFP(dstr) = IoIFP(dstr);
11027 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11028 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11029 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11030 /* I have no idea why fake dirp (rsfps)
11031 should be treated differently but otherwise
11032 we end up with leaks -- sky*/
11033 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11034 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11035 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11037 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11038 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11039 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11040 if (IoDIRP(dstr)) {
11041 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11044 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11047 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11048 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11049 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11052 /* avoid cloning an empty array */
11053 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11054 SV **dst_ary, **src_ary;
11055 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11057 src_ary = AvARRAY((const AV *)sstr);
11058 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11059 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11060 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11061 AvALLOC((const AV *)dstr) = dst_ary;
11062 if (AvREAL((const AV *)sstr)) {
11063 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11067 while (items-- > 0)
11068 *dst_ary++ = sv_dup(*src_ary++, param);
11070 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11071 while (items-- > 0) {
11072 *dst_ary++ = &PL_sv_undef;
11076 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11077 AvALLOC((const AV *)dstr) = (SV**)NULL;
11078 AvMAX( (const AV *)dstr) = -1;
11079 AvFILLp((const AV *)dstr) = -1;
11083 if (HvARRAY((const HV *)sstr)) {
11085 const bool sharekeys = !!HvSHAREKEYS(sstr);
11086 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11087 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11089 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11090 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11092 HvARRAY(dstr) = (HE**)darray;
11093 while (i <= sxhv->xhv_max) {
11094 const HE * const source = HvARRAY(sstr)[i];
11095 HvARRAY(dstr)[i] = source
11096 ? he_dup(source, sharekeys, param) : 0;
11101 const struct xpvhv_aux * const saux = HvAUX(sstr);
11102 struct xpvhv_aux * const daux = HvAUX(dstr);
11103 /* This flag isn't copied. */
11104 /* SvOOK_on(hv) attacks the IV flags. */
11105 SvFLAGS(dstr) |= SVf_OOK;
11107 hvname = saux->xhv_name;
11108 daux->xhv_name = hek_dup(hvname, param);
11110 daux->xhv_riter = saux->xhv_riter;
11111 daux->xhv_eiter = saux->xhv_eiter
11112 ? he_dup(saux->xhv_eiter,
11113 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11114 /* backref array needs refcnt=2; see sv_add_backref */
11115 daux->xhv_backreferences =
11116 saux->xhv_backreferences
11117 ? MUTABLE_AV(SvREFCNT_inc(
11118 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11121 daux->xhv_mro_meta = saux->xhv_mro_meta
11122 ? mro_meta_dup(saux->xhv_mro_meta, param)
11125 /* Record stashes for possible cloning in Perl_clone(). */
11127 av_push(param->stashes, dstr);
11131 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11134 if (!(param->flags & CLONEf_COPY_STACKS)) {
11138 /* NOTE: not refcounted */
11139 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11141 if (!CvISXSUB(dstr))
11142 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11144 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11145 CvXSUBANY(dstr).any_ptr =
11146 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11148 /* don't dup if copying back - CvGV isn't refcounted, so the
11149 * duped GV may never be freed. A bit of a hack! DAPM */
11150 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11151 NULL : gv_dup(CvGV(dstr), param) ;
11152 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11154 CvWEAKOUTSIDE(sstr)
11155 ? cv_dup( CvOUTSIDE(dstr), param)
11156 : cv_dup_inc(CvOUTSIDE(dstr), param);
11157 if (!CvISXSUB(dstr))
11158 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11164 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11170 /* duplicate a context */
11173 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11175 PERL_CONTEXT *ncxs;
11177 PERL_ARGS_ASSERT_CX_DUP;
11180 return (PERL_CONTEXT*)NULL;
11182 /* look for it in the table first */
11183 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11187 /* create anew and remember what it is */
11188 Newx(ncxs, max + 1, PERL_CONTEXT);
11189 ptr_table_store(PL_ptr_table, cxs, ncxs);
11190 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11193 PERL_CONTEXT * const ncx = &ncxs[ix];
11194 if (CxTYPE(ncx) == CXt_SUBST) {
11195 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11198 switch (CxTYPE(ncx)) {
11200 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11201 ? cv_dup_inc(ncx->blk_sub.cv, param)
11202 : cv_dup(ncx->blk_sub.cv,param));
11203 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11204 ? av_dup_inc(ncx->blk_sub.argarray,
11207 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11209 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11210 ncx->blk_sub.oldcomppad);
11213 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11215 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11217 case CXt_LOOP_LAZYSV:
11218 ncx->blk_loop.state_u.lazysv.end
11219 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11220 /* We are taking advantage of av_dup_inc and sv_dup_inc
11221 actually being the same function, and order equivalance of
11223 We can assert the later [but only at run time :-(] */
11224 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11225 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11227 ncx->blk_loop.state_u.ary.ary
11228 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11229 case CXt_LOOP_LAZYIV:
11230 case CXt_LOOP_PLAIN:
11231 if (CxPADLOOP(ncx)) {
11232 ncx->blk_loop.oldcomppad
11233 = (PAD*)ptr_table_fetch(PL_ptr_table,
11234 ncx->blk_loop.oldcomppad);
11236 ncx->blk_loop.oldcomppad
11237 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11242 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11243 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11244 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11257 /* duplicate a stack info structure */
11260 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11264 PERL_ARGS_ASSERT_SI_DUP;
11267 return (PERL_SI*)NULL;
11269 /* look for it in the table first */
11270 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11274 /* create anew and remember what it is */
11275 Newxz(nsi, 1, PERL_SI);
11276 ptr_table_store(PL_ptr_table, si, nsi);
11278 nsi->si_stack = av_dup_inc(si->si_stack, param);
11279 nsi->si_cxix = si->si_cxix;
11280 nsi->si_cxmax = si->si_cxmax;
11281 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11282 nsi->si_type = si->si_type;
11283 nsi->si_prev = si_dup(si->si_prev, param);
11284 nsi->si_next = si_dup(si->si_next, param);
11285 nsi->si_markoff = si->si_markoff;
11290 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11291 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11292 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11293 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11294 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11295 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11296 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11297 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11298 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11299 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11300 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11301 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11302 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11303 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11306 #define pv_dup_inc(p) SAVEPV(p)
11307 #define pv_dup(p) SAVEPV(p)
11308 #define svp_dup_inc(p,pp) any_dup(p,pp)
11310 /* map any object to the new equivent - either something in the
11311 * ptr table, or something in the interpreter structure
11315 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11319 PERL_ARGS_ASSERT_ANY_DUP;
11322 return (void*)NULL;
11324 /* look for it in the table first */
11325 ret = ptr_table_fetch(PL_ptr_table, v);
11329 /* see if it is part of the interpreter structure */
11330 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11331 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11339 /* duplicate the save stack */
11342 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11345 ANY * const ss = proto_perl->Isavestack;
11346 const I32 max = proto_perl->Isavestack_max;
11347 I32 ix = proto_perl->Isavestack_ix;
11360 void (*dptr) (void*);
11361 void (*dxptr) (pTHX_ void*);
11363 PERL_ARGS_ASSERT_SS_DUP;
11365 Newxz(nss, max, ANY);
11368 const I32 type = POPINT(ss,ix);
11369 TOPINT(nss,ix) = type;
11371 case SAVEt_HELEM: /* hash element */
11372 sv = (const SV *)POPPTR(ss,ix);
11373 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11375 case SAVEt_ITEM: /* normal string */
11376 case SAVEt_SV: /* scalar reference */
11377 sv = (const SV *)POPPTR(ss,ix);
11378 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11381 case SAVEt_MORTALIZESV:
11382 sv = (const SV *)POPPTR(ss,ix);
11383 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11385 case SAVEt_SHARED_PVREF: /* char* in shared space */
11386 c = (char*)POPPTR(ss,ix);
11387 TOPPTR(nss,ix) = savesharedpv(c);
11388 ptr = POPPTR(ss,ix);
11389 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11391 case SAVEt_GENERIC_SVREF: /* generic sv */
11392 case SAVEt_SVREF: /* scalar reference */
11393 sv = (const SV *)POPPTR(ss,ix);
11394 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11395 ptr = POPPTR(ss,ix);
11396 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11398 case SAVEt_HV: /* hash reference */
11399 case SAVEt_AV: /* array reference */
11400 sv = (const SV *) POPPTR(ss,ix);
11401 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11403 case SAVEt_COMPPAD:
11405 sv = (const SV *) POPPTR(ss,ix);
11406 TOPPTR(nss,ix) = sv_dup(sv, param);
11408 case SAVEt_INT: /* int reference */
11409 ptr = POPPTR(ss,ix);
11410 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11411 intval = (int)POPINT(ss,ix);
11412 TOPINT(nss,ix) = intval;
11414 case SAVEt_LONG: /* long reference */
11415 ptr = POPPTR(ss,ix);
11416 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11418 case SAVEt_CLEARSV:
11419 longval = (long)POPLONG(ss,ix);
11420 TOPLONG(nss,ix) = longval;
11422 case SAVEt_I32: /* I32 reference */
11423 case SAVEt_I16: /* I16 reference */
11424 case SAVEt_I8: /* I8 reference */
11425 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11426 ptr = POPPTR(ss,ix);
11427 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11429 TOPINT(nss,ix) = i;
11431 case SAVEt_IV: /* IV reference */
11432 ptr = POPPTR(ss,ix);
11433 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11435 TOPIV(nss,ix) = iv;
11437 case SAVEt_HPTR: /* HV* reference */
11438 case SAVEt_APTR: /* AV* reference */
11439 case SAVEt_SPTR: /* SV* reference */
11440 ptr = POPPTR(ss,ix);
11441 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11442 sv = (const SV *)POPPTR(ss,ix);
11443 TOPPTR(nss,ix) = sv_dup(sv, param);
11445 case SAVEt_VPTR: /* random* reference */
11446 ptr = POPPTR(ss,ix);
11447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11448 ptr = POPPTR(ss,ix);
11449 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11451 case SAVEt_GENERIC_PVREF: /* generic char* */
11452 case SAVEt_PPTR: /* char* reference */
11453 ptr = POPPTR(ss,ix);
11454 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11455 c = (char*)POPPTR(ss,ix);
11456 TOPPTR(nss,ix) = pv_dup(c);
11458 case SAVEt_GP: /* scalar reference */
11459 gp = (GP*)POPPTR(ss,ix);
11460 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11461 (void)GpREFCNT_inc(gp);
11462 gv = (const GV *)POPPTR(ss,ix);
11463 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11466 ptr = POPPTR(ss,ix);
11467 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11468 /* these are assumed to be refcounted properly */
11470 switch (((OP*)ptr)->op_type) {
11472 case OP_LEAVESUBLV:
11476 case OP_LEAVEWRITE:
11477 TOPPTR(nss,ix) = ptr;
11480 (void) OpREFCNT_inc(o);
11484 TOPPTR(nss,ix) = NULL;
11489 TOPPTR(nss,ix) = NULL;
11492 hv = (const HV *)POPPTR(ss,ix);
11493 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11495 TOPINT(nss,ix) = i;
11498 c = (char*)POPPTR(ss,ix);
11499 TOPPTR(nss,ix) = pv_dup_inc(c);
11501 case SAVEt_STACK_POS: /* Position on Perl stack */
11503 TOPINT(nss,ix) = i;
11505 case SAVEt_DESTRUCTOR:
11506 ptr = POPPTR(ss,ix);
11507 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11508 dptr = POPDPTR(ss,ix);
11509 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11510 any_dup(FPTR2DPTR(void *, dptr),
11513 case SAVEt_DESTRUCTOR_X:
11514 ptr = POPPTR(ss,ix);
11515 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11516 dxptr = POPDXPTR(ss,ix);
11517 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11518 any_dup(FPTR2DPTR(void *, dxptr),
11521 case SAVEt_REGCONTEXT:
11524 TOPINT(nss,ix) = i;
11527 case SAVEt_AELEM: /* array element */
11528 sv = (const SV *)POPPTR(ss,ix);
11529 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11531 TOPINT(nss,ix) = i;
11532 av = (const AV *)POPPTR(ss,ix);
11533 TOPPTR(nss,ix) = av_dup_inc(av, param);
11536 ptr = POPPTR(ss,ix);
11537 TOPPTR(nss,ix) = ptr;
11540 ptr = POPPTR(ss,ix);
11543 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11544 HINTS_REFCNT_UNLOCK;
11546 TOPPTR(nss,ix) = ptr;
11548 TOPINT(nss,ix) = i;
11549 if (i & HINT_LOCALIZE_HH) {
11550 hv = (const HV *)POPPTR(ss,ix);
11551 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11554 case SAVEt_PADSV_AND_MORTALIZE:
11555 longval = (long)POPLONG(ss,ix);
11556 TOPLONG(nss,ix) = longval;
11557 ptr = POPPTR(ss,ix);
11558 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11559 sv = (const SV *)POPPTR(ss,ix);
11560 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11563 ptr = POPPTR(ss,ix);
11564 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11565 longval = (long)POPBOOL(ss,ix);
11566 TOPBOOL(nss,ix) = (bool)longval;
11568 case SAVEt_SET_SVFLAGS:
11570 TOPINT(nss,ix) = i;
11572 TOPINT(nss,ix) = i;
11573 sv = (const SV *)POPPTR(ss,ix);
11574 TOPPTR(nss,ix) = sv_dup(sv, param);
11576 case SAVEt_RE_STATE:
11578 const struct re_save_state *const old_state
11579 = (struct re_save_state *)
11580 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11581 struct re_save_state *const new_state
11582 = (struct re_save_state *)
11583 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11585 Copy(old_state, new_state, 1, struct re_save_state);
11586 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11588 new_state->re_state_bostr
11589 = pv_dup(old_state->re_state_bostr);
11590 new_state->re_state_reginput
11591 = pv_dup(old_state->re_state_reginput);
11592 new_state->re_state_regeol
11593 = pv_dup(old_state->re_state_regeol);
11594 new_state->re_state_regoffs
11595 = (regexp_paren_pair*)
11596 any_dup(old_state->re_state_regoffs, proto_perl);
11597 new_state->re_state_reglastparen
11598 = (U32*) any_dup(old_state->re_state_reglastparen,
11600 new_state->re_state_reglastcloseparen
11601 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11603 /* XXX This just has to be broken. The old save_re_context
11604 code did SAVEGENERICPV(PL_reg_start_tmp);
11605 PL_reg_start_tmp is char **.
11606 Look above to what the dup code does for
11607 SAVEt_GENERIC_PVREF
11608 It can never have worked.
11609 So this is merely a faithful copy of the exiting bug: */
11610 new_state->re_state_reg_start_tmp
11611 = (char **) pv_dup((char *)
11612 old_state->re_state_reg_start_tmp);
11613 /* I assume that it only ever "worked" because no-one called
11614 (pseudo)fork while the regexp engine had re-entered itself.
11616 #ifdef PERL_OLD_COPY_ON_WRITE
11617 new_state->re_state_nrs
11618 = sv_dup(old_state->re_state_nrs, param);
11620 new_state->re_state_reg_magic
11621 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11623 new_state->re_state_reg_oldcurpm
11624 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11626 new_state->re_state_reg_curpm
11627 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11629 new_state->re_state_reg_oldsaved
11630 = pv_dup(old_state->re_state_reg_oldsaved);
11631 new_state->re_state_reg_poscache
11632 = pv_dup(old_state->re_state_reg_poscache);
11633 new_state->re_state_reg_starttry
11634 = pv_dup(old_state->re_state_reg_starttry);
11637 case SAVEt_COMPILE_WARNINGS:
11638 ptr = POPPTR(ss,ix);
11639 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11642 ptr = POPPTR(ss,ix);
11643 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11647 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11655 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11656 * flag to the result. This is done for each stash before cloning starts,
11657 * so we know which stashes want their objects cloned */
11660 do_mark_cloneable_stash(pTHX_ SV *const sv)
11662 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11664 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11665 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11666 if (cloner && GvCV(cloner)) {
11673 mXPUSHs(newSVhek(hvname));
11675 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11682 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11690 =for apidoc perl_clone
11692 Create and return a new interpreter by cloning the current one.
11694 perl_clone takes these flags as parameters:
11696 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11697 without it we only clone the data and zero the stacks,
11698 with it we copy the stacks and the new perl interpreter is
11699 ready to run at the exact same point as the previous one.
11700 The pseudo-fork code uses COPY_STACKS while the
11701 threads->create doesn't.
11703 CLONEf_KEEP_PTR_TABLE
11704 perl_clone keeps a ptr_table with the pointer of the old
11705 variable as a key and the new variable as a value,
11706 this allows it to check if something has been cloned and not
11707 clone it again but rather just use the value and increase the
11708 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11709 the ptr_table using the function
11710 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11711 reason to keep it around is if you want to dup some of your own
11712 variable who are outside the graph perl scans, example of this
11713 code is in threads.xs create
11716 This is a win32 thing, it is ignored on unix, it tells perls
11717 win32host code (which is c++) to clone itself, this is needed on
11718 win32 if you want to run two threads at the same time,
11719 if you just want to do some stuff in a separate perl interpreter
11720 and then throw it away and return to the original one,
11721 you don't need to do anything.
11726 /* XXX the above needs expanding by someone who actually understands it ! */
11727 EXTERN_C PerlInterpreter *
11728 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11731 perl_clone(PerlInterpreter *proto_perl, UV flags)
11734 #ifdef PERL_IMPLICIT_SYS
11736 PERL_ARGS_ASSERT_PERL_CLONE;
11738 /* perlhost.h so we need to call into it
11739 to clone the host, CPerlHost should have a c interface, sky */
11741 if (flags & CLONEf_CLONE_HOST) {
11742 return perl_clone_host(proto_perl,flags);
11744 return perl_clone_using(proto_perl, flags,
11746 proto_perl->IMemShared,
11747 proto_perl->IMemParse,
11749 proto_perl->IStdIO,
11753 proto_perl->IProc);
11757 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11758 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11759 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11760 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11761 struct IPerlDir* ipD, struct IPerlSock* ipS,
11762 struct IPerlProc* ipP)
11764 /* XXX many of the string copies here can be optimized if they're
11765 * constants; they need to be allocated as common memory and just
11766 * their pointers copied. */
11769 CLONE_PARAMS clone_params;
11770 CLONE_PARAMS* const param = &clone_params;
11772 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11774 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11776 /* for each stash, determine whether its objects should be cloned */
11777 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11778 PERL_SET_THX(my_perl);
11781 PoisonNew(my_perl, 1, PerlInterpreter);
11787 PL_savestack_ix = 0;
11788 PL_savestack_max = -1;
11789 PL_sig_pending = 0;
11791 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11792 # else /* !DEBUGGING */
11793 Zero(my_perl, 1, PerlInterpreter);
11794 # endif /* DEBUGGING */
11796 /* host pointers */
11798 PL_MemShared = ipMS;
11799 PL_MemParse = ipMP;
11806 #else /* !PERL_IMPLICIT_SYS */
11808 CLONE_PARAMS clone_params;
11809 CLONE_PARAMS* param = &clone_params;
11810 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11812 PERL_ARGS_ASSERT_PERL_CLONE;
11814 /* for each stash, determine whether its objects should be cloned */
11815 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11816 PERL_SET_THX(my_perl);
11819 PoisonNew(my_perl, 1, PerlInterpreter);
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);
12285 /* NOTE: si_dup() looks at PL_markstack */
12286 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12288 /* PL_curstack = PL_curstackinfo->si_stack; */
12289 PL_curstack = av_dup(proto_perl->Icurstack, param);
12290 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12292 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12293 PL_stack_base = AvARRAY(PL_curstack);
12294 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12295 - proto_perl->Istack_base);
12296 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12298 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12299 * NOTE: unlike the others! */
12300 PL_savestack_ix = proto_perl->Isavestack_ix;
12301 PL_savestack_max = proto_perl->Isavestack_max;
12302 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12303 PL_savestack = ss_dup(proto_perl, param);
12307 ENTER; /* perl_destruct() wants to LEAVE; */
12309 /* although we're not duplicating the tmps stack, we should still
12310 * add entries for any SVs on the tmps stack that got cloned by a
12311 * non-refcount means (eg a temp in @_); otherwise they will be
12314 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12315 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12316 proto_perl->Itmps_stack[i]));
12317 if (nsv && !SvREFCNT(nsv)) {
12318 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12323 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12324 PL_top_env = &PL_start_env;
12326 PL_op = proto_perl->Iop;
12329 PL_Xpv = (XPV*)NULL;
12330 my_perl->Ina = proto_perl->Ina;
12332 PL_statbuf = proto_perl->Istatbuf;
12333 PL_statcache = proto_perl->Istatcache;
12334 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12335 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12337 PL_timesbuf = proto_perl->Itimesbuf;
12340 PL_tainted = proto_perl->Itainted;
12341 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12342 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12343 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12344 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12345 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12346 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12347 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12348 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12350 PL_restartop = proto_perl->Irestartop;
12351 PL_in_eval = proto_perl->Iin_eval;
12352 PL_delaymagic = proto_perl->Idelaymagic;
12353 PL_dirty = proto_perl->Idirty;
12354 PL_localizing = proto_perl->Ilocalizing;
12356 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12357 PL_hv_fetch_ent_mh = NULL;
12358 PL_modcount = proto_perl->Imodcount;
12359 PL_lastgotoprobe = NULL;
12360 PL_dumpindent = proto_perl->Idumpindent;
12362 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12363 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12364 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12365 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12366 PL_efloatbuf = NULL; /* reinits on demand */
12367 PL_efloatsize = 0; /* reinits on demand */
12371 PL_screamfirst = NULL;
12372 PL_screamnext = NULL;
12373 PL_maxscream = -1; /* reinits on demand */
12374 PL_lastscream = NULL;
12377 PL_regdummy = proto_perl->Iregdummy;
12378 PL_colorset = 0; /* reinits PL_colors[] */
12379 /*PL_colors[6] = {0,0,0,0,0,0};*/
12383 /* Pluggable optimizer */
12384 PL_peepp = proto_perl->Ipeepp;
12385 /* op_free() hook */
12386 PL_opfreehook = proto_perl->Iopfreehook;
12388 PL_stashcache = newHV();
12390 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12391 proto_perl->Iwatchaddr);
12392 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12393 if (PL_debug && PL_watchaddr) {
12394 PerlIO_printf(Perl_debug_log,
12395 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12396 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12397 PTR2UV(PL_watchok));
12400 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12402 /* Call the ->CLONE method, if it exists, for each of the stashes
12403 identified by sv_dup() above.
12405 while(av_len(param->stashes) != -1) {
12406 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12407 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12408 if (cloner && GvCV(cloner)) {
12413 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12415 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12421 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12422 ptr_table_free(PL_ptr_table);
12423 PL_ptr_table = NULL;
12427 SvREFCNT_dec(param->stashes);
12429 /* orphaned? eg threads->new inside BEGIN or use */
12430 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12431 SvREFCNT_inc_simple_void(PL_compcv);
12432 SAVEFREESV(PL_compcv);
12438 #endif /* USE_ITHREADS */
12441 =head1 Unicode Support
12443 =for apidoc sv_recode_to_utf8
12445 The encoding is assumed to be an Encode object, on entry the PV
12446 of the sv is assumed to be octets in that encoding, and the sv
12447 will be converted into Unicode (and UTF-8).
12449 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12450 is not a reference, nothing is done to the sv. If the encoding is not
12451 an C<Encode::XS> Encoding object, bad things will happen.
12452 (See F<lib/encoding.pm> and L<Encode>).
12454 The PV of the sv is returned.
12459 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12463 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12465 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12479 Passing sv_yes is wrong - it needs to be or'ed set of constants
12480 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12481 remove converted chars from source.
12483 Both will default the value - let them.
12485 XPUSHs(&PL_sv_yes);
12488 call_method("decode", G_SCALAR);
12492 s = SvPV_const(uni, len);
12493 if (s != SvPVX_const(sv)) {
12494 SvGROW(sv, len + 1);
12495 Move(s, SvPVX(sv), len + 1, char);
12496 SvCUR_set(sv, len);
12503 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12507 =for apidoc sv_cat_decode
12509 The encoding is assumed to be an Encode object, the PV of the ssv is
12510 assumed to be octets in that encoding and decoding the input starts
12511 from the position which (PV + *offset) pointed to. The dsv will be
12512 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12513 when the string tstr appears in decoding output or the input ends on
12514 the PV of the ssv. The value which the offset points will be modified
12515 to the last input position on the ssv.
12517 Returns TRUE if the terminator was found, else returns FALSE.
12522 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12523 SV *ssv, int *offset, char *tstr, int tlen)
12528 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12530 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12541 offsv = newSViv(*offset);
12543 mXPUSHp(tstr, tlen);
12545 call_method("cat_decode", G_SCALAR);
12547 ret = SvTRUE(TOPs);
12548 *offset = SvIV(offsv);
12554 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12559 /* ---------------------------------------------------------------------
12561 * support functions for report_uninit()
12564 /* the maxiumum size of array or hash where we will scan looking
12565 * for the undefined element that triggered the warning */
12567 #define FUV_MAX_SEARCH_SIZE 1000
12569 /* Look for an entry in the hash whose value has the same SV as val;
12570 * If so, return a mortal copy of the key. */
12573 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12576 register HE **array;
12579 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12581 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12582 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12585 array = HvARRAY(hv);
12587 for (i=HvMAX(hv); i>0; i--) {
12588 register HE *entry;
12589 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12590 if (HeVAL(entry) != val)
12592 if ( HeVAL(entry) == &PL_sv_undef ||
12593 HeVAL(entry) == &PL_sv_placeholder)
12597 if (HeKLEN(entry) == HEf_SVKEY)
12598 return sv_mortalcopy(HeKEY_sv(entry));
12599 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12605 /* Look for an entry in the array whose value has the same SV as val;
12606 * If so, return the index, otherwise return -1. */
12609 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12613 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12615 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12616 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12619 if (val != &PL_sv_undef) {
12620 SV ** const svp = AvARRAY(av);
12623 for (i=AvFILLp(av); i>=0; i--)
12630 /* S_varname(): return the name of a variable, optionally with a subscript.
12631 * If gv is non-zero, use the name of that global, along with gvtype (one
12632 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12633 * targ. Depending on the value of the subscript_type flag, return:
12636 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12637 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12638 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12639 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12642 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12643 const SV *const keyname, I32 aindex, int subscript_type)
12646 SV * const name = sv_newmortal();
12649 buffer[0] = gvtype;
12652 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12654 gv_fullname4(name, gv, buffer, 0);
12656 if ((unsigned int)SvPVX(name)[1] <= 26) {
12658 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12660 /* Swap the 1 unprintable control character for the 2 byte pretty
12661 version - ie substr($name, 1, 1) = $buffer; */
12662 sv_insert(name, 1, 1, buffer, 2);
12666 CV * const cv = find_runcv(NULL);
12670 if (!cv || !CvPADLIST(cv))
12672 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12673 sv = *av_fetch(av, targ, FALSE);
12674 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12677 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12678 SV * const sv = newSV(0);
12679 *SvPVX(name) = '$';
12680 Perl_sv_catpvf(aTHX_ name, "{%s}",
12681 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12684 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12685 *SvPVX(name) = '$';
12686 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12688 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12689 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12690 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12698 =for apidoc find_uninit_var
12700 Find the name of the undefined variable (if any) that caused the operator o
12701 to issue a "Use of uninitialized value" warning.
12702 If match is true, only return a name if it's value matches uninit_sv.
12703 So roughly speaking, if a unary operator (such as OP_COS) generates a
12704 warning, then following the direct child of the op may yield an
12705 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12706 other hand, with OP_ADD there are two branches to follow, so we only print
12707 the variable name if we get an exact match.
12709 The name is returned as a mortal SV.
12711 Assumes that PL_op is the op that originally triggered the error, and that
12712 PL_comppad/PL_curpad points to the currently executing pad.
12718 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12724 const OP *o, *o2, *kid;
12726 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12727 uninit_sv == &PL_sv_placeholder)))
12730 switch (obase->op_type) {
12737 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12738 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12741 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12743 if (pad) { /* @lex, %lex */
12744 sv = PAD_SVl(obase->op_targ);
12748 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12749 /* @global, %global */
12750 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12753 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12755 else /* @{expr}, %{expr} */
12756 return find_uninit_var(cUNOPx(obase)->op_first,
12760 /* attempt to find a match within the aggregate */
12762 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12764 subscript_type = FUV_SUBSCRIPT_HASH;
12767 index = find_array_subscript((const AV *)sv, uninit_sv);
12769 subscript_type = FUV_SUBSCRIPT_ARRAY;
12772 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12775 return varname(gv, hash ? '%' : '@', obase->op_targ,
12776 keysv, index, subscript_type);
12780 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12782 return varname(NULL, '$', obase->op_targ,
12783 NULL, 0, FUV_SUBSCRIPT_NONE);
12786 gv = cGVOPx_gv(obase);
12787 if (!gv || (match && GvSV(gv) != uninit_sv))
12789 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12792 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12795 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12796 if (!av || SvRMAGICAL(av))
12798 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12799 if (!svp || *svp != uninit_sv)
12802 return varname(NULL, '$', obase->op_targ,
12803 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12806 gv = cGVOPx_gv(obase);
12811 AV *const av = GvAV(gv);
12812 if (!av || SvRMAGICAL(av))
12814 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12815 if (!svp || *svp != uninit_sv)
12818 return varname(gv, '$', 0,
12819 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12824 o = cUNOPx(obase)->op_first;
12825 if (!o || o->op_type != OP_NULL ||
12826 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12828 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12832 if (PL_op == obase)
12833 /* $a[uninit_expr] or $h{uninit_expr} */
12834 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12837 o = cBINOPx(obase)->op_first;
12838 kid = cBINOPx(obase)->op_last;
12840 /* get the av or hv, and optionally the gv */
12842 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12843 sv = PAD_SV(o->op_targ);
12845 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12846 && cUNOPo->op_first->op_type == OP_GV)
12848 gv = cGVOPx_gv(cUNOPo->op_first);
12852 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12857 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12858 /* index is constant */
12862 if (obase->op_type == OP_HELEM) {
12863 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12864 if (!he || HeVAL(he) != uninit_sv)
12868 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12869 if (!svp || *svp != uninit_sv)
12873 if (obase->op_type == OP_HELEM)
12874 return varname(gv, '%', o->op_targ,
12875 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12877 return varname(gv, '@', o->op_targ, NULL,
12878 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12881 /* index is an expression;
12882 * attempt to find a match within the aggregate */
12883 if (obase->op_type == OP_HELEM) {
12884 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12886 return varname(gv, '%', o->op_targ,
12887 keysv, 0, FUV_SUBSCRIPT_HASH);
12891 = find_array_subscript((const AV *)sv, uninit_sv);
12893 return varname(gv, '@', o->op_targ,
12894 NULL, index, FUV_SUBSCRIPT_ARRAY);
12899 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12901 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12906 /* only examine RHS */
12907 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12910 o = cUNOPx(obase)->op_first;
12911 if (o->op_type == OP_PUSHMARK)
12914 if (!o->op_sibling) {
12915 /* one-arg version of open is highly magical */
12917 if (o->op_type == OP_GV) { /* open FOO; */
12919 if (match && GvSV(gv) != uninit_sv)
12921 return varname(gv, '$', 0,
12922 NULL, 0, FUV_SUBSCRIPT_NONE);
12924 /* other possibilities not handled are:
12925 * open $x; or open my $x; should return '${*$x}'
12926 * open expr; should return '$'.expr ideally
12932 /* ops where $_ may be an implicit arg */
12936 if ( !(obase->op_flags & OPf_STACKED)) {
12937 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12938 ? PAD_SVl(obase->op_targ)
12941 sv = sv_newmortal();
12942 sv_setpvs(sv, "$_");
12951 match = 1; /* print etc can return undef on defined args */
12952 /* skip filehandle as it can't produce 'undef' warning */
12953 o = cUNOPx(obase)->op_first;
12954 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12955 o = o->op_sibling->op_sibling;
12959 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12961 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12963 /* the following ops are capable of returning PL_sv_undef even for
12964 * defined arg(s) */
12983 case OP_GETPEERNAME:
13031 case OP_SMARTMATCH:
13040 /* XXX tmp hack: these two may call an XS sub, and currently
13041 XS subs don't have a SUB entry on the context stack, so CV and
13042 pad determination goes wrong, and BAD things happen. So, just
13043 don't try to determine the value under those circumstances.
13044 Need a better fix at dome point. DAPM 11/2007 */
13050 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13051 if (gv && GvSV(gv) == uninit_sv)
13052 return newSVpvs_flags("$.", SVs_TEMP);
13057 /* def-ness of rval pos() is independent of the def-ness of its arg */
13058 if ( !(obase->op_flags & OPf_MOD))
13063 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13064 return newSVpvs_flags("${$/}", SVs_TEMP);
13069 if (!(obase->op_flags & OPf_KIDS))
13071 o = cUNOPx(obase)->op_first;
13077 /* if all except one arg are constant, or have no side-effects,
13078 * or are optimized away, then it's unambiguous */
13080 for (kid=o; kid; kid = kid->op_sibling) {
13082 const OPCODE type = kid->op_type;
13083 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13084 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13085 || (type == OP_PUSHMARK)
13089 if (o2) { /* more than one found */
13096 return find_uninit_var(o2, uninit_sv, match);
13098 /* scan all args */
13100 sv = find_uninit_var(o, uninit_sv, 1);
13112 =for apidoc report_uninit
13114 Print appropriate "Use of uninitialized variable" warning
13120 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13124 SV* varname = NULL;
13126 varname = find_uninit_var(PL_op, uninit_sv,0);
13128 sv_insert(varname, 0, 0, " ", 1);
13130 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13131 varname ? SvPV_nolen_const(varname) : "",
13132 " in ", OP_DESC(PL_op));
13135 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13141 * c-indentation-style: bsd
13142 * c-basic-offset: 4
13143 * indent-tabs-mode: t
13146 * ex: set ts=8 sts=4 sw=4 noet: