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);
3897 sv_upgrade(dstr, SVt_PV);
3900 if (dtype < SVt_PVIV)
3901 sv_upgrade(dstr, SVt_PVIV);
3904 if (dtype < SVt_PVNV)
3905 sv_upgrade(dstr, SVt_PVNV);
3909 const char * const type = sv_reftype(sstr,0);
3911 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3913 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3917 /* case SVt_BIND: */
3920 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3921 glob_assign_glob(dstr, sstr, dtype);
3924 /* SvVALID means that this PVGV is playing at being an FBM. */
3928 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3930 if (SvTYPE(sstr) != stype) {
3931 stype = SvTYPE(sstr);
3932 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3933 glob_assign_glob(dstr, sstr, dtype);
3938 if (stype == SVt_PVLV)
3939 SvUPGRADE(dstr, SVt_PVNV);
3941 SvUPGRADE(dstr, (svtype)stype);
3943 end_of_first_switch:
3945 /* dstr may have been upgraded. */
3946 dtype = SvTYPE(dstr);
3947 sflags = SvFLAGS(sstr);
3949 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3950 /* Assigning to a subroutine sets the prototype. */
3953 const char *const ptr = SvPV_const(sstr, len);
3955 SvGROW(dstr, len + 1);
3956 Copy(ptr, SvPVX(dstr), len + 1, char);
3957 SvCUR_set(dstr, len);
3959 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3963 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3964 const char * const type = sv_reftype(dstr,0);
3966 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3968 Perl_croak(aTHX_ "Cannot copy to %s", type);
3969 } else if (sflags & SVf_ROK) {
3970 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3971 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3974 if (GvIMPORTED(dstr) != GVf_IMPORTED
3975 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3977 GvIMPORTED_on(dstr);
3982 glob_assign_glob(dstr, sstr, dtype);
3986 if (dtype >= SVt_PV) {
3987 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3988 glob_assign_ref(dstr, sstr);
3991 if (SvPVX_const(dstr)) {
3997 (void)SvOK_off(dstr);
3998 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3999 SvFLAGS(dstr) |= sflags & SVf_ROK;
4000 assert(!(sflags & SVp_NOK));
4001 assert(!(sflags & SVp_IOK));
4002 assert(!(sflags & SVf_NOK));
4003 assert(!(sflags & SVf_IOK));
4005 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4006 if (!(sflags & SVf_OK)) {
4007 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4008 "Undefined value assigned to typeglob");
4011 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4012 if (dstr != (const SV *)gv) {
4014 gp_free(MUTABLE_GV(dstr));
4015 GvGP(dstr) = gp_ref(GvGP(gv));
4019 else if (sflags & SVp_POK) {
4023 * Check to see if we can just swipe the string. If so, it's a
4024 * possible small lose on short strings, but a big win on long ones.
4025 * It might even be a win on short strings if SvPVX_const(dstr)
4026 * has to be allocated and SvPVX_const(sstr) has to be freed.
4027 * Likewise if we can set up COW rather than doing an actual copy, we
4028 * drop to the else clause, as the swipe code and the COW setup code
4029 * have much in common.
4032 /* Whichever path we take through the next code, we want this true,
4033 and doing it now facilitates the COW check. */
4034 (void)SvPOK_only(dstr);
4037 /* If we're already COW then this clause is not true, and if COW
4038 is allowed then we drop down to the else and make dest COW
4039 with us. If caller hasn't said that we're allowed to COW
4040 shared hash keys then we don't do the COW setup, even if the
4041 source scalar is a shared hash key scalar. */
4042 (((flags & SV_COW_SHARED_HASH_KEYS)
4043 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4044 : 1 /* If making a COW copy is forbidden then the behaviour we
4045 desire is as if the source SV isn't actually already
4046 COW, even if it is. So we act as if the source flags
4047 are not COW, rather than actually testing them. */
4049 #ifndef PERL_OLD_COPY_ON_WRITE
4050 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4051 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4052 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4053 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4054 but in turn, it's somewhat dead code, never expected to go
4055 live, but more kept as a placeholder on how to do it better
4056 in a newer implementation. */
4057 /* If we are COW and dstr is a suitable target then we drop down
4058 into the else and make dest a COW of us. */
4059 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4064 (sflags & SVs_TEMP) && /* slated for free anyway? */
4065 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4066 (!(flags & SV_NOSTEAL)) &&
4067 /* and we're allowed to steal temps */
4068 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4069 SvLEN(sstr) && /* and really is a string */
4070 /* and won't be needed again, potentially */
4071 !(PL_op && PL_op->op_type == OP_AASSIGN))
4072 #ifdef PERL_OLD_COPY_ON_WRITE
4073 && ((flags & SV_COW_SHARED_HASH_KEYS)
4074 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4075 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4076 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4080 /* Failed the swipe test, and it's not a shared hash key either.
4081 Have to copy the string. */
4082 STRLEN len = SvCUR(sstr);
4083 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4084 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4085 SvCUR_set(dstr, len);
4086 *SvEND(dstr) = '\0';
4088 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4090 /* Either it's a shared hash key, or it's suitable for
4091 copy-on-write or we can swipe the string. */
4093 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4097 #ifdef PERL_OLD_COPY_ON_WRITE
4099 if ((sflags & (SVf_FAKE | SVf_READONLY))
4100 != (SVf_FAKE | SVf_READONLY)) {
4101 SvREADONLY_on(sstr);
4103 /* Make the source SV into a loop of 1.
4104 (about to become 2) */
4105 SV_COW_NEXT_SV_SET(sstr, sstr);
4109 /* Initial code is common. */
4110 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4115 /* making another shared SV. */
4116 STRLEN cur = SvCUR(sstr);
4117 STRLEN len = SvLEN(sstr);
4118 #ifdef PERL_OLD_COPY_ON_WRITE
4120 assert (SvTYPE(dstr) >= SVt_PVIV);
4121 /* SvIsCOW_normal */
4122 /* splice us in between source and next-after-source. */
4123 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4124 SV_COW_NEXT_SV_SET(sstr, dstr);
4125 SvPV_set(dstr, SvPVX_mutable(sstr));
4129 /* SvIsCOW_shared_hash */
4130 DEBUG_C(PerlIO_printf(Perl_debug_log,
4131 "Copy on write: Sharing hash\n"));
4133 assert (SvTYPE(dstr) >= SVt_PV);
4135 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4137 SvLEN_set(dstr, len);
4138 SvCUR_set(dstr, cur);
4139 SvREADONLY_on(dstr);
4143 { /* Passes the swipe test. */
4144 SvPV_set(dstr, SvPVX_mutable(sstr));
4145 SvLEN_set(dstr, SvLEN(sstr));
4146 SvCUR_set(dstr, SvCUR(sstr));
4149 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4150 SvPV_set(sstr, NULL);
4156 if (sflags & SVp_NOK) {
4157 SvNV_set(dstr, SvNVX(sstr));
4159 if (sflags & SVp_IOK) {
4160 SvIV_set(dstr, SvIVX(sstr));
4161 /* Must do this otherwise some other overloaded use of 0x80000000
4162 gets confused. I guess SVpbm_VALID */
4163 if (sflags & SVf_IVisUV)
4166 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4168 const MAGIC * const smg = SvVSTRING_mg(sstr);
4170 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4171 smg->mg_ptr, smg->mg_len);
4172 SvRMAGICAL_on(dstr);
4176 else if (sflags & (SVp_IOK|SVp_NOK)) {
4177 (void)SvOK_off(dstr);
4178 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4179 if (sflags & SVp_IOK) {
4180 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4181 SvIV_set(dstr, SvIVX(sstr));
4183 if (sflags & SVp_NOK) {
4184 SvNV_set(dstr, SvNVX(sstr));
4188 if (isGV_with_GP(sstr)) {
4189 /* This stringification rule for globs is spread in 3 places.
4190 This feels bad. FIXME. */
4191 const U32 wasfake = sflags & SVf_FAKE;
4193 /* FAKE globs can get coerced, so need to turn this off
4194 temporarily if it is on. */
4196 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4197 SvFLAGS(sstr) |= wasfake;
4200 (void)SvOK_off(dstr);
4202 if (SvTAINTED(sstr))
4207 =for apidoc sv_setsv_mg
4209 Like C<sv_setsv>, but also handles 'set' magic.
4215 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4217 PERL_ARGS_ASSERT_SV_SETSV_MG;
4219 sv_setsv(dstr,sstr);
4223 #ifdef PERL_OLD_COPY_ON_WRITE
4225 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4227 STRLEN cur = SvCUR(sstr);
4228 STRLEN len = SvLEN(sstr);
4229 register char *new_pv;
4231 PERL_ARGS_ASSERT_SV_SETSV_COW;
4234 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4235 (void*)sstr, (void*)dstr);
4242 if (SvTHINKFIRST(dstr))
4243 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4244 else if (SvPVX_const(dstr))
4245 Safefree(SvPVX_const(dstr));
4249 SvUPGRADE(dstr, SVt_PVIV);
4251 assert (SvPOK(sstr));
4252 assert (SvPOKp(sstr));
4253 assert (!SvIOK(sstr));
4254 assert (!SvIOKp(sstr));
4255 assert (!SvNOK(sstr));
4256 assert (!SvNOKp(sstr));
4258 if (SvIsCOW(sstr)) {
4260 if (SvLEN(sstr) == 0) {
4261 /* source is a COW shared hash key. */
4262 DEBUG_C(PerlIO_printf(Perl_debug_log,
4263 "Fast copy on write: Sharing hash\n"));
4264 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4267 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4269 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4270 SvUPGRADE(sstr, SVt_PVIV);
4271 SvREADONLY_on(sstr);
4273 DEBUG_C(PerlIO_printf(Perl_debug_log,
4274 "Fast copy on write: Converting sstr to COW\n"));
4275 SV_COW_NEXT_SV_SET(dstr, sstr);
4277 SV_COW_NEXT_SV_SET(sstr, dstr);
4278 new_pv = SvPVX_mutable(sstr);
4281 SvPV_set(dstr, new_pv);
4282 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4285 SvLEN_set(dstr, len);
4286 SvCUR_set(dstr, cur);
4295 =for apidoc sv_setpvn
4297 Copies a string into an SV. The C<len> parameter indicates the number of
4298 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4299 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4305 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4308 register char *dptr;
4310 PERL_ARGS_ASSERT_SV_SETPVN;
4312 SV_CHECK_THINKFIRST_COW_DROP(sv);
4318 /* len is STRLEN which is unsigned, need to copy to signed */
4321 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4323 SvUPGRADE(sv, SVt_PV);
4325 dptr = SvGROW(sv, len + 1);
4326 Move(ptr,dptr,len,char);
4329 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4334 =for apidoc sv_setpvn_mg
4336 Like C<sv_setpvn>, but also handles 'set' magic.
4342 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4344 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4346 sv_setpvn(sv,ptr,len);
4351 =for apidoc sv_setpv
4353 Copies a string into an SV. The string must be null-terminated. Does not
4354 handle 'set' magic. See C<sv_setpv_mg>.
4360 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4363 register STRLEN len;
4365 PERL_ARGS_ASSERT_SV_SETPV;
4367 SV_CHECK_THINKFIRST_COW_DROP(sv);
4373 SvUPGRADE(sv, SVt_PV);
4375 SvGROW(sv, len + 1);
4376 Move(ptr,SvPVX(sv),len+1,char);
4378 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4383 =for apidoc sv_setpv_mg
4385 Like C<sv_setpv>, but also handles 'set' magic.
4391 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4393 PERL_ARGS_ASSERT_SV_SETPV_MG;
4400 =for apidoc sv_usepvn_flags
4402 Tells an SV to use C<ptr> to find its string value. Normally the
4403 string is stored inside the SV but sv_usepvn allows the SV to use an
4404 outside string. The C<ptr> should point to memory that was allocated
4405 by C<malloc>. The string length, C<len>, must be supplied. By default
4406 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4407 so that pointer should not be freed or used by the programmer after
4408 giving it to sv_usepvn, and neither should any pointers from "behind"
4409 that pointer (e.g. ptr + 1) be used.
4411 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4412 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4413 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4414 C<len>, and already meets the requirements for storing in C<SvPVX>)
4420 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4425 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4427 SV_CHECK_THINKFIRST_COW_DROP(sv);
4428 SvUPGRADE(sv, SVt_PV);
4431 if (flags & SV_SMAGIC)
4435 if (SvPVX_const(sv))
4439 if (flags & SV_HAS_TRAILING_NUL)
4440 assert(ptr[len] == '\0');
4443 allocate = (flags & SV_HAS_TRAILING_NUL)
4445 #ifdef Perl_safesysmalloc_size
4448 PERL_STRLEN_ROUNDUP(len + 1);
4450 if (flags & SV_HAS_TRAILING_NUL) {
4451 /* It's long enough - do nothing.
4452 Specfically Perl_newCONSTSUB is relying on this. */
4455 /* Force a move to shake out bugs in callers. */
4456 char *new_ptr = (char*)safemalloc(allocate);
4457 Copy(ptr, new_ptr, len, char);
4458 PoisonFree(ptr,len,char);
4462 ptr = (char*) saferealloc (ptr, allocate);
4465 #ifdef Perl_safesysmalloc_size
4466 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4468 SvLEN_set(sv, allocate);
4472 if (!(flags & SV_HAS_TRAILING_NUL)) {
4475 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4477 if (flags & SV_SMAGIC)
4481 #ifdef PERL_OLD_COPY_ON_WRITE
4482 /* Need to do this *after* making the SV normal, as we need the buffer
4483 pointer to remain valid until after we've copied it. If we let go too early,
4484 another thread could invalidate it by unsharing last of the same hash key
4485 (which it can do by means other than releasing copy-on-write Svs)
4486 or by changing the other copy-on-write SVs in the loop. */
4488 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4490 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4492 { /* this SV was SvIsCOW_normal(sv) */
4493 /* we need to find the SV pointing to us. */
4494 SV *current = SV_COW_NEXT_SV(after);
4496 if (current == sv) {
4497 /* The SV we point to points back to us (there were only two of us
4499 Hence other SV is no longer copy on write either. */
4501 SvREADONLY_off(after);
4503 /* We need to follow the pointers around the loop. */
4505 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4508 /* don't loop forever if the structure is bust, and we have
4509 a pointer into a closed loop. */
4510 assert (current != after);
4511 assert (SvPVX_const(current) == pvx);
4513 /* Make the SV before us point to the SV after us. */
4514 SV_COW_NEXT_SV_SET(current, after);
4520 =for apidoc sv_force_normal_flags
4522 Undo various types of fakery on an SV: if the PV is a shared string, make
4523 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4524 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4525 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4526 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4527 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4528 set to some other value.) In addition, the C<flags> parameter gets passed to
4529 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4530 with flags set to 0.
4536 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4540 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4542 #ifdef PERL_OLD_COPY_ON_WRITE
4543 if (SvREADONLY(sv)) {
4545 const char * const pvx = SvPVX_const(sv);
4546 const STRLEN len = SvLEN(sv);
4547 const STRLEN cur = SvCUR(sv);
4548 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4549 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4550 we'll fail an assertion. */
4551 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4554 PerlIO_printf(Perl_debug_log,
4555 "Copy on write: Force normal %ld\n",
4561 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4564 if (flags & SV_COW_DROP_PV) {
4565 /* OK, so we don't need to copy our buffer. */
4568 SvGROW(sv, cur + 1);
4569 Move(pvx,SvPVX(sv),cur,char);
4574 sv_release_COW(sv, pvx, next);
4576 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4582 else if (IN_PERL_RUNTIME)
4583 Perl_croak(aTHX_ "%s", PL_no_modify);
4586 if (SvREADONLY(sv)) {
4588 const char * const pvx = SvPVX_const(sv);
4589 const STRLEN len = SvCUR(sv);
4594 SvGROW(sv, len + 1);
4595 Move(pvx,SvPVX(sv),len,char);
4597 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4599 else if (IN_PERL_RUNTIME)
4600 Perl_croak(aTHX_ "%s", PL_no_modify);
4604 sv_unref_flags(sv, flags);
4605 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4612 Efficient removal of characters from the beginning of the string buffer.
4613 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4614 the string buffer. The C<ptr> becomes the first character of the adjusted
4615 string. Uses the "OOK hack".
4616 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4617 refer to the same chunk of data.
4623 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4629 const U8 *real_start;
4633 PERL_ARGS_ASSERT_SV_CHOP;
4635 if (!ptr || !SvPOKp(sv))
4637 delta = ptr - SvPVX_const(sv);
4639 /* Nothing to do. */
4642 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4643 nothing uses the value of ptr any more. */
4644 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4645 if (ptr <= SvPVX_const(sv))
4646 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4647 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4648 SV_CHECK_THINKFIRST(sv);
4649 if (delta > max_delta)
4650 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4651 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4652 SvPVX_const(sv) + max_delta);
4655 if (!SvLEN(sv)) { /* make copy of shared string */
4656 const char *pvx = SvPVX_const(sv);
4657 const STRLEN len = SvCUR(sv);
4658 SvGROW(sv, len + 1);
4659 Move(pvx,SvPVX(sv),len,char);
4662 SvFLAGS(sv) |= SVf_OOK;
4665 SvOOK_offset(sv, old_delta);
4667 SvLEN_set(sv, SvLEN(sv) - delta);
4668 SvCUR_set(sv, SvCUR(sv) - delta);
4669 SvPV_set(sv, SvPVX(sv) + delta);
4671 p = (U8 *)SvPVX_const(sv);
4676 real_start = p - delta;
4680 if (delta < 0x100) {
4684 p -= sizeof(STRLEN);
4685 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4689 /* Fill the preceding buffer with sentinals to verify that no-one is
4691 while (p > real_start) {
4699 =for apidoc sv_catpvn
4701 Concatenates the string onto the end of the string which is in the SV. The
4702 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4703 status set, then the bytes appended should be valid UTF-8.
4704 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4706 =for apidoc sv_catpvn_flags
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 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4712 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4713 in terms of this function.
4719 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4723 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4725 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4727 SvGROW(dsv, dlen + slen + 1);
4729 sstr = SvPVX_const(dsv);
4730 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4731 SvCUR_set(dsv, SvCUR(dsv) + slen);
4733 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4735 if (flags & SV_SMAGIC)
4740 =for apidoc sv_catsv
4742 Concatenates the string from SV C<ssv> onto the end of the string in
4743 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4744 not 'set' magic. See C<sv_catsv_mg>.
4746 =for apidoc sv_catsv_flags
4748 Concatenates the string from SV C<ssv> onto the end of the string in
4749 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4750 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4751 and C<sv_catsv_nomg> are implemented in terms of this function.
4756 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4760 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4764 const char *spv = SvPV_const(ssv, slen);
4766 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4767 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4768 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4769 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4770 dsv->sv_flags doesn't have that bit set.
4771 Andy Dougherty 12 Oct 2001
4773 const I32 sutf8 = DO_UTF8(ssv);
4776 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4778 dutf8 = DO_UTF8(dsv);
4780 if (dutf8 != sutf8) {
4782 /* Not modifying source SV, so taking a temporary copy. */
4783 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4785 sv_utf8_upgrade(csv);
4786 spv = SvPV_const(csv, slen);
4789 /* Leave enough space for the cat that's about to happen */
4790 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4792 sv_catpvn_nomg(dsv, spv, slen);
4795 if (flags & SV_SMAGIC)
4800 =for apidoc sv_catpv
4802 Concatenates the string onto the end of the string which is in the SV.
4803 If the SV has the UTF-8 status set, then the bytes appended should be
4804 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4809 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4812 register STRLEN len;
4816 PERL_ARGS_ASSERT_SV_CATPV;
4820 junk = SvPV_force(sv, tlen);
4822 SvGROW(sv, tlen + len + 1);
4824 ptr = SvPVX_const(sv);
4825 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4826 SvCUR_set(sv, SvCUR(sv) + len);
4827 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4832 =for apidoc sv_catpv_mg
4834 Like C<sv_catpv>, but also handles 'set' magic.
4840 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4842 PERL_ARGS_ASSERT_SV_CATPV_MG;
4851 Creates a new SV. A non-zero C<len> parameter indicates the number of
4852 bytes of preallocated string space the SV should have. An extra byte for a
4853 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4854 space is allocated.) The reference count for the new SV is set to 1.
4856 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4857 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4858 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4859 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4860 modules supporting older perls.
4866 Perl_newSV(pTHX_ const STRLEN len)
4873 sv_upgrade(sv, SVt_PV);
4874 SvGROW(sv, len + 1);
4879 =for apidoc sv_magicext
4881 Adds magic to an SV, upgrading it if necessary. Applies the
4882 supplied vtable and returns a pointer to the magic added.
4884 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4885 In particular, you can add magic to SvREADONLY SVs, and add more than
4886 one instance of the same 'how'.
4888 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4889 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4890 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4891 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4893 (This is now used as a subroutine by C<sv_magic>.)
4898 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4899 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4904 PERL_ARGS_ASSERT_SV_MAGICEXT;
4906 SvUPGRADE(sv, SVt_PVMG);
4907 Newxz(mg, 1, MAGIC);
4908 mg->mg_moremagic = SvMAGIC(sv);
4909 SvMAGIC_set(sv, mg);
4911 /* Sometimes a magic contains a reference loop, where the sv and
4912 object refer to each other. To prevent a reference loop that
4913 would prevent such objects being freed, we look for such loops
4914 and if we find one we avoid incrementing the object refcount.
4916 Note we cannot do this to avoid self-tie loops as intervening RV must
4917 have its REFCNT incremented to keep it in existence.
4920 if (!obj || obj == sv ||
4921 how == PERL_MAGIC_arylen ||
4922 how == PERL_MAGIC_symtab ||
4923 (SvTYPE(obj) == SVt_PVGV &&
4924 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4925 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4926 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4931 mg->mg_obj = SvREFCNT_inc_simple(obj);
4932 mg->mg_flags |= MGf_REFCOUNTED;
4935 /* Normal self-ties simply pass a null object, and instead of
4936 using mg_obj directly, use the SvTIED_obj macro to produce a
4937 new RV as needed. For glob "self-ties", we are tieing the PVIO
4938 with an RV obj pointing to the glob containing the PVIO. In
4939 this case, to avoid a reference loop, we need to weaken the
4943 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4944 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4950 mg->mg_len = namlen;
4953 mg->mg_ptr = savepvn(name, namlen);
4954 else if (namlen == HEf_SVKEY) {
4955 /* Yes, this is casting away const. This is only for the case of
4956 HEf_SVKEY. I think we need to document this abberation of the
4957 constness of the API, rather than making name non-const, as
4958 that change propagating outwards a long way. */
4959 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4961 mg->mg_ptr = (char *) name;
4963 mg->mg_virtual = (MGVTBL *) vtable;
4967 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4972 =for apidoc sv_magic
4974 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4975 then adds a new magic item of type C<how> to the head of the magic list.
4977 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4978 handling of the C<name> and C<namlen> arguments.
4980 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4981 to add more than one instance of the same 'how'.
4987 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4988 const char *const name, const I32 namlen)
4991 const MGVTBL *vtable;
4994 PERL_ARGS_ASSERT_SV_MAGIC;
4996 #ifdef PERL_OLD_COPY_ON_WRITE
4998 sv_force_normal_flags(sv, 0);
5000 if (SvREADONLY(sv)) {
5002 /* its okay to attach magic to shared strings; the subsequent
5003 * upgrade to PVMG will unshare the string */
5004 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5007 && how != PERL_MAGIC_regex_global
5008 && how != PERL_MAGIC_bm
5009 && how != PERL_MAGIC_fm
5010 && how != PERL_MAGIC_sv
5011 && how != PERL_MAGIC_backref
5014 Perl_croak(aTHX_ "%s", PL_no_modify);
5017 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5018 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5019 /* sv_magic() refuses to add a magic of the same 'how' as an
5022 if (how == PERL_MAGIC_taint) {
5024 /* Any scalar which already had taint magic on which someone
5025 (erroneously?) did SvIOK_on() or similar will now be
5026 incorrectly sporting public "OK" flags. */
5027 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5035 vtable = &PL_vtbl_sv;
5037 case PERL_MAGIC_overload:
5038 vtable = &PL_vtbl_amagic;
5040 case PERL_MAGIC_overload_elem:
5041 vtable = &PL_vtbl_amagicelem;
5043 case PERL_MAGIC_overload_table:
5044 vtable = &PL_vtbl_ovrld;
5047 vtable = &PL_vtbl_bm;
5049 case PERL_MAGIC_regdata:
5050 vtable = &PL_vtbl_regdata;
5052 case PERL_MAGIC_regdatum:
5053 vtable = &PL_vtbl_regdatum;
5055 case PERL_MAGIC_env:
5056 vtable = &PL_vtbl_env;
5059 vtable = &PL_vtbl_fm;
5061 case PERL_MAGIC_envelem:
5062 vtable = &PL_vtbl_envelem;
5064 case PERL_MAGIC_regex_global:
5065 vtable = &PL_vtbl_mglob;
5067 case PERL_MAGIC_isa:
5068 vtable = &PL_vtbl_isa;
5070 case PERL_MAGIC_isaelem:
5071 vtable = &PL_vtbl_isaelem;
5073 case PERL_MAGIC_nkeys:
5074 vtable = &PL_vtbl_nkeys;
5076 case PERL_MAGIC_dbfile:
5079 case PERL_MAGIC_dbline:
5080 vtable = &PL_vtbl_dbline;
5082 #ifdef USE_LOCALE_COLLATE
5083 case PERL_MAGIC_collxfrm:
5084 vtable = &PL_vtbl_collxfrm;
5086 #endif /* USE_LOCALE_COLLATE */
5087 case PERL_MAGIC_tied:
5088 vtable = &PL_vtbl_pack;
5090 case PERL_MAGIC_tiedelem:
5091 case PERL_MAGIC_tiedscalar:
5092 vtable = &PL_vtbl_packelem;
5095 vtable = &PL_vtbl_regexp;
5097 case PERL_MAGIC_sig:
5098 vtable = &PL_vtbl_sig;
5100 case PERL_MAGIC_sigelem:
5101 vtable = &PL_vtbl_sigelem;
5103 case PERL_MAGIC_taint:
5104 vtable = &PL_vtbl_taint;
5106 case PERL_MAGIC_uvar:
5107 vtable = &PL_vtbl_uvar;
5109 case PERL_MAGIC_vec:
5110 vtable = &PL_vtbl_vec;
5112 case PERL_MAGIC_arylen_p:
5113 case PERL_MAGIC_rhash:
5114 case PERL_MAGIC_symtab:
5115 case PERL_MAGIC_vstring:
5118 case PERL_MAGIC_utf8:
5119 vtable = &PL_vtbl_utf8;
5121 case PERL_MAGIC_substr:
5122 vtable = &PL_vtbl_substr;
5124 case PERL_MAGIC_defelem:
5125 vtable = &PL_vtbl_defelem;
5127 case PERL_MAGIC_arylen:
5128 vtable = &PL_vtbl_arylen;
5130 case PERL_MAGIC_pos:
5131 vtable = &PL_vtbl_pos;
5133 case PERL_MAGIC_backref:
5134 vtable = &PL_vtbl_backref;
5136 case PERL_MAGIC_hintselem:
5137 vtable = &PL_vtbl_hintselem;
5139 case PERL_MAGIC_hints:
5140 vtable = &PL_vtbl_hints;
5142 case PERL_MAGIC_ext:
5143 /* Reserved for use by extensions not perl internals. */
5144 /* Useful for attaching extension internal data to perl vars. */
5145 /* Note that multiple extensions may clash if magical scalars */
5146 /* etc holding private data from one are passed to another. */
5150 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5153 /* Rest of work is done else where */
5154 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5157 case PERL_MAGIC_taint:
5160 case PERL_MAGIC_ext:
5161 case PERL_MAGIC_dbfile:
5168 =for apidoc sv_unmagic
5170 Removes all magic of type C<type> from an SV.
5176 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5181 PERL_ARGS_ASSERT_SV_UNMAGIC;
5183 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5185 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5186 for (mg = *mgp; mg; mg = *mgp) {
5187 if (mg->mg_type == type) {
5188 const MGVTBL* const vtbl = mg->mg_virtual;
5189 *mgp = mg->mg_moremagic;
5190 if (vtbl && vtbl->svt_free)
5191 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5192 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5194 Safefree(mg->mg_ptr);
5195 else if (mg->mg_len == HEf_SVKEY)
5196 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5197 else if (mg->mg_type == PERL_MAGIC_utf8)
5198 Safefree(mg->mg_ptr);
5200 if (mg->mg_flags & MGf_REFCOUNTED)
5201 SvREFCNT_dec(mg->mg_obj);
5205 mgp = &mg->mg_moremagic;
5209 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5210 SvMAGIC_set(sv, NULL);
5217 =for apidoc sv_rvweaken
5219 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5220 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5221 push a back-reference to this RV onto the array of backreferences
5222 associated with that magic. If the RV is magical, set magic will be
5223 called after the RV is cleared.
5229 Perl_sv_rvweaken(pTHX_ SV *const sv)
5233 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5235 if (!SvOK(sv)) /* let undefs pass */
5238 Perl_croak(aTHX_ "Can't weaken a nonreference");
5239 else if (SvWEAKREF(sv)) {
5240 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5244 Perl_sv_add_backref(aTHX_ tsv, sv);
5250 /* Give tsv backref magic if it hasn't already got it, then push a
5251 * back-reference to sv onto the array associated with the backref magic.
5254 /* A discussion about the backreferences array and its refcount:
5256 * The AV holding the backreferences is pointed to either as the mg_obj of
5257 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5258 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5259 * have the standard magic instead.) The array is created with a refcount
5260 * of 2. This means that if during global destruction the array gets
5261 * picked on first to have its refcount decremented by the random zapper,
5262 * it won't actually be freed, meaning it's still theere for when its
5263 * parent gets freed.
5264 * When the parent SV is freed, in the case of magic, the magic is freed,
5265 * Perl_magic_killbackrefs is called which decrements one refcount, then
5266 * mg_obj is freed which kills the second count.
5267 * In the vase of a HV being freed, one ref is removed by
5268 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5273 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5278 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5280 if (SvTYPE(tsv) == SVt_PVHV) {
5281 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5285 /* There is no AV in the offical place - try a fixup. */
5286 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5289 /* Aha. They've got it stowed in magic. Bring it back. */
5290 av = MUTABLE_AV(mg->mg_obj);
5291 /* Stop mg_free decreasing the refernce count. */
5293 /* Stop mg_free even calling the destructor, given that
5294 there's no AV to free up. */
5296 sv_unmagic(tsv, PERL_MAGIC_backref);
5300 SvREFCNT_inc_simple_void(av); /* see discussion above */
5305 const MAGIC *const mg
5306 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5308 av = MUTABLE_AV(mg->mg_obj);
5312 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5313 /* av now has a refcnt of 2; see discussion above */
5316 if (AvFILLp(av) >= AvMAX(av)) {
5317 av_extend(av, AvFILLp(av)+1);
5319 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5322 /* delete a back-reference to ourselves from the backref magic associated
5323 * with the SV we point to.
5327 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5334 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5336 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5337 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5338 /* We mustn't attempt to "fix up" the hash here by moving the
5339 backreference array back to the hv_aux structure, as that is stored
5340 in the main HvARRAY(), and hfreentries assumes that no-one
5341 reallocates HvARRAY() while it is running. */
5344 const MAGIC *const mg
5345 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5347 av = MUTABLE_AV(mg->mg_obj);
5351 Perl_croak(aTHX_ "panic: del_backref");
5353 assert(!SvIS_FREED(av));
5356 /* We shouldn't be in here more than once, but for paranoia reasons lets
5358 for (i = AvFILLp(av); i >= 0; i--) {
5360 const SSize_t fill = AvFILLp(av);
5362 /* We weren't the last entry.
5363 An unordered list has this property that you can take the
5364 last element off the end to fill the hole, and it's still
5365 an unordered list :-)
5370 AvFILLp(av) = fill - 1;
5376 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5378 SV **svp = AvARRAY(av);
5380 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5381 PERL_UNUSED_ARG(sv);
5383 assert(!svp || !SvIS_FREED(av));
5385 SV *const *const last = svp + AvFILLp(av);
5387 while (svp <= last) {
5389 SV *const referrer = *svp;
5390 if (SvWEAKREF(referrer)) {
5391 /* XXX Should we check that it hasn't changed? */
5392 SvRV_set(referrer, 0);
5394 SvWEAKREF_off(referrer);
5395 SvSETMAGIC(referrer);
5396 } else if (SvTYPE(referrer) == SVt_PVGV ||
5397 SvTYPE(referrer) == SVt_PVLV) {
5398 /* You lookin' at me? */
5399 assert(GvSTASH(referrer));
5400 assert(GvSTASH(referrer) == (const HV *)sv);
5401 GvSTASH(referrer) = 0;
5404 "panic: magic_killbackrefs (flags=%"UVxf")",
5405 (UV)SvFLAGS(referrer));
5413 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5418 =for apidoc sv_insert
5420 Inserts a string at the specified offset/length within the SV. Similar to
5421 the Perl substr() function. Handles get magic.
5423 =for apidoc sv_insert_flags
5425 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5431 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5436 register char *midend;
5437 register char *bigend;
5441 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5444 Perl_croak(aTHX_ "Can't modify non-existent substring");
5445 SvPV_force_flags(bigstr, curlen, flags);
5446 (void)SvPOK_only_UTF8(bigstr);
5447 if (offset + len > curlen) {
5448 SvGROW(bigstr, offset+len+1);
5449 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5450 SvCUR_set(bigstr, offset+len);
5454 i = littlelen - len;
5455 if (i > 0) { /* string might grow */
5456 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5457 mid = big + offset + len;
5458 midend = bigend = big + SvCUR(bigstr);
5461 while (midend > mid) /* shove everything down */
5462 *--bigend = *--midend;
5463 Move(little,big+offset,littlelen,char);
5464 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5469 Move(little,SvPVX(bigstr)+offset,len,char);
5474 big = SvPVX(bigstr);
5477 bigend = big + SvCUR(bigstr);
5479 if (midend > bigend)
5480 Perl_croak(aTHX_ "panic: sv_insert");
5482 if (mid - big > bigend - midend) { /* faster to shorten from end */
5484 Move(little, mid, littlelen,char);
5487 i = bigend - midend;
5489 Move(midend, mid, i,char);
5493 SvCUR_set(bigstr, mid - big);
5495 else if ((i = mid - big)) { /* faster from front */
5496 midend -= littlelen;
5498 Move(big, midend - i, i, char);
5499 sv_chop(bigstr,midend-i);
5501 Move(little, mid, littlelen,char);
5503 else if (littlelen) {
5504 midend -= littlelen;
5505 sv_chop(bigstr,midend);
5506 Move(little,midend,littlelen,char);
5509 sv_chop(bigstr,midend);
5515 =for apidoc sv_replace
5517 Make the first argument a copy of the second, then delete the original.
5518 The target SV physically takes over ownership of the body of the source SV
5519 and inherits its flags; however, the target keeps any magic it owns,
5520 and any magic in the source is discarded.
5521 Note that this is a rather specialist SV copying operation; most of the
5522 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5528 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5531 const U32 refcnt = SvREFCNT(sv);
5533 PERL_ARGS_ASSERT_SV_REPLACE;
5535 SV_CHECK_THINKFIRST_COW_DROP(sv);
5536 if (SvREFCNT(nsv) != 1) {
5537 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5538 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5540 if (SvMAGICAL(sv)) {
5544 sv_upgrade(nsv, SVt_PVMG);
5545 SvMAGIC_set(nsv, SvMAGIC(sv));
5546 SvFLAGS(nsv) |= SvMAGICAL(sv);
5548 SvMAGIC_set(sv, NULL);
5552 assert(!SvREFCNT(sv));
5553 #ifdef DEBUG_LEAKING_SCALARS
5554 sv->sv_flags = nsv->sv_flags;
5555 sv->sv_any = nsv->sv_any;
5556 sv->sv_refcnt = nsv->sv_refcnt;
5557 sv->sv_u = nsv->sv_u;
5559 StructCopy(nsv,sv,SV);
5561 if(SvTYPE(sv) == SVt_IV) {
5563 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5567 #ifdef PERL_OLD_COPY_ON_WRITE
5568 if (SvIsCOW_normal(nsv)) {
5569 /* We need to follow the pointers around the loop to make the
5570 previous SV point to sv, rather than nsv. */
5573 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5576 assert(SvPVX_const(current) == SvPVX_const(nsv));
5578 /* Make the SV before us point to the SV after us. */
5580 PerlIO_printf(Perl_debug_log, "previous is\n");
5582 PerlIO_printf(Perl_debug_log,
5583 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5584 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5586 SV_COW_NEXT_SV_SET(current, sv);
5589 SvREFCNT(sv) = refcnt;
5590 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5596 =for apidoc sv_clear
5598 Clear an SV: call any destructors, free up any memory used by the body,
5599 and free the body itself. The SV's head is I<not> freed, although
5600 its type is set to all 1's so that it won't inadvertently be assumed
5601 to be live during global destruction etc.
5602 This function should only be called when REFCNT is zero. Most of the time
5603 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5610 Perl_sv_clear(pTHX_ register SV *const sv)
5613 const U32 type = SvTYPE(sv);
5614 const struct body_details *const sv_type_details
5615 = bodies_by_type + type;
5618 PERL_ARGS_ASSERT_SV_CLEAR;
5619 assert(SvREFCNT(sv) == 0);
5620 assert(SvTYPE(sv) != SVTYPEMASK);
5622 if (type <= SVt_IV) {
5623 /* See the comment in sv.h about the collusion between this early
5624 return and the overloading of the NULL and IV slots in the size
5627 SV * const target = SvRV(sv);
5629 sv_del_backref(target, sv);
5631 SvREFCNT_dec(target);
5633 SvFLAGS(sv) &= SVf_BREAK;
5634 SvFLAGS(sv) |= SVTYPEMASK;
5639 if (PL_defstash && /* Still have a symbol table? */
5646 stash = SvSTASH(sv);
5647 destructor = StashHANDLER(stash,DESTROY);
5649 /* A constant subroutine can have no side effects, so
5650 don't bother calling it. */
5651 && !CvCONST(destructor)
5652 /* Don't bother calling an empty destructor */
5653 && (CvISXSUB(destructor)
5654 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5656 SV* const tmpref = newRV(sv);
5657 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5659 PUSHSTACKi(PERLSI_DESTROY);
5664 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5670 if(SvREFCNT(tmpref) < 2) {
5671 /* tmpref is not kept alive! */
5673 SvRV_set(tmpref, NULL);
5676 SvREFCNT_dec(tmpref);
5678 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5682 if (PL_in_clean_objs)
5683 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5685 /* DESTROY gave object new lease on life */
5691 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5692 SvOBJECT_off(sv); /* Curse the object. */
5693 if (type != SVt_PVIO)
5694 --PL_sv_objcount; /* XXX Might want something more general */
5697 if (type >= SVt_PVMG) {
5698 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5699 SvREFCNT_dec(SvOURSTASH(sv));
5700 } else if (SvMAGIC(sv))
5702 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5703 SvREFCNT_dec(SvSTASH(sv));
5706 /* case SVt_BIND: */
5709 IoIFP(sv) != PerlIO_stdin() &&
5710 IoIFP(sv) != PerlIO_stdout() &&
5711 IoIFP(sv) != PerlIO_stderr())
5713 io_close(MUTABLE_IO(sv), FALSE);
5715 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5716 PerlDir_close(IoDIRP(sv));
5717 IoDIRP(sv) = (DIR*)NULL;
5718 Safefree(IoTOP_NAME(sv));
5719 Safefree(IoFMT_NAME(sv));
5720 Safefree(IoBOTTOM_NAME(sv));
5723 /* FIXME for plugins */
5724 pregfree2((REGEXP*) sv);
5728 cv_undef(MUTABLE_CV(sv));
5731 if (PL_last_swash_hv == (const HV *)sv) {
5732 PL_last_swash_hv = NULL;
5734 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5735 hv_undef(MUTABLE_HV(sv));
5738 if (PL_comppad == MUTABLE_AV(sv)) {
5742 av_undef(MUTABLE_AV(sv));
5745 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5746 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5747 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5748 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5750 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5751 SvREFCNT_dec(LvTARG(sv));
5753 if (isGV_with_GP(sv)) {
5754 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5755 && HvNAME_get(stash))
5756 mro_method_changed_in(stash);
5757 gp_free(MUTABLE_GV(sv));
5759 unshare_hek(GvNAME_HEK(sv));
5760 /* If we're in a stash, we don't own a reference to it. However it does
5761 have a back reference to us, which needs to be cleared. */
5762 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5763 sv_del_backref(MUTABLE_SV(stash), sv);
5765 /* FIXME. There are probably more unreferenced pointers to SVs in the
5766 interpreter struct that we should check and tidy in a similar
5768 if ((const GV *)sv == PL_last_in_gv)
5769 PL_last_in_gv = NULL;
5775 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5778 SvOOK_offset(sv, offset);
5779 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5780 /* Don't even bother with turning off the OOK flag. */
5783 SV * const target = SvRV(sv);
5785 sv_del_backref(target, sv);
5787 SvREFCNT_dec(target);
5789 #ifdef PERL_OLD_COPY_ON_WRITE
5790 else if (SvPVX_const(sv)) {
5793 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5797 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5799 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5803 } else if (SvLEN(sv)) {
5804 Safefree(SvPVX_const(sv));
5808 else if (SvPVX_const(sv) && SvLEN(sv))
5809 Safefree(SvPVX_mutable(sv));
5810 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5811 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5820 SvFLAGS(sv) &= SVf_BREAK;
5821 SvFLAGS(sv) |= SVTYPEMASK;
5823 if (sv_type_details->arena) {
5824 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5825 &PL_body_roots[type]);
5827 else if (sv_type_details->body_size) {
5828 my_safefree(SvANY(sv));
5833 =for apidoc sv_newref
5835 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5842 Perl_sv_newref(pTHX_ SV *const sv)
5844 PERL_UNUSED_CONTEXT;
5853 Decrement an SV's reference count, and if it drops to zero, call
5854 C<sv_clear> to invoke destructors and free up any memory used by
5855 the body; finally, deallocate the SV's head itself.
5856 Normally called via a wrapper macro C<SvREFCNT_dec>.
5862 Perl_sv_free(pTHX_ SV *const sv)
5867 if (SvREFCNT(sv) == 0) {
5868 if (SvFLAGS(sv) & SVf_BREAK)
5869 /* this SV's refcnt has been artificially decremented to
5870 * trigger cleanup */
5872 if (PL_in_clean_all) /* All is fair */
5874 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5875 /* make sure SvREFCNT(sv)==0 happens very seldom */
5876 SvREFCNT(sv) = (~(U32)0)/2;
5879 if (ckWARN_d(WARN_INTERNAL)) {
5880 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5881 Perl_dump_sv_child(aTHX_ sv);
5883 #ifdef DEBUG_LEAKING_SCALARS
5886 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5887 if (PL_warnhook == PERL_WARNHOOK_FATAL
5888 || ckDEAD(packWARN(WARN_INTERNAL))) {
5889 /* Don't let Perl_warner cause us to escape our fate: */
5893 /* This may not return: */
5894 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5895 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5896 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5899 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5904 if (--(SvREFCNT(sv)) > 0)
5906 Perl_sv_free2(aTHX_ sv);
5910 Perl_sv_free2(pTHX_ SV *const sv)
5914 PERL_ARGS_ASSERT_SV_FREE2;
5918 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5919 "Attempt to free temp prematurely: SV 0x%"UVxf
5920 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5924 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5925 /* make sure SvREFCNT(sv)==0 happens very seldom */
5926 SvREFCNT(sv) = (~(U32)0)/2;
5937 Returns the length of the string in the SV. Handles magic and type
5938 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5944 Perl_sv_len(pTHX_ register SV *const sv)
5952 len = mg_length(sv);
5954 (void)SvPV_const(sv, len);
5959 =for apidoc sv_len_utf8
5961 Returns the number of characters in the string in an SV, counting wide
5962 UTF-8 bytes as a single character. Handles magic and type coercion.
5968 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5969 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5970 * (Note that the mg_len is not the length of the mg_ptr field.
5971 * This allows the cache to store the character length of the string without
5972 * needing to malloc() extra storage to attach to the mg_ptr.)
5977 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5983 return mg_length(sv);
5987 const U8 *s = (U8*)SvPV_const(sv, len);
5991 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5993 if (mg && mg->mg_len != -1) {
5995 if (PL_utf8cache < 0) {
5996 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5998 /* Need to turn the assertions off otherwise we may
5999 recurse infinitely while printing error messages.
6001 SAVEI8(PL_utf8cache);
6003 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6004 " real %"UVuf" for %"SVf,
6005 (UV) ulen, (UV) real, SVfARG(sv));
6010 ulen = Perl_utf8_length(aTHX_ s, s + len);
6011 if (!SvREADONLY(sv)) {
6013 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6014 &PL_vtbl_utf8, 0, 0);
6022 return Perl_utf8_length(aTHX_ s, s + len);
6026 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6029 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6032 const U8 *s = start;
6034 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6036 while (s < send && uoffset--)
6039 /* This is the existing behaviour. Possibly it should be a croak, as
6040 it's actually a bounds error */
6046 /* Given the length of the string in both bytes and UTF-8 characters, decide
6047 whether to walk forwards or backwards to find the byte corresponding to
6048 the passed in UTF-8 offset. */
6050 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6051 const STRLEN uoffset, const STRLEN uend)
6053 STRLEN backw = uend - uoffset;
6055 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6057 if (uoffset < 2 * backw) {
6058 /* The assumption is that going forwards is twice the speed of going
6059 forward (that's where the 2 * backw comes from).
6060 (The real figure of course depends on the UTF-8 data.) */
6061 return sv_pos_u2b_forwards(start, send, uoffset);
6066 while (UTF8_IS_CONTINUATION(*send))
6069 return send - start;
6072 /* For the string representation of the given scalar, find the byte
6073 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6074 give another position in the string, *before* the sought offset, which
6075 (which is always true, as 0, 0 is a valid pair of positions), which should
6076 help reduce the amount of linear searching.
6077 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6078 will be used to reduce the amount of linear searching. The cache will be
6079 created if necessary, and the found value offered to it for update. */
6081 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6082 const U8 *const send, const STRLEN uoffset,
6083 STRLEN uoffset0, STRLEN boffset0)
6085 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6088 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6090 assert (uoffset >= uoffset0);
6092 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6093 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6094 if ((*mgp)->mg_ptr) {
6095 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6096 if (cache[0] == uoffset) {
6097 /* An exact match. */
6100 if (cache[2] == uoffset) {
6101 /* An exact match. */
6105 if (cache[0] < uoffset) {
6106 /* The cache already knows part of the way. */
6107 if (cache[0] > uoffset0) {
6108 /* The cache knows more than the passed in pair */
6109 uoffset0 = cache[0];
6110 boffset0 = cache[1];
6112 if ((*mgp)->mg_len != -1) {
6113 /* And we know the end too. */
6115 + sv_pos_u2b_midway(start + boffset0, send,
6117 (*mgp)->mg_len - uoffset0);
6120 + sv_pos_u2b_forwards(start + boffset0,
6121 send, uoffset - uoffset0);
6124 else if (cache[2] < uoffset) {
6125 /* We're between the two cache entries. */
6126 if (cache[2] > uoffset0) {
6127 /* and the cache knows more than the passed in pair */
6128 uoffset0 = cache[2];
6129 boffset0 = cache[3];
6133 + sv_pos_u2b_midway(start + boffset0,
6136 cache[0] - uoffset0);
6139 + sv_pos_u2b_midway(start + boffset0,
6142 cache[2] - uoffset0);
6146 else if ((*mgp)->mg_len != -1) {
6147 /* If we can take advantage of a passed in offset, do so. */
6148 /* In fact, offset0 is either 0, or less than offset, so don't
6149 need to worry about the other possibility. */
6151 + sv_pos_u2b_midway(start + boffset0, send,
6153 (*mgp)->mg_len - uoffset0);
6158 if (!found || PL_utf8cache < 0) {
6159 const STRLEN real_boffset
6160 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6161 send, uoffset - uoffset0);
6163 if (found && PL_utf8cache < 0) {
6164 if (real_boffset != boffset) {
6165 /* Need to turn the assertions off otherwise we may recurse
6166 infinitely while printing error messages. */
6167 SAVEI8(PL_utf8cache);
6169 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6170 " real %"UVuf" for %"SVf,
6171 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6174 boffset = real_boffset;
6178 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6184 =for apidoc sv_pos_u2b
6186 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6187 the start of the string, to a count of the equivalent number of bytes; if
6188 lenp is non-zero, it does the same to lenp, but this time starting from
6189 the offset, rather than from the start of the string. Handles magic and
6196 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6197 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6198 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6203 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6208 PERL_ARGS_ASSERT_SV_POS_U2B;
6213 start = (U8*)SvPV_const(sv, len);
6215 STRLEN uoffset = (STRLEN) *offsetp;
6216 const U8 * const send = start + len;
6218 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6221 *offsetp = (I32) boffset;
6224 /* Convert the relative offset to absolute. */
6225 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6226 const STRLEN boffset2
6227 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6228 uoffset, boffset) - boffset;
6242 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6243 byte length pairing. The (byte) length of the total SV is passed in too,
6244 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6245 may not have updated SvCUR, so we can't rely on reading it directly.
6247 The proffered utf8/byte length pairing isn't used if the cache already has
6248 two pairs, and swapping either for the proffered pair would increase the
6249 RMS of the intervals between known byte offsets.
6251 The cache itself consists of 4 STRLEN values
6252 0: larger UTF-8 offset
6253 1: corresponding byte offset
6254 2: smaller UTF-8 offset
6255 3: corresponding byte offset
6257 Unused cache pairs have the value 0, 0.
6258 Keeping the cache "backwards" means that the invariant of
6259 cache[0] >= cache[2] is maintained even with empty slots, which means that
6260 the code that uses it doesn't need to worry if only 1 entry has actually
6261 been set to non-zero. It also makes the "position beyond the end of the
6262 cache" logic much simpler, as the first slot is always the one to start
6266 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6267 const STRLEN utf8, const STRLEN blen)
6271 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6277 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6279 (*mgp)->mg_len = -1;
6283 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6284 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6285 (*mgp)->mg_ptr = (char *) cache;
6289 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6290 /* SvPOKp() because it's possible that sv has string overloading, and
6291 therefore is a reference, hence SvPVX() is actually a pointer.
6292 This cures the (very real) symptoms of RT 69422, but I'm not actually
6293 sure whether we should even be caching the results of UTF-8
6294 operations on overloading, given that nothing stops overloading
6295 returning a different value every time it's called. */
6296 const U8 *start = (const U8 *) SvPVX_const(sv);
6297 const STRLEN realutf8 = utf8_length(start, start + byte);
6299 if (realutf8 != utf8) {
6300 /* Need to turn the assertions off otherwise we may recurse
6301 infinitely while printing error messages. */
6302 SAVEI8(PL_utf8cache);
6304 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6305 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6309 /* Cache is held with the later position first, to simplify the code
6310 that deals with unbounded ends. */
6312 ASSERT_UTF8_CACHE(cache);
6313 if (cache[1] == 0) {
6314 /* Cache is totally empty */
6317 } else if (cache[3] == 0) {
6318 if (byte > cache[1]) {
6319 /* New one is larger, so goes first. */
6320 cache[2] = cache[0];
6321 cache[3] = cache[1];
6329 #define THREEWAY_SQUARE(a,b,c,d) \
6330 ((float)((d) - (c))) * ((float)((d) - (c))) \
6331 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6332 + ((float)((b) - (a))) * ((float)((b) - (a)))
6334 /* Cache has 2 slots in use, and we know three potential pairs.
6335 Keep the two that give the lowest RMS distance. Do the
6336 calcualation in bytes simply because we always know the byte
6337 length. squareroot has the same ordering as the positive value,
6338 so don't bother with the actual square root. */
6339 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6340 if (byte > cache[1]) {
6341 /* New position is after the existing pair of pairs. */
6342 const float keep_earlier
6343 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6344 const float keep_later
6345 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6347 if (keep_later < keep_earlier) {
6348 if (keep_later < existing) {
6349 cache[2] = cache[0];
6350 cache[3] = cache[1];
6356 if (keep_earlier < existing) {
6362 else if (byte > cache[3]) {
6363 /* New position is between the existing pair of pairs. */
6364 const float keep_earlier
6365 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6366 const float keep_later
6367 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6369 if (keep_later < keep_earlier) {
6370 if (keep_later < existing) {
6376 if (keep_earlier < existing) {
6383 /* New position is before the existing pair of pairs. */
6384 const float keep_earlier
6385 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6386 const float keep_later
6387 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6389 if (keep_later < keep_earlier) {
6390 if (keep_later < existing) {
6396 if (keep_earlier < existing) {
6397 cache[0] = cache[2];
6398 cache[1] = cache[3];
6405 ASSERT_UTF8_CACHE(cache);
6408 /* We already know all of the way, now we may be able to walk back. The same
6409 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6410 backward is half the speed of walking forward. */
6412 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6413 const U8 *end, STRLEN endu)
6415 const STRLEN forw = target - s;
6416 STRLEN backw = end - target;
6418 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6420 if (forw < 2 * backw) {
6421 return utf8_length(s, target);
6424 while (end > target) {
6426 while (UTF8_IS_CONTINUATION(*end)) {
6435 =for apidoc sv_pos_b2u
6437 Converts the value pointed to by offsetp from a count of bytes from the
6438 start of the string, to a count of the equivalent number of UTF-8 chars.
6439 Handles magic and type coercion.
6445 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6446 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6451 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6454 const STRLEN byte = *offsetp;
6455 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6461 PERL_ARGS_ASSERT_SV_POS_B2U;
6466 s = (const U8*)SvPV_const(sv, blen);
6469 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6473 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6474 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6476 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6477 if (cache[1] == byte) {
6478 /* An exact match. */
6479 *offsetp = cache[0];
6482 if (cache[3] == byte) {
6483 /* An exact match. */
6484 *offsetp = cache[2];
6488 if (cache[1] < byte) {
6489 /* We already know part of the way. */
6490 if (mg->mg_len != -1) {
6491 /* Actually, we know the end too. */
6493 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6494 s + blen, mg->mg_len - cache[0]);
6496 len = cache[0] + utf8_length(s + cache[1], send);
6499 else if (cache[3] < byte) {
6500 /* We're between the two cached pairs, so we do the calculation
6501 offset by the byte/utf-8 positions for the earlier pair,
6502 then add the utf-8 characters from the string start to
6504 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6505 s + cache[1], cache[0] - cache[2])
6509 else { /* cache[3] > byte */
6510 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6514 ASSERT_UTF8_CACHE(cache);
6516 } else if (mg->mg_len != -1) {
6517 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6521 if (!found || PL_utf8cache < 0) {
6522 const STRLEN real_len = utf8_length(s, send);
6524 if (found && PL_utf8cache < 0) {
6525 if (len != real_len) {
6526 /* Need to turn the assertions off otherwise we may recurse
6527 infinitely while printing error messages. */
6528 SAVEI8(PL_utf8cache);
6530 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6531 " real %"UVuf" for %"SVf,
6532 (UV) len, (UV) real_len, SVfARG(sv));
6540 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6546 Returns a boolean indicating whether the strings in the two SVs are
6547 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6548 coerce its args to strings if necessary.
6554 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6563 SV* svrecode = NULL;
6570 /* if pv1 and pv2 are the same, second SvPV_const call may
6571 * invalidate pv1, so we may need to make a copy */
6572 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6573 pv1 = SvPV_const(sv1, cur1);
6574 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6576 pv1 = SvPV_const(sv1, cur1);
6584 pv2 = SvPV_const(sv2, cur2);
6586 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6587 /* Differing utf8ness.
6588 * Do not UTF8size the comparands as a side-effect. */
6591 svrecode = newSVpvn(pv2, cur2);
6592 sv_recode_to_utf8(svrecode, PL_encoding);
6593 pv2 = SvPV_const(svrecode, cur2);
6596 svrecode = newSVpvn(pv1, cur1);
6597 sv_recode_to_utf8(svrecode, PL_encoding);
6598 pv1 = SvPV_const(svrecode, cur1);
6600 /* Now both are in UTF-8. */
6602 SvREFCNT_dec(svrecode);
6607 bool is_utf8 = TRUE;
6610 /* sv1 is the UTF-8 one,
6611 * if is equal it must be downgrade-able */
6612 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6618 /* sv2 is the UTF-8 one,
6619 * if is equal it must be downgrade-able */
6620 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6626 /* Downgrade not possible - cannot be eq */
6634 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6636 SvREFCNT_dec(svrecode);
6646 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6647 string in C<sv1> is less than, equal to, or greater than the string in
6648 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6649 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6655 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6659 const char *pv1, *pv2;
6662 SV *svrecode = NULL;
6669 pv1 = SvPV_const(sv1, cur1);
6676 pv2 = SvPV_const(sv2, cur2);
6678 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6679 /* Differing utf8ness.
6680 * Do not UTF8size the comparands as a side-effect. */
6683 svrecode = newSVpvn(pv2, cur2);
6684 sv_recode_to_utf8(svrecode, PL_encoding);
6685 pv2 = SvPV_const(svrecode, cur2);
6688 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6693 svrecode = newSVpvn(pv1, cur1);
6694 sv_recode_to_utf8(svrecode, PL_encoding);
6695 pv1 = SvPV_const(svrecode, cur1);
6698 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6704 cmp = cur2 ? -1 : 0;
6708 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6711 cmp = retval < 0 ? -1 : 1;
6712 } else if (cur1 == cur2) {
6715 cmp = cur1 < cur2 ? -1 : 1;
6719 SvREFCNT_dec(svrecode);
6727 =for apidoc sv_cmp_locale
6729 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6730 'use bytes' aware, handles get magic, and will coerce its args to strings
6731 if necessary. See also C<sv_cmp>.
6737 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6740 #ifdef USE_LOCALE_COLLATE
6746 if (PL_collation_standard)
6750 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6752 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6754 if (!pv1 || !len1) {
6765 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6768 return retval < 0 ? -1 : 1;
6771 * When the result of collation is equality, that doesn't mean
6772 * that there are no differences -- some locales exclude some
6773 * characters from consideration. So to avoid false equalities,
6774 * we use the raw string as a tiebreaker.
6780 #endif /* USE_LOCALE_COLLATE */
6782 return sv_cmp(sv1, sv2);
6786 #ifdef USE_LOCALE_COLLATE
6789 =for apidoc sv_collxfrm
6791 Add Collate Transform magic to an SV if it doesn't already have it.
6793 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6794 scalar data of the variable, but transformed to such a format that a normal
6795 memory comparison can be used to compare the data according to the locale
6802 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6807 PERL_ARGS_ASSERT_SV_COLLXFRM;
6809 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6810 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6816 Safefree(mg->mg_ptr);
6817 s = SvPV_const(sv, len);
6818 if ((xf = mem_collxfrm(s, len, &xlen))) {
6820 #ifdef PERL_OLD_COPY_ON_WRITE
6822 sv_force_normal_flags(sv, 0);
6824 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6838 if (mg && mg->mg_ptr) {
6840 return mg->mg_ptr + sizeof(PL_collation_ix);
6848 #endif /* USE_LOCALE_COLLATE */
6853 Get a line from the filehandle and store it into the SV, optionally
6854 appending to the currently-stored string.
6860 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6865 register STDCHAR rslast;
6866 register STDCHAR *bp;
6871 PERL_ARGS_ASSERT_SV_GETS;
6873 if (SvTHINKFIRST(sv))
6874 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6875 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6877 However, perlbench says it's slower, because the existing swipe code
6878 is faster than copy on write.
6879 Swings and roundabouts. */
6880 SvUPGRADE(sv, SVt_PV);
6885 if (PerlIO_isutf8(fp)) {
6887 sv_utf8_upgrade_nomg(sv);
6888 sv_pos_u2b(sv,&append,0);
6890 } else if (SvUTF8(sv)) {
6891 SV * const tsv = newSV(0);
6892 sv_gets(tsv, fp, 0);
6893 sv_utf8_upgrade_nomg(tsv);
6894 SvCUR_set(sv,append);
6897 goto return_string_or_null;
6902 if (PerlIO_isutf8(fp))
6905 if (IN_PERL_COMPILETIME) {
6906 /* we always read code in line mode */
6910 else if (RsSNARF(PL_rs)) {
6911 /* If it is a regular disk file use size from stat() as estimate
6912 of amount we are going to read -- may result in mallocing
6913 more memory than we really need if the layers below reduce
6914 the size we read (e.g. CRLF or a gzip layer).
6917 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6918 const Off_t offset = PerlIO_tell(fp);
6919 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6920 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6926 else if (RsRECORD(PL_rs)) {
6934 /* Grab the size of the record we're getting */
6935 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6936 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6939 /* VMS wants read instead of fread, because fread doesn't respect */
6940 /* RMS record boundaries. This is not necessarily a good thing to be */
6941 /* doing, but we've got no other real choice - except avoid stdio
6942 as implementation - perhaps write a :vms layer ?
6944 fd = PerlIO_fileno(fp);
6945 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6946 bytesread = PerlIO_read(fp, buffer, recsize);
6949 bytesread = PerlLIO_read(fd, buffer, recsize);
6952 bytesread = PerlIO_read(fp, buffer, recsize);
6956 SvCUR_set(sv, bytesread + append);
6957 buffer[bytesread] = '\0';
6958 goto return_string_or_null;
6960 else if (RsPARA(PL_rs)) {
6966 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6967 if (PerlIO_isutf8(fp)) {
6968 rsptr = SvPVutf8(PL_rs, rslen);
6971 if (SvUTF8(PL_rs)) {
6972 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6973 Perl_croak(aTHX_ "Wide character in $/");
6976 rsptr = SvPV_const(PL_rs, rslen);
6980 rslast = rslen ? rsptr[rslen - 1] : '\0';
6982 if (rspara) { /* have to do this both before and after */
6983 do { /* to make sure file boundaries work right */
6986 i = PerlIO_getc(fp);
6990 PerlIO_ungetc(fp,i);
6996 /* See if we know enough about I/O mechanism to cheat it ! */
6998 /* This used to be #ifdef test - it is made run-time test for ease
6999 of abstracting out stdio interface. One call should be cheap
7000 enough here - and may even be a macro allowing compile
7004 if (PerlIO_fast_gets(fp)) {
7007 * We're going to steal some values from the stdio struct
7008 * and put EVERYTHING in the innermost loop into registers.
7010 register STDCHAR *ptr;
7014 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7015 /* An ungetc()d char is handled separately from the regular
7016 * buffer, so we getc() it back out and stuff it in the buffer.
7018 i = PerlIO_getc(fp);
7019 if (i == EOF) return 0;
7020 *(--((*fp)->_ptr)) = (unsigned char) i;
7024 /* Here is some breathtakingly efficient cheating */
7026 cnt = PerlIO_get_cnt(fp); /* get count into register */
7027 /* make sure we have the room */
7028 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7029 /* Not room for all of it
7030 if we are looking for a separator and room for some
7032 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7033 /* just process what we have room for */
7034 shortbuffered = cnt - SvLEN(sv) + append + 1;
7035 cnt -= shortbuffered;
7039 /* remember that cnt can be negative */
7040 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7045 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7046 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7047 DEBUG_P(PerlIO_printf(Perl_debug_log,
7048 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7049 DEBUG_P(PerlIO_printf(Perl_debug_log,
7050 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7051 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7052 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7057 while (cnt > 0) { /* this | eat */
7059 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7060 goto thats_all_folks; /* screams | sed :-) */
7064 Copy(ptr, bp, cnt, char); /* this | eat */
7065 bp += cnt; /* screams | dust */
7066 ptr += cnt; /* louder | sed :-) */
7071 if (shortbuffered) { /* oh well, must extend */
7072 cnt = shortbuffered;
7074 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7076 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7077 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7081 DEBUG_P(PerlIO_printf(Perl_debug_log,
7082 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7083 PTR2UV(ptr),(long)cnt));
7084 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7086 DEBUG_P(PerlIO_printf(Perl_debug_log,
7087 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7088 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7089 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7091 /* This used to call 'filbuf' in stdio form, but as that behaves like
7092 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7093 another abstraction. */
7094 i = PerlIO_getc(fp); /* get more characters */
7096 DEBUG_P(PerlIO_printf(Perl_debug_log,
7097 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7098 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7099 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7101 cnt = PerlIO_get_cnt(fp);
7102 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7103 DEBUG_P(PerlIO_printf(Perl_debug_log,
7104 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7106 if (i == EOF) /* all done for ever? */
7107 goto thats_really_all_folks;
7109 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7111 SvGROW(sv, bpx + cnt + 2);
7112 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7114 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7116 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7117 goto thats_all_folks;
7121 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7122 memNE((char*)bp - rslen, rsptr, rslen))
7123 goto screamer; /* go back to the fray */
7124 thats_really_all_folks:
7126 cnt += shortbuffered;
7127 DEBUG_P(PerlIO_printf(Perl_debug_log,
7128 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7129 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7130 DEBUG_P(PerlIO_printf(Perl_debug_log,
7131 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7132 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7133 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7135 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7136 DEBUG_P(PerlIO_printf(Perl_debug_log,
7137 "Screamer: done, len=%ld, string=|%.*s|\n",
7138 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7142 /*The big, slow, and stupid way. */
7143 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7144 STDCHAR *buf = NULL;
7145 Newx(buf, 8192, STDCHAR);
7153 register const STDCHAR * const bpe = buf + sizeof(buf);
7155 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7156 ; /* keep reading */
7160 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7161 /* Accomodate broken VAXC compiler, which applies U8 cast to
7162 * both args of ?: operator, causing EOF to change into 255
7165 i = (U8)buf[cnt - 1];
7171 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7173 sv_catpvn(sv, (char *) buf, cnt);
7175 sv_setpvn(sv, (char *) buf, cnt);
7177 if (i != EOF && /* joy */
7179 SvCUR(sv) < rslen ||
7180 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7184 * If we're reading from a TTY and we get a short read,
7185 * indicating that the user hit his EOF character, we need
7186 * to notice it now, because if we try to read from the TTY
7187 * again, the EOF condition will disappear.
7189 * The comparison of cnt to sizeof(buf) is an optimization
7190 * that prevents unnecessary calls to feof().
7194 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7198 #ifdef USE_HEAP_INSTEAD_OF_STACK
7203 if (rspara) { /* have to do this both before and after */
7204 while (i != EOF) { /* to make sure file boundaries work right */
7205 i = PerlIO_getc(fp);
7207 PerlIO_ungetc(fp,i);
7213 return_string_or_null:
7214 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7220 Auto-increment of the value in the SV, doing string to numeric conversion
7221 if necessary. Handles 'get' magic.
7227 Perl_sv_inc(pTHX_ register SV *const sv)
7236 if (SvTHINKFIRST(sv)) {
7238 sv_force_normal_flags(sv, 0);
7239 if (SvREADONLY(sv)) {
7240 if (IN_PERL_RUNTIME)
7241 Perl_croak(aTHX_ "%s", PL_no_modify);
7245 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7247 i = PTR2IV(SvRV(sv));
7252 flags = SvFLAGS(sv);
7253 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7254 /* It's (privately or publicly) a float, but not tested as an
7255 integer, so test it to see. */
7257 flags = SvFLAGS(sv);
7259 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7260 /* It's publicly an integer, or privately an integer-not-float */
7261 #ifdef PERL_PRESERVE_IVUV
7265 if (SvUVX(sv) == UV_MAX)
7266 sv_setnv(sv, UV_MAX_P1);
7268 (void)SvIOK_only_UV(sv);
7269 SvUV_set(sv, SvUVX(sv) + 1);
7271 if (SvIVX(sv) == IV_MAX)
7272 sv_setuv(sv, (UV)IV_MAX + 1);
7274 (void)SvIOK_only(sv);
7275 SvIV_set(sv, SvIVX(sv) + 1);
7280 if (flags & SVp_NOK) {
7281 const NV was = SvNVX(sv);
7282 if (NV_OVERFLOWS_INTEGERS_AT &&
7283 was >= NV_OVERFLOWS_INTEGERS_AT) {
7284 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7285 "Lost precision when incrementing %" NVff " by 1",
7288 (void)SvNOK_only(sv);
7289 SvNV_set(sv, was + 1.0);
7293 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7294 if ((flags & SVTYPEMASK) < SVt_PVIV)
7295 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7296 (void)SvIOK_only(sv);
7301 while (isALPHA(*d)) d++;
7302 while (isDIGIT(*d)) d++;
7303 if (d < SvEND(sv)) {
7304 #ifdef PERL_PRESERVE_IVUV
7305 /* Got to punt this as an integer if needs be, but we don't issue
7306 warnings. Probably ought to make the sv_iv_please() that does
7307 the conversion if possible, and silently. */
7308 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7309 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7310 /* Need to try really hard to see if it's an integer.
7311 9.22337203685478e+18 is an integer.
7312 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7313 so $a="9.22337203685478e+18"; $a+0; $a++
7314 needs to be the same as $a="9.22337203685478e+18"; $a++
7321 /* sv_2iv *should* have made this an NV */
7322 if (flags & SVp_NOK) {
7323 (void)SvNOK_only(sv);
7324 SvNV_set(sv, SvNVX(sv) + 1.0);
7327 /* I don't think we can get here. Maybe I should assert this
7328 And if we do get here I suspect that sv_setnv will croak. NWC
7330 #if defined(USE_LONG_DOUBLE)
7331 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",
7332 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7334 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7335 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7338 #endif /* PERL_PRESERVE_IVUV */
7339 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7343 while (d >= SvPVX_const(sv)) {
7351 /* MKS: The original code here died if letters weren't consecutive.
7352 * at least it didn't have to worry about non-C locales. The
7353 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7354 * arranged in order (although not consecutively) and that only
7355 * [A-Za-z] are accepted by isALPHA in the C locale.
7357 if (*d != 'z' && *d != 'Z') {
7358 do { ++*d; } while (!isALPHA(*d));
7361 *(d--) -= 'z' - 'a';
7366 *(d--) -= 'z' - 'a' + 1;
7370 /* oh,oh, the number grew */
7371 SvGROW(sv, SvCUR(sv) + 2);
7372 SvCUR_set(sv, SvCUR(sv) + 1);
7373 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7384 Auto-decrement of the value in the SV, doing string to numeric conversion
7385 if necessary. Handles 'get' magic.
7391 Perl_sv_dec(pTHX_ register SV *const sv)
7399 if (SvTHINKFIRST(sv)) {
7401 sv_force_normal_flags(sv, 0);
7402 if (SvREADONLY(sv)) {
7403 if (IN_PERL_RUNTIME)
7404 Perl_croak(aTHX_ "%s", PL_no_modify);
7408 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7410 i = PTR2IV(SvRV(sv));
7415 /* Unlike sv_inc we don't have to worry about string-never-numbers
7416 and keeping them magic. But we mustn't warn on punting */
7417 flags = SvFLAGS(sv);
7418 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7419 /* It's publicly an integer, or privately an integer-not-float */
7420 #ifdef PERL_PRESERVE_IVUV
7424 if (SvUVX(sv) == 0) {
7425 (void)SvIOK_only(sv);
7429 (void)SvIOK_only_UV(sv);
7430 SvUV_set(sv, SvUVX(sv) - 1);
7433 if (SvIVX(sv) == IV_MIN) {
7434 sv_setnv(sv, (NV)IV_MIN);
7438 (void)SvIOK_only(sv);
7439 SvIV_set(sv, SvIVX(sv) - 1);
7444 if (flags & SVp_NOK) {
7447 const NV was = SvNVX(sv);
7448 if (NV_OVERFLOWS_INTEGERS_AT &&
7449 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7450 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7451 "Lost precision when decrementing %" NVff " by 1",
7454 (void)SvNOK_only(sv);
7455 SvNV_set(sv, was - 1.0);
7459 if (!(flags & SVp_POK)) {
7460 if ((flags & SVTYPEMASK) < SVt_PVIV)
7461 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7463 (void)SvIOK_only(sv);
7466 #ifdef PERL_PRESERVE_IVUV
7468 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7469 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7470 /* Need to try really hard to see if it's an integer.
7471 9.22337203685478e+18 is an integer.
7472 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7473 so $a="9.22337203685478e+18"; $a+0; $a--
7474 needs to be the same as $a="9.22337203685478e+18"; $a--
7481 /* sv_2iv *should* have made this an NV */
7482 if (flags & SVp_NOK) {
7483 (void)SvNOK_only(sv);
7484 SvNV_set(sv, SvNVX(sv) - 1.0);
7487 /* I don't think we can get here. Maybe I should assert this
7488 And if we do get here I suspect that sv_setnv will croak. NWC
7490 #if defined(USE_LONG_DOUBLE)
7491 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",
7492 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7494 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7495 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7499 #endif /* PERL_PRESERVE_IVUV */
7500 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7503 /* this define is used to eliminate a chunk of duplicated but shared logic
7504 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7505 * used anywhere but here - yves
7507 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7510 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7514 =for apidoc sv_mortalcopy
7516 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7517 The new SV is marked as mortal. It will be destroyed "soon", either by an
7518 explicit call to FREETMPS, or by an implicit call at places such as
7519 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7524 /* Make a string that will exist for the duration of the expression
7525 * evaluation. Actually, it may have to last longer than that, but
7526 * hopefully we won't free it until it has been assigned to a
7527 * permanent location. */
7530 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7536 sv_setsv(sv,oldstr);
7537 PUSH_EXTEND_MORTAL__SV_C(sv);
7543 =for apidoc sv_newmortal
7545 Creates a new null SV which is mortal. The reference count of the SV is
7546 set to 1. It will be destroyed "soon", either by an explicit call to
7547 FREETMPS, or by an implicit call at places such as statement boundaries.
7548 See also C<sv_mortalcopy> and C<sv_2mortal>.
7554 Perl_sv_newmortal(pTHX)
7560 SvFLAGS(sv) = SVs_TEMP;
7561 PUSH_EXTEND_MORTAL__SV_C(sv);
7567 =for apidoc newSVpvn_flags
7569 Creates a new SV and copies a string into it. The reference count for the
7570 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7571 string. You are responsible for ensuring that the source string is at least
7572 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7573 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7574 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7575 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7576 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7578 #define newSVpvn_utf8(s, len, u) \
7579 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7585 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7590 /* All the flags we don't support must be zero.
7591 And we're new code so I'm going to assert this from the start. */
7592 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7594 sv_setpvn(sv,s,len);
7596 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7597 * and do what it does outselves here.
7598 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7599 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7600 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7601 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7604 SvFLAGS(sv) |= flags;
7606 if(flags & SVs_TEMP){
7607 PUSH_EXTEND_MORTAL__SV_C(sv);
7614 =for apidoc sv_2mortal
7616 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7617 by an explicit call to FREETMPS, or by an implicit call at places such as
7618 statement boundaries. SvTEMP() is turned on which means that the SV's
7619 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7620 and C<sv_mortalcopy>.
7626 Perl_sv_2mortal(pTHX_ register SV *const sv)
7631 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7633 PUSH_EXTEND_MORTAL__SV_C(sv);
7641 Creates a new SV and copies a string into it. The reference count for the
7642 SV is set to 1. If C<len> is zero, Perl will compute the length using
7643 strlen(). For efficiency, consider using C<newSVpvn> instead.
7649 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7655 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7660 =for apidoc newSVpvn
7662 Creates a new SV and copies a string into it. The reference count for the
7663 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7664 string. You are responsible for ensuring that the source string is at least
7665 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7671 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7677 sv_setpvn(sv,s,len);
7682 =for apidoc newSVhek
7684 Creates a new SV from the hash key structure. It will generate scalars that
7685 point to the shared string table where possible. Returns a new (undefined)
7686 SV if the hek is NULL.
7692 Perl_newSVhek(pTHX_ const HEK *const hek)
7702 if (HEK_LEN(hek) == HEf_SVKEY) {
7703 return newSVsv(*(SV**)HEK_KEY(hek));
7705 const int flags = HEK_FLAGS(hek);
7706 if (flags & HVhek_WASUTF8) {
7708 Andreas would like keys he put in as utf8 to come back as utf8
7710 STRLEN utf8_len = HEK_LEN(hek);
7711 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7712 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7715 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7717 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7718 /* We don't have a pointer to the hv, so we have to replicate the
7719 flag into every HEK. This hv is using custom a hasing
7720 algorithm. Hence we can't return a shared string scalar, as
7721 that would contain the (wrong) hash value, and might get passed
7722 into an hv routine with a regular hash.
7723 Similarly, a hash that isn't using shared hash keys has to have
7724 the flag in every key so that we know not to try to call
7725 share_hek_kek on it. */
7727 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7732 /* This will be overwhelminly the most common case. */
7734 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7735 more efficient than sharepvn(). */
7739 sv_upgrade(sv, SVt_PV);
7740 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7741 SvCUR_set(sv, HEK_LEN(hek));
7754 =for apidoc newSVpvn_share
7756 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7757 table. If the string does not already exist in the table, it is created
7758 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7759 value is used; otherwise the hash is computed. The string's hash can be later
7760 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7761 that as the string table is used for shared hash keys these strings will have
7762 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7768 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7772 bool is_utf8 = FALSE;
7773 const char *const orig_src = src;
7776 STRLEN tmplen = -len;
7778 /* See the note in hv.c:hv_fetch() --jhi */
7779 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7783 PERL_HASH(hash, src, len);
7785 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7786 changes here, update it there too. */
7787 sv_upgrade(sv, SVt_PV);
7788 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7796 if (src != orig_src)
7802 #if defined(PERL_IMPLICIT_CONTEXT)
7804 /* pTHX_ magic can't cope with varargs, so this is a no-context
7805 * version of the main function, (which may itself be aliased to us).
7806 * Don't access this version directly.
7810 Perl_newSVpvf_nocontext(const char *const pat, ...)
7816 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7818 va_start(args, pat);
7819 sv = vnewSVpvf(pat, &args);
7826 =for apidoc newSVpvf
7828 Creates a new SV and initializes it with the string formatted like
7835 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7840 PERL_ARGS_ASSERT_NEWSVPVF;
7842 va_start(args, pat);
7843 sv = vnewSVpvf(pat, &args);
7848 /* backend for newSVpvf() and newSVpvf_nocontext() */
7851 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7856 PERL_ARGS_ASSERT_VNEWSVPVF;
7859 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7866 Creates a new SV and copies a floating point value into it.
7867 The reference count for the SV is set to 1.
7873 Perl_newSVnv(pTHX_ const NV n)
7886 Creates a new SV and copies an integer into it. The reference count for the
7893 Perl_newSViv(pTHX_ const IV i)
7906 Creates a new SV and copies an unsigned integer into it.
7907 The reference count for the SV is set to 1.
7913 Perl_newSVuv(pTHX_ const UV u)
7924 =for apidoc newSV_type
7926 Creates a new SV, of the type specified. The reference count for the new SV
7933 Perl_newSV_type(pTHX_ const svtype type)
7938 sv_upgrade(sv, type);
7943 =for apidoc newRV_noinc
7945 Creates an RV wrapper for an SV. The reference count for the original
7946 SV is B<not> incremented.
7952 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7955 register SV *sv = newSV_type(SVt_IV);
7957 PERL_ARGS_ASSERT_NEWRV_NOINC;
7960 SvRV_set(sv, tmpRef);
7965 /* newRV_inc is the official function name to use now.
7966 * newRV_inc is in fact #defined to newRV in sv.h
7970 Perl_newRV(pTHX_ SV *const sv)
7974 PERL_ARGS_ASSERT_NEWRV;
7976 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7982 Creates a new SV which is an exact duplicate of the original SV.
7989 Perl_newSVsv(pTHX_ register SV *const old)
7996 if (SvTYPE(old) == SVTYPEMASK) {
7997 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8001 /* SV_GMAGIC is the default for sv_setv()
8002 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8003 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8004 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8009 =for apidoc sv_reset
8011 Underlying implementation for the C<reset> Perl function.
8012 Note that the perl-level function is vaguely deprecated.
8018 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8021 char todo[PERL_UCHAR_MAX+1];
8023 PERL_ARGS_ASSERT_SV_RESET;
8028 if (!*s) { /* reset ?? searches */
8029 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8031 const U32 count = mg->mg_len / sizeof(PMOP**);
8032 PMOP **pmp = (PMOP**) mg->mg_ptr;
8033 PMOP *const *const end = pmp + count;
8037 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8039 (*pmp)->op_pmflags &= ~PMf_USED;
8047 /* reset variables */
8049 if (!HvARRAY(stash))
8052 Zero(todo, 256, char);
8055 I32 i = (unsigned char)*s;
8059 max = (unsigned char)*s++;
8060 for ( ; i <= max; i++) {
8063 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8065 for (entry = HvARRAY(stash)[i];
8067 entry = HeNEXT(entry))
8072 if (!todo[(U8)*HeKEY(entry)])
8074 gv = MUTABLE_GV(HeVAL(entry));
8077 if (SvTHINKFIRST(sv)) {
8078 if (!SvREADONLY(sv) && SvROK(sv))
8080 /* XXX Is this continue a bug? Why should THINKFIRST
8081 exempt us from resetting arrays and hashes? */
8085 if (SvTYPE(sv) >= SVt_PV) {
8087 if (SvPVX_const(sv) != NULL)
8095 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8097 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8100 # if defined(USE_ENVIRON_ARRAY)
8103 # endif /* USE_ENVIRON_ARRAY */
8114 Using various gambits, try to get an IO from an SV: the IO slot if its a
8115 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8116 named after the PV if we're a string.
8122 Perl_sv_2io(pTHX_ SV *const sv)
8127 PERL_ARGS_ASSERT_SV_2IO;
8129 switch (SvTYPE(sv)) {
8131 io = MUTABLE_IO(sv);
8134 if (isGV_with_GP(sv)) {
8135 gv = MUTABLE_GV(sv);
8138 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8144 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8146 return sv_2io(SvRV(sv));
8147 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8153 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8162 Using various gambits, try to get a CV from an SV; in addition, try if
8163 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8164 The flags in C<lref> are passed to gv_fetchsv.
8170 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8176 PERL_ARGS_ASSERT_SV_2CV;
8183 switch (SvTYPE(sv)) {
8187 return MUTABLE_CV(sv);
8194 if (isGV_with_GP(sv)) {
8195 gv = MUTABLE_GV(sv);
8204 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8206 tryAMAGICunDEREF(to_cv);
8209 if (SvTYPE(sv) == SVt_PVCV) {
8210 cv = MUTABLE_CV(sv);
8215 else if(isGV_with_GP(sv))
8216 gv = MUTABLE_GV(sv);
8218 Perl_croak(aTHX_ "Not a subroutine reference");
8220 else if (isGV_with_GP(sv)) {
8222 gv = MUTABLE_GV(sv);
8225 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8231 /* Some flags to gv_fetchsv mean don't really create the GV */
8232 if (!isGV_with_GP(gv)) {
8238 if (lref && !GvCVu(gv)) {
8242 gv_efullname3(tmpsv, gv, NULL);
8243 /* XXX this is probably not what they think they're getting.
8244 * It has the same effect as "sub name;", i.e. just a forward
8246 newSUB(start_subparse(FALSE, 0),
8247 newSVOP(OP_CONST, 0, tmpsv),
8251 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8252 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8261 Returns true if the SV has a true value by Perl's rules.
8262 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8263 instead use an in-line version.
8269 Perl_sv_true(pTHX_ register SV *const sv)
8274 register const XPV* const tXpv = (XPV*)SvANY(sv);
8276 (tXpv->xpv_cur > 1 ||
8277 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8284 return SvIVX(sv) != 0;
8287 return SvNVX(sv) != 0.0;
8289 return sv_2bool(sv);
8295 =for apidoc sv_pvn_force
8297 Get a sensible string out of the SV somehow.
8298 A private implementation of the C<SvPV_force> macro for compilers which
8299 can't cope with complex macro expressions. Always use the macro instead.
8301 =for apidoc sv_pvn_force_flags
8303 Get a sensible string out of the SV somehow.
8304 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8305 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8306 implemented in terms of this function.
8307 You normally want to use the various wrapper macros instead: see
8308 C<SvPV_force> and C<SvPV_force_nomg>
8314 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8318 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8320 if (SvTHINKFIRST(sv) && !SvROK(sv))
8321 sv_force_normal_flags(sv, 0);
8331 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8332 const char * const ref = sv_reftype(sv,0);
8334 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8335 ref, OP_NAME(PL_op));
8337 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8339 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8340 || isGV_with_GP(sv))
8341 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8343 s = sv_2pv_flags(sv, &len, flags);
8347 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8350 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8351 SvGROW(sv, len + 1);
8352 Move(s,SvPVX(sv),len,char);
8354 SvPVX(sv)[len] = '\0';
8357 SvPOK_on(sv); /* validate pointer */
8359 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8360 PTR2UV(sv),SvPVX_const(sv)));
8363 return SvPVX_mutable(sv);
8367 =for apidoc sv_pvbyten_force
8369 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8375 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8377 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8379 sv_pvn_force(sv,lp);
8380 sv_utf8_downgrade(sv,0);
8386 =for apidoc sv_pvutf8n_force
8388 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8394 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8396 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8398 sv_pvn_force(sv,lp);
8399 sv_utf8_upgrade(sv);
8405 =for apidoc sv_reftype
8407 Returns a string describing what the SV is a reference to.
8413 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8415 PERL_ARGS_ASSERT_SV_REFTYPE;
8417 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8418 inside return suggests a const propagation bug in g++. */
8419 if (ob && SvOBJECT(sv)) {
8420 char * const name = HvNAME_get(SvSTASH(sv));
8421 return name ? name : (char *) "__ANON__";
8424 switch (SvTYPE(sv)) {
8439 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8440 /* tied lvalues should appear to be
8441 * scalars for backwards compatitbility */
8442 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8443 ? "SCALAR" : "LVALUE");
8444 case SVt_PVAV: return "ARRAY";
8445 case SVt_PVHV: return "HASH";
8446 case SVt_PVCV: return "CODE";
8447 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8448 ? "GLOB" : "SCALAR");
8449 case SVt_PVFM: return "FORMAT";
8450 case SVt_PVIO: return "IO";
8451 case SVt_BIND: return "BIND";
8452 case SVt_REGEXP: return "REGEXP";
8453 default: return "UNKNOWN";
8459 =for apidoc sv_isobject
8461 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8462 object. If the SV is not an RV, or if the object is not blessed, then this
8469 Perl_sv_isobject(pTHX_ SV *sv)
8485 Returns a boolean indicating whether the SV is blessed into the specified
8486 class. This does not check for subtypes; use C<sv_derived_from> to verify
8487 an inheritance relationship.
8493 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8497 PERL_ARGS_ASSERT_SV_ISA;
8507 hvname = HvNAME_get(SvSTASH(sv));
8511 return strEQ(hvname, name);
8517 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8518 it will be upgraded to one. If C<classname> is non-null then the new SV will
8519 be blessed in the specified package. The new SV is returned and its
8520 reference count is 1.
8526 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8531 PERL_ARGS_ASSERT_NEWSVRV;
8535 SV_CHECK_THINKFIRST_COW_DROP(rv);
8536 (void)SvAMAGIC_off(rv);
8538 if (SvTYPE(rv) >= SVt_PVMG) {
8539 const U32 refcnt = SvREFCNT(rv);
8543 SvREFCNT(rv) = refcnt;
8545 sv_upgrade(rv, SVt_IV);
8546 } else if (SvROK(rv)) {
8547 SvREFCNT_dec(SvRV(rv));
8549 prepare_SV_for_RV(rv);
8557 HV* const stash = gv_stashpv(classname, GV_ADD);
8558 (void)sv_bless(rv, stash);
8564 =for apidoc sv_setref_pv
8566 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8567 argument will be upgraded to an RV. That RV will be modified to point to
8568 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8569 into the SV. The C<classname> argument indicates the package for the
8570 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8571 will have a reference count of 1, and the RV will be returned.
8573 Do not use with other Perl types such as HV, AV, SV, CV, because those
8574 objects will become corrupted by the pointer copy process.
8576 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8582 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8586 PERL_ARGS_ASSERT_SV_SETREF_PV;
8589 sv_setsv(rv, &PL_sv_undef);
8593 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8598 =for apidoc sv_setref_iv
8600 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8601 argument will be upgraded to an RV. That RV will be modified to point to
8602 the new SV. The C<classname> argument indicates the package for the
8603 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8604 will have a reference count of 1, and the RV will be returned.
8610 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8612 PERL_ARGS_ASSERT_SV_SETREF_IV;
8614 sv_setiv(newSVrv(rv,classname), iv);
8619 =for apidoc sv_setref_uv
8621 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8622 argument will be upgraded to an RV. That RV will be modified to point to
8623 the new SV. The C<classname> argument indicates the package for the
8624 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8625 will have a reference count of 1, and the RV will be returned.
8631 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8633 PERL_ARGS_ASSERT_SV_SETREF_UV;
8635 sv_setuv(newSVrv(rv,classname), uv);
8640 =for apidoc sv_setref_nv
8642 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8643 argument will be upgraded to an RV. That RV will be modified to point to
8644 the new SV. The C<classname> argument indicates the package for the
8645 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8646 will have a reference count of 1, and the RV will be returned.
8652 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8654 PERL_ARGS_ASSERT_SV_SETREF_NV;
8656 sv_setnv(newSVrv(rv,classname), nv);
8661 =for apidoc sv_setref_pvn
8663 Copies a string into a new SV, optionally blessing the SV. The length of the
8664 string must be specified with C<n>. The C<rv> argument will be upgraded to
8665 an RV. That RV will be modified to point to the new SV. The C<classname>
8666 argument indicates the package for the blessing. Set C<classname> to
8667 C<NULL> to avoid the blessing. The new SV will have a reference count
8668 of 1, and the RV will be returned.
8670 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8676 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8677 const char *const pv, const STRLEN n)
8679 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8681 sv_setpvn(newSVrv(rv,classname), pv, n);
8686 =for apidoc sv_bless
8688 Blesses an SV into a specified package. The SV must be an RV. The package
8689 must be designated by its stash (see C<gv_stashpv()>). The reference count
8690 of the SV is unaffected.
8696 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8701 PERL_ARGS_ASSERT_SV_BLESS;
8704 Perl_croak(aTHX_ "Can't bless non-reference value");
8706 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8707 if (SvIsCOW(tmpRef))
8708 sv_force_normal_flags(tmpRef, 0);
8709 if (SvREADONLY(tmpRef))
8710 Perl_croak(aTHX_ "%s", PL_no_modify);
8711 if (SvOBJECT(tmpRef)) {
8712 if (SvTYPE(tmpRef) != SVt_PVIO)
8714 SvREFCNT_dec(SvSTASH(tmpRef));
8717 SvOBJECT_on(tmpRef);
8718 if (SvTYPE(tmpRef) != SVt_PVIO)
8720 SvUPGRADE(tmpRef, SVt_PVMG);
8721 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8726 (void)SvAMAGIC_off(sv);
8728 if(SvSMAGICAL(tmpRef))
8729 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8737 /* Downgrades a PVGV to a PVMG.
8741 S_sv_unglob(pTHX_ SV *const sv)
8746 SV * const temp = sv_newmortal();
8748 PERL_ARGS_ASSERT_SV_UNGLOB;
8750 assert(SvTYPE(sv) == SVt_PVGV);
8752 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8755 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8756 && HvNAME_get(stash))
8757 mro_method_changed_in(stash);
8758 gp_free(MUTABLE_GV(sv));
8761 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8765 if (GvNAME_HEK(sv)) {
8766 unshare_hek(GvNAME_HEK(sv));
8768 isGV_with_GP_off(sv);
8770 /* need to keep SvANY(sv) in the right arena */
8771 xpvmg = new_XPVMG();
8772 StructCopy(SvANY(sv), xpvmg, XPVMG);
8773 del_XPVGV(SvANY(sv));
8776 SvFLAGS(sv) &= ~SVTYPEMASK;
8777 SvFLAGS(sv) |= SVt_PVMG;
8779 /* Intentionally not calling any local SET magic, as this isn't so much a
8780 set operation as merely an internal storage change. */
8781 sv_setsv_flags(sv, temp, 0);
8785 =for apidoc sv_unref_flags
8787 Unsets the RV status of the SV, and decrements the reference count of
8788 whatever was being referenced by the RV. This can almost be thought of
8789 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8790 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8791 (otherwise the decrementing is conditional on the reference count being
8792 different from one or the reference being a readonly SV).
8799 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8801 SV* const target = SvRV(ref);
8803 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8805 if (SvWEAKREF(ref)) {
8806 sv_del_backref(target, ref);
8808 SvRV_set(ref, NULL);
8811 SvRV_set(ref, NULL);
8813 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8814 assigned to as BEGIN {$a = \"Foo"} will fail. */
8815 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8816 SvREFCNT_dec(target);
8817 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8818 sv_2mortal(target); /* Schedule for freeing later */
8822 =for apidoc sv_untaint
8824 Untaint an SV. Use C<SvTAINTED_off> instead.
8829 Perl_sv_untaint(pTHX_ SV *const sv)
8831 PERL_ARGS_ASSERT_SV_UNTAINT;
8833 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8834 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8841 =for apidoc sv_tainted
8843 Test an SV for taintedness. Use C<SvTAINTED> instead.
8848 Perl_sv_tainted(pTHX_ SV *const sv)
8850 PERL_ARGS_ASSERT_SV_TAINTED;
8852 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8853 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8854 if (mg && (mg->mg_len & 1) )
8861 =for apidoc sv_setpviv
8863 Copies an integer into the given SV, also updating its string value.
8864 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8870 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8872 char buf[TYPE_CHARS(UV)];
8874 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8876 PERL_ARGS_ASSERT_SV_SETPVIV;
8878 sv_setpvn(sv, ptr, ebuf - ptr);
8882 =for apidoc sv_setpviv_mg
8884 Like C<sv_setpviv>, but also handles 'set' magic.
8890 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8892 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8898 #if defined(PERL_IMPLICIT_CONTEXT)
8900 /* pTHX_ magic can't cope with varargs, so this is a no-context
8901 * version of the main function, (which may itself be aliased to us).
8902 * Don't access this version directly.
8906 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8911 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8913 va_start(args, pat);
8914 sv_vsetpvf(sv, pat, &args);
8918 /* pTHX_ magic can't cope with varargs, so this is a no-context
8919 * version of the main function, (which may itself be aliased to us).
8920 * Don't access this version directly.
8924 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8929 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8931 va_start(args, pat);
8932 sv_vsetpvf_mg(sv, pat, &args);
8938 =for apidoc sv_setpvf
8940 Works like C<sv_catpvf> but copies the text into the SV instead of
8941 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8947 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8951 PERL_ARGS_ASSERT_SV_SETPVF;
8953 va_start(args, pat);
8954 sv_vsetpvf(sv, pat, &args);
8959 =for apidoc sv_vsetpvf
8961 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8962 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8964 Usually used via its frontend C<sv_setpvf>.
8970 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8972 PERL_ARGS_ASSERT_SV_VSETPVF;
8974 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8978 =for apidoc sv_setpvf_mg
8980 Like C<sv_setpvf>, but also handles 'set' magic.
8986 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8990 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8992 va_start(args, pat);
8993 sv_vsetpvf_mg(sv, pat, &args);
8998 =for apidoc sv_vsetpvf_mg
9000 Like C<sv_vsetpvf>, but also handles 'set' magic.
9002 Usually used via its frontend C<sv_setpvf_mg>.
9008 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9010 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9012 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9016 #if defined(PERL_IMPLICIT_CONTEXT)
9018 /* pTHX_ magic can't cope with varargs, so this is a no-context
9019 * version of the main function, (which may itself be aliased to us).
9020 * Don't access this version directly.
9024 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9029 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9031 va_start(args, pat);
9032 sv_vcatpvf(sv, pat, &args);
9036 /* pTHX_ magic can't cope with varargs, so this is a no-context
9037 * version of the main function, (which may itself be aliased to us).
9038 * Don't access this version directly.
9042 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9047 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9049 va_start(args, pat);
9050 sv_vcatpvf_mg(sv, pat, &args);
9056 =for apidoc sv_catpvf
9058 Processes its arguments like C<sprintf> and appends the formatted
9059 output to an SV. If the appended data contains "wide" characters
9060 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9061 and characters >255 formatted with %c), the original SV might get
9062 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9063 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9064 valid UTF-8; if the original SV was bytes, the pattern should be too.
9069 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9073 PERL_ARGS_ASSERT_SV_CATPVF;
9075 va_start(args, pat);
9076 sv_vcatpvf(sv, pat, &args);
9081 =for apidoc sv_vcatpvf
9083 Processes its arguments like C<vsprintf> and appends the formatted output
9084 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9086 Usually used via its frontend C<sv_catpvf>.
9092 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9094 PERL_ARGS_ASSERT_SV_VCATPVF;
9096 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9100 =for apidoc sv_catpvf_mg
9102 Like C<sv_catpvf>, but also handles 'set' magic.
9108 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9112 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9114 va_start(args, pat);
9115 sv_vcatpvf_mg(sv, pat, &args);
9120 =for apidoc sv_vcatpvf_mg
9122 Like C<sv_vcatpvf>, but also handles 'set' magic.
9124 Usually used via its frontend C<sv_catpvf_mg>.
9130 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9132 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9134 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9139 =for apidoc sv_vsetpvfn
9141 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9144 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9150 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9151 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9153 PERL_ARGS_ASSERT_SV_VSETPVFN;
9156 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9160 S_expect_number(pTHX_ char **const pattern)
9165 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9167 switch (**pattern) {
9168 case '1': case '2': case '3':
9169 case '4': case '5': case '6':
9170 case '7': case '8': case '9':
9171 var = *(*pattern)++ - '0';
9172 while (isDIGIT(**pattern)) {
9173 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9175 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9183 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9185 const int neg = nv < 0;
9188 PERL_ARGS_ASSERT_F0CONVERT;
9196 if (uv & 1 && uv == nv)
9197 uv--; /* Round to even */
9199 const unsigned dig = uv % 10;
9212 =for apidoc sv_vcatpvfn
9214 Processes its arguments like C<vsprintf> and appends the formatted output
9215 to an SV. Uses an array of SVs if the C style variable argument list is
9216 missing (NULL). When running with taint checks enabled, indicates via
9217 C<maybe_tainted> if results are untrustworthy (often due to the use of
9220 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9226 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9227 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9228 vec_utf8 = DO_UTF8(vecsv);
9230 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9233 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9234 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9242 static const char nullstr[] = "(null)";
9244 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9245 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9247 /* Times 4: a decimal digit takes more than 3 binary digits.
9248 * NV_DIG: mantissa takes than many decimal digits.
9249 * Plus 32: Playing safe. */
9250 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9251 /* large enough for "%#.#f" --chip */
9252 /* what about long double NVs? --jhi */
9254 PERL_ARGS_ASSERT_SV_VCATPVFN;
9255 PERL_UNUSED_ARG(maybe_tainted);
9257 /* no matter what, this is a string now */
9258 (void)SvPV_force(sv, origlen);
9260 /* special-case "", "%s", and "%-p" (SVf - see below) */
9263 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9265 const char * const s = va_arg(*args, char*);
9266 sv_catpv(sv, s ? s : nullstr);
9268 else if (svix < svmax) {
9269 sv_catsv(sv, *svargs);
9273 if (args && patlen == 3 && pat[0] == '%' &&
9274 pat[1] == '-' && pat[2] == 'p') {
9275 argsv = MUTABLE_SV(va_arg(*args, void*));
9276 sv_catsv(sv, argsv);
9280 #ifndef USE_LONG_DOUBLE
9281 /* special-case "%.<number>[gf]" */
9282 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9283 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9284 unsigned digits = 0;
9288 while (*pp >= '0' && *pp <= '9')
9289 digits = 10 * digits + (*pp++ - '0');
9290 if (pp - pat == (int)patlen - 1) {
9298 /* Add check for digits != 0 because it seems that some
9299 gconverts are buggy in this case, and we don't yet have
9300 a Configure test for this. */
9301 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9302 /* 0, point, slack */
9303 Gconvert(nv, (int)digits, 0, ebuf);
9305 if (*ebuf) /* May return an empty string for digits==0 */
9308 } else if (!digits) {
9311 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9312 sv_catpvn(sv, p, l);
9318 #endif /* !USE_LONG_DOUBLE */
9320 if (!args && svix < svmax && DO_UTF8(*svargs))
9323 patend = (char*)pat + patlen;
9324 for (p = (char*)pat; p < patend; p = q) {
9327 bool vectorize = FALSE;
9328 bool vectorarg = FALSE;
9329 bool vec_utf8 = FALSE;
9335 bool has_precis = FALSE;
9337 const I32 osvix = svix;
9338 bool is_utf8 = FALSE; /* is this item utf8? */
9339 #ifdef HAS_LDBL_SPRINTF_BUG
9340 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9341 with sfio - Allen <allens@cpan.org> */
9342 bool fix_ldbl_sprintf_bug = FALSE;
9346 U8 utf8buf[UTF8_MAXBYTES+1];
9347 STRLEN esignlen = 0;
9349 const char *eptr = NULL;
9350 const char *fmtstart;
9353 const U8 *vecstr = NULL;
9360 /* we need a long double target in case HAS_LONG_DOUBLE but
9363 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9371 const char *dotstr = ".";
9372 STRLEN dotstrlen = 1;
9373 I32 efix = 0; /* explicit format parameter index */
9374 I32 ewix = 0; /* explicit width index */
9375 I32 epix = 0; /* explicit precision index */
9376 I32 evix = 0; /* explicit vector index */
9377 bool asterisk = FALSE;
9379 /* echo everything up to the next format specification */
9380 for (q = p; q < patend && *q != '%'; ++q) ;
9382 if (has_utf8 && !pat_utf8)
9383 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9385 sv_catpvn(sv, p, q - p);
9394 We allow format specification elements in this order:
9395 \d+\$ explicit format parameter index
9397 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9398 0 flag (as above): repeated to allow "v02"
9399 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9400 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9402 [%bcdefginopsuxDFOUX] format (mandatory)
9407 As of perl5.9.3, printf format checking is on by default.
9408 Internally, perl uses %p formats to provide an escape to
9409 some extended formatting. This block deals with those
9410 extensions: if it does not match, (char*)q is reset and
9411 the normal format processing code is used.
9413 Currently defined extensions are:
9414 %p include pointer address (standard)
9415 %-p (SVf) include an SV (previously %_)
9416 %-<num>p include an SV with precision <num>
9417 %<num>p reserved for future extensions
9419 Robin Barker 2005-07-14
9421 %1p (VDf) removed. RMB 2007-10-19
9428 n = expect_number(&q);
9435 argsv = MUTABLE_SV(va_arg(*args, void*));
9436 eptr = SvPV_const(argsv, elen);
9442 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9443 "internal %%<num>p might conflict with future printf extensions");
9449 if ( (width = expect_number(&q)) ) {
9464 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9493 if ( (ewix = expect_number(&q)) )
9502 if ((vectorarg = asterisk)) {
9515 width = expect_number(&q);
9521 vecsv = va_arg(*args, SV*);
9523 vecsv = (evix > 0 && evix <= svmax)
9524 ? svargs[evix-1] : &PL_sv_undef;
9526 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9528 dotstr = SvPV_const(vecsv, dotstrlen);
9529 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9530 bad with tied or overloaded values that return UTF8. */
9533 else if (has_utf8) {
9534 vecsv = sv_mortalcopy(vecsv);
9535 sv_utf8_upgrade(vecsv);
9536 dotstr = SvPV_const(vecsv, dotstrlen);
9543 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9544 vecsv = svargs[efix ? efix-1 : svix++];
9545 vecstr = (U8*)SvPV_const(vecsv,veclen);
9546 vec_utf8 = DO_UTF8(vecsv);
9548 /* if this is a version object, we need to convert
9549 * back into v-string notation and then let the
9550 * vectorize happen normally
9552 if (sv_derived_from(vecsv, "version")) {
9553 char *version = savesvpv(vecsv);
9554 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9555 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9556 "vector argument not supported with alpha versions");
9559 vecsv = sv_newmortal();
9560 scan_vstring(version, version + veclen, vecsv);
9561 vecstr = (U8*)SvPV_const(vecsv, veclen);
9562 vec_utf8 = DO_UTF8(vecsv);
9574 i = va_arg(*args, int);
9576 i = (ewix ? ewix <= svmax : svix < svmax) ?
9577 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9579 width = (i < 0) ? -i : i;
9589 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9591 /* XXX: todo, support specified precision parameter */
9595 i = va_arg(*args, int);
9597 i = (ewix ? ewix <= svmax : svix < svmax)
9598 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9600 has_precis = !(i < 0);
9605 precis = precis * 10 + (*q++ - '0');
9614 case 'I': /* Ix, I32x, and I64x */
9616 if (q[1] == '6' && q[2] == '4') {
9622 if (q[1] == '3' && q[2] == '2') {
9632 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9643 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9644 if (*(q + 1) == 'l') { /* lld, llf */
9670 if (!vectorize && !args) {
9672 const I32 i = efix-1;
9673 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9675 argsv = (svix >= 0 && svix < svmax)
9676 ? svargs[svix++] : &PL_sv_undef;
9687 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9689 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9691 eptr = (char*)utf8buf;
9692 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9706 eptr = va_arg(*args, char*);
9708 elen = strlen(eptr);
9710 eptr = (char *)nullstr;
9711 elen = sizeof nullstr - 1;
9715 eptr = SvPV_const(argsv, elen);
9716 if (DO_UTF8(argsv)) {
9717 STRLEN old_precis = precis;
9718 if (has_precis && precis < elen) {
9719 STRLEN ulen = sv_len_utf8(argsv);
9720 I32 p = precis > ulen ? ulen : precis;
9721 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9724 if (width) { /* fudge width (can't fudge elen) */
9725 if (has_precis && precis < elen)
9726 width += precis - old_precis;
9728 width += elen - sv_len_utf8(argsv);
9735 if (has_precis && precis < elen)
9742 if (alt || vectorize)
9744 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9765 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9774 esignbuf[esignlen++] = plus;
9778 case 'h': iv = (short)va_arg(*args, int); break;
9779 case 'l': iv = va_arg(*args, long); break;
9780 case 'V': iv = va_arg(*args, IV); break;
9781 default: iv = va_arg(*args, int); break;
9784 iv = va_arg(*args, Quad_t); break;
9791 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9793 case 'h': iv = (short)tiv; break;
9794 case 'l': iv = (long)tiv; break;
9796 default: iv = tiv; break;
9799 iv = (Quad_t)tiv; break;
9805 if ( !vectorize ) /* we already set uv above */
9810 esignbuf[esignlen++] = plus;
9814 esignbuf[esignlen++] = '-';
9858 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9869 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9870 case 'l': uv = va_arg(*args, unsigned long); break;
9871 case 'V': uv = va_arg(*args, UV); break;
9872 default: uv = va_arg(*args, unsigned); break;
9875 uv = va_arg(*args, Uquad_t); break;
9882 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9884 case 'h': uv = (unsigned short)tuv; break;
9885 case 'l': uv = (unsigned long)tuv; break;
9887 default: uv = tuv; break;
9890 uv = (Uquad_t)tuv; break;
9899 char *ptr = ebuf + sizeof ebuf;
9900 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9906 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9912 esignbuf[esignlen++] = '0';
9913 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9921 if (alt && *ptr != '0')
9930 esignbuf[esignlen++] = '0';
9931 esignbuf[esignlen++] = c;
9934 default: /* it had better be ten or less */
9938 } while (uv /= base);
9941 elen = (ebuf + sizeof ebuf) - ptr;
9945 zeros = precis - elen;
9946 else if (precis == 0 && elen == 1 && *eptr == '0'
9947 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9950 /* a precision nullifies the 0 flag. */
9957 /* FLOATING POINT */
9960 c = 'f'; /* maybe %F isn't supported here */
9968 /* This is evil, but floating point is even more evil */
9970 /* for SV-style calling, we can only get NV
9971 for C-style calling, we assume %f is double;
9972 for simplicity we allow any of %Lf, %llf, %qf for long double
9976 #if defined(USE_LONG_DOUBLE)
9980 /* [perl #20339] - we should accept and ignore %lf rather than die */
9984 #if defined(USE_LONG_DOUBLE)
9985 intsize = args ? 0 : 'q';
9989 #if defined(HAS_LONG_DOUBLE)
9998 /* now we need (long double) if intsize == 'q', else (double) */
10000 #if LONG_DOUBLESIZE > DOUBLESIZE
10002 va_arg(*args, long double) :
10003 va_arg(*args, double)
10005 va_arg(*args, double)
10010 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10011 else. frexp() has some unspecified behaviour for those three */
10012 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10014 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10015 will cast our (long double) to (double) */
10016 (void)Perl_frexp(nv, &i);
10017 if (i == PERL_INT_MIN)
10018 Perl_die(aTHX_ "panic: frexp");
10020 need = BIT_DIGITS(i);
10022 need += has_precis ? precis : 6; /* known default */
10027 #ifdef HAS_LDBL_SPRINTF_BUG
10028 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10029 with sfio - Allen <allens@cpan.org> */
10032 # define MY_DBL_MAX DBL_MAX
10033 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10034 # if DOUBLESIZE >= 8
10035 # define MY_DBL_MAX 1.7976931348623157E+308L
10037 # define MY_DBL_MAX 3.40282347E+38L
10041 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10042 # define MY_DBL_MAX_BUG 1L
10044 # define MY_DBL_MAX_BUG MY_DBL_MAX
10048 # define MY_DBL_MIN DBL_MIN
10049 # else /* XXX guessing! -Allen */
10050 # if DOUBLESIZE >= 8
10051 # define MY_DBL_MIN 2.2250738585072014E-308L
10053 # define MY_DBL_MIN 1.17549435E-38L
10057 if ((intsize == 'q') && (c == 'f') &&
10058 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10059 (need < DBL_DIG)) {
10060 /* it's going to be short enough that
10061 * long double precision is not needed */
10063 if ((nv <= 0L) && (nv >= -0L))
10064 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10066 /* would use Perl_fp_class as a double-check but not
10067 * functional on IRIX - see perl.h comments */
10069 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10070 /* It's within the range that a double can represent */
10071 #if defined(DBL_MAX) && !defined(DBL_MIN)
10072 if ((nv >= ((long double)1/DBL_MAX)) ||
10073 (nv <= (-(long double)1/DBL_MAX)))
10075 fix_ldbl_sprintf_bug = TRUE;
10078 if (fix_ldbl_sprintf_bug == TRUE) {
10088 # undef MY_DBL_MAX_BUG
10091 #endif /* HAS_LDBL_SPRINTF_BUG */
10093 need += 20; /* fudge factor */
10094 if (PL_efloatsize < need) {
10095 Safefree(PL_efloatbuf);
10096 PL_efloatsize = need + 20; /* more fudge */
10097 Newx(PL_efloatbuf, PL_efloatsize, char);
10098 PL_efloatbuf[0] = '\0';
10101 if ( !(width || left || plus || alt) && fill != '0'
10102 && has_precis && intsize != 'q' ) { /* Shortcuts */
10103 /* See earlier comment about buggy Gconvert when digits,
10105 if ( c == 'g' && precis) {
10106 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10107 /* May return an empty string for digits==0 */
10108 if (*PL_efloatbuf) {
10109 elen = strlen(PL_efloatbuf);
10110 goto float_converted;
10112 } else if ( c == 'f' && !precis) {
10113 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10118 char *ptr = ebuf + sizeof ebuf;
10121 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10122 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10123 if (intsize == 'q') {
10124 /* Copy the one or more characters in a long double
10125 * format before the 'base' ([efgEFG]) character to
10126 * the format string. */
10127 static char const prifldbl[] = PERL_PRIfldbl;
10128 char const *p = prifldbl + sizeof(prifldbl) - 3;
10129 while (p >= prifldbl) { *--ptr = *p--; }
10134 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10139 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10151 /* No taint. Otherwise we are in the strange situation
10152 * where printf() taints but print($float) doesn't.
10154 #if defined(HAS_LONG_DOUBLE)
10155 elen = ((intsize == 'q')
10156 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10157 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10159 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10163 eptr = PL_efloatbuf;
10171 i = SvCUR(sv) - origlen;
10174 case 'h': *(va_arg(*args, short*)) = i; break;
10175 default: *(va_arg(*args, int*)) = i; break;
10176 case 'l': *(va_arg(*args, long*)) = i; break;
10177 case 'V': *(va_arg(*args, IV*)) = i; break;
10180 *(va_arg(*args, Quad_t*)) = i; break;
10187 sv_setuv_mg(argsv, (UV)i);
10188 continue; /* not "break" */
10195 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10196 && ckWARN(WARN_PRINTF))
10198 SV * const msg = sv_newmortal();
10199 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10200 (PL_op->op_type == OP_PRTF) ? "" : "s");
10201 if (fmtstart < patend) {
10202 const char * const fmtend = q < patend ? q : patend;
10204 sv_catpvs(msg, "\"%");
10205 for (f = fmtstart; f < fmtend; f++) {
10207 sv_catpvn(msg, f, 1);
10209 Perl_sv_catpvf(aTHX_ msg,
10210 "\\%03"UVof, (UV)*f & 0xFF);
10213 sv_catpvs(msg, "\"");
10215 sv_catpvs(msg, "end of string");
10217 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10220 /* output mangled stuff ... */
10226 /* ... right here, because formatting flags should not apply */
10227 SvGROW(sv, SvCUR(sv) + elen + 1);
10229 Copy(eptr, p, elen, char);
10232 SvCUR_set(sv, p - SvPVX_const(sv));
10234 continue; /* not "break" */
10237 if (is_utf8 != has_utf8) {
10240 sv_utf8_upgrade(sv);
10243 const STRLEN old_elen = elen;
10244 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10245 sv_utf8_upgrade(nsv);
10246 eptr = SvPVX_const(nsv);
10249 if (width) { /* fudge width (can't fudge elen) */
10250 width += elen - old_elen;
10256 have = esignlen + zeros + elen;
10258 Perl_croak_nocontext("%s", PL_memory_wrap);
10260 need = (have > width ? have : width);
10263 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10264 Perl_croak_nocontext("%s", PL_memory_wrap);
10265 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10267 if (esignlen && fill == '0') {
10269 for (i = 0; i < (int)esignlen; i++)
10270 *p++ = esignbuf[i];
10272 if (gap && !left) {
10273 memset(p, fill, gap);
10276 if (esignlen && fill != '0') {
10278 for (i = 0; i < (int)esignlen; i++)
10279 *p++ = esignbuf[i];
10283 for (i = zeros; i; i--)
10287 Copy(eptr, p, elen, char);
10291 memset(p, ' ', gap);
10296 Copy(dotstr, p, dotstrlen, char);
10300 vectorize = FALSE; /* done iterating over vecstr */
10307 SvCUR_set(sv, p - SvPVX_const(sv));
10315 /* =========================================================================
10317 =head1 Cloning an interpreter
10319 All the macros and functions in this section are for the private use of
10320 the main function, perl_clone().
10322 The foo_dup() functions make an exact copy of an existing foo thingy.
10323 During the course of a cloning, a hash table is used to map old addresses
10324 to new addresses. The table is created and manipulated with the
10325 ptr_table_* functions.
10329 * =========================================================================*/
10332 #if defined(USE_ITHREADS)
10334 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10335 #ifndef GpREFCNT_inc
10336 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10340 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10341 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10342 If this changes, please unmerge ss_dup.
10343 Likewise, sv_dup_inc_multiple() relies on this fact. */
10344 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10345 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10346 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10347 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10348 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10349 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10350 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10351 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10352 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10353 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10354 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10355 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10356 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10357 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10359 /* clone a parser */
10362 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10366 PERL_ARGS_ASSERT_PARSER_DUP;
10371 /* look for it in the table first */
10372 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10376 /* create anew and remember what it is */
10377 Newxz(parser, 1, yy_parser);
10378 ptr_table_store(PL_ptr_table, proto, parser);
10380 parser->yyerrstatus = 0;
10381 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10383 /* XXX these not yet duped */
10384 parser->old_parser = NULL;
10385 parser->stack = NULL;
10387 parser->stack_size = 0;
10388 /* XXX parser->stack->state = 0; */
10390 /* XXX eventually, just Copy() most of the parser struct ? */
10392 parser->lex_brackets = proto->lex_brackets;
10393 parser->lex_casemods = proto->lex_casemods;
10394 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10395 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10396 parser->lex_casestack = savepvn(proto->lex_casestack,
10397 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10398 parser->lex_defer = proto->lex_defer;
10399 parser->lex_dojoin = proto->lex_dojoin;
10400 parser->lex_expect = proto->lex_expect;
10401 parser->lex_formbrack = proto->lex_formbrack;
10402 parser->lex_inpat = proto->lex_inpat;
10403 parser->lex_inwhat = proto->lex_inwhat;
10404 parser->lex_op = proto->lex_op;
10405 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10406 parser->lex_starts = proto->lex_starts;
10407 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10408 parser->multi_close = proto->multi_close;
10409 parser->multi_open = proto->multi_open;
10410 parser->multi_start = proto->multi_start;
10411 parser->multi_end = proto->multi_end;
10412 parser->pending_ident = proto->pending_ident;
10413 parser->preambled = proto->preambled;
10414 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10415 parser->linestr = sv_dup_inc(proto->linestr, param);
10416 parser->expect = proto->expect;
10417 parser->copline = proto->copline;
10418 parser->last_lop_op = proto->last_lop_op;
10419 parser->lex_state = proto->lex_state;
10420 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10421 /* rsfp_filters entries have fake IoDIRP() */
10422 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10423 parser->in_my = proto->in_my;
10424 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10425 parser->error_count = proto->error_count;
10428 parser->linestr = sv_dup_inc(proto->linestr, param);
10431 char * const ols = SvPVX(proto->linestr);
10432 char * const ls = SvPVX(parser->linestr);
10434 parser->bufptr = ls + (proto->bufptr >= ols ?
10435 proto->bufptr - ols : 0);
10436 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10437 proto->oldbufptr - ols : 0);
10438 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10439 proto->oldoldbufptr - ols : 0);
10440 parser->linestart = ls + (proto->linestart >= ols ?
10441 proto->linestart - ols : 0);
10442 parser->last_uni = ls + (proto->last_uni >= ols ?
10443 proto->last_uni - ols : 0);
10444 parser->last_lop = ls + (proto->last_lop >= ols ?
10445 proto->last_lop - ols : 0);
10447 parser->bufend = ls + SvCUR(parser->linestr);
10450 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10454 parser->endwhite = proto->endwhite;
10455 parser->faketokens = proto->faketokens;
10456 parser->lasttoke = proto->lasttoke;
10457 parser->nextwhite = proto->nextwhite;
10458 parser->realtokenstart = proto->realtokenstart;
10459 parser->skipwhite = proto->skipwhite;
10460 parser->thisclose = proto->thisclose;
10461 parser->thismad = proto->thismad;
10462 parser->thisopen = proto->thisopen;
10463 parser->thisstuff = proto->thisstuff;
10464 parser->thistoken = proto->thistoken;
10465 parser->thiswhite = proto->thiswhite;
10467 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10468 parser->curforce = proto->curforce;
10470 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10471 Copy(proto->nexttype, parser->nexttype, 5, I32);
10472 parser->nexttoke = proto->nexttoke;
10475 /* XXX should clone saved_curcop here, but we aren't passed
10476 * proto_perl; so do it in perl_clone_using instead */
10482 /* duplicate a file handle */
10485 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10489 PERL_ARGS_ASSERT_FP_DUP;
10490 PERL_UNUSED_ARG(type);
10493 return (PerlIO*)NULL;
10495 /* look for it in the table first */
10496 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10500 /* create anew and remember what it is */
10501 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10502 ptr_table_store(PL_ptr_table, fp, ret);
10506 /* duplicate a directory handle */
10509 Perl_dirp_dup(pTHX_ DIR *const dp)
10511 PERL_UNUSED_CONTEXT;
10518 /* duplicate a typeglob */
10521 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10525 PERL_ARGS_ASSERT_GP_DUP;
10529 /* look for it in the table first */
10530 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10534 /* create anew and remember what it is */
10536 ptr_table_store(PL_ptr_table, gp, ret);
10539 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10540 on Newxz() to do this for us. */
10541 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10542 ret->gp_io = io_dup_inc(gp->gp_io, param);
10543 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10544 ret->gp_av = av_dup_inc(gp->gp_av, param);
10545 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10546 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10547 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10548 ret->gp_cvgen = gp->gp_cvgen;
10549 ret->gp_line = gp->gp_line;
10550 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10554 /* duplicate a chain of magic */
10557 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10559 MAGIC *mgret = NULL;
10560 MAGIC **mgprev_p = &mgret;
10562 PERL_ARGS_ASSERT_MG_DUP;
10564 for (; mg; mg = mg->mg_moremagic) {
10566 Newx(nmg, 1, MAGIC);
10568 mgprev_p = &(nmg->mg_moremagic);
10570 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10571 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10572 from the original commit adding Perl_mg_dup() - revision 4538.
10573 Similarly there is the annotation "XXX random ptr?" next to the
10574 assignment to nmg->mg_ptr. */
10577 /* FIXME for plugins
10578 if (nmg->mg_type == PERL_MAGIC_qr) {
10579 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10583 if(nmg->mg_type == PERL_MAGIC_backref) {
10584 /* The backref AV has its reference count deliberately bumped by
10587 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10590 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10591 ? sv_dup_inc(nmg->mg_obj, param)
10592 : sv_dup(nmg->mg_obj, param);
10595 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10596 if (nmg->mg_len > 0) {
10597 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10598 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10599 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10601 AMT * const namtp = (AMT*)nmg->mg_ptr;
10602 sv_dup_inc_multiple((SV**)(namtp->table),
10603 (SV**)(namtp->table), NofAMmeth, param);
10606 else if (nmg->mg_len == HEf_SVKEY)
10607 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10609 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10610 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10616 #endif /* USE_ITHREADS */
10618 /* create a new pointer-mapping table */
10621 Perl_ptr_table_new(pTHX)
10624 PERL_UNUSED_CONTEXT;
10626 Newx(tbl, 1, PTR_TBL_t);
10627 tbl->tbl_max = 511;
10628 tbl->tbl_items = 0;
10629 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10633 #define PTR_TABLE_HASH(ptr) \
10634 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10637 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10638 following define) and at call to new_body_inline made below in
10639 Perl_ptr_table_store()
10642 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10644 /* map an existing pointer using a table */
10646 STATIC PTR_TBL_ENT_t *
10647 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10649 PTR_TBL_ENT_t *tblent;
10650 const UV hash = PTR_TABLE_HASH(sv);
10652 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10654 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10655 for (; tblent; tblent = tblent->next) {
10656 if (tblent->oldval == sv)
10663 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10665 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10667 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10668 PERL_UNUSED_CONTEXT;
10670 return tblent ? tblent->newval : NULL;
10673 /* add a new entry to a pointer-mapping table */
10676 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10678 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10680 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10681 PERL_UNUSED_CONTEXT;
10684 tblent->newval = newsv;
10686 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10688 new_body_inline(tblent, PTE_SVSLOT);
10690 tblent->oldval = oldsv;
10691 tblent->newval = newsv;
10692 tblent->next = tbl->tbl_ary[entry];
10693 tbl->tbl_ary[entry] = tblent;
10695 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10696 ptr_table_split(tbl);
10700 /* double the hash bucket size of an existing ptr table */
10703 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10705 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10706 const UV oldsize = tbl->tbl_max + 1;
10707 UV newsize = oldsize * 2;
10710 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10711 PERL_UNUSED_CONTEXT;
10713 Renew(ary, newsize, PTR_TBL_ENT_t*);
10714 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10715 tbl->tbl_max = --newsize;
10716 tbl->tbl_ary = ary;
10717 for (i=0; i < oldsize; i++, ary++) {
10718 PTR_TBL_ENT_t **curentp, **entp, *ent;
10721 curentp = ary + oldsize;
10722 for (entp = ary, ent = *ary; ent; ent = *entp) {
10723 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10725 ent->next = *curentp;
10735 /* remove all the entries from a ptr table */
10738 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10740 if (tbl && tbl->tbl_items) {
10741 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10742 UV riter = tbl->tbl_max;
10745 PTR_TBL_ENT_t *entry = array[riter];
10748 PTR_TBL_ENT_t * const oentry = entry;
10749 entry = entry->next;
10754 tbl->tbl_items = 0;
10758 /* clear and free a ptr table */
10761 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10766 ptr_table_clear(tbl);
10767 Safefree(tbl->tbl_ary);
10771 #if defined(USE_ITHREADS)
10774 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10776 PERL_ARGS_ASSERT_RVPV_DUP;
10779 SvRV_set(dstr, SvWEAKREF(sstr)
10780 ? sv_dup(SvRV_const(sstr), param)
10781 : sv_dup_inc(SvRV_const(sstr), param));
10784 else if (SvPVX_const(sstr)) {
10785 /* Has something there */
10787 /* Normal PV - clone whole allocated space */
10788 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10789 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10790 /* Not that normal - actually sstr is copy on write.
10791 But we are a true, independant SV, so: */
10792 SvREADONLY_off(dstr);
10797 /* Special case - not normally malloced for some reason */
10798 if (isGV_with_GP(sstr)) {
10799 /* Don't need to do anything here. */
10801 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10802 /* A "shared" PV - clone it as "shared" PV */
10804 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10808 /* Some other special case - random pointer */
10809 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10814 /* Copy the NULL */
10815 SvPV_set(dstr, NULL);
10819 /* duplicate a list of SVs. source and dest may point to the same memory. */
10821 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10822 SSize_t items, CLONE_PARAMS *const param)
10824 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10826 while (items-- > 0) {
10827 *dest++ = sv_dup_inc(*source++, param);
10833 /* duplicate an SV of any type (including AV, HV etc) */
10836 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10841 PERL_ARGS_ASSERT_SV_DUP;
10845 if (SvTYPE(sstr) == SVTYPEMASK) {
10846 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10851 /* look for it in the table first */
10852 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10856 if(param->flags & CLONEf_JOIN_IN) {
10857 /** We are joining here so we don't want do clone
10858 something that is bad **/
10859 if (SvTYPE(sstr) == SVt_PVHV) {
10860 const HEK * const hvname = HvNAME_HEK(sstr);
10862 /** don't clone stashes if they already exist **/
10863 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10867 /* create anew and remember what it is */
10870 #ifdef DEBUG_LEAKING_SCALARS
10871 dstr->sv_debug_optype = sstr->sv_debug_optype;
10872 dstr->sv_debug_line = sstr->sv_debug_line;
10873 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10874 dstr->sv_debug_cloned = 1;
10875 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10878 ptr_table_store(PL_ptr_table, sstr, dstr);
10881 SvFLAGS(dstr) = SvFLAGS(sstr);
10882 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10883 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10886 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10887 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10888 (void*)PL_watch_pvx, SvPVX_const(sstr));
10891 /* don't clone objects whose class has asked us not to */
10892 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10897 switch (SvTYPE(sstr)) {
10899 SvANY(dstr) = NULL;
10902 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10904 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10906 SvIV_set(dstr, SvIVX(sstr));
10910 SvANY(dstr) = new_XNV();
10911 SvNV_set(dstr, SvNVX(sstr));
10913 /* case SVt_BIND: */
10916 /* These are all the types that need complex bodies allocating. */
10918 const svtype sv_type = SvTYPE(sstr);
10919 const struct body_details *const sv_type_details
10920 = bodies_by_type + sv_type;
10924 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10939 assert(sv_type_details->body_size);
10940 if (sv_type_details->arena) {
10941 new_body_inline(new_body, sv_type);
10943 = (void*)((char*)new_body - sv_type_details->offset);
10945 new_body = new_NOARENA(sv_type_details);
10949 SvANY(dstr) = new_body;
10952 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10953 ((char*)SvANY(dstr)) + sv_type_details->offset,
10954 sv_type_details->copy, char);
10956 Copy(((char*)SvANY(sstr)),
10957 ((char*)SvANY(dstr)),
10958 sv_type_details->body_size + sv_type_details->offset, char);
10961 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10962 && !isGV_with_GP(dstr))
10963 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10965 /* The Copy above means that all the source (unduplicated) pointers
10966 are now in the destination. We can check the flags and the
10967 pointers in either, but it's possible that there's less cache
10968 missing by always going for the destination.
10969 FIXME - instrument and check that assumption */
10970 if (sv_type >= SVt_PVMG) {
10971 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10972 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10973 } else if (SvMAGIC(dstr))
10974 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10976 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10979 /* The cast silences a GCC warning about unhandled types. */
10980 switch ((int)sv_type) {
10990 /* FIXME for plugins */
10991 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10994 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10995 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10996 LvTARG(dstr) = dstr;
10997 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10998 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11000 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11002 if(isGV_with_GP(sstr)) {
11003 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11004 /* Don't call sv_add_backref here as it's going to be
11005 created as part of the magic cloning of the symbol
11007 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11008 at the point of this comment. */
11009 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11010 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11011 (void)GpREFCNT_inc(GvGP(dstr));
11013 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11016 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11017 if (IoOFP(dstr) == IoIFP(sstr))
11018 IoOFP(dstr) = IoIFP(dstr);
11020 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11021 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11022 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11023 /* I have no idea why fake dirp (rsfps)
11024 should be treated differently but otherwise
11025 we end up with leaks -- sky*/
11026 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11027 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11028 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11030 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11031 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11032 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11033 if (IoDIRP(dstr)) {
11034 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11037 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11040 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11041 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11042 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11045 /* avoid cloning an empty array */
11046 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11047 SV **dst_ary, **src_ary;
11048 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11050 src_ary = AvARRAY((const AV *)sstr);
11051 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11052 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11053 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11054 AvALLOC((const AV *)dstr) = dst_ary;
11055 if (AvREAL((const AV *)sstr)) {
11056 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11060 while (items-- > 0)
11061 *dst_ary++ = sv_dup(*src_ary++, param);
11063 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11064 while (items-- > 0) {
11065 *dst_ary++ = &PL_sv_undef;
11069 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11070 AvALLOC((const AV *)dstr) = (SV**)NULL;
11071 AvMAX( (const AV *)dstr) = -1;
11072 AvFILLp((const AV *)dstr) = -1;
11076 if (HvARRAY((const HV *)sstr)) {
11078 const bool sharekeys = !!HvSHAREKEYS(sstr);
11079 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11080 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11082 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11083 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11085 HvARRAY(dstr) = (HE**)darray;
11086 while (i <= sxhv->xhv_max) {
11087 const HE * const source = HvARRAY(sstr)[i];
11088 HvARRAY(dstr)[i] = source
11089 ? he_dup(source, sharekeys, param) : 0;
11094 const struct xpvhv_aux * const saux = HvAUX(sstr);
11095 struct xpvhv_aux * const daux = HvAUX(dstr);
11096 /* This flag isn't copied. */
11097 /* SvOOK_on(hv) attacks the IV flags. */
11098 SvFLAGS(dstr) |= SVf_OOK;
11100 hvname = saux->xhv_name;
11101 daux->xhv_name = hek_dup(hvname, param);
11103 daux->xhv_riter = saux->xhv_riter;
11104 daux->xhv_eiter = saux->xhv_eiter
11105 ? he_dup(saux->xhv_eiter,
11106 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11107 /* backref array needs refcnt=2; see sv_add_backref */
11108 daux->xhv_backreferences =
11109 saux->xhv_backreferences
11110 ? MUTABLE_AV(SvREFCNT_inc(
11111 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11114 daux->xhv_mro_meta = saux->xhv_mro_meta
11115 ? mro_meta_dup(saux->xhv_mro_meta, param)
11118 /* Record stashes for possible cloning in Perl_clone(). */
11120 av_push(param->stashes, dstr);
11124 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11127 if (!(param->flags & CLONEf_COPY_STACKS)) {
11131 /* NOTE: not refcounted */
11132 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11134 if (!CvISXSUB(dstr))
11135 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11137 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11138 CvXSUBANY(dstr).any_ptr =
11139 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11141 /* don't dup if copying back - CvGV isn't refcounted, so the
11142 * duped GV may never be freed. A bit of a hack! DAPM */
11143 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11144 NULL : gv_dup(CvGV(dstr), param) ;
11145 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11147 CvWEAKOUTSIDE(sstr)
11148 ? cv_dup( CvOUTSIDE(dstr), param)
11149 : cv_dup_inc(CvOUTSIDE(dstr), param);
11150 if (!CvISXSUB(dstr))
11151 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11157 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11163 /* duplicate a context */
11166 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11168 PERL_CONTEXT *ncxs;
11170 PERL_ARGS_ASSERT_CX_DUP;
11173 return (PERL_CONTEXT*)NULL;
11175 /* look for it in the table first */
11176 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11180 /* create anew and remember what it is */
11181 Newx(ncxs, max + 1, PERL_CONTEXT);
11182 ptr_table_store(PL_ptr_table, cxs, ncxs);
11183 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11186 PERL_CONTEXT * const ncx = &ncxs[ix];
11187 if (CxTYPE(ncx) == CXt_SUBST) {
11188 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11191 switch (CxTYPE(ncx)) {
11193 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11194 ? cv_dup_inc(ncx->blk_sub.cv, param)
11195 : cv_dup(ncx->blk_sub.cv,param));
11196 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11197 ? av_dup_inc(ncx->blk_sub.argarray,
11200 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11202 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11203 ncx->blk_sub.oldcomppad);
11206 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11208 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11210 case CXt_LOOP_LAZYSV:
11211 ncx->blk_loop.state_u.lazysv.end
11212 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11213 /* We are taking advantage of av_dup_inc and sv_dup_inc
11214 actually being the same function, and order equivalance of
11216 We can assert the later [but only at run time :-(] */
11217 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11218 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11220 ncx->blk_loop.state_u.ary.ary
11221 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11222 case CXt_LOOP_LAZYIV:
11223 case CXt_LOOP_PLAIN:
11224 if (CxPADLOOP(ncx)) {
11225 ncx->blk_loop.oldcomppad
11226 = (PAD*)ptr_table_fetch(PL_ptr_table,
11227 ncx->blk_loop.oldcomppad);
11229 ncx->blk_loop.oldcomppad
11230 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11235 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11236 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11237 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11250 /* duplicate a stack info structure */
11253 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11257 PERL_ARGS_ASSERT_SI_DUP;
11260 return (PERL_SI*)NULL;
11262 /* look for it in the table first */
11263 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11267 /* create anew and remember what it is */
11268 Newxz(nsi, 1, PERL_SI);
11269 ptr_table_store(PL_ptr_table, si, nsi);
11271 nsi->si_stack = av_dup_inc(si->si_stack, param);
11272 nsi->si_cxix = si->si_cxix;
11273 nsi->si_cxmax = si->si_cxmax;
11274 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11275 nsi->si_type = si->si_type;
11276 nsi->si_prev = si_dup(si->si_prev, param);
11277 nsi->si_next = si_dup(si->si_next, param);
11278 nsi->si_markoff = si->si_markoff;
11283 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11284 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11285 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11286 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11287 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11288 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11289 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11290 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11291 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11292 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11293 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11294 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11295 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11296 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11299 #define pv_dup_inc(p) SAVEPV(p)
11300 #define pv_dup(p) SAVEPV(p)
11301 #define svp_dup_inc(p,pp) any_dup(p,pp)
11303 /* map any object to the new equivent - either something in the
11304 * ptr table, or something in the interpreter structure
11308 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11312 PERL_ARGS_ASSERT_ANY_DUP;
11315 return (void*)NULL;
11317 /* look for it in the table first */
11318 ret = ptr_table_fetch(PL_ptr_table, v);
11322 /* see if it is part of the interpreter structure */
11323 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11324 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11332 /* duplicate the save stack */
11335 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11338 ANY * const ss = proto_perl->Isavestack;
11339 const I32 max = proto_perl->Isavestack_max;
11340 I32 ix = proto_perl->Isavestack_ix;
11353 void (*dptr) (void*);
11354 void (*dxptr) (pTHX_ void*);
11356 PERL_ARGS_ASSERT_SS_DUP;
11358 Newxz(nss, max, ANY);
11361 const I32 type = POPINT(ss,ix);
11362 TOPINT(nss,ix) = type;
11364 case SAVEt_HELEM: /* hash element */
11365 sv = (const SV *)POPPTR(ss,ix);
11366 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11368 case SAVEt_ITEM: /* normal string */
11369 case SAVEt_SV: /* scalar reference */
11370 sv = (const SV *)POPPTR(ss,ix);
11371 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11374 case SAVEt_MORTALIZESV:
11375 sv = (const SV *)POPPTR(ss,ix);
11376 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11378 case SAVEt_SHARED_PVREF: /* char* in shared space */
11379 c = (char*)POPPTR(ss,ix);
11380 TOPPTR(nss,ix) = savesharedpv(c);
11381 ptr = POPPTR(ss,ix);
11382 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11384 case SAVEt_GENERIC_SVREF: /* generic sv */
11385 case SAVEt_SVREF: /* scalar reference */
11386 sv = (const SV *)POPPTR(ss,ix);
11387 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11388 ptr = POPPTR(ss,ix);
11389 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11391 case SAVEt_HV: /* hash reference */
11392 case SAVEt_AV: /* array reference */
11393 sv = (const SV *) POPPTR(ss,ix);
11394 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11396 case SAVEt_COMPPAD:
11398 sv = (const SV *) POPPTR(ss,ix);
11399 TOPPTR(nss,ix) = sv_dup(sv, param);
11401 case SAVEt_INT: /* int reference */
11402 ptr = POPPTR(ss,ix);
11403 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11404 intval = (int)POPINT(ss,ix);
11405 TOPINT(nss,ix) = intval;
11407 case SAVEt_LONG: /* long reference */
11408 ptr = POPPTR(ss,ix);
11409 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11411 case SAVEt_CLEARSV:
11412 longval = (long)POPLONG(ss,ix);
11413 TOPLONG(nss,ix) = longval;
11415 case SAVEt_I32: /* I32 reference */
11416 case SAVEt_I16: /* I16 reference */
11417 case SAVEt_I8: /* I8 reference */
11418 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11419 ptr = POPPTR(ss,ix);
11420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11422 TOPINT(nss,ix) = i;
11424 case SAVEt_IV: /* IV reference */
11425 ptr = POPPTR(ss,ix);
11426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11428 TOPIV(nss,ix) = iv;
11430 case SAVEt_HPTR: /* HV* reference */
11431 case SAVEt_APTR: /* AV* reference */
11432 case SAVEt_SPTR: /* SV* reference */
11433 ptr = POPPTR(ss,ix);
11434 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11435 sv = (const SV *)POPPTR(ss,ix);
11436 TOPPTR(nss,ix) = sv_dup(sv, param);
11438 case SAVEt_VPTR: /* random* reference */
11439 ptr = POPPTR(ss,ix);
11440 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11441 ptr = POPPTR(ss,ix);
11442 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11444 case SAVEt_GENERIC_PVREF: /* generic char* */
11445 case SAVEt_PPTR: /* char* reference */
11446 ptr = POPPTR(ss,ix);
11447 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11448 c = (char*)POPPTR(ss,ix);
11449 TOPPTR(nss,ix) = pv_dup(c);
11451 case SAVEt_GP: /* scalar reference */
11452 gp = (GP*)POPPTR(ss,ix);
11453 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11454 (void)GpREFCNT_inc(gp);
11455 gv = (const GV *)POPPTR(ss,ix);
11456 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11459 ptr = POPPTR(ss,ix);
11460 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11461 /* these are assumed to be refcounted properly */
11463 switch (((OP*)ptr)->op_type) {
11465 case OP_LEAVESUBLV:
11469 case OP_LEAVEWRITE:
11470 TOPPTR(nss,ix) = ptr;
11473 (void) OpREFCNT_inc(o);
11477 TOPPTR(nss,ix) = NULL;
11482 TOPPTR(nss,ix) = NULL;
11485 hv = (const HV *)POPPTR(ss,ix);
11486 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11488 TOPINT(nss,ix) = i;
11491 c = (char*)POPPTR(ss,ix);
11492 TOPPTR(nss,ix) = pv_dup_inc(c);
11494 case SAVEt_STACK_POS: /* Position on Perl stack */
11496 TOPINT(nss,ix) = i;
11498 case SAVEt_DESTRUCTOR:
11499 ptr = POPPTR(ss,ix);
11500 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11501 dptr = POPDPTR(ss,ix);
11502 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11503 any_dup(FPTR2DPTR(void *, dptr),
11506 case SAVEt_DESTRUCTOR_X:
11507 ptr = POPPTR(ss,ix);
11508 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11509 dxptr = POPDXPTR(ss,ix);
11510 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11511 any_dup(FPTR2DPTR(void *, dxptr),
11514 case SAVEt_REGCONTEXT:
11517 TOPINT(nss,ix) = i;
11520 case SAVEt_AELEM: /* array element */
11521 sv = (const SV *)POPPTR(ss,ix);
11522 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11524 TOPINT(nss,ix) = i;
11525 av = (const AV *)POPPTR(ss,ix);
11526 TOPPTR(nss,ix) = av_dup_inc(av, param);
11529 ptr = POPPTR(ss,ix);
11530 TOPPTR(nss,ix) = ptr;
11533 ptr = POPPTR(ss,ix);
11536 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11537 HINTS_REFCNT_UNLOCK;
11539 TOPPTR(nss,ix) = ptr;
11541 TOPINT(nss,ix) = i;
11542 if (i & HINT_LOCALIZE_HH) {
11543 hv = (const HV *)POPPTR(ss,ix);
11544 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11547 case SAVEt_PADSV_AND_MORTALIZE:
11548 longval = (long)POPLONG(ss,ix);
11549 TOPLONG(nss,ix) = longval;
11550 ptr = POPPTR(ss,ix);
11551 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11552 sv = (const SV *)POPPTR(ss,ix);
11553 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11556 ptr = POPPTR(ss,ix);
11557 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11558 longval = (long)POPBOOL(ss,ix);
11559 TOPBOOL(nss,ix) = (bool)longval;
11561 case SAVEt_SET_SVFLAGS:
11563 TOPINT(nss,ix) = i;
11565 TOPINT(nss,ix) = i;
11566 sv = (const SV *)POPPTR(ss,ix);
11567 TOPPTR(nss,ix) = sv_dup(sv, param);
11569 case SAVEt_RE_STATE:
11571 const struct re_save_state *const old_state
11572 = (struct re_save_state *)
11573 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11574 struct re_save_state *const new_state
11575 = (struct re_save_state *)
11576 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11578 Copy(old_state, new_state, 1, struct re_save_state);
11579 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11581 new_state->re_state_bostr
11582 = pv_dup(old_state->re_state_bostr);
11583 new_state->re_state_reginput
11584 = pv_dup(old_state->re_state_reginput);
11585 new_state->re_state_regeol
11586 = pv_dup(old_state->re_state_regeol);
11587 new_state->re_state_regoffs
11588 = (regexp_paren_pair*)
11589 any_dup(old_state->re_state_regoffs, proto_perl);
11590 new_state->re_state_reglastparen
11591 = (U32*) any_dup(old_state->re_state_reglastparen,
11593 new_state->re_state_reglastcloseparen
11594 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11596 /* XXX This just has to be broken. The old save_re_context
11597 code did SAVEGENERICPV(PL_reg_start_tmp);
11598 PL_reg_start_tmp is char **.
11599 Look above to what the dup code does for
11600 SAVEt_GENERIC_PVREF
11601 It can never have worked.
11602 So this is merely a faithful copy of the exiting bug: */
11603 new_state->re_state_reg_start_tmp
11604 = (char **) pv_dup((char *)
11605 old_state->re_state_reg_start_tmp);
11606 /* I assume that it only ever "worked" because no-one called
11607 (pseudo)fork while the regexp engine had re-entered itself.
11609 #ifdef PERL_OLD_COPY_ON_WRITE
11610 new_state->re_state_nrs
11611 = sv_dup(old_state->re_state_nrs, param);
11613 new_state->re_state_reg_magic
11614 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11616 new_state->re_state_reg_oldcurpm
11617 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11619 new_state->re_state_reg_curpm
11620 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11622 new_state->re_state_reg_oldsaved
11623 = pv_dup(old_state->re_state_reg_oldsaved);
11624 new_state->re_state_reg_poscache
11625 = pv_dup(old_state->re_state_reg_poscache);
11626 new_state->re_state_reg_starttry
11627 = pv_dup(old_state->re_state_reg_starttry);
11630 case SAVEt_COMPILE_WARNINGS:
11631 ptr = POPPTR(ss,ix);
11632 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11635 ptr = POPPTR(ss,ix);
11636 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11640 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11648 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11649 * flag to the result. This is done for each stash before cloning starts,
11650 * so we know which stashes want their objects cloned */
11653 do_mark_cloneable_stash(pTHX_ SV *const sv)
11655 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11657 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11658 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11659 if (cloner && GvCV(cloner)) {
11666 mXPUSHs(newSVhek(hvname));
11668 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11675 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11683 =for apidoc perl_clone
11685 Create and return a new interpreter by cloning the current one.
11687 perl_clone takes these flags as parameters:
11689 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11690 without it we only clone the data and zero the stacks,
11691 with it we copy the stacks and the new perl interpreter is
11692 ready to run at the exact same point as the previous one.
11693 The pseudo-fork code uses COPY_STACKS while the
11694 threads->create doesn't.
11696 CLONEf_KEEP_PTR_TABLE
11697 perl_clone keeps a ptr_table with the pointer of the old
11698 variable as a key and the new variable as a value,
11699 this allows it to check if something has been cloned and not
11700 clone it again but rather just use the value and increase the
11701 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11702 the ptr_table using the function
11703 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11704 reason to keep it around is if you want to dup some of your own
11705 variable who are outside the graph perl scans, example of this
11706 code is in threads.xs create
11709 This is a win32 thing, it is ignored on unix, it tells perls
11710 win32host code (which is c++) to clone itself, this is needed on
11711 win32 if you want to run two threads at the same time,
11712 if you just want to do some stuff in a separate perl interpreter
11713 and then throw it away and return to the original one,
11714 you don't need to do anything.
11719 /* XXX the above needs expanding by someone who actually understands it ! */
11720 EXTERN_C PerlInterpreter *
11721 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11724 perl_clone(PerlInterpreter *proto_perl, UV flags)
11727 #ifdef PERL_IMPLICIT_SYS
11729 PERL_ARGS_ASSERT_PERL_CLONE;
11731 /* perlhost.h so we need to call into it
11732 to clone the host, CPerlHost should have a c interface, sky */
11734 if (flags & CLONEf_CLONE_HOST) {
11735 return perl_clone_host(proto_perl,flags);
11737 return perl_clone_using(proto_perl, flags,
11739 proto_perl->IMemShared,
11740 proto_perl->IMemParse,
11742 proto_perl->IStdIO,
11746 proto_perl->IProc);
11750 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11751 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11752 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11753 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11754 struct IPerlDir* ipD, struct IPerlSock* ipS,
11755 struct IPerlProc* ipP)
11757 /* XXX many of the string copies here can be optimized if they're
11758 * constants; they need to be allocated as common memory and just
11759 * their pointers copied. */
11762 CLONE_PARAMS clone_params;
11763 CLONE_PARAMS* const param = &clone_params;
11765 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11767 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11769 /* for each stash, determine whether its objects should be cloned */
11770 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11771 PERL_SET_THX(my_perl);
11774 PoisonNew(my_perl, 1, PerlInterpreter);
11780 PL_savestack_ix = 0;
11781 PL_savestack_max = -1;
11782 PL_sig_pending = 0;
11784 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11785 # else /* !DEBUGGING */
11786 Zero(my_perl, 1, PerlInterpreter);
11787 # endif /* DEBUGGING */
11789 /* host pointers */
11791 PL_MemShared = ipMS;
11792 PL_MemParse = ipMP;
11799 #else /* !PERL_IMPLICIT_SYS */
11801 CLONE_PARAMS clone_params;
11802 CLONE_PARAMS* param = &clone_params;
11803 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11805 PERL_ARGS_ASSERT_PERL_CLONE;
11807 /* for each stash, determine whether its objects should be cloned */
11808 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11809 PERL_SET_THX(my_perl);
11812 PoisonNew(my_perl, 1, PerlInterpreter);
11818 PL_savestack_ix = 0;
11819 PL_savestack_max = -1;
11820 PL_sig_pending = 0;
11822 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11823 # else /* !DEBUGGING */
11824 Zero(my_perl, 1, PerlInterpreter);
11825 # endif /* DEBUGGING */
11826 #endif /* PERL_IMPLICIT_SYS */
11827 param->flags = flags;
11828 param->proto_perl = proto_perl;
11830 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11832 PL_body_arenas = NULL;
11833 Zero(&PL_body_roots, 1, PL_body_roots);
11835 PL_nice_chunk = NULL;
11836 PL_nice_chunk_size = 0;
11838 PL_sv_objcount = 0;
11840 PL_sv_arenaroot = NULL;
11842 PL_debug = proto_perl->Idebug;
11844 PL_hash_seed = proto_perl->Ihash_seed;
11845 PL_rehash_seed = proto_perl->Irehash_seed;
11847 #ifdef USE_REENTRANT_API
11848 /* XXX: things like -Dm will segfault here in perlio, but doing
11849 * PERL_SET_CONTEXT(proto_perl);
11850 * breaks too many other things
11852 Perl_reentrant_init(aTHX);
11855 /* create SV map for pointer relocation */
11856 PL_ptr_table = ptr_table_new();
11858 /* initialize these special pointers as early as possible */
11859 SvANY(&PL_sv_undef) = NULL;
11860 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11861 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11862 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11864 SvANY(&PL_sv_no) = new_XPVNV();
11865 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11866 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11867 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11868 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11869 SvCUR_set(&PL_sv_no, 0);
11870 SvLEN_set(&PL_sv_no, 1);
11871 SvIV_set(&PL_sv_no, 0);
11872 SvNV_set(&PL_sv_no, 0);
11873 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11875 SvANY(&PL_sv_yes) = new_XPVNV();
11876 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11877 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11878 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11879 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11880 SvCUR_set(&PL_sv_yes, 1);
11881 SvLEN_set(&PL_sv_yes, 2);
11882 SvIV_set(&PL_sv_yes, 1);
11883 SvNV_set(&PL_sv_yes, 1);
11884 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11886 /* create (a non-shared!) shared string table */
11887 PL_strtab = newHV();
11888 HvSHAREKEYS_off(PL_strtab);
11889 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11890 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11892 PL_compiling = proto_perl->Icompiling;
11894 /* These two PVs will be free'd special way so must set them same way op.c does */
11895 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11896 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11898 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11899 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11901 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11902 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11903 if (PL_compiling.cop_hints_hash) {
11905 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11906 HINTS_REFCNT_UNLOCK;
11908 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11909 #ifdef PERL_DEBUG_READONLY_OPS
11914 /* pseudo environmental stuff */
11915 PL_origargc = proto_perl->Iorigargc;
11916 PL_origargv = proto_perl->Iorigargv;
11918 param->stashes = newAV(); /* Setup array of objects to call clone on */
11920 /* Set tainting stuff before PerlIO_debug can possibly get called */
11921 PL_tainting = proto_perl->Itainting;
11922 PL_taint_warn = proto_perl->Itaint_warn;
11924 #ifdef PERLIO_LAYERS
11925 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11926 PerlIO_clone(aTHX_ proto_perl, param);
11929 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11930 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11931 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11932 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11933 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11934 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11937 PL_minus_c = proto_perl->Iminus_c;
11938 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11939 PL_localpatches = proto_perl->Ilocalpatches;
11940 PL_splitstr = proto_perl->Isplitstr;
11941 PL_minus_n = proto_perl->Iminus_n;
11942 PL_minus_p = proto_perl->Iminus_p;
11943 PL_minus_l = proto_perl->Iminus_l;
11944 PL_minus_a = proto_perl->Iminus_a;
11945 PL_minus_E = proto_perl->Iminus_E;
11946 PL_minus_F = proto_perl->Iminus_F;
11947 PL_doswitches = proto_perl->Idoswitches;
11948 PL_dowarn = proto_perl->Idowarn;
11949 PL_doextract = proto_perl->Idoextract;
11950 PL_sawampersand = proto_perl->Isawampersand;
11951 PL_unsafe = proto_perl->Iunsafe;
11952 PL_inplace = SAVEPV(proto_perl->Iinplace);
11953 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11954 PL_perldb = proto_perl->Iperldb;
11955 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11956 PL_exit_flags = proto_perl->Iexit_flags;
11958 /* magical thingies */
11959 /* XXX time(&PL_basetime) when asked for? */
11960 PL_basetime = proto_perl->Ibasetime;
11961 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11963 PL_maxsysfd = proto_perl->Imaxsysfd;
11964 PL_statusvalue = proto_perl->Istatusvalue;
11966 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11968 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11970 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11972 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11973 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11974 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11977 /* RE engine related */
11978 Zero(&PL_reg_state, 1, struct re_save_state);
11979 PL_reginterp_cnt = 0;
11980 PL_regmatch_slab = NULL;
11982 /* Clone the regex array */
11983 /* ORANGE FIXME for plugins, probably in the SV dup code.
11984 newSViv(PTR2IV(CALLREGDUPE(
11985 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11987 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11988 PL_regex_pad = AvARRAY(PL_regex_padav);
11990 /* shortcuts to various I/O objects */
11991 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11992 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11993 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11994 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11995 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11996 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11997 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
11999 /* shortcuts to regexp stuff */
12000 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12002 /* shortcuts to misc objects */
12003 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12005 /* shortcuts to debugging objects */
12006 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12007 PL_DBline = gv_dup(proto_perl->IDBline, param);
12008 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12009 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12010 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12011 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12012 PL_dbargs = av_dup(proto_perl->Idbargs, param);
12014 /* symbol tables */
12015 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12016 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12017 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12018 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12019 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12021 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12022 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12023 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12024 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12025 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12026 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12027 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12028 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12030 PL_sub_generation = proto_perl->Isub_generation;
12031 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12033 /* funky return mechanisms */
12034 PL_forkprocess = proto_perl->Iforkprocess;
12036 /* subprocess state */
12037 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12039 /* internal state */
12040 PL_maxo = proto_perl->Imaxo;
12041 if (proto_perl->Iop_mask)
12042 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12045 /* PL_asserting = proto_perl->Iasserting; */
12047 /* current interpreter roots */
12048 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12050 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12052 PL_main_start = proto_perl->Imain_start;
12053 PL_eval_root = proto_perl->Ieval_root;
12054 PL_eval_start = proto_perl->Ieval_start;
12056 /* runtime control stuff */
12057 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12059 PL_filemode = proto_perl->Ifilemode;
12060 PL_lastfd = proto_perl->Ilastfd;
12061 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12064 PL_gensym = proto_perl->Igensym;
12065 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12066 PL_laststatval = proto_perl->Ilaststatval;
12067 PL_laststype = proto_perl->Ilaststype;
12070 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12072 /* interpreter atexit processing */
12073 PL_exitlistlen = proto_perl->Iexitlistlen;
12074 if (PL_exitlistlen) {
12075 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12076 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12079 PL_exitlist = (PerlExitListEntry*)NULL;
12081 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12082 if (PL_my_cxt_size) {
12083 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12084 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12085 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12086 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12087 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12091 PL_my_cxt_list = (void**)NULL;
12092 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12093 PL_my_cxt_keys = (const char**)NULL;
12096 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12097 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12098 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12100 PL_profiledata = NULL;
12102 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12104 PAD_CLONE_VARS(proto_perl, param);
12106 #ifdef HAVE_INTERP_INTERN
12107 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12110 /* more statics moved here */
12111 PL_generation = proto_perl->Igeneration;
12112 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12114 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12115 PL_in_clean_all = proto_perl->Iin_clean_all;
12117 PL_uid = proto_perl->Iuid;
12118 PL_euid = proto_perl->Ieuid;
12119 PL_gid = proto_perl->Igid;
12120 PL_egid = proto_perl->Iegid;
12121 PL_nomemok = proto_perl->Inomemok;
12122 PL_an = proto_perl->Ian;
12123 PL_evalseq = proto_perl->Ievalseq;
12124 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12125 PL_origalen = proto_perl->Iorigalen;
12126 #ifdef PERL_USES_PL_PIDSTATUS
12127 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12129 PL_osname = SAVEPV(proto_perl->Iosname);
12130 PL_sighandlerp = proto_perl->Isighandlerp;
12132 PL_runops = proto_perl->Irunops;
12134 PL_parser = parser_dup(proto_perl->Iparser, param);
12136 /* XXX this only works if the saved cop has already been cloned */
12137 if (proto_perl->Iparser) {
12138 PL_parser->saved_curcop = (COP*)any_dup(
12139 proto_perl->Iparser->saved_curcop,
12143 PL_subline = proto_perl->Isubline;
12144 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12147 PL_cryptseen = proto_perl->Icryptseen;
12150 PL_hints = proto_perl->Ihints;
12152 PL_amagic_generation = proto_perl->Iamagic_generation;
12154 #ifdef USE_LOCALE_COLLATE
12155 PL_collation_ix = proto_perl->Icollation_ix;
12156 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12157 PL_collation_standard = proto_perl->Icollation_standard;
12158 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12159 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12160 #endif /* USE_LOCALE_COLLATE */
12162 #ifdef USE_LOCALE_NUMERIC
12163 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12164 PL_numeric_standard = proto_perl->Inumeric_standard;
12165 PL_numeric_local = proto_perl->Inumeric_local;
12166 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12167 #endif /* !USE_LOCALE_NUMERIC */
12169 /* utf8 character classes */
12170 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12171 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12172 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12173 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12174 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12175 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12176 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12177 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12178 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12179 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12180 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12181 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12182 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12183 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12184 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12185 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12186 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12187 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12188 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12190 /* Did the locale setup indicate UTF-8? */
12191 PL_utf8locale = proto_perl->Iutf8locale;
12192 /* Unicode features (see perlrun/-C) */
12193 PL_unicode = proto_perl->Iunicode;
12195 /* Pre-5.8 signals control */
12196 PL_signals = proto_perl->Isignals;
12198 /* times() ticks per second */
12199 PL_clocktick = proto_perl->Iclocktick;
12201 /* Recursion stopper for PerlIO_find_layer */
12202 PL_in_load_module = proto_perl->Iin_load_module;
12204 /* sort() routine */
12205 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12207 /* Not really needed/useful since the reenrant_retint is "volatile",
12208 * but do it for consistency's sake. */
12209 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12211 /* Hooks to shared SVs and locks. */
12212 PL_sharehook = proto_perl->Isharehook;
12213 PL_lockhook = proto_perl->Ilockhook;
12214 PL_unlockhook = proto_perl->Iunlockhook;
12215 PL_threadhook = proto_perl->Ithreadhook;
12216 PL_destroyhook = proto_perl->Idestroyhook;
12218 #ifdef THREADS_HAVE_PIDS
12219 PL_ppid = proto_perl->Ippid;
12223 PL_last_swash_hv = NULL; /* reinits on demand */
12224 PL_last_swash_klen = 0;
12225 PL_last_swash_key[0]= '\0';
12226 PL_last_swash_tmps = (U8*)NULL;
12227 PL_last_swash_slen = 0;
12229 PL_glob_index = proto_perl->Iglob_index;
12230 PL_srand_called = proto_perl->Isrand_called;
12232 if (proto_perl->Ipsig_pend) {
12233 Newxz(PL_psig_pend, SIG_SIZE, int);
12236 PL_psig_pend = (int*)NULL;
12239 if (proto_perl->Ipsig_name) {
12240 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12241 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12243 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12246 PL_psig_ptr = (SV**)NULL;
12247 PL_psig_name = (SV**)NULL;
12250 /* intrpvar.h stuff */
12252 if (flags & CLONEf_COPY_STACKS) {
12253 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12254 PL_tmps_ix = proto_perl->Itmps_ix;
12255 PL_tmps_max = proto_perl->Itmps_max;
12256 PL_tmps_floor = proto_perl->Itmps_floor;
12257 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12258 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack, PL_tmps_ix,
12261 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12262 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12263 Newxz(PL_markstack, i, I32);
12264 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12265 - proto_perl->Imarkstack);
12266 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12267 - proto_perl->Imarkstack);
12268 Copy(proto_perl->Imarkstack, PL_markstack,
12269 PL_markstack_ptr - PL_markstack + 1, I32);
12271 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12272 * NOTE: unlike the others! */
12273 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12274 PL_scopestack_max = proto_perl->Iscopestack_max;
12275 Newxz(PL_scopestack, PL_scopestack_max, I32);
12276 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12278 /* NOTE: si_dup() looks at PL_markstack */
12279 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12281 /* PL_curstack = PL_curstackinfo->si_stack; */
12282 PL_curstack = av_dup(proto_perl->Icurstack, param);
12283 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12285 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12286 PL_stack_base = AvARRAY(PL_curstack);
12287 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12288 - proto_perl->Istack_base);
12289 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12291 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12292 * NOTE: unlike the others! */
12293 PL_savestack_ix = proto_perl->Isavestack_ix;
12294 PL_savestack_max = proto_perl->Isavestack_max;
12295 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12296 PL_savestack = ss_dup(proto_perl, param);
12300 ENTER; /* perl_destruct() wants to LEAVE; */
12302 /* although we're not duplicating the tmps stack, we should still
12303 * add entries for any SVs on the tmps stack that got cloned by a
12304 * non-refcount means (eg a temp in @_); otherwise they will be
12307 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12308 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12309 proto_perl->Itmps_stack[i]));
12310 if (nsv && !SvREFCNT(nsv)) {
12311 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12316 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12317 PL_top_env = &PL_start_env;
12319 PL_op = proto_perl->Iop;
12322 PL_Xpv = (XPV*)NULL;
12323 my_perl->Ina = proto_perl->Ina;
12325 PL_statbuf = proto_perl->Istatbuf;
12326 PL_statcache = proto_perl->Istatcache;
12327 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12328 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12330 PL_timesbuf = proto_perl->Itimesbuf;
12333 PL_tainted = proto_perl->Itainted;
12334 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12335 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12336 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12337 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12338 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12339 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12340 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12341 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12343 PL_restartop = proto_perl->Irestartop;
12344 PL_in_eval = proto_perl->Iin_eval;
12345 PL_delaymagic = proto_perl->Idelaymagic;
12346 PL_dirty = proto_perl->Idirty;
12347 PL_localizing = proto_perl->Ilocalizing;
12349 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12350 PL_hv_fetch_ent_mh = NULL;
12351 PL_modcount = proto_perl->Imodcount;
12352 PL_lastgotoprobe = NULL;
12353 PL_dumpindent = proto_perl->Idumpindent;
12355 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12356 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12357 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12358 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12359 PL_efloatbuf = NULL; /* reinits on demand */
12360 PL_efloatsize = 0; /* reinits on demand */
12364 PL_screamfirst = NULL;
12365 PL_screamnext = NULL;
12366 PL_maxscream = -1; /* reinits on demand */
12367 PL_lastscream = NULL;
12370 PL_regdummy = proto_perl->Iregdummy;
12371 PL_colorset = 0; /* reinits PL_colors[] */
12372 /*PL_colors[6] = {0,0,0,0,0,0};*/
12376 /* Pluggable optimizer */
12377 PL_peepp = proto_perl->Ipeepp;
12378 /* op_free() hook */
12379 PL_opfreehook = proto_perl->Iopfreehook;
12381 PL_stashcache = newHV();
12383 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12384 proto_perl->Iwatchaddr);
12385 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12386 if (PL_debug && PL_watchaddr) {
12387 PerlIO_printf(Perl_debug_log,
12388 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12389 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12390 PTR2UV(PL_watchok));
12393 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12395 /* Call the ->CLONE method, if it exists, for each of the stashes
12396 identified by sv_dup() above.
12398 while(av_len(param->stashes) != -1) {
12399 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12400 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12401 if (cloner && GvCV(cloner)) {
12406 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12408 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12414 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12415 ptr_table_free(PL_ptr_table);
12416 PL_ptr_table = NULL;
12420 SvREFCNT_dec(param->stashes);
12422 /* orphaned? eg threads->new inside BEGIN or use */
12423 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12424 SvREFCNT_inc_simple_void(PL_compcv);
12425 SAVEFREESV(PL_compcv);
12431 #endif /* USE_ITHREADS */
12434 =head1 Unicode Support
12436 =for apidoc sv_recode_to_utf8
12438 The encoding is assumed to be an Encode object, on entry the PV
12439 of the sv is assumed to be octets in that encoding, and the sv
12440 will be converted into Unicode (and UTF-8).
12442 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12443 is not a reference, nothing is done to the sv. If the encoding is not
12444 an C<Encode::XS> Encoding object, bad things will happen.
12445 (See F<lib/encoding.pm> and L<Encode>).
12447 The PV of the sv is returned.
12452 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12456 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12458 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12472 Passing sv_yes is wrong - it needs to be or'ed set of constants
12473 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12474 remove converted chars from source.
12476 Both will default the value - let them.
12478 XPUSHs(&PL_sv_yes);
12481 call_method("decode", G_SCALAR);
12485 s = SvPV_const(uni, len);
12486 if (s != SvPVX_const(sv)) {
12487 SvGROW(sv, len + 1);
12488 Move(s, SvPVX(sv), len + 1, char);
12489 SvCUR_set(sv, len);
12496 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12500 =for apidoc sv_cat_decode
12502 The encoding is assumed to be an Encode object, the PV of the ssv is
12503 assumed to be octets in that encoding and decoding the input starts
12504 from the position which (PV + *offset) pointed to. The dsv will be
12505 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12506 when the string tstr appears in decoding output or the input ends on
12507 the PV of the ssv. The value which the offset points will be modified
12508 to the last input position on the ssv.
12510 Returns TRUE if the terminator was found, else returns FALSE.
12515 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12516 SV *ssv, int *offset, char *tstr, int tlen)
12521 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12523 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12534 offsv = newSViv(*offset);
12536 mXPUSHp(tstr, tlen);
12538 call_method("cat_decode", G_SCALAR);
12540 ret = SvTRUE(TOPs);
12541 *offset = SvIV(offsv);
12547 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12552 /* ---------------------------------------------------------------------
12554 * support functions for report_uninit()
12557 /* the maxiumum size of array or hash where we will scan looking
12558 * for the undefined element that triggered the warning */
12560 #define FUV_MAX_SEARCH_SIZE 1000
12562 /* Look for an entry in the hash whose value has the same SV as val;
12563 * If so, return a mortal copy of the key. */
12566 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12569 register HE **array;
12572 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12574 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12575 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12578 array = HvARRAY(hv);
12580 for (i=HvMAX(hv); i>0; i--) {
12581 register HE *entry;
12582 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12583 if (HeVAL(entry) != val)
12585 if ( HeVAL(entry) == &PL_sv_undef ||
12586 HeVAL(entry) == &PL_sv_placeholder)
12590 if (HeKLEN(entry) == HEf_SVKEY)
12591 return sv_mortalcopy(HeKEY_sv(entry));
12592 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12598 /* Look for an entry in the array whose value has the same SV as val;
12599 * If so, return the index, otherwise return -1. */
12602 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12606 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12608 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12609 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12612 if (val != &PL_sv_undef) {
12613 SV ** const svp = AvARRAY(av);
12616 for (i=AvFILLp(av); i>=0; i--)
12623 /* S_varname(): return the name of a variable, optionally with a subscript.
12624 * If gv is non-zero, use the name of that global, along with gvtype (one
12625 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12626 * targ. Depending on the value of the subscript_type flag, return:
12629 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12630 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12631 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12632 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12635 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12636 const SV *const keyname, I32 aindex, int subscript_type)
12639 SV * const name = sv_newmortal();
12642 buffer[0] = gvtype;
12645 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12647 gv_fullname4(name, gv, buffer, 0);
12649 if ((unsigned int)SvPVX(name)[1] <= 26) {
12651 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12653 /* Swap the 1 unprintable control character for the 2 byte pretty
12654 version - ie substr($name, 1, 1) = $buffer; */
12655 sv_insert(name, 1, 1, buffer, 2);
12659 CV * const cv = find_runcv(NULL);
12663 if (!cv || !CvPADLIST(cv))
12665 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12666 sv = *av_fetch(av, targ, FALSE);
12667 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12670 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12671 SV * const sv = newSV(0);
12672 *SvPVX(name) = '$';
12673 Perl_sv_catpvf(aTHX_ name, "{%s}",
12674 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12677 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12678 *SvPVX(name) = '$';
12679 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12681 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12682 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12683 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12691 =for apidoc find_uninit_var
12693 Find the name of the undefined variable (if any) that caused the operator o
12694 to issue a "Use of uninitialized value" warning.
12695 If match is true, only return a name if it's value matches uninit_sv.
12696 So roughly speaking, if a unary operator (such as OP_COS) generates a
12697 warning, then following the direct child of the op may yield an
12698 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12699 other hand, with OP_ADD there are two branches to follow, so we only print
12700 the variable name if we get an exact match.
12702 The name is returned as a mortal SV.
12704 Assumes that PL_op is the op that originally triggered the error, and that
12705 PL_comppad/PL_curpad points to the currently executing pad.
12711 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12717 const OP *o, *o2, *kid;
12719 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12720 uninit_sv == &PL_sv_placeholder)))
12723 switch (obase->op_type) {
12730 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12731 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12734 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12736 if (pad) { /* @lex, %lex */
12737 sv = PAD_SVl(obase->op_targ);
12741 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12742 /* @global, %global */
12743 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12746 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12748 else /* @{expr}, %{expr} */
12749 return find_uninit_var(cUNOPx(obase)->op_first,
12753 /* attempt to find a match within the aggregate */
12755 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12757 subscript_type = FUV_SUBSCRIPT_HASH;
12760 index = find_array_subscript((const AV *)sv, uninit_sv);
12762 subscript_type = FUV_SUBSCRIPT_ARRAY;
12765 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12768 return varname(gv, hash ? '%' : '@', obase->op_targ,
12769 keysv, index, subscript_type);
12773 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12775 return varname(NULL, '$', obase->op_targ,
12776 NULL, 0, FUV_SUBSCRIPT_NONE);
12779 gv = cGVOPx_gv(obase);
12780 if (!gv || (match && GvSV(gv) != uninit_sv))
12782 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12785 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12788 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12789 if (!av || SvRMAGICAL(av))
12791 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12792 if (!svp || *svp != uninit_sv)
12795 return varname(NULL, '$', obase->op_targ,
12796 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12799 gv = cGVOPx_gv(obase);
12804 AV *const av = GvAV(gv);
12805 if (!av || SvRMAGICAL(av))
12807 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12808 if (!svp || *svp != uninit_sv)
12811 return varname(gv, '$', 0,
12812 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12817 o = cUNOPx(obase)->op_first;
12818 if (!o || o->op_type != OP_NULL ||
12819 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12821 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12825 if (PL_op == obase)
12826 /* $a[uninit_expr] or $h{uninit_expr} */
12827 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12830 o = cBINOPx(obase)->op_first;
12831 kid = cBINOPx(obase)->op_last;
12833 /* get the av or hv, and optionally the gv */
12835 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12836 sv = PAD_SV(o->op_targ);
12838 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12839 && cUNOPo->op_first->op_type == OP_GV)
12841 gv = cGVOPx_gv(cUNOPo->op_first);
12845 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12850 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12851 /* index is constant */
12855 if (obase->op_type == OP_HELEM) {
12856 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12857 if (!he || HeVAL(he) != uninit_sv)
12861 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12862 if (!svp || *svp != uninit_sv)
12866 if (obase->op_type == OP_HELEM)
12867 return varname(gv, '%', o->op_targ,
12868 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12870 return varname(gv, '@', o->op_targ, NULL,
12871 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12874 /* index is an expression;
12875 * attempt to find a match within the aggregate */
12876 if (obase->op_type == OP_HELEM) {
12877 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12879 return varname(gv, '%', o->op_targ,
12880 keysv, 0, FUV_SUBSCRIPT_HASH);
12884 = find_array_subscript((const AV *)sv, uninit_sv);
12886 return varname(gv, '@', o->op_targ,
12887 NULL, index, FUV_SUBSCRIPT_ARRAY);
12892 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12894 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12899 /* only examine RHS */
12900 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12903 o = cUNOPx(obase)->op_first;
12904 if (o->op_type == OP_PUSHMARK)
12907 if (!o->op_sibling) {
12908 /* one-arg version of open is highly magical */
12910 if (o->op_type == OP_GV) { /* open FOO; */
12912 if (match && GvSV(gv) != uninit_sv)
12914 return varname(gv, '$', 0,
12915 NULL, 0, FUV_SUBSCRIPT_NONE);
12917 /* other possibilities not handled are:
12918 * open $x; or open my $x; should return '${*$x}'
12919 * open expr; should return '$'.expr ideally
12925 /* ops where $_ may be an implicit arg */
12929 if ( !(obase->op_flags & OPf_STACKED)) {
12930 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12931 ? PAD_SVl(obase->op_targ)
12934 sv = sv_newmortal();
12935 sv_setpvs(sv, "$_");
12944 match = 1; /* print etc can return undef on defined args */
12945 /* skip filehandle as it can't produce 'undef' warning */
12946 o = cUNOPx(obase)->op_first;
12947 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12948 o = o->op_sibling->op_sibling;
12952 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12954 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12956 /* the following ops are capable of returning PL_sv_undef even for
12957 * defined arg(s) */
12976 case OP_GETPEERNAME:
13024 case OP_SMARTMATCH:
13033 /* XXX tmp hack: these two may call an XS sub, and currently
13034 XS subs don't have a SUB entry on the context stack, so CV and
13035 pad determination goes wrong, and BAD things happen. So, just
13036 don't try to determine the value under those circumstances.
13037 Need a better fix at dome point. DAPM 11/2007 */
13043 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13044 if (gv && GvSV(gv) == uninit_sv)
13045 return newSVpvs_flags("$.", SVs_TEMP);
13050 /* def-ness of rval pos() is independent of the def-ness of its arg */
13051 if ( !(obase->op_flags & OPf_MOD))
13056 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13057 return newSVpvs_flags("${$/}", SVs_TEMP);
13062 if (!(obase->op_flags & OPf_KIDS))
13064 o = cUNOPx(obase)->op_first;
13070 /* if all except one arg are constant, or have no side-effects,
13071 * or are optimized away, then it's unambiguous */
13073 for (kid=o; kid; kid = kid->op_sibling) {
13075 const OPCODE type = kid->op_type;
13076 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13077 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13078 || (type == OP_PUSHMARK)
13082 if (o2) { /* more than one found */
13089 return find_uninit_var(o2, uninit_sv, match);
13091 /* scan all args */
13093 sv = find_uninit_var(o, uninit_sv, 1);
13105 =for apidoc report_uninit
13107 Print appropriate "Use of uninitialized variable" warning
13113 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13117 SV* varname = NULL;
13119 varname = find_uninit_var(PL_op, uninit_sv,0);
13121 sv_insert(varname, 0, 0, " ", 1);
13123 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13124 varname ? SvPV_nolen_const(varname) : "",
13125 " in ", OP_DESC(PL_op));
13128 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13134 * c-indentation-style: bsd
13135 * c-basic-offset: 4
13136 * indent-tabs-mode: t
13139 * ex: set ts=8 sts=4 sw=4 noet:
https://perl5.git.perl.org / perl5.git / blob