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 svtype utype; /* bodytype stored in arena */
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 svtype bodytype)
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;
752 adesc->utype = bodytype;
753 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
754 curr, (void*)adesc->arena, (UV)arena_size));
760 /* return a thing to the free list */
762 #define del_body(thing, root) \
764 void ** const thing_copy = (void **)thing;\
765 *thing_copy = *root; \
766 *root = (void*)thing_copy; \
771 =head1 SV-Body Allocation
773 Allocation of SV-bodies is similar to SV-heads, differing as follows;
774 the allocation mechanism is used for many body types, so is somewhat
775 more complicated, it uses arena-sets, and has no need for still-live
778 At the outermost level, (new|del)_X*V macros return bodies of the
779 appropriate type. These macros call either (new|del)_body_type or
780 (new|del)_body_allocated macro pairs, depending on specifics of the
781 type. Most body types use the former pair, the latter pair is used to
782 allocate body types with "ghost fields".
784 "ghost fields" are fields that are unused in certain types, and
785 consequently don't need to actually exist. They are declared because
786 they're part of a "base type", which allows use of functions as
787 methods. The simplest examples are AVs and HVs, 2 aggregate types
788 which don't use the fields which support SCALAR semantics.
790 For these types, the arenas are carved up into appropriately sized
791 chunks, we thus avoid wasted memory for those unaccessed members.
792 When bodies are allocated, we adjust the pointer back in memory by the
793 size of the part not allocated, so it's as if we allocated the full
794 structure. (But things will all go boom if you write to the part that
795 is "not there", because you'll be overwriting the last members of the
796 preceding structure in memory.)
798 We calculate the correction using the STRUCT_OFFSET macro on the first
799 member present. If the allocated structure is smaller (no initial NV
800 actually allocated) then the net effect is to subtract the size of the NV
801 from the pointer, to return a new pointer as if an initial NV were actually
802 allocated. (We were using structures named *_allocated for this, but
803 this turned out to be a subtle bug, because a structure without an NV
804 could have a lower alignment constraint, but the compiler is allowed to
805 optimised accesses based on the alignment constraint of the actual pointer
806 to the full structure, for example, using a single 64 bit load instruction
807 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
809 This is the same trick as was used for NV and IV bodies. Ironically it
810 doesn't need to be used for NV bodies any more, because NV is now at
811 the start of the structure. IV bodies don't need it either, because
812 they are no longer allocated.
814 In turn, the new_body_* allocators call S_new_body(), which invokes
815 new_body_inline macro, which takes a lock, and takes a body off the
816 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
817 necessary to refresh an empty list. Then the lock is released, and
818 the body is returned.
820 S_more_bodies calls get_arena(), and carves it up into an array of N
821 bodies, which it strings into a linked list. It looks up arena-size
822 and body-size from the body_details table described below, thus
823 supporting the multiple body-types.
825 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
826 the (new|del)_X*V macros are mapped directly to malloc/free.
832 For each sv-type, struct body_details bodies_by_type[] carries
833 parameters which control these aspects of SV handling:
835 Arena_size determines whether arenas are used for this body type, and if
836 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
837 zero, forcing individual mallocs and frees.
839 Body_size determines how big a body is, and therefore how many fit into
840 each arena. Offset carries the body-pointer adjustment needed for
841 "ghost fields", and is used in *_allocated macros.
843 But its main purpose is to parameterize info needed in
844 Perl_sv_upgrade(). The info here dramatically simplifies the function
845 vs the implementation in 5.8.8, making it table-driven. All fields
846 are used for this, except for arena_size.
848 For the sv-types that have no bodies, arenas are not used, so those
849 PL_body_roots[sv_type] are unused, and can be overloaded. In
850 something of a special case, SVt_NULL is borrowed for HE arenas;
851 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
852 bodies_by_type[SVt_NULL] slot is not used, as the table is not
855 PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
856 they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
857 just use the same allocation semantics. At first, PTEs were also
858 overloaded to a non-body sv-type, but this yielded hard-to-find malloc
859 bugs, so was simplified by claiming a new slot. This choice has no
860 consequence at this time.
864 struct body_details {
865 U8 body_size; /* Size to allocate */
866 U8 copy; /* Size of structure to copy (may be shorter) */
868 unsigned int type : 4; /* We have space for a sanity check. */
869 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
870 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
871 unsigned int arena : 1; /* Allocated from an arena */
872 size_t arena_size; /* Size of arena to allocate */
880 /* With -DPURFIY we allocate everything directly, and don't use arenas.
881 This seems a rather elegant way to simplify some of the code below. */
882 #define HASARENA FALSE
884 #define HASARENA TRUE
886 #define NOARENA FALSE
888 /* Size the arenas to exactly fit a given number of bodies. A count
889 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
890 simplifying the default. If count > 0, the arena is sized to fit
891 only that many bodies, allowing arenas to be used for large, rare
892 bodies (XPVFM, XPVIO) without undue waste. The arena size is
893 limited by PERL_ARENA_SIZE, so we can safely oversize the
896 #define FIT_ARENA0(body_size) \
897 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
898 #define FIT_ARENAn(count,body_size) \
899 ( count * body_size <= PERL_ARENA_SIZE) \
900 ? count * body_size \
901 : FIT_ARENA0 (body_size)
902 #define FIT_ARENA(count,body_size) \
904 ? FIT_ARENAn (count, body_size) \
905 : FIT_ARENA0 (body_size)
907 /* Calculate the length to copy. Specifically work out the length less any
908 final padding the compiler needed to add. See the comment in sv_upgrade
909 for why copying the padding proved to be a bug. */
911 #define copy_length(type, last_member) \
912 STRUCT_OFFSET(type, last_member) \
913 + sizeof (((type*)SvANY((const SV *)0))->last_member)
915 static const struct body_details bodies_by_type[] = {
916 { sizeof(HE), 0, 0, SVt_NULL,
917 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
919 /* The bind placeholder pretends to be an RV for now.
920 Also it's marked as "can't upgrade" to stop anyone using it before it's
922 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
924 /* IVs are in the head, so the allocation size is 0.
925 However, the slot is overloaded for PTEs. */
926 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
927 sizeof(IV), /* This is used to copy out the IV body. */
928 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
929 NOARENA /* IVS don't need an arena */,
930 /* But PTEs need to know the size of their arena */
931 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
934 /* 8 bytes on most ILP32 with IEEE doubles */
935 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
936 FIT_ARENA(0, sizeof(NV)) },
938 /* 8 bytes on most ILP32 with IEEE doubles */
939 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
940 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
941 + STRUCT_OFFSET(XPV, xpv_cur),
942 SVt_PV, FALSE, NONV, HASARENA,
943 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
946 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
947 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
948 + STRUCT_OFFSET(XPVIV, xpv_cur),
949 SVt_PVIV, FALSE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
953 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
954 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
957 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
958 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
961 { sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
962 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
963 + STRUCT_OFFSET(regexp, xpv_cur),
964 SVt_REGEXP, FALSE, NONV, HASARENA,
965 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
969 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
970 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
973 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
974 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
976 { sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill),
977 copy_length(XPVAV, xmg_stash) - STRUCT_OFFSET(XPVAV, xav_fill),
978 + STRUCT_OFFSET(XPVAV, xav_fill),
979 SVt_PVAV, TRUE, NONV, HASARENA,
980 FIT_ARENA(0, sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill)) },
982 { sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill),
983 copy_length(XPVHV, xmg_stash) - STRUCT_OFFSET(XPVHV, xhv_fill),
984 + STRUCT_OFFSET(XPVHV, xhv_fill),
985 SVt_PVHV, TRUE, NONV, HASARENA,
986 FIT_ARENA(0, sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill)) },
989 { sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
990 sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
991 + STRUCT_OFFSET(XPVCV, xpv_cur),
992 SVt_PVCV, TRUE, NONV, HASARENA,
993 FIT_ARENA(0, sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur)) },
995 { sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
996 sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
997 + STRUCT_OFFSET(XPVFM, xpv_cur),
998 SVt_PVFM, TRUE, NONV, NOARENA,
999 FIT_ARENA(20, sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur)) },
1001 /* XPVIO is 84 bytes, fits 48x */
1002 { sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1003 sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1004 + STRUCT_OFFSET(XPVIO, xpv_cur),
1005 SVt_PVIO, TRUE, NONV, HASARENA,
1006 FIT_ARENA(24, sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur)) },
1009 #define new_body_type(sv_type) \
1010 (void *)((char *)S_new_body(aTHX_ sv_type))
1012 #define del_body_type(p, sv_type) \
1013 del_body(p, &PL_body_roots[sv_type])
1016 #define new_body_allocated(sv_type) \
1017 (void *)((char *)S_new_body(aTHX_ sv_type) \
1018 - bodies_by_type[sv_type].offset)
1020 #define del_body_allocated(p, sv_type) \
1021 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1024 #define my_safemalloc(s) (void*)safemalloc(s)
1025 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1026 #define my_safefree(p) safefree((char*)p)
1030 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1031 #define del_XNV(p) my_safefree(p)
1033 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1034 #define del_XPVNV(p) my_safefree(p)
1036 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1037 #define del_XPVAV(p) my_safefree(p)
1039 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1040 #define del_XPVHV(p) my_safefree(p)
1042 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1043 #define del_XPVMG(p) my_safefree(p)
1045 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1046 #define del_XPVGV(p) my_safefree(p)
1050 #define new_XNV() new_body_type(SVt_NV)
1051 #define del_XNV(p) del_body_type(p, SVt_NV)
1053 #define new_XPVNV() new_body_type(SVt_PVNV)
1054 #define del_XPVNV(p) del_body_type(p, SVt_PVNV)
1056 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1057 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1059 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1060 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1062 #define new_XPVMG() new_body_type(SVt_PVMG)
1063 #define del_XPVMG(p) del_body_type(p, SVt_PVMG)
1065 #define new_XPVGV() new_body_type(SVt_PVGV)
1066 #define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1070 /* no arena for you! */
1072 #define new_NOARENA(details) \
1073 my_safemalloc((details)->body_size + (details)->offset)
1074 #define new_NOARENAZ(details) \
1075 my_safecalloc((details)->body_size + (details)->offset)
1078 S_more_bodies (pTHX_ const svtype sv_type)
1081 void ** const root = &PL_body_roots[sv_type];
1082 const struct body_details * const bdp = &bodies_by_type[sv_type];
1083 const size_t body_size = bdp->body_size;
1086 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1087 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1088 static bool done_sanity_check;
1090 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1091 * variables like done_sanity_check. */
1092 if (!done_sanity_check) {
1093 unsigned int i = SVt_LAST;
1095 done_sanity_check = TRUE;
1098 assert (bodies_by_type[i].type == i);
1102 assert(bdp->arena_size);
1104 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1106 end = start + arena_size - 2 * body_size;
1108 /* computed count doesnt reflect the 1st slot reservation */
1109 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1110 DEBUG_m(PerlIO_printf(Perl_debug_log,
1111 "arena %p end %p arena-size %d (from %d) type %d "
1113 (void*)start, (void*)end, (int)arena_size,
1114 (int)bdp->arena_size, sv_type, (int)body_size,
1115 (int)arena_size / (int)body_size));
1117 DEBUG_m(PerlIO_printf(Perl_debug_log,
1118 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1119 (void*)start, (void*)end,
1120 (int)bdp->arena_size, sv_type, (int)body_size,
1121 (int)bdp->arena_size / (int)body_size));
1123 *root = (void *)start;
1125 while (start <= end) {
1126 char * const next = start + body_size;
1127 *(void**) start = (void *)next;
1130 *(void **)start = 0;
1135 /* grab a new thing from the free list, allocating more if necessary.
1136 The inline version is used for speed in hot routines, and the
1137 function using it serves the rest (unless PURIFY).
1139 #define new_body_inline(xpv, sv_type) \
1141 void ** const r3wt = &PL_body_roots[sv_type]; \
1142 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1143 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1144 *(r3wt) = *(void**)(xpv); \
1150 S_new_body(pTHX_ const svtype sv_type)
1154 new_body_inline(xpv, sv_type);
1160 static const struct body_details fake_rv =
1161 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1164 =for apidoc sv_upgrade
1166 Upgrade an SV to a more complex form. Generally adds a new body type to the
1167 SV, then copies across as much information as possible from the old body.
1168 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1174 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1179 const svtype old_type = SvTYPE(sv);
1180 const struct body_details *new_type_details;
1181 const struct body_details *old_type_details
1182 = bodies_by_type + old_type;
1183 SV *referant = NULL;
1185 PERL_ARGS_ASSERT_SV_UPGRADE;
1187 if (old_type == new_type)
1190 /* This clause was purposefully added ahead of the early return above to
1191 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1192 inference by Nick I-S that it would fix other troublesome cases. See
1193 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1195 Given that shared hash key scalars are no longer PVIV, but PV, there is
1196 no longer need to unshare so as to free up the IVX slot for its proper
1197 purpose. So it's safe to move the early return earlier. */
1199 if (new_type != SVt_PV && SvIsCOW(sv)) {
1200 sv_force_normal_flags(sv, 0);
1203 old_body = SvANY(sv);
1205 /* Copying structures onto other structures that have been neatly zeroed
1206 has a subtle gotcha. Consider XPVMG
1208 +------+------+------+------+------+-------+-------+
1209 | NV | CUR | LEN | IV | MAGIC | STASH |
1210 +------+------+------+------+------+-------+-------+
1211 0 4 8 12 16 20 24 28
1213 where NVs are aligned to 8 bytes, so that sizeof that structure is
1214 actually 32 bytes long, with 4 bytes of padding at the end:
1216 +------+------+------+------+------+-------+-------+------+
1217 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1218 +------+------+------+------+------+-------+-------+------+
1219 0 4 8 12 16 20 24 28 32
1221 so what happens if you allocate memory for this structure:
1223 +------+------+------+------+------+-------+-------+------+------+...
1224 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1225 +------+------+------+------+------+-------+-------+------+------+...
1226 0 4 8 12 16 20 24 28 32 36
1228 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1229 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1230 started out as zero once, but it's quite possible that it isn't. So now,
1231 rather than a nicely zeroed GP, you have it pointing somewhere random.
1234 (In fact, GP ends up pointing at a previous GP structure, because the
1235 principle cause of the padding in XPVMG getting garbage is a copy of
1236 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1237 this happens to be moot because XPVGV has been re-ordered, with GP
1238 no longer after STASH)
1240 So we are careful and work out the size of used parts of all the
1248 referant = SvRV(sv);
1249 old_type_details = &fake_rv;
1250 if (new_type == SVt_NV)
1251 new_type = SVt_PVNV;
1253 if (new_type < SVt_PVIV) {
1254 new_type = (new_type == SVt_NV)
1255 ? SVt_PVNV : SVt_PVIV;
1260 if (new_type < SVt_PVNV) {
1261 new_type = SVt_PVNV;
1265 assert(new_type > SVt_PV);
1266 assert(SVt_IV < SVt_PV);
1267 assert(SVt_NV < SVt_PV);
1274 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1275 there's no way that it can be safely upgraded, because perl.c
1276 expects to Safefree(SvANY(PL_mess_sv)) */
1277 assert(sv != PL_mess_sv);
1278 /* This flag bit is used to mean other things in other scalar types.
1279 Given that it only has meaning inside the pad, it shouldn't be set
1280 on anything that can get upgraded. */
1281 assert(!SvPAD_TYPED(sv));
1284 if (old_type_details->cant_upgrade)
1285 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1286 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1289 if (old_type > new_type)
1290 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1291 (int)old_type, (int)new_type);
1293 new_type_details = bodies_by_type + new_type;
1295 SvFLAGS(sv) &= ~SVTYPEMASK;
1296 SvFLAGS(sv) |= new_type;
1298 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1299 the return statements above will have triggered. */
1300 assert (new_type != SVt_NULL);
1303 assert(old_type == SVt_NULL);
1304 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1308 assert(old_type == SVt_NULL);
1309 SvANY(sv) = new_XNV();
1314 assert(new_type_details->body_size);
1317 assert(new_type_details->arena);
1318 assert(new_type_details->arena_size);
1319 /* This points to the start of the allocated area. */
1320 new_body_inline(new_body, new_type);
1321 Zero(new_body, new_type_details->body_size, char);
1322 new_body = ((char *)new_body) - new_type_details->offset;
1324 /* We always allocated the full length item with PURIFY. To do this
1325 we fake things so that arena is false for all 16 types.. */
1326 new_body = new_NOARENAZ(new_type_details);
1328 SvANY(sv) = new_body;
1329 if (new_type == SVt_PVAV) {
1333 if (old_type_details->body_size) {
1336 /* It will have been zeroed when the new body was allocated.
1337 Lets not write to it, in case it confuses a write-back
1343 #ifndef NODEFAULT_SHAREKEYS
1344 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1346 HvMAX(sv) = 7; /* (start with 8 buckets) */
1347 if (old_type_details->body_size) {
1350 /* It will have been zeroed when the new body was allocated.
1351 Lets not write to it, in case it confuses a write-back
1356 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1357 The target created by newSVrv also is, and it can have magic.
1358 However, it never has SvPVX set.
1360 if (old_type == SVt_IV) {
1362 } else if (old_type >= SVt_PV) {
1363 assert(SvPVX_const(sv) == 0);
1366 if (old_type >= SVt_PVMG) {
1367 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1368 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1370 sv->sv_u.svu_array = NULL; /* or svu_hash */
1376 /* XXX Is this still needed? Was it ever needed? Surely as there is
1377 no route from NV to PVIV, NOK can never be true */
1378 assert(!SvNOKp(sv));
1390 assert(new_type_details->body_size);
1391 /* We always allocated the full length item with PURIFY. To do this
1392 we fake things so that arena is false for all 16 types.. */
1393 if(new_type_details->arena) {
1394 /* This points to the start of the allocated area. */
1395 new_body_inline(new_body, new_type);
1396 Zero(new_body, new_type_details->body_size, char);
1397 new_body = ((char *)new_body) - new_type_details->offset;
1399 new_body = new_NOARENAZ(new_type_details);
1401 SvANY(sv) = new_body;
1403 if (old_type_details->copy) {
1404 /* There is now the potential for an upgrade from something without
1405 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1406 int offset = old_type_details->offset;
1407 int length = old_type_details->copy;
1409 if (new_type_details->offset > old_type_details->offset) {
1410 const int difference
1411 = new_type_details->offset - old_type_details->offset;
1412 offset += difference;
1413 length -= difference;
1415 assert (length >= 0);
1417 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1421 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1422 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1423 * correct 0.0 for us. Otherwise, if the old body didn't have an
1424 * NV slot, but the new one does, then we need to initialise the
1425 * freshly created NV slot with whatever the correct bit pattern is
1427 if (old_type_details->zero_nv && !new_type_details->zero_nv
1428 && !isGV_with_GP(sv))
1432 if (new_type == SVt_PVIO) {
1433 IO * const io = MUTABLE_IO(sv);
1434 GV *iogv = gv_fetchpvs("IO::Handle::", GV_ADD, SVt_PVHV);
1437 /* Clear the stashcache because a new IO could overrule a package
1439 hv_clear(PL_stashcache);
1441 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1442 IoPAGE_LEN(sv) = 60;
1444 if (old_type < SVt_PV) {
1445 /* referant will be NULL unless the old type was SVt_IV emulating
1447 sv->sv_u.svu_rv = referant;
1451 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1452 (unsigned long)new_type);
1455 if (old_type > SVt_IV) { /* SVt_IVs are overloaded for PTEs */
1457 my_safefree(old_body);
1459 /* Note that there is an assumption that all bodies of types that
1460 can be upgraded came from arenas. Only the more complex non-
1461 upgradable types are allowed to be directly malloc()ed. */
1462 assert(old_type_details->arena);
1463 del_body((void*)((char*)old_body + old_type_details->offset),
1464 &PL_body_roots[old_type]);
1470 =for apidoc sv_backoff
1472 Remove any string offset. You should normally use the C<SvOOK_off> macro
1479 Perl_sv_backoff(pTHX_ register SV *const sv)
1482 const char * const s = SvPVX_const(sv);
1484 PERL_ARGS_ASSERT_SV_BACKOFF;
1485 PERL_UNUSED_CONTEXT;
1488 assert(SvTYPE(sv) != SVt_PVHV);
1489 assert(SvTYPE(sv) != SVt_PVAV);
1491 SvOOK_offset(sv, delta);
1493 SvLEN_set(sv, SvLEN(sv) + delta);
1494 SvPV_set(sv, SvPVX(sv) - delta);
1495 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1496 SvFLAGS(sv) &= ~SVf_OOK;
1503 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1504 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1505 Use the C<SvGROW> wrapper instead.
1511 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1515 PERL_ARGS_ASSERT_SV_GROW;
1517 if (PL_madskills && newlen >= 0x100000) {
1518 PerlIO_printf(Perl_debug_log,
1519 "Allocation too large: %"UVxf"\n", (UV)newlen);
1521 #ifdef HAS_64K_LIMIT
1522 if (newlen >= 0x10000) {
1523 PerlIO_printf(Perl_debug_log,
1524 "Allocation too large: %"UVxf"\n", (UV)newlen);
1527 #endif /* HAS_64K_LIMIT */
1530 if (SvTYPE(sv) < SVt_PV) {
1531 sv_upgrade(sv, SVt_PV);
1532 s = SvPVX_mutable(sv);
1534 else if (SvOOK(sv)) { /* pv is offset? */
1536 s = SvPVX_mutable(sv);
1537 if (newlen > SvLEN(sv))
1538 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1539 #ifdef HAS_64K_LIMIT
1540 if (newlen >= 0x10000)
1545 s = SvPVX_mutable(sv);
1547 if (newlen > SvLEN(sv)) { /* need more room? */
1548 #ifndef Perl_safesysmalloc_size
1549 newlen = PERL_STRLEN_ROUNDUP(newlen);
1551 if (SvLEN(sv) && s) {
1552 s = (char*)saferealloc(s, newlen);
1555 s = (char*)safemalloc(newlen);
1556 if (SvPVX_const(sv) && SvCUR(sv)) {
1557 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1561 #ifdef Perl_safesysmalloc_size
1562 /* Do this here, do it once, do it right, and then we will never get
1563 called back into sv_grow() unless there really is some growing
1565 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1567 SvLEN_set(sv, newlen);
1574 =for apidoc sv_setiv
1576 Copies an integer into the given SV, upgrading first if necessary.
1577 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1583 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1587 PERL_ARGS_ASSERT_SV_SETIV;
1589 SV_CHECK_THINKFIRST_COW_DROP(sv);
1590 switch (SvTYPE(sv)) {
1593 sv_upgrade(sv, SVt_IV);
1596 sv_upgrade(sv, SVt_PVIV);
1600 if (!isGV_with_GP(sv))
1607 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1611 (void)SvIOK_only(sv); /* validate number */
1617 =for apidoc sv_setiv_mg
1619 Like C<sv_setiv>, but also handles 'set' magic.
1625 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1627 PERL_ARGS_ASSERT_SV_SETIV_MG;
1634 =for apidoc sv_setuv
1636 Copies an unsigned integer into the given SV, upgrading first if necessary.
1637 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1643 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1645 PERL_ARGS_ASSERT_SV_SETUV;
1647 /* With these two if statements:
1648 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1651 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1653 If you wish to remove them, please benchmark to see what the effect is
1655 if (u <= (UV)IV_MAX) {
1656 sv_setiv(sv, (IV)u);
1665 =for apidoc sv_setuv_mg
1667 Like C<sv_setuv>, but also handles 'set' magic.
1673 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1675 PERL_ARGS_ASSERT_SV_SETUV_MG;
1682 =for apidoc sv_setnv
1684 Copies a double into the given SV, upgrading first if necessary.
1685 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1691 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1695 PERL_ARGS_ASSERT_SV_SETNV;
1697 SV_CHECK_THINKFIRST_COW_DROP(sv);
1698 switch (SvTYPE(sv)) {
1701 sv_upgrade(sv, SVt_NV);
1705 sv_upgrade(sv, SVt_PVNV);
1709 if (!isGV_with_GP(sv))
1716 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1721 (void)SvNOK_only(sv); /* validate number */
1726 =for apidoc sv_setnv_mg
1728 Like C<sv_setnv>, but also handles 'set' magic.
1734 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1736 PERL_ARGS_ASSERT_SV_SETNV_MG;
1742 /* Print an "isn't numeric" warning, using a cleaned-up,
1743 * printable version of the offending string
1747 S_not_a_number(pTHX_ SV *const sv)
1754 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1757 dsv = newSVpvs_flags("", SVs_TEMP);
1758 pv = sv_uni_display(dsv, sv, 10, 0);
1761 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1762 /* each *s can expand to 4 chars + "...\0",
1763 i.e. need room for 8 chars */
1765 const char *s = SvPVX_const(sv);
1766 const char * const end = s + SvCUR(sv);
1767 for ( ; s < end && d < limit; s++ ) {
1769 if (ch & 128 && !isPRINT_LC(ch)) {
1778 else if (ch == '\r') {
1782 else if (ch == '\f') {
1786 else if (ch == '\\') {
1790 else if (ch == '\0') {
1794 else if (isPRINT_LC(ch))
1811 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1812 "Argument \"%s\" isn't numeric in %s", pv,
1815 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1816 "Argument \"%s\" isn't numeric", pv);
1820 =for apidoc looks_like_number
1822 Test if the content of an SV looks like a number (or is a number).
1823 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1824 non-numeric warning), even if your atof() doesn't grok them.
1830 Perl_looks_like_number(pTHX_ SV *const sv)
1832 register const char *sbegin;
1835 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1838 sbegin = SvPVX_const(sv);
1841 else if (SvPOKp(sv))
1842 sbegin = SvPV_const(sv, len);
1844 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1845 return grok_number(sbegin, len, NULL);
1849 S_glob_2number(pTHX_ GV * const gv)
1851 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1852 SV *const buffer = sv_newmortal();
1854 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1856 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1859 gv_efullname3(buffer, gv, "*");
1860 SvFLAGS(gv) |= wasfake;
1862 /* We know that all GVs stringify to something that is not-a-number,
1863 so no need to test that. */
1864 if (ckWARN(WARN_NUMERIC))
1865 not_a_number(buffer);
1866 /* We just want something true to return, so that S_sv_2iuv_common
1867 can tail call us and return true. */
1871 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1872 until proven guilty, assume that things are not that bad... */
1877 As 64 bit platforms often have an NV that doesn't preserve all bits of
1878 an IV (an assumption perl has been based on to date) it becomes necessary
1879 to remove the assumption that the NV always carries enough precision to
1880 recreate the IV whenever needed, and that the NV is the canonical form.
1881 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1882 precision as a side effect of conversion (which would lead to insanity
1883 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1884 1) to distinguish between IV/UV/NV slots that have cached a valid
1885 conversion where precision was lost and IV/UV/NV slots that have a
1886 valid conversion which has lost no precision
1887 2) to ensure that if a numeric conversion to one form is requested that
1888 would lose precision, the precise conversion (or differently
1889 imprecise conversion) is also performed and cached, to prevent
1890 requests for different numeric formats on the same SV causing
1891 lossy conversion chains. (lossless conversion chains are perfectly
1896 SvIOKp is true if the IV slot contains a valid value
1897 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1898 SvNOKp is true if the NV slot contains a valid value
1899 SvNOK is true only if the NV value is accurate
1902 while converting from PV to NV, check to see if converting that NV to an
1903 IV(or UV) would lose accuracy over a direct conversion from PV to
1904 IV(or UV). If it would, cache both conversions, return NV, but mark
1905 SV as IOK NOKp (ie not NOK).
1907 While converting from PV to IV, check to see if converting that IV to an
1908 NV would lose accuracy over a direct conversion from PV to NV. If it
1909 would, cache both conversions, flag similarly.
1911 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1912 correctly because if IV & NV were set NV *always* overruled.
1913 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1914 changes - now IV and NV together means that the two are interchangeable:
1915 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1917 The benefit of this is that operations such as pp_add know that if
1918 SvIOK is true for both left and right operands, then integer addition
1919 can be used instead of floating point (for cases where the result won't
1920 overflow). Before, floating point was always used, which could lead to
1921 loss of precision compared with integer addition.
1923 * making IV and NV equal status should make maths accurate on 64 bit
1925 * may speed up maths somewhat if pp_add and friends start to use
1926 integers when possible instead of fp. (Hopefully the overhead in
1927 looking for SvIOK and checking for overflow will not outweigh the
1928 fp to integer speedup)
1929 * will slow down integer operations (callers of SvIV) on "inaccurate"
1930 values, as the change from SvIOK to SvIOKp will cause a call into
1931 sv_2iv each time rather than a macro access direct to the IV slot
1932 * should speed up number->string conversion on integers as IV is
1933 favoured when IV and NV are equally accurate
1935 ####################################################################
1936 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1937 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1938 On the other hand, SvUOK is true iff UV.
1939 ####################################################################
1941 Your mileage will vary depending your CPU's relative fp to integer
1945 #ifndef NV_PRESERVES_UV
1946 # define IS_NUMBER_UNDERFLOW_IV 1
1947 # define IS_NUMBER_UNDERFLOW_UV 2
1948 # define IS_NUMBER_IV_AND_UV 2
1949 # define IS_NUMBER_OVERFLOW_IV 4
1950 # define IS_NUMBER_OVERFLOW_UV 5
1952 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1954 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1956 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1964 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1966 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1967 if (SvNVX(sv) < (NV)IV_MIN) {
1968 (void)SvIOKp_on(sv);
1970 SvIV_set(sv, IV_MIN);
1971 return IS_NUMBER_UNDERFLOW_IV;
1973 if (SvNVX(sv) > (NV)UV_MAX) {
1974 (void)SvIOKp_on(sv);
1977 SvUV_set(sv, UV_MAX);
1978 return IS_NUMBER_OVERFLOW_UV;
1980 (void)SvIOKp_on(sv);
1982 /* Can't use strtol etc to convert this string. (See truth table in
1984 if (SvNVX(sv) <= (UV)IV_MAX) {
1985 SvIV_set(sv, I_V(SvNVX(sv)));
1986 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1987 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1989 /* Integer is imprecise. NOK, IOKp */
1991 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1994 SvUV_set(sv, U_V(SvNVX(sv)));
1995 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1996 if (SvUVX(sv) == UV_MAX) {
1997 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1998 possibly be preserved by NV. Hence, it must be overflow.
2000 return IS_NUMBER_OVERFLOW_UV;
2002 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2004 /* Integer is imprecise. NOK, IOKp */
2006 return IS_NUMBER_OVERFLOW_IV;
2008 #endif /* !NV_PRESERVES_UV*/
2011 S_sv_2iuv_common(pTHX_ SV *const sv)
2015 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2018 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2019 * without also getting a cached IV/UV from it at the same time
2020 * (ie PV->NV conversion should detect loss of accuracy and cache
2021 * IV or UV at same time to avoid this. */
2022 /* IV-over-UV optimisation - choose to cache IV if possible */
2024 if (SvTYPE(sv) == SVt_NV)
2025 sv_upgrade(sv, SVt_PVNV);
2027 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2028 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2029 certainly cast into the IV range at IV_MAX, whereas the correct
2030 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2032 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2033 if (Perl_isnan(SvNVX(sv))) {
2039 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2040 SvIV_set(sv, I_V(SvNVX(sv)));
2041 if (SvNVX(sv) == (NV) SvIVX(sv)
2042 #ifndef NV_PRESERVES_UV
2043 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2044 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2045 /* Don't flag it as "accurately an integer" if the number
2046 came from a (by definition imprecise) NV operation, and
2047 we're outside the range of NV integer precision */
2051 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2053 /* scalar has trailing garbage, eg "42a" */
2055 DEBUG_c(PerlIO_printf(Perl_debug_log,
2056 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2062 /* IV not precise. No need to convert from PV, as NV
2063 conversion would already have cached IV if it detected
2064 that PV->IV would be better than PV->NV->IV
2065 flags already correct - don't set public IOK. */
2066 DEBUG_c(PerlIO_printf(Perl_debug_log,
2067 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2072 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2073 but the cast (NV)IV_MIN rounds to a the value less (more
2074 negative) than IV_MIN which happens to be equal to SvNVX ??
2075 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2076 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2077 (NV)UVX == NVX are both true, but the values differ. :-(
2078 Hopefully for 2s complement IV_MIN is something like
2079 0x8000000000000000 which will be exact. NWC */
2082 SvUV_set(sv, U_V(SvNVX(sv)));
2084 (SvNVX(sv) == (NV) SvUVX(sv))
2085 #ifndef NV_PRESERVES_UV
2086 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2087 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2088 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2089 /* Don't flag it as "accurately an integer" if the number
2090 came from a (by definition imprecise) NV operation, and
2091 we're outside the range of NV integer precision */
2097 DEBUG_c(PerlIO_printf(Perl_debug_log,
2098 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2104 else if (SvPOKp(sv) && SvLEN(sv)) {
2106 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2107 /* We want to avoid a possible problem when we cache an IV/ a UV which
2108 may be later translated to an NV, and the resulting NV is not
2109 the same as the direct translation of the initial string
2110 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2111 be careful to ensure that the value with the .456 is around if the
2112 NV value is requested in the future).
2114 This means that if we cache such an IV/a UV, we need to cache the
2115 NV as well. Moreover, we trade speed for space, and do not
2116 cache the NV if we are sure it's not needed.
2119 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2120 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2121 == IS_NUMBER_IN_UV) {
2122 /* It's definitely an integer, only upgrade to PVIV */
2123 if (SvTYPE(sv) < SVt_PVIV)
2124 sv_upgrade(sv, SVt_PVIV);
2126 } else if (SvTYPE(sv) < SVt_PVNV)
2127 sv_upgrade(sv, SVt_PVNV);
2129 /* If NVs preserve UVs then we only use the UV value if we know that
2130 we aren't going to call atof() below. If NVs don't preserve UVs
2131 then the value returned may have more precision than atof() will
2132 return, even though value isn't perfectly accurate. */
2133 if ((numtype & (IS_NUMBER_IN_UV
2134 #ifdef NV_PRESERVES_UV
2137 )) == IS_NUMBER_IN_UV) {
2138 /* This won't turn off the public IOK flag if it was set above */
2139 (void)SvIOKp_on(sv);
2141 if (!(numtype & IS_NUMBER_NEG)) {
2143 if (value <= (UV)IV_MAX) {
2144 SvIV_set(sv, (IV)value);
2146 /* it didn't overflow, and it was positive. */
2147 SvUV_set(sv, value);
2151 /* 2s complement assumption */
2152 if (value <= (UV)IV_MIN) {
2153 SvIV_set(sv, -(IV)value);
2155 /* Too negative for an IV. This is a double upgrade, but
2156 I'm assuming it will be rare. */
2157 if (SvTYPE(sv) < SVt_PVNV)
2158 sv_upgrade(sv, SVt_PVNV);
2162 SvNV_set(sv, -(NV)value);
2163 SvIV_set(sv, IV_MIN);
2167 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2168 will be in the previous block to set the IV slot, and the next
2169 block to set the NV slot. So no else here. */
2171 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2172 != IS_NUMBER_IN_UV) {
2173 /* It wasn't an (integer that doesn't overflow the UV). */
2174 SvNV_set(sv, Atof(SvPVX_const(sv)));
2176 if (! numtype && ckWARN(WARN_NUMERIC))
2179 #if defined(USE_LONG_DOUBLE)
2180 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2181 PTR2UV(sv), SvNVX(sv)));
2183 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2184 PTR2UV(sv), SvNVX(sv)));
2187 #ifdef NV_PRESERVES_UV
2188 (void)SvIOKp_on(sv);
2190 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2195 NOOP; /* Integer is imprecise. NOK, IOKp */
2197 /* UV will not work better than IV */
2199 if (SvNVX(sv) > (NV)UV_MAX) {
2201 /* Integer is inaccurate. NOK, IOKp, is UV */
2202 SvUV_set(sv, UV_MAX);
2204 SvUV_set(sv, U_V(SvNVX(sv)));
2205 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2206 NV preservse UV so can do correct comparison. */
2207 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2210 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2215 #else /* NV_PRESERVES_UV */
2216 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2217 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2218 /* The IV/UV slot will have been set from value returned by
2219 grok_number above. The NV slot has just been set using
2222 assert (SvIOKp(sv));
2224 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2225 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2226 /* Small enough to preserve all bits. */
2227 (void)SvIOKp_on(sv);
2229 SvIV_set(sv, I_V(SvNVX(sv)));
2230 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2232 /* Assumption: first non-preserved integer is < IV_MAX,
2233 this NV is in the preserved range, therefore: */
2234 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2236 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2240 0 0 already failed to read UV.
2241 0 1 already failed to read UV.
2242 1 0 you won't get here in this case. IV/UV
2243 slot set, public IOK, Atof() unneeded.
2244 1 1 already read UV.
2245 so there's no point in sv_2iuv_non_preserve() attempting
2246 to use atol, strtol, strtoul etc. */
2248 sv_2iuv_non_preserve (sv, numtype);
2250 sv_2iuv_non_preserve (sv);
2254 #endif /* NV_PRESERVES_UV */
2255 /* It might be more code efficient to go through the entire logic above
2256 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2257 gets complex and potentially buggy, so more programmer efficient
2258 to do it this way, by turning off the public flags: */
2260 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2264 if (isGV_with_GP(sv))
2265 return glob_2number(MUTABLE_GV(sv));
2267 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2268 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2271 if (SvTYPE(sv) < SVt_IV)
2272 /* Typically the caller expects that sv_any is not NULL now. */
2273 sv_upgrade(sv, SVt_IV);
2274 /* Return 0 from the caller. */
2281 =for apidoc sv_2iv_flags
2283 Return the integer value of an SV, doing any necessary string
2284 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2285 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2291 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2296 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2297 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2298 cache IVs just in case. In practice it seems that they never
2299 actually anywhere accessible by user Perl code, let alone get used
2300 in anything other than a string context. */
2301 if (flags & SV_GMAGIC)
2306 return I_V(SvNVX(sv));
2308 if (SvPOKp(sv) && SvLEN(sv)) {
2311 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2313 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2314 == IS_NUMBER_IN_UV) {
2315 /* It's definitely an integer */
2316 if (numtype & IS_NUMBER_NEG) {
2317 if (value < (UV)IV_MIN)
2320 if (value < (UV)IV_MAX)
2325 if (ckWARN(WARN_NUMERIC))
2328 return I_V(Atof(SvPVX_const(sv)));
2333 assert(SvTYPE(sv) >= SVt_PVMG);
2334 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2335 } else if (SvTHINKFIRST(sv)) {
2339 SV * const tmpstr=AMG_CALLun(sv,numer);
2340 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2341 return SvIV(tmpstr);
2344 return PTR2IV(SvRV(sv));
2347 sv_force_normal_flags(sv, 0);
2349 if (SvREADONLY(sv) && !SvOK(sv)) {
2350 if (ckWARN(WARN_UNINITIALIZED))
2356 if (S_sv_2iuv_common(aTHX_ sv))
2359 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2360 PTR2UV(sv),SvIVX(sv)));
2361 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2365 =for apidoc sv_2uv_flags
2367 Return the unsigned integer value of an SV, doing any necessary string
2368 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2369 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2375 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2380 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2381 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2382 cache IVs just in case. */
2383 if (flags & SV_GMAGIC)
2388 return U_V(SvNVX(sv));
2389 if (SvPOKp(sv) && SvLEN(sv)) {
2392 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2394 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2395 == IS_NUMBER_IN_UV) {
2396 /* It's definitely an integer */
2397 if (!(numtype & IS_NUMBER_NEG))
2401 if (ckWARN(WARN_NUMERIC))
2404 return U_V(Atof(SvPVX_const(sv)));
2409 assert(SvTYPE(sv) >= SVt_PVMG);
2410 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2411 } else if (SvTHINKFIRST(sv)) {
2415 SV *const tmpstr = AMG_CALLun(sv,numer);
2416 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2417 return SvUV(tmpstr);
2420 return PTR2UV(SvRV(sv));
2423 sv_force_normal_flags(sv, 0);
2425 if (SvREADONLY(sv) && !SvOK(sv)) {
2426 if (ckWARN(WARN_UNINITIALIZED))
2432 if (S_sv_2iuv_common(aTHX_ sv))
2436 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2437 PTR2UV(sv),SvUVX(sv)));
2438 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2444 Return the num value of an SV, doing any necessary string or integer
2445 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2452 Perl_sv_2nv(pTHX_ register SV *const sv)
2457 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2458 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2459 cache IVs just in case. */
2463 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2464 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2465 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2467 return Atof(SvPVX_const(sv));
2471 return (NV)SvUVX(sv);
2473 return (NV)SvIVX(sv);
2478 assert(SvTYPE(sv) >= SVt_PVMG);
2479 /* This falls through to the report_uninit near the end of the
2481 } else if (SvTHINKFIRST(sv)) {
2485 SV *const tmpstr = AMG_CALLun(sv,numer);
2486 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2487 return SvNV(tmpstr);
2490 return PTR2NV(SvRV(sv));
2493 sv_force_normal_flags(sv, 0);
2495 if (SvREADONLY(sv) && !SvOK(sv)) {
2496 if (ckWARN(WARN_UNINITIALIZED))
2501 if (SvTYPE(sv) < SVt_NV) {
2502 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2503 sv_upgrade(sv, SVt_NV);
2504 #ifdef USE_LONG_DOUBLE
2506 STORE_NUMERIC_LOCAL_SET_STANDARD();
2507 PerlIO_printf(Perl_debug_log,
2508 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2509 PTR2UV(sv), SvNVX(sv));
2510 RESTORE_NUMERIC_LOCAL();
2514 STORE_NUMERIC_LOCAL_SET_STANDARD();
2515 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2516 PTR2UV(sv), SvNVX(sv));
2517 RESTORE_NUMERIC_LOCAL();
2521 else if (SvTYPE(sv) < SVt_PVNV)
2522 sv_upgrade(sv, SVt_PVNV);
2527 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2528 #ifdef NV_PRESERVES_UV
2534 /* Only set the public NV OK flag if this NV preserves the IV */
2535 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2537 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2538 : (SvIVX(sv) == I_V(SvNVX(sv))))
2544 else if (SvPOKp(sv) && SvLEN(sv)) {
2546 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2547 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2549 #ifdef NV_PRESERVES_UV
2550 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2551 == IS_NUMBER_IN_UV) {
2552 /* It's definitely an integer */
2553 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2555 SvNV_set(sv, Atof(SvPVX_const(sv)));
2561 SvNV_set(sv, Atof(SvPVX_const(sv)));
2562 /* Only set the public NV OK flag if this NV preserves the value in
2563 the PV at least as well as an IV/UV would.
2564 Not sure how to do this 100% reliably. */
2565 /* if that shift count is out of range then Configure's test is
2566 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2568 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2569 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2570 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2571 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2572 /* Can't use strtol etc to convert this string, so don't try.
2573 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2576 /* value has been set. It may not be precise. */
2577 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2578 /* 2s complement assumption for (UV)IV_MIN */
2579 SvNOK_on(sv); /* Integer is too negative. */
2584 if (numtype & IS_NUMBER_NEG) {
2585 SvIV_set(sv, -(IV)value);
2586 } else if (value <= (UV)IV_MAX) {
2587 SvIV_set(sv, (IV)value);
2589 SvUV_set(sv, value);
2593 if (numtype & IS_NUMBER_NOT_INT) {
2594 /* I believe that even if the original PV had decimals,
2595 they are lost beyond the limit of the FP precision.
2596 However, neither is canonical, so both only get p
2597 flags. NWC, 2000/11/25 */
2598 /* Both already have p flags, so do nothing */
2600 const NV nv = SvNVX(sv);
2601 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2602 if (SvIVX(sv) == I_V(nv)) {
2605 /* It had no "." so it must be integer. */
2609 /* between IV_MAX and NV(UV_MAX).
2610 Could be slightly > UV_MAX */
2612 if (numtype & IS_NUMBER_NOT_INT) {
2613 /* UV and NV both imprecise. */
2615 const UV nv_as_uv = U_V(nv);
2617 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2626 /* It might be more code efficient to go through the entire logic above
2627 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2628 gets complex and potentially buggy, so more programmer efficient
2629 to do it this way, by turning off the public flags: */
2631 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2632 #endif /* NV_PRESERVES_UV */
2635 if (isGV_with_GP(sv)) {
2636 glob_2number(MUTABLE_GV(sv));
2640 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2642 assert (SvTYPE(sv) >= SVt_NV);
2643 /* Typically the caller expects that sv_any is not NULL now. */
2644 /* XXX Ilya implies that this is a bug in callers that assume this
2645 and ideally should be fixed. */
2648 #if defined(USE_LONG_DOUBLE)
2650 STORE_NUMERIC_LOCAL_SET_STANDARD();
2651 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2652 PTR2UV(sv), SvNVX(sv));
2653 RESTORE_NUMERIC_LOCAL();
2657 STORE_NUMERIC_LOCAL_SET_STANDARD();
2658 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2659 PTR2UV(sv), SvNVX(sv));
2660 RESTORE_NUMERIC_LOCAL();
2669 Return an SV with the numeric value of the source SV, doing any necessary
2670 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2671 access this function.
2677 Perl_sv_2num(pTHX_ register SV *const sv)
2679 PERL_ARGS_ASSERT_SV_2NUM;
2684 SV * const tmpsv = AMG_CALLun(sv,numer);
2685 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2686 return sv_2num(tmpsv);
2688 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2691 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2692 * UV as a string towards the end of buf, and return pointers to start and
2695 * We assume that buf is at least TYPE_CHARS(UV) long.
2699 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2701 char *ptr = buf + TYPE_CHARS(UV);
2702 char * const ebuf = ptr;
2705 PERL_ARGS_ASSERT_UIV_2BUF;
2717 *--ptr = '0' + (char)(uv % 10);
2726 =for apidoc sv_2pv_flags
2728 Returns a pointer to the string value of an SV, and sets *lp to its length.
2729 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2731 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2732 usually end up here too.
2738 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2748 if (SvGMAGICAL(sv)) {
2749 if (flags & SV_GMAGIC)
2754 if (flags & SV_MUTABLE_RETURN)
2755 return SvPVX_mutable(sv);
2756 if (flags & SV_CONST_RETURN)
2757 return (char *)SvPVX_const(sv);
2760 if (SvIOKp(sv) || SvNOKp(sv)) {
2761 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2766 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2767 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2769 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2776 #ifdef FIXNEGATIVEZERO
2777 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2783 SvUPGRADE(sv, SVt_PV);
2786 s = SvGROW_mutable(sv, len + 1);
2789 return (char*)memcpy(s, tbuf, len + 1);
2795 assert(SvTYPE(sv) >= SVt_PVMG);
2796 /* This falls through to the report_uninit near the end of the
2798 } else if (SvTHINKFIRST(sv)) {
2802 SV *const tmpstr = AMG_CALLun(sv,string);
2803 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2805 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2809 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2810 if (flags & SV_CONST_RETURN) {
2811 pv = (char *) SvPVX_const(tmpstr);
2813 pv = (flags & SV_MUTABLE_RETURN)
2814 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2817 *lp = SvCUR(tmpstr);
2819 pv = sv_2pv_flags(tmpstr, lp, flags);
2832 SV *const referent = SvRV(sv);
2836 retval = buffer = savepvn("NULLREF", len);
2837 } else if (SvTYPE(referent) == SVt_REGEXP) {
2838 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2843 /* If the regex is UTF-8 we want the containing scalar to
2844 have an UTF-8 flag too */
2850 if ((seen_evals = RX_SEEN_EVALS(re)))
2851 PL_reginterp_cnt += seen_evals;
2854 *lp = RX_WRAPLEN(re);
2856 return RX_WRAPPED(re);
2858 const char *const typestr = sv_reftype(referent, 0);
2859 const STRLEN typelen = strlen(typestr);
2860 UV addr = PTR2UV(referent);
2861 const char *stashname = NULL;
2862 STRLEN stashnamelen = 0; /* hush, gcc */
2863 const char *buffer_end;
2865 if (SvOBJECT(referent)) {
2866 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2869 stashname = HEK_KEY(name);
2870 stashnamelen = HEK_LEN(name);
2872 if (HEK_UTF8(name)) {
2878 stashname = "__ANON__";
2881 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2882 + 2 * sizeof(UV) + 2 /* )\0 */;
2884 len = typelen + 3 /* (0x */
2885 + 2 * sizeof(UV) + 2 /* )\0 */;
2888 Newx(buffer, len, char);
2889 buffer_end = retval = buffer + len;
2891 /* Working backwards */
2895 *--retval = PL_hexdigit[addr & 15];
2896 } while (addr >>= 4);
2902 memcpy(retval, typestr, typelen);
2906 retval -= stashnamelen;
2907 memcpy(retval, stashname, stashnamelen);
2909 /* retval may not neccesarily have reached the start of the
2911 assert (retval >= buffer);
2913 len = buffer_end - retval - 1; /* -1 for that \0 */
2921 if (SvREADONLY(sv) && !SvOK(sv)) {
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2926 if (ckWARN(WARN_UNINITIALIZED))
2931 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2932 /* I'm assuming that if both IV and NV are equally valid then
2933 converting the IV is going to be more efficient */
2934 const U32 isUIOK = SvIsUV(sv);
2935 char buf[TYPE_CHARS(UV)];
2939 if (SvTYPE(sv) < SVt_PVIV)
2940 sv_upgrade(sv, SVt_PVIV);
2941 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2943 /* inlined from sv_setpvn */
2944 s = SvGROW_mutable(sv, len + 1);
2945 Move(ptr, s, len, char);
2949 else if (SvNOKp(sv)) {
2951 if (SvTYPE(sv) < SVt_PVNV)
2952 sv_upgrade(sv, SVt_PVNV);
2953 /* The +20 is pure guesswork. Configure test needed. --jhi */
2954 s = SvGROW_mutable(sv, NV_DIG + 20);
2955 /* some Xenix systems wipe out errno here */
2957 if (SvNVX(sv) == 0.0)
2958 my_strlcpy(s, "0", SvLEN(sv));
2962 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2965 #ifdef FIXNEGATIVEZERO
2966 if (*s == '-' && s[1] == '0' && !s[2]) {
2978 if (isGV_with_GP(sv)) {
2979 GV *const gv = MUTABLE_GV(sv);
2980 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2981 SV *const buffer = sv_newmortal();
2983 /* FAKE globs can get coerced, so need to turn this off temporarily
2986 gv_efullname3(buffer, gv, "*");
2987 SvFLAGS(gv) |= wasfake;
2989 assert(SvPOK(buffer));
2991 *lp = SvCUR(buffer);
2993 return SvPVX(buffer);
2998 if (flags & SV_UNDEF_RETURNS_NULL)
3000 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3002 if (SvTYPE(sv) < SVt_PV)
3003 /* Typically the caller expects that sv_any is not NULL now. */
3004 sv_upgrade(sv, SVt_PV);
3008 const STRLEN len = s - SvPVX_const(sv);
3014 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3015 PTR2UV(sv),SvPVX_const(sv)));
3016 if (flags & SV_CONST_RETURN)
3017 return (char *)SvPVX_const(sv);
3018 if (flags & SV_MUTABLE_RETURN)
3019 return SvPVX_mutable(sv);
3024 =for apidoc sv_copypv
3026 Copies a stringified representation of the source SV into the
3027 destination SV. Automatically performs any necessary mg_get and
3028 coercion of numeric values into strings. Guaranteed to preserve
3029 UTF8 flag even from overloaded objects. Similar in nature to
3030 sv_2pv[_flags] but operates directly on an SV instead of just the
3031 string. Mostly uses sv_2pv_flags to do its work, except when that
3032 would lose the UTF-8'ness of the PV.
3038 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3041 const char * const s = SvPV_const(ssv,len);
3043 PERL_ARGS_ASSERT_SV_COPYPV;
3045 sv_setpvn(dsv,s,len);
3053 =for apidoc sv_2pvbyte
3055 Return a pointer to the byte-encoded representation of the SV, and set *lp
3056 to its length. May cause the SV to be downgraded from UTF-8 as a
3059 Usually accessed via the C<SvPVbyte> macro.
3065 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3067 PERL_ARGS_ASSERT_SV_2PVBYTE;
3069 sv_utf8_downgrade(sv,0);
3070 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3074 =for apidoc sv_2pvutf8
3076 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3077 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3079 Usually accessed via the C<SvPVutf8> macro.
3085 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3087 PERL_ARGS_ASSERT_SV_2PVUTF8;
3089 sv_utf8_upgrade(sv);
3090 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3095 =for apidoc sv_2bool
3097 This function is only called on magical items, and is only used by
3098 sv_true() or its macro equivalent.
3104 Perl_sv_2bool(pTHX_ register SV *const sv)
3108 PERL_ARGS_ASSERT_SV_2BOOL;
3116 SV * const tmpsv = AMG_CALLun(sv,bool_);
3117 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3118 return (bool)SvTRUE(tmpsv);
3120 return SvRV(sv) != 0;
3123 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3125 (*sv->sv_u.svu_pv > '0' ||
3126 Xpvtmp->xpv_cur > 1 ||
3127 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3134 return SvIVX(sv) != 0;
3137 return SvNVX(sv) != 0.0;
3139 if (isGV_with_GP(sv))
3149 =for apidoc sv_utf8_upgrade
3151 Converts the PV of an SV to its UTF-8-encoded form.
3152 Forces the SV to string form if it is not already.
3153 Will C<mg_get> on C<sv> if appropriate.
3154 Always sets the SvUTF8 flag to avoid future validity checks even
3155 if the whole string is the same in UTF-8 as not.
3156 Returns the number of bytes in the converted string
3158 This is not as a general purpose byte encoding to Unicode interface:
3159 use the Encode extension for that.
3161 =for apidoc sv_utf8_upgrade_nomg
3163 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3165 =for apidoc sv_utf8_upgrade_flags
3167 Converts the PV of an SV to its UTF-8-encoded form.
3168 Forces the SV to string form if it is not already.
3169 Always sets the SvUTF8 flag to avoid future validity checks even
3170 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3171 will C<mg_get> on C<sv> if appropriate, else not.
3172 Returns the number of bytes in the converted string
3173 C<sv_utf8_upgrade> and
3174 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3176 This is not as a general purpose byte encoding to Unicode interface:
3177 use the Encode extension for that.
3181 The grow version is currently not externally documented. It adds a parameter,
3182 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3183 have free after it upon return. This allows the caller to reserve extra space
3184 that it intends to fill, to avoid extra grows.
3186 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3187 which can be used to tell this function to not first check to see if there are
3188 any characters that are different in UTF-8 (variant characters) which would
3189 force it to allocate a new string to sv, but to assume there are. Typically
3190 this flag is used by a routine that has already parsed the string to find that
3191 there are such characters, and passes this information on so that the work
3192 doesn't have to be repeated.
3194 (One might think that the calling routine could pass in the position of the
3195 first such variant, so it wouldn't have to be found again. But that is not the
3196 case, because typically when the caller is likely to use this flag, it won't be
3197 calling this routine unless it finds something that won't fit into a byte.
3198 Otherwise it tries to not upgrade and just use bytes. But some things that
3199 do fit into a byte are variants in utf8, and the caller may not have been
3200 keeping track of these.)
3202 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3203 isn't guaranteed due to having other routines do the work in some input cases,
3204 or if the input is already flagged as being in utf8.
3206 The speed of this could perhaps be improved for many cases if someone wanted to
3207 write a fast function that counts the number of variant characters in a string,
3208 especially if it could return the position of the first one.
3213 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3217 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3219 if (sv == &PL_sv_undef)
3223 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3224 (void) sv_2pv_flags(sv,&len, flags);
3226 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3230 (void) SvPV_force(sv,len);
3235 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3240 sv_force_normal_flags(sv, 0);
3243 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3244 sv_recode_to_utf8(sv, PL_encoding);
3245 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3249 if (SvCUR(sv) == 0) {
3250 if (extra) SvGROW(sv, extra);
3251 } else { /* Assume Latin-1/EBCDIC */
3252 /* This function could be much more efficient if we
3253 * had a FLAG in SVs to signal if there are any variant
3254 * chars in the PV. Given that there isn't such a flag
3255 * make the loop as fast as possible (although there are certainly ways
3256 * to speed this up, eg. through vectorization) */
3257 U8 * s = (U8 *) SvPVX_const(sv);
3258 U8 * e = (U8 *) SvEND(sv);
3260 STRLEN two_byte_count = 0;
3262 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3264 /* See if really will need to convert to utf8. We mustn't rely on our
3265 * incoming SV being well formed and having a trailing '\0', as certain
3266 * code in pp_formline can send us partially built SVs. */
3270 if (NATIVE_IS_INVARIANT(ch)) continue;
3272 t--; /* t already incremented; re-point to first variant */
3277 /* utf8 conversion not needed because all are invariants. Mark as
3278 * UTF-8 even if no variant - saves scanning loop */
3284 /* Here, the string should be converted to utf8, either because of an
3285 * input flag (two_byte_count = 0), or because a character that
3286 * requires 2 bytes was found (two_byte_count = 1). t points either to
3287 * the beginning of the string (if we didn't examine anything), or to
3288 * the first variant. In either case, everything from s to t - 1 will
3289 * occupy only 1 byte each on output.
3291 * There are two main ways to convert. One is to create a new string
3292 * and go through the input starting from the beginning, appending each
3293 * converted value onto the new string as we go along. It's probably
3294 * best to allocate enough space in the string for the worst possible
3295 * case rather than possibly running out of space and having to
3296 * reallocate and then copy what we've done so far. Since everything
3297 * from s to t - 1 is invariant, the destination can be initialized
3298 * with these using a fast memory copy
3300 * The other way is to figure out exactly how big the string should be
3301 * by parsing the entire input. Then you don't have to make it big
3302 * enough to handle the worst possible case, and more importantly, if
3303 * the string you already have is large enough, you don't have to
3304 * allocate a new string, you can copy the last character in the input
3305 * string to the final position(s) that will be occupied by the
3306 * converted string and go backwards, stopping at t, since everything
3307 * before that is invariant.
3309 * There are advantages and disadvantages to each method.
3311 * In the first method, we can allocate a new string, do the memory
3312 * copy from the s to t - 1, and then proceed through the rest of the
3313 * string byte-by-byte.
3315 * In the second method, we proceed through the rest of the input
3316 * string just calculating how big the converted string will be. Then
3317 * there are two cases:
3318 * 1) if the string has enough extra space to handle the converted
3319 * value. We go backwards through the string, converting until we
3320 * get to the position we are at now, and then stop. If this
3321 * position is far enough along in the string, this method is
3322 * faster than the other method. If the memory copy were the same
3323 * speed as the byte-by-byte loop, that position would be about
3324 * half-way, as at the half-way mark, parsing to the end and back
3325 * is one complete string's parse, the same amount as starting
3326 * over and going all the way through. Actually, it would be
3327 * somewhat less than half-way, as it's faster to just count bytes
3328 * than to also copy, and we don't have the overhead of allocating
3329 * a new string, changing the scalar to use it, and freeing the
3330 * existing one. But if the memory copy is fast, the break-even
3331 * point is somewhere after half way. The counting loop could be
3332 * sped up by vectorization, etc, to move the break-even point
3333 * further towards the beginning.
3334 * 2) if the string doesn't have enough space to handle the converted
3335 * value. A new string will have to be allocated, and one might
3336 * as well, given that, start from the beginning doing the first
3337 * method. We've spent extra time parsing the string and in
3338 * exchange all we've gotten is that we know precisely how big to
3339 * make the new one. Perl is more optimized for time than space,
3340 * so this case is a loser.
3341 * So what I've decided to do is not use the 2nd method unless it is
3342 * guaranteed that a new string won't have to be allocated, assuming
3343 * the worst case. I also decided not to put any more conditions on it
3344 * than this, for now. It seems likely that, since the worst case is
3345 * twice as big as the unknown portion of the string (plus 1), we won't
3346 * be guaranteed enough space, causing us to go to the first method,
3347 * unless the string is short, or the first variant character is near
3348 * the end of it. In either of these cases, it seems best to use the
3349 * 2nd method. The only circumstance I can think of where this would
3350 * be really slower is if the string had once had much more data in it
3351 * than it does now, but there is still a substantial amount in it */
3354 STRLEN invariant_head = t - s;
3355 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3356 if (SvLEN(sv) < size) {
3358 /* Here, have decided to allocate a new string */
3363 Newx(dst, size, U8);
3365 /* If no known invariants at the beginning of the input string,
3366 * set so starts from there. Otherwise, can use memory copy to
3367 * get up to where we are now, and then start from here */
3369 if (invariant_head <= 0) {
3372 Copy(s, dst, invariant_head, char);
3373 d = dst + invariant_head;
3377 const UV uv = NATIVE8_TO_UNI(*t++);
3378 if (UNI_IS_INVARIANT(uv))
3379 *d++ = (U8)UNI_TO_NATIVE(uv);
3381 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3382 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3386 SvPV_free(sv); /* No longer using pre-existing string */
3387 SvPV_set(sv, (char*)dst);
3388 SvCUR_set(sv, d - dst);
3389 SvLEN_set(sv, size);
3392 /* Here, have decided to get the exact size of the string.
3393 * Currently this happens only when we know that there is
3394 * guaranteed enough space to fit the converted string, so
3395 * don't have to worry about growing. If two_byte_count is 0,
3396 * then t points to the first byte of the string which hasn't
3397 * been examined yet. Otherwise two_byte_count is 1, and t
3398 * points to the first byte in the string that will expand to
3399 * two. Depending on this, start examining at t or 1 after t.
3402 U8 *d = t + two_byte_count;
3405 /* Count up the remaining bytes that expand to two */
3408 const U8 chr = *d++;
3409 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3412 /* The string will expand by just the number of bytes that
3413 * occupy two positions. But we are one afterwards because of
3414 * the increment just above. This is the place to put the
3415 * trailing NUL, and to set the length before we decrement */
3417 d += two_byte_count;
3418 SvCUR_set(sv, d - s);
3422 /* Having decremented d, it points to the position to put the
3423 * very last byte of the expanded string. Go backwards through
3424 * the string, copying and expanding as we go, stopping when we
3425 * get to the part that is invariant the rest of the way down */
3429 const U8 ch = NATIVE8_TO_UNI(*e--);
3430 if (UNI_IS_INVARIANT(ch)) {
3431 *d-- = UNI_TO_NATIVE(ch);
3433 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3434 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3441 /* Mark as UTF-8 even if no variant - saves scanning loop */
3447 =for apidoc sv_utf8_downgrade
3449 Attempts to convert the PV of an SV from characters to bytes.
3450 If the PV contains a character that cannot fit
3451 in a byte, this conversion will fail;
3452 in this case, either returns false or, if C<fail_ok> is not
3455 This is not as a general purpose Unicode to byte encoding interface:
3456 use the Encode extension for that.
3462 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3466 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3468 if (SvPOKp(sv) && SvUTF8(sv)) {
3474 sv_force_normal_flags(sv, 0);
3476 s = (U8 *) SvPV(sv, len);
3477 if (!utf8_to_bytes(s, &len)) {
3482 Perl_croak(aTHX_ "Wide character in %s",
3485 Perl_croak(aTHX_ "Wide character");
3496 =for apidoc sv_utf8_encode
3498 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3499 flag off so that it looks like octets again.
3505 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3507 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3510 sv_force_normal_flags(sv, 0);
3512 if (SvREADONLY(sv)) {
3513 Perl_croak(aTHX_ "%s", PL_no_modify);
3515 (void) sv_utf8_upgrade(sv);
3520 =for apidoc sv_utf8_decode
3522 If the PV of the SV is an octet sequence in UTF-8
3523 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3524 so that it looks like a character. If the PV contains only single-byte
3525 characters, the C<SvUTF8> flag stays being off.
3526 Scans PV for validity and returns false if the PV is invalid UTF-8.
3532 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3534 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3540 /* The octets may have got themselves encoded - get them back as
3543 if (!sv_utf8_downgrade(sv, TRUE))
3546 /* it is actually just a matter of turning the utf8 flag on, but
3547 * we want to make sure everything inside is valid utf8 first.
3549 c = (const U8 *) SvPVX_const(sv);
3550 if (!is_utf8_string(c, SvCUR(sv)+1))
3552 e = (const U8 *) SvEND(sv);
3555 if (!UTF8_IS_INVARIANT(ch)) {
3565 =for apidoc sv_setsv
3567 Copies the contents of the source SV C<ssv> into the destination SV
3568 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3569 function if the source SV needs to be reused. Does not handle 'set' magic.
3570 Loosely speaking, it performs a copy-by-value, obliterating any previous
3571 content of the destination.
3573 You probably want to use one of the assortment of wrappers, such as
3574 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3575 C<SvSetMagicSV_nosteal>.
3577 =for apidoc sv_setsv_flags
3579 Copies the contents of the source SV C<ssv> into the destination SV
3580 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3581 function if the source SV needs to be reused. Does not handle 'set' magic.
3582 Loosely speaking, it performs a copy-by-value, obliterating any previous
3583 content of the destination.
3584 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3585 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3586 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3587 and C<sv_setsv_nomg> are implemented in terms of this function.
3589 You probably want to use one of the assortment of wrappers, such as
3590 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3591 C<SvSetMagicSV_nosteal>.
3593 This is the primary function for copying scalars, and most other
3594 copy-ish functions and macros use this underneath.
3600 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3602 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3604 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3606 if (dtype != SVt_PVGV) {
3607 const char * const name = GvNAME(sstr);
3608 const STRLEN len = GvNAMELEN(sstr);
3610 if (dtype >= SVt_PV) {
3616 SvUPGRADE(dstr, SVt_PVGV);
3617 (void)SvOK_off(dstr);
3618 /* FIXME - why are we doing this, then turning it off and on again
3620 isGV_with_GP_on(dstr);
3622 GvSTASH(dstr) = GvSTASH(sstr);
3624 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3625 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3626 SvFAKE_on(dstr); /* can coerce to non-glob */
3629 if(GvGP(MUTABLE_GV(sstr))) {
3630 /* If source has method cache entry, clear it */
3632 SvREFCNT_dec(GvCV(sstr));
3636 /* If source has a real method, then a method is
3638 else if(GvCV((const GV *)sstr)) {
3643 /* If dest already had a real method, that's a change as well */
3644 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3648 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3651 gp_free(MUTABLE_GV(dstr));
3652 isGV_with_GP_off(dstr);
3653 (void)SvOK_off(dstr);
3654 isGV_with_GP_on(dstr);
3655 GvINTRO_off(dstr); /* one-shot flag */
3656 GvGP(dstr) = gp_ref(GvGP(sstr));
3657 if (SvTAINTED(sstr))
3659 if (GvIMPORTED(dstr) != GVf_IMPORTED
3660 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3662 GvIMPORTED_on(dstr);
3665 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3666 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3671 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3673 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3675 const int intro = GvINTRO(dstr);
3678 const U32 stype = SvTYPE(sref);
3679 bool mro_changes = FALSE;
3681 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3684 GvINTRO_off(dstr); /* one-shot flag */
3685 GvLINE(dstr) = CopLINE(PL_curcop);
3686 GvEGV(dstr) = MUTABLE_GV(dstr);
3691 location = (SV **) &GvCV(dstr);
3692 import_flag = GVf_IMPORTED_CV;
3695 location = (SV **) &GvHV(dstr);
3696 import_flag = GVf_IMPORTED_HV;
3699 location = (SV **) &GvAV(dstr);
3700 if (strEQ(GvNAME((GV*)dstr), "ISA"))
3702 import_flag = GVf_IMPORTED_AV;
3705 location = (SV **) &GvIOp(dstr);
3708 location = (SV **) &GvFORM(dstr);
3711 location = &GvSV(dstr);
3712 import_flag = GVf_IMPORTED_SV;
3715 if (stype == SVt_PVCV) {
3716 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3717 if (GvCVGEN(dstr)) {
3718 SvREFCNT_dec(GvCV(dstr));
3720 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3723 SAVEGENERICSV(*location);
3727 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3728 CV* const cv = MUTABLE_CV(*location);
3730 if (!GvCVGEN((const GV *)dstr) &&
3731 (CvROOT(cv) || CvXSUB(cv)))
3733 /* Redefining a sub - warning is mandatory if
3734 it was a const and its value changed. */
3735 if (CvCONST(cv) && CvCONST((const CV *)sref)
3737 == cv_const_sv((const CV *)sref)) {
3739 /* They are 2 constant subroutines generated from
3740 the same constant. This probably means that
3741 they are really the "same" proxy subroutine
3742 instantiated in 2 places. Most likely this is
3743 when a constant is exported twice. Don't warn.
3746 else if (ckWARN(WARN_REDEFINE)
3748 && (!CvCONST((const CV *)sref)
3749 || sv_cmp(cv_const_sv(cv),
3750 cv_const_sv((const CV *)
3752 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3755 ? "Constant subroutine %s::%s redefined"
3756 : "Subroutine %s::%s redefined"),
3757 HvNAME_get(GvSTASH((const GV *)dstr)),
3758 GvENAME(MUTABLE_GV(dstr)));
3762 cv_ckproto_len(cv, (const GV *)dstr,
3763 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3764 SvPOK(sref) ? SvCUR(sref) : 0);
3766 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3767 GvASSUMECV_on(dstr);
3768 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3771 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3772 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3773 GvFLAGS(dstr) |= import_flag;
3778 if (SvTAINTED(sstr))
3780 if (mro_changes) mro_isa_changed_in(GvSTASH(dstr));
3785 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3788 register U32 sflags;
3790 register svtype stype;
3792 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3797 if (SvIS_FREED(dstr)) {
3798 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3799 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3801 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3803 sstr = &PL_sv_undef;
3804 if (SvIS_FREED(sstr)) {
3805 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3806 (void*)sstr, (void*)dstr);
3808 stype = SvTYPE(sstr);
3809 dtype = SvTYPE(dstr);
3811 (void)SvAMAGIC_off(dstr);
3814 /* need to nuke the magic */
3818 /* There's a lot of redundancy below but we're going for speed here */
3823 if (dtype != SVt_PVGV) {
3824 (void)SvOK_off(dstr);
3832 sv_upgrade(dstr, SVt_IV);
3836 sv_upgrade(dstr, SVt_PVIV);
3839 goto end_of_first_switch;
3841 (void)SvIOK_only(dstr);
3842 SvIV_set(dstr, SvIVX(sstr));
3845 /* SvTAINTED can only be true if the SV has taint magic, which in
3846 turn means that the SV type is PVMG (or greater). This is the
3847 case statement for SVt_IV, so this cannot be true (whatever gcov
3849 assert(!SvTAINTED(sstr));
3854 if (dtype < SVt_PV && dtype != SVt_IV)
3855 sv_upgrade(dstr, SVt_IV);
3863 sv_upgrade(dstr, SVt_NV);
3867 sv_upgrade(dstr, SVt_PVNV);
3870 goto end_of_first_switch;
3872 SvNV_set(dstr, SvNVX(sstr));
3873 (void)SvNOK_only(dstr);
3874 /* SvTAINTED can only be true if the SV has taint magic, which in
3875 turn means that the SV type is PVMG (or greater). This is the
3876 case statement for SVt_NV, so this cannot be true (whatever gcov
3878 assert(!SvTAINTED(sstr));
3884 #ifdef PERL_OLD_COPY_ON_WRITE
3885 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3886 if (dtype < SVt_PVIV)
3887 sv_upgrade(dstr, SVt_PVIV);
3894 sv_upgrade(dstr, SVt_PV);
3897 if (dtype < SVt_PVIV)
3898 sv_upgrade(dstr, SVt_PVIV);
3901 if (dtype < SVt_PVNV)
3902 sv_upgrade(dstr, SVt_PVNV);
3906 const char * const type = sv_reftype(sstr,0);
3908 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3910 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3915 if (dtype < SVt_REGEXP)
3916 sv_upgrade(dstr, SVt_REGEXP);
3919 /* case SVt_BIND: */
3922 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3923 glob_assign_glob(dstr, sstr, dtype);
3926 /* SvVALID means that this PVGV is playing at being an FBM. */
3930 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3932 if (SvTYPE(sstr) != stype) {
3933 stype = SvTYPE(sstr);
3934 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3935 glob_assign_glob(dstr, sstr, dtype);
3940 if (stype == SVt_PVLV)
3941 SvUPGRADE(dstr, SVt_PVNV);
3943 SvUPGRADE(dstr, (svtype)stype);
3945 end_of_first_switch:
3947 /* dstr may have been upgraded. */
3948 dtype = SvTYPE(dstr);
3949 sflags = SvFLAGS(sstr);
3951 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3952 /* Assigning to a subroutine sets the prototype. */
3955 const char *const ptr = SvPV_const(sstr, len);
3957 SvGROW(dstr, len + 1);
3958 Copy(ptr, SvPVX(dstr), len + 1, char);
3959 SvCUR_set(dstr, len);
3961 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3965 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3966 const char * const type = sv_reftype(dstr,0);
3968 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3970 Perl_croak(aTHX_ "Cannot copy to %s", type);
3971 } else if (sflags & SVf_ROK) {
3972 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3973 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3976 if (GvIMPORTED(dstr) != GVf_IMPORTED
3977 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3979 GvIMPORTED_on(dstr);
3984 glob_assign_glob(dstr, sstr, dtype);
3988 if (dtype >= SVt_PV) {
3989 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3990 glob_assign_ref(dstr, sstr);
3993 if (SvPVX_const(dstr)) {
3999 (void)SvOK_off(dstr);
4000 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4001 SvFLAGS(dstr) |= sflags & SVf_ROK;
4002 assert(!(sflags & SVp_NOK));
4003 assert(!(sflags & SVp_IOK));
4004 assert(!(sflags & SVf_NOK));
4005 assert(!(sflags & SVf_IOK));
4007 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4008 if (!(sflags & SVf_OK)) {
4009 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4010 "Undefined value assigned to typeglob");
4013 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4014 if (dstr != (const SV *)gv) {
4016 gp_free(MUTABLE_GV(dstr));
4017 GvGP(dstr) = gp_ref(GvGP(gv));
4021 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4022 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4024 else if (sflags & SVp_POK) {
4028 * Check to see if we can just swipe the string. If so, it's a
4029 * possible small lose on short strings, but a big win on long ones.
4030 * It might even be a win on short strings if SvPVX_const(dstr)
4031 * has to be allocated and SvPVX_const(sstr) has to be freed.
4032 * Likewise if we can set up COW rather than doing an actual copy, we
4033 * drop to the else clause, as the swipe code and the COW setup code
4034 * have much in common.
4037 /* Whichever path we take through the next code, we want this true,
4038 and doing it now facilitates the COW check. */
4039 (void)SvPOK_only(dstr);
4042 /* If we're already COW then this clause is not true, and if COW
4043 is allowed then we drop down to the else and make dest COW
4044 with us. If caller hasn't said that we're allowed to COW
4045 shared hash keys then we don't do the COW setup, even if the
4046 source scalar is a shared hash key scalar. */
4047 (((flags & SV_COW_SHARED_HASH_KEYS)
4048 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4049 : 1 /* If making a COW copy is forbidden then the behaviour we
4050 desire is as if the source SV isn't actually already
4051 COW, even if it is. So we act as if the source flags
4052 are not COW, rather than actually testing them. */
4054 #ifndef PERL_OLD_COPY_ON_WRITE
4055 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4056 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4057 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4058 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4059 but in turn, it's somewhat dead code, never expected to go
4060 live, but more kept as a placeholder on how to do it better
4061 in a newer implementation. */
4062 /* If we are COW and dstr is a suitable target then we drop down
4063 into the else and make dest a COW of us. */
4064 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4069 (sflags & SVs_TEMP) && /* slated for free anyway? */
4070 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4071 (!(flags & SV_NOSTEAL)) &&
4072 /* and we're allowed to steal temps */
4073 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4074 SvLEN(sstr) && /* and really is a string */
4075 /* and won't be needed again, potentially */
4076 !(PL_op && PL_op->op_type == OP_AASSIGN))
4077 #ifdef PERL_OLD_COPY_ON_WRITE
4078 && ((flags & SV_COW_SHARED_HASH_KEYS)
4079 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4080 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4081 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4085 /* Failed the swipe test, and it's not a shared hash key either.
4086 Have to copy the string. */
4087 STRLEN len = SvCUR(sstr);
4088 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4089 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4090 SvCUR_set(dstr, len);
4091 *SvEND(dstr) = '\0';
4093 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4095 /* Either it's a shared hash key, or it's suitable for
4096 copy-on-write or we can swipe the string. */
4098 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4102 #ifdef PERL_OLD_COPY_ON_WRITE
4104 if ((sflags & (SVf_FAKE | SVf_READONLY))
4105 != (SVf_FAKE | SVf_READONLY)) {
4106 SvREADONLY_on(sstr);
4108 /* Make the source SV into a loop of 1.
4109 (about to become 2) */
4110 SV_COW_NEXT_SV_SET(sstr, sstr);
4114 /* Initial code is common. */
4115 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4120 /* making another shared SV. */
4121 STRLEN cur = SvCUR(sstr);
4122 STRLEN len = SvLEN(sstr);
4123 #ifdef PERL_OLD_COPY_ON_WRITE
4125 assert (SvTYPE(dstr) >= SVt_PVIV);
4126 /* SvIsCOW_normal */
4127 /* splice us in between source and next-after-source. */
4128 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4129 SV_COW_NEXT_SV_SET(sstr, dstr);
4130 SvPV_set(dstr, SvPVX_mutable(sstr));
4134 /* SvIsCOW_shared_hash */
4135 DEBUG_C(PerlIO_printf(Perl_debug_log,
4136 "Copy on write: Sharing hash\n"));
4138 assert (SvTYPE(dstr) >= SVt_PV);
4140 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4142 SvLEN_set(dstr, len);
4143 SvCUR_set(dstr, cur);
4144 SvREADONLY_on(dstr);
4148 { /* Passes the swipe test. */
4149 SvPV_set(dstr, SvPVX_mutable(sstr));
4150 SvLEN_set(dstr, SvLEN(sstr));
4151 SvCUR_set(dstr, SvCUR(sstr));
4154 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4155 SvPV_set(sstr, NULL);
4161 if (sflags & SVp_NOK) {
4162 SvNV_set(dstr, SvNVX(sstr));
4164 if (sflags & SVp_IOK) {
4165 SvIV_set(dstr, SvIVX(sstr));
4166 /* Must do this otherwise some other overloaded use of 0x80000000
4167 gets confused. I guess SVpbm_VALID */
4168 if (sflags & SVf_IVisUV)
4171 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4173 const MAGIC * const smg = SvVSTRING_mg(sstr);
4175 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4176 smg->mg_ptr, smg->mg_len);
4177 SvRMAGICAL_on(dstr);
4181 else if (sflags & (SVp_IOK|SVp_NOK)) {
4182 (void)SvOK_off(dstr);
4183 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4184 if (sflags & SVp_IOK) {
4185 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4186 SvIV_set(dstr, SvIVX(sstr));
4188 if (sflags & SVp_NOK) {
4189 SvNV_set(dstr, SvNVX(sstr));
4193 if (isGV_with_GP(sstr)) {
4194 /* This stringification rule for globs is spread in 3 places.
4195 This feels bad. FIXME. */
4196 const U32 wasfake = sflags & SVf_FAKE;
4198 /* FAKE globs can get coerced, so need to turn this off
4199 temporarily if it is on. */
4201 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4202 SvFLAGS(sstr) |= wasfake;
4205 (void)SvOK_off(dstr);
4207 if (SvTAINTED(sstr))
4212 =for apidoc sv_setsv_mg
4214 Like C<sv_setsv>, but also handles 'set' magic.
4220 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4222 PERL_ARGS_ASSERT_SV_SETSV_MG;
4224 sv_setsv(dstr,sstr);
4228 #ifdef PERL_OLD_COPY_ON_WRITE
4230 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4232 STRLEN cur = SvCUR(sstr);
4233 STRLEN len = SvLEN(sstr);
4234 register char *new_pv;
4236 PERL_ARGS_ASSERT_SV_SETSV_COW;
4239 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4240 (void*)sstr, (void*)dstr);
4247 if (SvTHINKFIRST(dstr))
4248 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4249 else if (SvPVX_const(dstr))
4250 Safefree(SvPVX_const(dstr));
4254 SvUPGRADE(dstr, SVt_PVIV);
4256 assert (SvPOK(sstr));
4257 assert (SvPOKp(sstr));
4258 assert (!SvIOK(sstr));
4259 assert (!SvIOKp(sstr));
4260 assert (!SvNOK(sstr));
4261 assert (!SvNOKp(sstr));
4263 if (SvIsCOW(sstr)) {
4265 if (SvLEN(sstr) == 0) {
4266 /* source is a COW shared hash key. */
4267 DEBUG_C(PerlIO_printf(Perl_debug_log,
4268 "Fast copy on write: Sharing hash\n"));
4269 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4272 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4274 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4275 SvUPGRADE(sstr, SVt_PVIV);
4276 SvREADONLY_on(sstr);
4278 DEBUG_C(PerlIO_printf(Perl_debug_log,
4279 "Fast copy on write: Converting sstr to COW\n"));
4280 SV_COW_NEXT_SV_SET(dstr, sstr);
4282 SV_COW_NEXT_SV_SET(sstr, dstr);
4283 new_pv = SvPVX_mutable(sstr);
4286 SvPV_set(dstr, new_pv);
4287 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4290 SvLEN_set(dstr, len);
4291 SvCUR_set(dstr, cur);
4300 =for apidoc sv_setpvn
4302 Copies a string into an SV. The C<len> parameter indicates the number of
4303 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4304 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4310 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4313 register char *dptr;
4315 PERL_ARGS_ASSERT_SV_SETPVN;
4317 SV_CHECK_THINKFIRST_COW_DROP(sv);
4323 /* len is STRLEN which is unsigned, need to copy to signed */
4326 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4328 SvUPGRADE(sv, SVt_PV);
4330 dptr = SvGROW(sv, len + 1);
4331 Move(ptr,dptr,len,char);
4334 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4339 =for apidoc sv_setpvn_mg
4341 Like C<sv_setpvn>, but also handles 'set' magic.
4347 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4349 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4351 sv_setpvn(sv,ptr,len);
4356 =for apidoc sv_setpv
4358 Copies a string into an SV. The string must be null-terminated. Does not
4359 handle 'set' magic. See C<sv_setpv_mg>.
4365 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4368 register STRLEN len;
4370 PERL_ARGS_ASSERT_SV_SETPV;
4372 SV_CHECK_THINKFIRST_COW_DROP(sv);
4378 SvUPGRADE(sv, SVt_PV);
4380 SvGROW(sv, len + 1);
4381 Move(ptr,SvPVX(sv),len+1,char);
4383 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4388 =for apidoc sv_setpv_mg
4390 Like C<sv_setpv>, but also handles 'set' magic.
4396 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4398 PERL_ARGS_ASSERT_SV_SETPV_MG;
4405 =for apidoc sv_usepvn_flags
4407 Tells an SV to use C<ptr> to find its string value. Normally the
4408 string is stored inside the SV but sv_usepvn allows the SV to use an
4409 outside string. The C<ptr> should point to memory that was allocated
4410 by C<malloc>. The string length, C<len>, must be supplied. By default
4411 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4412 so that pointer should not be freed or used by the programmer after
4413 giving it to sv_usepvn, and neither should any pointers from "behind"
4414 that pointer (e.g. ptr + 1) be used.
4416 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4417 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4418 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4419 C<len>, and already meets the requirements for storing in C<SvPVX>)
4425 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4430 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4432 SV_CHECK_THINKFIRST_COW_DROP(sv);
4433 SvUPGRADE(sv, SVt_PV);
4436 if (flags & SV_SMAGIC)
4440 if (SvPVX_const(sv))
4444 if (flags & SV_HAS_TRAILING_NUL)
4445 assert(ptr[len] == '\0');
4448 allocate = (flags & SV_HAS_TRAILING_NUL)
4450 #ifdef Perl_safesysmalloc_size
4453 PERL_STRLEN_ROUNDUP(len + 1);
4455 if (flags & SV_HAS_TRAILING_NUL) {
4456 /* It's long enough - do nothing.
4457 Specfically Perl_newCONSTSUB is relying on this. */
4460 /* Force a move to shake out bugs in callers. */
4461 char *new_ptr = (char*)safemalloc(allocate);
4462 Copy(ptr, new_ptr, len, char);
4463 PoisonFree(ptr,len,char);
4467 ptr = (char*) saferealloc (ptr, allocate);
4470 #ifdef Perl_safesysmalloc_size
4471 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4473 SvLEN_set(sv, allocate);
4477 if (!(flags & SV_HAS_TRAILING_NUL)) {
4480 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4482 if (flags & SV_SMAGIC)
4486 #ifdef PERL_OLD_COPY_ON_WRITE
4487 /* Need to do this *after* making the SV normal, as we need the buffer
4488 pointer to remain valid until after we've copied it. If we let go too early,
4489 another thread could invalidate it by unsharing last of the same hash key
4490 (which it can do by means other than releasing copy-on-write Svs)
4491 or by changing the other copy-on-write SVs in the loop. */
4493 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4495 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4497 { /* this SV was SvIsCOW_normal(sv) */
4498 /* we need to find the SV pointing to us. */
4499 SV *current = SV_COW_NEXT_SV(after);
4501 if (current == sv) {
4502 /* The SV we point to points back to us (there were only two of us
4504 Hence other SV is no longer copy on write either. */
4506 SvREADONLY_off(after);
4508 /* We need to follow the pointers around the loop. */
4510 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4513 /* don't loop forever if the structure is bust, and we have
4514 a pointer into a closed loop. */
4515 assert (current != after);
4516 assert (SvPVX_const(current) == pvx);
4518 /* Make the SV before us point to the SV after us. */
4519 SV_COW_NEXT_SV_SET(current, after);
4525 =for apidoc sv_force_normal_flags
4527 Undo various types of fakery on an SV: if the PV is a shared string, make
4528 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4529 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4530 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4531 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4532 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4533 set to some other value.) In addition, the C<flags> parameter gets passed to
4534 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4535 with flags set to 0.
4541 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4545 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4547 #ifdef PERL_OLD_COPY_ON_WRITE
4548 if (SvREADONLY(sv)) {
4550 const char * const pvx = SvPVX_const(sv);
4551 const STRLEN len = SvLEN(sv);
4552 const STRLEN cur = SvCUR(sv);
4553 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4554 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4555 we'll fail an assertion. */
4556 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4559 PerlIO_printf(Perl_debug_log,
4560 "Copy on write: Force normal %ld\n",
4566 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4569 if (flags & SV_COW_DROP_PV) {
4570 /* OK, so we don't need to copy our buffer. */
4573 SvGROW(sv, cur + 1);
4574 Move(pvx,SvPVX(sv),cur,char);
4579 sv_release_COW(sv, pvx, next);
4581 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4587 else if (IN_PERL_RUNTIME)
4588 Perl_croak(aTHX_ "%s", PL_no_modify);
4591 if (SvREADONLY(sv)) {
4593 const char * const pvx = SvPVX_const(sv);
4594 const STRLEN len = SvCUR(sv);
4599 SvGROW(sv, len + 1);
4600 Move(pvx,SvPVX(sv),len,char);
4602 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4604 else if (IN_PERL_RUNTIME)
4605 Perl_croak(aTHX_ "%s", PL_no_modify);
4609 sv_unref_flags(sv, flags);
4610 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4617 Efficient removal of characters from the beginning of the string buffer.
4618 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4619 the string buffer. The C<ptr> becomes the first character of the adjusted
4620 string. Uses the "OOK hack".
4621 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4622 refer to the same chunk of data.
4628 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4634 const U8 *real_start;
4638 PERL_ARGS_ASSERT_SV_CHOP;
4640 if (!ptr || !SvPOKp(sv))
4642 delta = ptr - SvPVX_const(sv);
4644 /* Nothing to do. */
4647 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4648 nothing uses the value of ptr any more. */
4649 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4650 if (ptr <= SvPVX_const(sv))
4651 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4652 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4653 SV_CHECK_THINKFIRST(sv);
4654 if (delta > max_delta)
4655 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4656 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4657 SvPVX_const(sv) + max_delta);
4660 if (!SvLEN(sv)) { /* make copy of shared string */
4661 const char *pvx = SvPVX_const(sv);
4662 const STRLEN len = SvCUR(sv);
4663 SvGROW(sv, len + 1);
4664 Move(pvx,SvPVX(sv),len,char);
4667 SvFLAGS(sv) |= SVf_OOK;
4670 SvOOK_offset(sv, old_delta);
4672 SvLEN_set(sv, SvLEN(sv) - delta);
4673 SvCUR_set(sv, SvCUR(sv) - delta);
4674 SvPV_set(sv, SvPVX(sv) + delta);
4676 p = (U8 *)SvPVX_const(sv);
4681 real_start = p - delta;
4685 if (delta < 0x100) {
4689 p -= sizeof(STRLEN);
4690 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4694 /* Fill the preceding buffer with sentinals to verify that no-one is
4696 while (p > real_start) {
4704 =for apidoc sv_catpvn
4706 Concatenates the string onto the end of the string which is in the SV. The
4707 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4708 status set, then the bytes appended should be valid UTF-8.
4709 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4711 =for apidoc sv_catpvn_flags
4713 Concatenates the string onto the end of the string which is in the SV. The
4714 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4715 status set, then the bytes appended should be valid UTF-8.
4716 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4717 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4718 in terms of this function.
4724 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4728 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4730 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4732 SvGROW(dsv, dlen + slen + 1);
4734 sstr = SvPVX_const(dsv);
4735 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4736 SvCUR_set(dsv, SvCUR(dsv) + slen);
4738 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4740 if (flags & SV_SMAGIC)
4745 =for apidoc sv_catsv
4747 Concatenates the string from SV C<ssv> onto the end of the string in
4748 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4749 not 'set' magic. See C<sv_catsv_mg>.
4751 =for apidoc sv_catsv_flags
4753 Concatenates the string from SV C<ssv> onto the end of the string in
4754 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4755 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4756 and C<sv_catsv_nomg> are implemented in terms of this function.
4761 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4765 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4769 const char *spv = SvPV_const(ssv, slen);
4771 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4772 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4773 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4774 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4775 dsv->sv_flags doesn't have that bit set.
4776 Andy Dougherty 12 Oct 2001
4778 const I32 sutf8 = DO_UTF8(ssv);
4781 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4783 dutf8 = DO_UTF8(dsv);
4785 if (dutf8 != sutf8) {
4787 /* Not modifying source SV, so taking a temporary copy. */
4788 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4790 sv_utf8_upgrade(csv);
4791 spv = SvPV_const(csv, slen);
4794 /* Leave enough space for the cat that's about to happen */
4795 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4797 sv_catpvn_nomg(dsv, spv, slen);
4800 if (flags & SV_SMAGIC)
4805 =for apidoc sv_catpv
4807 Concatenates the string onto the end of the string which is in the SV.
4808 If the SV has the UTF-8 status set, then the bytes appended should be
4809 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4814 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4817 register STRLEN len;
4821 PERL_ARGS_ASSERT_SV_CATPV;
4825 junk = SvPV_force(sv, tlen);
4827 SvGROW(sv, tlen + len + 1);
4829 ptr = SvPVX_const(sv);
4830 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4831 SvCUR_set(sv, SvCUR(sv) + len);
4832 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4837 =for apidoc sv_catpv_mg
4839 Like C<sv_catpv>, but also handles 'set' magic.
4845 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4847 PERL_ARGS_ASSERT_SV_CATPV_MG;
4856 Creates a new SV. A non-zero C<len> parameter indicates the number of
4857 bytes of preallocated string space the SV should have. An extra byte for a
4858 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4859 space is allocated.) The reference count for the new SV is set to 1.
4861 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4862 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4863 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4864 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4865 modules supporting older perls.
4871 Perl_newSV(pTHX_ const STRLEN len)
4878 sv_upgrade(sv, SVt_PV);
4879 SvGROW(sv, len + 1);
4884 =for apidoc sv_magicext
4886 Adds magic to an SV, upgrading it if necessary. Applies the
4887 supplied vtable and returns a pointer to the magic added.
4889 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4890 In particular, you can add magic to SvREADONLY SVs, and add more than
4891 one instance of the same 'how'.
4893 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4894 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4895 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4896 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4898 (This is now used as a subroutine by C<sv_magic>.)
4903 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4904 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4909 PERL_ARGS_ASSERT_SV_MAGICEXT;
4911 SvUPGRADE(sv, SVt_PVMG);
4912 Newxz(mg, 1, MAGIC);
4913 mg->mg_moremagic = SvMAGIC(sv);
4914 SvMAGIC_set(sv, mg);
4916 /* Sometimes a magic contains a reference loop, where the sv and
4917 object refer to each other. To prevent a reference loop that
4918 would prevent such objects being freed, we look for such loops
4919 and if we find one we avoid incrementing the object refcount.
4921 Note we cannot do this to avoid self-tie loops as intervening RV must
4922 have its REFCNT incremented to keep it in existence.
4925 if (!obj || obj == sv ||
4926 how == PERL_MAGIC_arylen ||
4927 how == PERL_MAGIC_symtab ||
4928 (SvTYPE(obj) == SVt_PVGV &&
4929 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4930 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4931 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4936 mg->mg_obj = SvREFCNT_inc_simple(obj);
4937 mg->mg_flags |= MGf_REFCOUNTED;
4940 /* Normal self-ties simply pass a null object, and instead of
4941 using mg_obj directly, use the SvTIED_obj macro to produce a
4942 new RV as needed. For glob "self-ties", we are tieing the PVIO
4943 with an RV obj pointing to the glob containing the PVIO. In
4944 this case, to avoid a reference loop, we need to weaken the
4948 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4949 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4955 mg->mg_len = namlen;
4958 mg->mg_ptr = savepvn(name, namlen);
4959 else if (namlen == HEf_SVKEY) {
4960 /* Yes, this is casting away const. This is only for the case of
4961 HEf_SVKEY. I think we need to document this abberation of the
4962 constness of the API, rather than making name non-const, as
4963 that change propagating outwards a long way. */
4964 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4966 mg->mg_ptr = (char *) name;
4968 mg->mg_virtual = (MGVTBL *) vtable;
4972 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4977 =for apidoc sv_magic
4979 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4980 then adds a new magic item of type C<how> to the head of the magic list.
4982 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4983 handling of the C<name> and C<namlen> arguments.
4985 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4986 to add more than one instance of the same 'how'.
4992 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4993 const char *const name, const I32 namlen)
4996 const MGVTBL *vtable;
4999 PERL_ARGS_ASSERT_SV_MAGIC;
5001 #ifdef PERL_OLD_COPY_ON_WRITE
5003 sv_force_normal_flags(sv, 0);
5005 if (SvREADONLY(sv)) {
5007 /* its okay to attach magic to shared strings; the subsequent
5008 * upgrade to PVMG will unshare the string */
5009 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5012 && how != PERL_MAGIC_regex_global
5013 && how != PERL_MAGIC_bm
5014 && how != PERL_MAGIC_fm
5015 && how != PERL_MAGIC_sv
5016 && how != PERL_MAGIC_backref
5019 Perl_croak(aTHX_ "%s", PL_no_modify);
5022 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5023 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5024 /* sv_magic() refuses to add a magic of the same 'how' as an
5027 if (how == PERL_MAGIC_taint) {
5029 /* Any scalar which already had taint magic on which someone
5030 (erroneously?) did SvIOK_on() or similar will now be
5031 incorrectly sporting public "OK" flags. */
5032 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5040 vtable = &PL_vtbl_sv;
5042 case PERL_MAGIC_overload:
5043 vtable = &PL_vtbl_amagic;
5045 case PERL_MAGIC_overload_elem:
5046 vtable = &PL_vtbl_amagicelem;
5048 case PERL_MAGIC_overload_table:
5049 vtable = &PL_vtbl_ovrld;
5052 vtable = &PL_vtbl_bm;
5054 case PERL_MAGIC_regdata:
5055 vtable = &PL_vtbl_regdata;
5057 case PERL_MAGIC_regdatum:
5058 vtable = &PL_vtbl_regdatum;
5060 case PERL_MAGIC_env:
5061 vtable = &PL_vtbl_env;
5064 vtable = &PL_vtbl_fm;
5066 case PERL_MAGIC_envelem:
5067 vtable = &PL_vtbl_envelem;
5069 case PERL_MAGIC_regex_global:
5070 vtable = &PL_vtbl_mglob;
5072 case PERL_MAGIC_isa:
5073 vtable = &PL_vtbl_isa;
5075 case PERL_MAGIC_isaelem:
5076 vtable = &PL_vtbl_isaelem;
5078 case PERL_MAGIC_nkeys:
5079 vtable = &PL_vtbl_nkeys;
5081 case PERL_MAGIC_dbfile:
5084 case PERL_MAGIC_dbline:
5085 vtable = &PL_vtbl_dbline;
5087 #ifdef USE_LOCALE_COLLATE
5088 case PERL_MAGIC_collxfrm:
5089 vtable = &PL_vtbl_collxfrm;
5091 #endif /* USE_LOCALE_COLLATE */
5092 case PERL_MAGIC_tied:
5093 vtable = &PL_vtbl_pack;
5095 case PERL_MAGIC_tiedelem:
5096 case PERL_MAGIC_tiedscalar:
5097 vtable = &PL_vtbl_packelem;
5100 vtable = &PL_vtbl_regexp;
5102 case PERL_MAGIC_sig:
5103 vtable = &PL_vtbl_sig;
5105 case PERL_MAGIC_sigelem:
5106 vtable = &PL_vtbl_sigelem;
5108 case PERL_MAGIC_taint:
5109 vtable = &PL_vtbl_taint;
5111 case PERL_MAGIC_uvar:
5112 vtable = &PL_vtbl_uvar;
5114 case PERL_MAGIC_vec:
5115 vtable = &PL_vtbl_vec;
5117 case PERL_MAGIC_arylen_p:
5118 case PERL_MAGIC_rhash:
5119 case PERL_MAGIC_symtab:
5120 case PERL_MAGIC_vstring:
5123 case PERL_MAGIC_utf8:
5124 vtable = &PL_vtbl_utf8;
5126 case PERL_MAGIC_substr:
5127 vtable = &PL_vtbl_substr;
5129 case PERL_MAGIC_defelem:
5130 vtable = &PL_vtbl_defelem;
5132 case PERL_MAGIC_arylen:
5133 vtable = &PL_vtbl_arylen;
5135 case PERL_MAGIC_pos:
5136 vtable = &PL_vtbl_pos;
5138 case PERL_MAGIC_backref:
5139 vtable = &PL_vtbl_backref;
5141 case PERL_MAGIC_hintselem:
5142 vtable = &PL_vtbl_hintselem;
5144 case PERL_MAGIC_hints:
5145 vtable = &PL_vtbl_hints;
5147 case PERL_MAGIC_ext:
5148 /* Reserved for use by extensions not perl internals. */
5149 /* Useful for attaching extension internal data to perl vars. */
5150 /* Note that multiple extensions may clash if magical scalars */
5151 /* etc holding private data from one are passed to another. */
5155 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5158 /* Rest of work is done else where */
5159 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5162 case PERL_MAGIC_taint:
5165 case PERL_MAGIC_ext:
5166 case PERL_MAGIC_dbfile:
5173 =for apidoc sv_unmagic
5175 Removes all magic of type C<type> from an SV.
5181 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5186 PERL_ARGS_ASSERT_SV_UNMAGIC;
5188 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5190 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5191 for (mg = *mgp; mg; mg = *mgp) {
5192 if (mg->mg_type == type) {
5193 const MGVTBL* const vtbl = mg->mg_virtual;
5194 *mgp = mg->mg_moremagic;
5195 if (vtbl && vtbl->svt_free)
5196 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5197 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5199 Safefree(mg->mg_ptr);
5200 else if (mg->mg_len == HEf_SVKEY)
5201 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5202 else if (mg->mg_type == PERL_MAGIC_utf8)
5203 Safefree(mg->mg_ptr);
5205 if (mg->mg_flags & MGf_REFCOUNTED)
5206 SvREFCNT_dec(mg->mg_obj);
5210 mgp = &mg->mg_moremagic;
5213 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5214 mg_magical(sv); /* else fix the flags now */
5218 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5224 =for apidoc sv_rvweaken
5226 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5227 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5228 push a back-reference to this RV onto the array of backreferences
5229 associated with that magic. If the RV is magical, set magic will be
5230 called after the RV is cleared.
5236 Perl_sv_rvweaken(pTHX_ SV *const sv)
5240 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5242 if (!SvOK(sv)) /* let undefs pass */
5245 Perl_croak(aTHX_ "Can't weaken a nonreference");
5246 else if (SvWEAKREF(sv)) {
5247 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5251 Perl_sv_add_backref(aTHX_ tsv, sv);
5257 /* Give tsv backref magic if it hasn't already got it, then push a
5258 * back-reference to sv onto the array associated with the backref magic.
5261 /* A discussion about the backreferences array and its refcount:
5263 * The AV holding the backreferences is pointed to either as the mg_obj of
5264 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5265 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5266 * have the standard magic instead.) The array is created with a refcount
5267 * of 2. This means that if during global destruction the array gets
5268 * picked on first to have its refcount decremented by the random zapper,
5269 * it won't actually be freed, meaning it's still theere for when its
5270 * parent gets freed.
5271 * When the parent SV is freed, in the case of magic, the magic is freed,
5272 * Perl_magic_killbackrefs is called which decrements one refcount, then
5273 * mg_obj is freed which kills the second count.
5274 * In the vase of a HV being freed, one ref is removed by
5275 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5280 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5285 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5287 if (SvTYPE(tsv) == SVt_PVHV) {
5288 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5292 /* There is no AV in the offical place - try a fixup. */
5293 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5296 /* Aha. They've got it stowed in magic. Bring it back. */
5297 av = MUTABLE_AV(mg->mg_obj);
5298 /* Stop mg_free decreasing the refernce count. */
5300 /* Stop mg_free even calling the destructor, given that
5301 there's no AV to free up. */
5303 sv_unmagic(tsv, PERL_MAGIC_backref);
5307 SvREFCNT_inc_simple_void(av); /* see discussion above */
5312 const MAGIC *const mg
5313 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5315 av = MUTABLE_AV(mg->mg_obj);
5319 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5320 /* av now has a refcnt of 2; see discussion above */
5323 if (AvFILLp(av) >= AvMAX(av)) {
5324 av_extend(av, AvFILLp(av)+1);
5326 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5329 /* delete a back-reference to ourselves from the backref magic associated
5330 * with the SV we point to.
5334 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5341 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5343 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5344 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5345 /* We mustn't attempt to "fix up" the hash here by moving the
5346 backreference array back to the hv_aux structure, as that is stored
5347 in the main HvARRAY(), and hfreentries assumes that no-one
5348 reallocates HvARRAY() while it is running. */
5351 const MAGIC *const mg
5352 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5354 av = MUTABLE_AV(mg->mg_obj);
5358 Perl_croak(aTHX_ "panic: del_backref");
5360 assert(!SvIS_FREED(av));
5363 /* We shouldn't be in here more than once, but for paranoia reasons lets
5365 for (i = AvFILLp(av); i >= 0; i--) {
5367 const SSize_t fill = AvFILLp(av);
5369 /* We weren't the last entry.
5370 An unordered list has this property that you can take the
5371 last element off the end to fill the hole, and it's still
5372 an unordered list :-)
5377 AvFILLp(av) = fill - 1;
5383 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5385 SV **svp = AvARRAY(av);
5387 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5388 PERL_UNUSED_ARG(sv);
5390 assert(!svp || !SvIS_FREED(av));
5392 SV *const *const last = svp + AvFILLp(av);
5394 while (svp <= last) {
5396 SV *const referrer = *svp;
5397 if (SvWEAKREF(referrer)) {
5398 /* XXX Should we check that it hasn't changed? */
5399 SvRV_set(referrer, 0);
5401 SvWEAKREF_off(referrer);
5402 SvSETMAGIC(referrer);
5403 } else if (SvTYPE(referrer) == SVt_PVGV ||
5404 SvTYPE(referrer) == SVt_PVLV) {
5405 /* You lookin' at me? */
5406 assert(GvSTASH(referrer));
5407 assert(GvSTASH(referrer) == (const HV *)sv);
5408 GvSTASH(referrer) = 0;
5411 "panic: magic_killbackrefs (flags=%"UVxf")",
5412 (UV)SvFLAGS(referrer));
5420 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5425 =for apidoc sv_insert
5427 Inserts a string at the specified offset/length within the SV. Similar to
5428 the Perl substr() function. Handles get magic.
5430 =for apidoc sv_insert_flags
5432 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5438 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5443 register char *midend;
5444 register char *bigend;
5448 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5451 Perl_croak(aTHX_ "Can't modify non-existent substring");
5452 SvPV_force_flags(bigstr, curlen, flags);
5453 (void)SvPOK_only_UTF8(bigstr);
5454 if (offset + len > curlen) {
5455 SvGROW(bigstr, offset+len+1);
5456 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5457 SvCUR_set(bigstr, offset+len);
5461 i = littlelen - len;
5462 if (i > 0) { /* string might grow */
5463 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5464 mid = big + offset + len;
5465 midend = bigend = big + SvCUR(bigstr);
5468 while (midend > mid) /* shove everything down */
5469 *--bigend = *--midend;
5470 Move(little,big+offset,littlelen,char);
5471 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5476 Move(little,SvPVX(bigstr)+offset,len,char);
5481 big = SvPVX(bigstr);
5484 bigend = big + SvCUR(bigstr);
5486 if (midend > bigend)
5487 Perl_croak(aTHX_ "panic: sv_insert");
5489 if (mid - big > bigend - midend) { /* faster to shorten from end */
5491 Move(little, mid, littlelen,char);
5494 i = bigend - midend;
5496 Move(midend, mid, i,char);
5500 SvCUR_set(bigstr, mid - big);
5502 else if ((i = mid - big)) { /* faster from front */
5503 midend -= littlelen;
5505 Move(big, midend - i, i, char);
5506 sv_chop(bigstr,midend-i);
5508 Move(little, mid, littlelen,char);
5510 else if (littlelen) {
5511 midend -= littlelen;
5512 sv_chop(bigstr,midend);
5513 Move(little,midend,littlelen,char);
5516 sv_chop(bigstr,midend);
5522 =for apidoc sv_replace
5524 Make the first argument a copy of the second, then delete the original.
5525 The target SV physically takes over ownership of the body of the source SV
5526 and inherits its flags; however, the target keeps any magic it owns,
5527 and any magic in the source is discarded.
5528 Note that this is a rather specialist SV copying operation; most of the
5529 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5535 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5538 const U32 refcnt = SvREFCNT(sv);
5540 PERL_ARGS_ASSERT_SV_REPLACE;
5542 SV_CHECK_THINKFIRST_COW_DROP(sv);
5543 if (SvREFCNT(nsv) != 1) {
5544 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5545 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5547 if (SvMAGICAL(sv)) {
5551 sv_upgrade(nsv, SVt_PVMG);
5552 SvMAGIC_set(nsv, SvMAGIC(sv));
5553 SvFLAGS(nsv) |= SvMAGICAL(sv);
5555 SvMAGIC_set(sv, NULL);
5559 assert(!SvREFCNT(sv));
5560 #ifdef DEBUG_LEAKING_SCALARS
5561 sv->sv_flags = nsv->sv_flags;
5562 sv->sv_any = nsv->sv_any;
5563 sv->sv_refcnt = nsv->sv_refcnt;
5564 sv->sv_u = nsv->sv_u;
5566 StructCopy(nsv,sv,SV);
5568 if(SvTYPE(sv) == SVt_IV) {
5570 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5574 #ifdef PERL_OLD_COPY_ON_WRITE
5575 if (SvIsCOW_normal(nsv)) {
5576 /* We need to follow the pointers around the loop to make the
5577 previous SV point to sv, rather than nsv. */
5580 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5583 assert(SvPVX_const(current) == SvPVX_const(nsv));
5585 /* Make the SV before us point to the SV after us. */
5587 PerlIO_printf(Perl_debug_log, "previous is\n");
5589 PerlIO_printf(Perl_debug_log,
5590 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5591 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5593 SV_COW_NEXT_SV_SET(current, sv);
5596 SvREFCNT(sv) = refcnt;
5597 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5603 =for apidoc sv_clear
5605 Clear an SV: call any destructors, free up any memory used by the body,
5606 and free the body itself. The SV's head is I<not> freed, although
5607 its type is set to all 1's so that it won't inadvertently be assumed
5608 to be live during global destruction etc.
5609 This function should only be called when REFCNT is zero. Most of the time
5610 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5617 Perl_sv_clear(pTHX_ register SV *const sv)
5620 const U32 type = SvTYPE(sv);
5621 const struct body_details *const sv_type_details
5622 = bodies_by_type + type;
5625 PERL_ARGS_ASSERT_SV_CLEAR;
5626 assert(SvREFCNT(sv) == 0);
5627 assert(SvTYPE(sv) != SVTYPEMASK);
5629 if (type <= SVt_IV) {
5630 /* See the comment in sv.h about the collusion between this early
5631 return and the overloading of the NULL and IV slots in the size
5634 SV * const target = SvRV(sv);
5636 sv_del_backref(target, sv);
5638 SvREFCNT_dec(target);
5640 SvFLAGS(sv) &= SVf_BREAK;
5641 SvFLAGS(sv) |= SVTYPEMASK;
5646 if (PL_defstash && /* Still have a symbol table? */
5653 stash = SvSTASH(sv);
5654 destructor = StashHANDLER(stash,DESTROY);
5656 /* A constant subroutine can have no side effects, so
5657 don't bother calling it. */
5658 && !CvCONST(destructor)
5659 /* Don't bother calling an empty destructor */
5660 && (CvISXSUB(destructor)
5661 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5663 SV* const tmpref = newRV(sv);
5664 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5666 PUSHSTACKi(PERLSI_DESTROY);
5671 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5677 if(SvREFCNT(tmpref) < 2) {
5678 /* tmpref is not kept alive! */
5680 SvRV_set(tmpref, NULL);
5683 SvREFCNT_dec(tmpref);
5685 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5689 if (PL_in_clean_objs)
5690 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5692 /* DESTROY gave object new lease on life */
5698 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5699 SvOBJECT_off(sv); /* Curse the object. */
5700 if (type != SVt_PVIO)
5701 --PL_sv_objcount; /* XXX Might want something more general */
5704 if (type >= SVt_PVMG) {
5705 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5706 SvREFCNT_dec(SvOURSTASH(sv));
5707 } else if (SvMAGIC(sv))
5709 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5710 SvREFCNT_dec(SvSTASH(sv));
5713 /* case SVt_BIND: */
5716 IoIFP(sv) != PerlIO_stdin() &&
5717 IoIFP(sv) != PerlIO_stdout() &&
5718 IoIFP(sv) != PerlIO_stderr())
5720 io_close(MUTABLE_IO(sv), FALSE);
5722 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5723 PerlDir_close(IoDIRP(sv));
5724 IoDIRP(sv) = (DIR*)NULL;
5725 Safefree(IoTOP_NAME(sv));
5726 Safefree(IoFMT_NAME(sv));
5727 Safefree(IoBOTTOM_NAME(sv));
5730 /* FIXME for plugins */
5731 pregfree2((REGEXP*) sv);
5735 cv_undef(MUTABLE_CV(sv));
5738 if (PL_last_swash_hv == (const HV *)sv) {
5739 PL_last_swash_hv = NULL;
5741 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5742 hv_undef(MUTABLE_HV(sv));
5745 if (PL_comppad == MUTABLE_AV(sv)) {
5749 av_undef(MUTABLE_AV(sv));
5752 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5753 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5754 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5755 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5757 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5758 SvREFCNT_dec(LvTARG(sv));
5760 if (isGV_with_GP(sv)) {
5761 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5762 && HvNAME_get(stash))
5763 mro_method_changed_in(stash);
5764 gp_free(MUTABLE_GV(sv));
5766 unshare_hek(GvNAME_HEK(sv));
5767 /* If we're in a stash, we don't own a reference to it. However it does
5768 have a back reference to us, which needs to be cleared. */
5769 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5770 sv_del_backref(MUTABLE_SV(stash), sv);
5772 /* FIXME. There are probably more unreferenced pointers to SVs in the
5773 interpreter struct that we should check and tidy in a similar
5775 if ((const GV *)sv == PL_last_in_gv)
5776 PL_last_in_gv = NULL;
5782 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5785 SvOOK_offset(sv, offset);
5786 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5787 /* Don't even bother with turning off the OOK flag. */
5790 SV * const target = SvRV(sv);
5792 sv_del_backref(target, sv);
5794 SvREFCNT_dec(target);
5796 #ifdef PERL_OLD_COPY_ON_WRITE
5797 else if (SvPVX_const(sv)) {
5800 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5804 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5806 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5810 } else if (SvLEN(sv)) {
5811 Safefree(SvPVX_const(sv));
5815 else if (SvPVX_const(sv) && SvLEN(sv))
5816 Safefree(SvPVX_mutable(sv));
5817 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5818 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5827 SvFLAGS(sv) &= SVf_BREAK;
5828 SvFLAGS(sv) |= SVTYPEMASK;
5830 if (sv_type_details->arena) {
5831 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5832 &PL_body_roots[type]);
5834 else if (sv_type_details->body_size) {
5835 my_safefree(SvANY(sv));
5840 =for apidoc sv_newref
5842 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5849 Perl_sv_newref(pTHX_ SV *const sv)
5851 PERL_UNUSED_CONTEXT;
5860 Decrement an SV's reference count, and if it drops to zero, call
5861 C<sv_clear> to invoke destructors and free up any memory used by
5862 the body; finally, deallocate the SV's head itself.
5863 Normally called via a wrapper macro C<SvREFCNT_dec>.
5869 Perl_sv_free(pTHX_ SV *const sv)
5874 if (SvREFCNT(sv) == 0) {
5875 if (SvFLAGS(sv) & SVf_BREAK)
5876 /* this SV's refcnt has been artificially decremented to
5877 * trigger cleanup */
5879 if (PL_in_clean_all) /* All is fair */
5881 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5882 /* make sure SvREFCNT(sv)==0 happens very seldom */
5883 SvREFCNT(sv) = (~(U32)0)/2;
5886 if (ckWARN_d(WARN_INTERNAL)) {
5887 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5888 Perl_dump_sv_child(aTHX_ sv);
5890 #ifdef DEBUG_LEAKING_SCALARS
5893 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5894 if (PL_warnhook == PERL_WARNHOOK_FATAL
5895 || ckDEAD(packWARN(WARN_INTERNAL))) {
5896 /* Don't let Perl_warner cause us to escape our fate: */
5900 /* This may not return: */
5901 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5902 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5903 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5906 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5911 if (--(SvREFCNT(sv)) > 0)
5913 Perl_sv_free2(aTHX_ sv);
5917 Perl_sv_free2(pTHX_ SV *const sv)
5921 PERL_ARGS_ASSERT_SV_FREE2;
5925 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5926 "Attempt to free temp prematurely: SV 0x%"UVxf
5927 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5931 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5932 /* make sure SvREFCNT(sv)==0 happens very seldom */
5933 SvREFCNT(sv) = (~(U32)0)/2;
5944 Returns the length of the string in the SV. Handles magic and type
5945 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5951 Perl_sv_len(pTHX_ register SV *const sv)
5959 len = mg_length(sv);
5961 (void)SvPV_const(sv, len);
5966 =for apidoc sv_len_utf8
5968 Returns the number of characters in the string in an SV, counting wide
5969 UTF-8 bytes as a single character. Handles magic and type coercion.
5975 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5976 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5977 * (Note that the mg_len is not the length of the mg_ptr field.
5978 * This allows the cache to store the character length of the string without
5979 * needing to malloc() extra storage to attach to the mg_ptr.)
5984 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5990 return mg_length(sv);
5994 const U8 *s = (U8*)SvPV_const(sv, len);
5998 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6000 if (mg && mg->mg_len != -1) {
6002 if (PL_utf8cache < 0) {
6003 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6005 /* Need to turn the assertions off otherwise we may
6006 recurse infinitely while printing error messages.
6008 SAVEI8(PL_utf8cache);
6010 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6011 " real %"UVuf" for %"SVf,
6012 (UV) ulen, (UV) real, SVfARG(sv));
6017 ulen = Perl_utf8_length(aTHX_ s, s + len);
6018 if (!SvREADONLY(sv)) {
6019 if (!mg && (SvTYPE(sv) < SVt_PVMG ||
6020 !(mg = mg_find(sv, PERL_MAGIC_utf8)))) {
6021 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6022 &PL_vtbl_utf8, 0, 0);
6030 return Perl_utf8_length(aTHX_ s, s + len);
6034 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6037 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6040 const U8 *s = start;
6042 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6044 while (s < send && uoffset--)
6047 /* This is the existing behaviour. Possibly it should be a croak, as
6048 it's actually a bounds error */
6054 /* Given the length of the string in both bytes and UTF-8 characters, decide
6055 whether to walk forwards or backwards to find the byte corresponding to
6056 the passed in UTF-8 offset. */
6058 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6059 const STRLEN uoffset, const STRLEN uend)
6061 STRLEN backw = uend - uoffset;
6063 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6065 if (uoffset < 2 * backw) {
6066 /* The assumption is that going forwards is twice the speed of going
6067 forward (that's where the 2 * backw comes from).
6068 (The real figure of course depends on the UTF-8 data.) */
6069 return sv_pos_u2b_forwards(start, send, uoffset);
6074 while (UTF8_IS_CONTINUATION(*send))
6077 return send - start;
6080 /* For the string representation of the given scalar, find the byte
6081 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6082 give another position in the string, *before* the sought offset, which
6083 (which is always true, as 0, 0 is a valid pair of positions), which should
6084 help reduce the amount of linear searching.
6085 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6086 will be used to reduce the amount of linear searching. The cache will be
6087 created if necessary, and the found value offered to it for update. */
6089 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6090 const U8 *const send, const STRLEN uoffset,
6091 STRLEN uoffset0, STRLEN boffset0)
6093 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6096 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6098 assert (uoffset >= uoffset0);
6102 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6103 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6104 if ((*mgp)->mg_ptr) {
6105 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6106 if (cache[0] == uoffset) {
6107 /* An exact match. */
6110 if (cache[2] == uoffset) {
6111 /* An exact match. */
6115 if (cache[0] < uoffset) {
6116 /* The cache already knows part of the way. */
6117 if (cache[0] > uoffset0) {
6118 /* The cache knows more than the passed in pair */
6119 uoffset0 = cache[0];
6120 boffset0 = cache[1];
6122 if ((*mgp)->mg_len != -1) {
6123 /* And we know the end too. */
6125 + sv_pos_u2b_midway(start + boffset0, send,
6127 (*mgp)->mg_len - uoffset0);
6130 + sv_pos_u2b_forwards(start + boffset0,
6131 send, uoffset - uoffset0);
6134 else if (cache[2] < uoffset) {
6135 /* We're between the two cache entries. */
6136 if (cache[2] > uoffset0) {
6137 /* and the cache knows more than the passed in pair */
6138 uoffset0 = cache[2];
6139 boffset0 = cache[3];
6143 + sv_pos_u2b_midway(start + boffset0,
6146 cache[0] - uoffset0);
6149 + sv_pos_u2b_midway(start + boffset0,
6152 cache[2] - uoffset0);
6156 else if ((*mgp)->mg_len != -1) {
6157 /* If we can take advantage of a passed in offset, do so. */
6158 /* In fact, offset0 is either 0, or less than offset, so don't
6159 need to worry about the other possibility. */
6161 + sv_pos_u2b_midway(start + boffset0, send,
6163 (*mgp)->mg_len - uoffset0);
6168 if (!found || PL_utf8cache < 0) {
6169 const STRLEN real_boffset
6170 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6171 send, uoffset - uoffset0);
6173 if (found && PL_utf8cache < 0) {
6174 if (real_boffset != boffset) {
6175 /* Need to turn the assertions off otherwise we may recurse
6176 infinitely while printing error messages. */
6177 SAVEI8(PL_utf8cache);
6179 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6180 " real %"UVuf" for %"SVf,
6181 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6184 boffset = real_boffset;
6188 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6194 =for apidoc sv_pos_u2b
6196 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6197 the start of the string, to a count of the equivalent number of bytes; if
6198 lenp is non-zero, it does the same to lenp, but this time starting from
6199 the offset, rather than from the start of the string. Handles magic and
6206 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6207 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6208 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6213 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6218 PERL_ARGS_ASSERT_SV_POS_U2B;
6223 start = (U8*)SvPV_const(sv, len);
6225 STRLEN uoffset = (STRLEN) *offsetp;
6226 const U8 * const send = start + len;
6228 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6231 *offsetp = (I32) boffset;
6234 /* Convert the relative offset to absolute. */
6235 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6236 const STRLEN boffset2
6237 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6238 uoffset, boffset) - boffset;
6252 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6253 byte length pairing. The (byte) length of the total SV is passed in too,
6254 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6255 may not have updated SvCUR, so we can't rely on reading it directly.
6257 The proffered utf8/byte length pairing isn't used if the cache already has
6258 two pairs, and swapping either for the proffered pair would increase the
6259 RMS of the intervals between known byte offsets.
6261 The cache itself consists of 4 STRLEN values
6262 0: larger UTF-8 offset
6263 1: corresponding byte offset
6264 2: smaller UTF-8 offset
6265 3: corresponding byte offset
6267 Unused cache pairs have the value 0, 0.
6268 Keeping the cache "backwards" means that the invariant of
6269 cache[0] >= cache[2] is maintained even with empty slots, which means that
6270 the code that uses it doesn't need to worry if only 1 entry has actually
6271 been set to non-zero. It also makes the "position beyond the end of the
6272 cache" logic much simpler, as the first slot is always the one to start
6276 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6277 const STRLEN utf8, const STRLEN blen)
6281 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6286 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6287 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6288 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6290 (*mgp)->mg_len = -1;
6294 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6295 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6296 (*mgp)->mg_ptr = (char *) cache;
6300 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6301 /* SvPOKp() because it's possible that sv has string overloading, and
6302 therefore is a reference, hence SvPVX() is actually a pointer.
6303 This cures the (very real) symptoms of RT 69422, but I'm not actually
6304 sure whether we should even be caching the results of UTF-8
6305 operations on overloading, given that nothing stops overloading
6306 returning a different value every time it's called. */
6307 const U8 *start = (const U8 *) SvPVX_const(sv);
6308 const STRLEN realutf8 = utf8_length(start, start + byte);
6310 if (realutf8 != utf8) {
6311 /* Need to turn the assertions off otherwise we may recurse
6312 infinitely while printing error messages. */
6313 SAVEI8(PL_utf8cache);
6315 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6316 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6320 /* Cache is held with the later position first, to simplify the code
6321 that deals with unbounded ends. */
6323 ASSERT_UTF8_CACHE(cache);
6324 if (cache[1] == 0) {
6325 /* Cache is totally empty */
6328 } else if (cache[3] == 0) {
6329 if (byte > cache[1]) {
6330 /* New one is larger, so goes first. */
6331 cache[2] = cache[0];
6332 cache[3] = cache[1];
6340 #define THREEWAY_SQUARE(a,b,c,d) \
6341 ((float)((d) - (c))) * ((float)((d) - (c))) \
6342 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6343 + ((float)((b) - (a))) * ((float)((b) - (a)))
6345 /* Cache has 2 slots in use, and we know three potential pairs.
6346 Keep the two that give the lowest RMS distance. Do the
6347 calcualation in bytes simply because we always know the byte
6348 length. squareroot has the same ordering as the positive value,
6349 so don't bother with the actual square root. */
6350 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6351 if (byte > cache[1]) {
6352 /* New position is after the existing pair of pairs. */
6353 const float keep_earlier
6354 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6355 const float keep_later
6356 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6358 if (keep_later < keep_earlier) {
6359 if (keep_later < existing) {
6360 cache[2] = cache[0];
6361 cache[3] = cache[1];
6367 if (keep_earlier < existing) {
6373 else if (byte > cache[3]) {
6374 /* New position is between the existing pair of pairs. */
6375 const float keep_earlier
6376 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6377 const float keep_later
6378 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6380 if (keep_later < keep_earlier) {
6381 if (keep_later < existing) {
6387 if (keep_earlier < existing) {
6394 /* New position is before the existing pair of pairs. */
6395 const float keep_earlier
6396 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6397 const float keep_later
6398 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6400 if (keep_later < keep_earlier) {
6401 if (keep_later < existing) {
6407 if (keep_earlier < existing) {
6408 cache[0] = cache[2];
6409 cache[1] = cache[3];
6416 ASSERT_UTF8_CACHE(cache);
6419 /* We already know all of the way, now we may be able to walk back. The same
6420 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6421 backward is half the speed of walking forward. */
6423 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6424 const U8 *end, STRLEN endu)
6426 const STRLEN forw = target - s;
6427 STRLEN backw = end - target;
6429 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6431 if (forw < 2 * backw) {
6432 return utf8_length(s, target);
6435 while (end > target) {
6437 while (UTF8_IS_CONTINUATION(*end)) {
6446 =for apidoc sv_pos_b2u
6448 Converts the value pointed to by offsetp from a count of bytes from the
6449 start of the string, to a count of the equivalent number of UTF-8 chars.
6450 Handles magic and type coercion.
6456 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6457 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6462 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6465 const STRLEN byte = *offsetp;
6466 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6472 PERL_ARGS_ASSERT_SV_POS_B2U;
6477 s = (const U8*)SvPV_const(sv, blen);
6480 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6486 && SvTYPE(sv) >= SVt_PVMG
6487 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6490 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6491 if (cache[1] == byte) {
6492 /* An exact match. */
6493 *offsetp = cache[0];
6496 if (cache[3] == byte) {
6497 /* An exact match. */
6498 *offsetp = cache[2];
6502 if (cache[1] < byte) {
6503 /* We already know part of the way. */
6504 if (mg->mg_len != -1) {
6505 /* Actually, we know the end too. */
6507 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6508 s + blen, mg->mg_len - cache[0]);
6510 len = cache[0] + utf8_length(s + cache[1], send);
6513 else if (cache[3] < byte) {
6514 /* We're between the two cached pairs, so we do the calculation
6515 offset by the byte/utf-8 positions for the earlier pair,
6516 then add the utf-8 characters from the string start to
6518 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6519 s + cache[1], cache[0] - cache[2])
6523 else { /* cache[3] > byte */
6524 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6528 ASSERT_UTF8_CACHE(cache);
6530 } else if (mg->mg_len != -1) {
6531 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6535 if (!found || PL_utf8cache < 0) {
6536 const STRLEN real_len = utf8_length(s, send);
6538 if (found && PL_utf8cache < 0) {
6539 if (len != real_len) {
6540 /* Need to turn the assertions off otherwise we may recurse
6541 infinitely while printing error messages. */
6542 SAVEI8(PL_utf8cache);
6544 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6545 " real %"UVuf" for %"SVf,
6546 (UV) len, (UV) real_len, SVfARG(sv));
6554 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6560 Returns a boolean indicating whether the strings in the two SVs are
6561 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6562 coerce its args to strings if necessary.
6568 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6577 SV* svrecode = NULL;
6584 /* if pv1 and pv2 are the same, second SvPV_const call may
6585 * invalidate pv1, so we may need to make a copy */
6586 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6587 pv1 = SvPV_const(sv1, cur1);
6588 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6590 pv1 = SvPV_const(sv1, cur1);
6598 pv2 = SvPV_const(sv2, cur2);
6600 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6601 /* Differing utf8ness.
6602 * Do not UTF8size the comparands as a side-effect. */
6605 svrecode = newSVpvn(pv2, cur2);
6606 sv_recode_to_utf8(svrecode, PL_encoding);
6607 pv2 = SvPV_const(svrecode, cur2);
6610 svrecode = newSVpvn(pv1, cur1);
6611 sv_recode_to_utf8(svrecode, PL_encoding);
6612 pv1 = SvPV_const(svrecode, cur1);
6614 /* Now both are in UTF-8. */
6616 SvREFCNT_dec(svrecode);
6621 bool is_utf8 = TRUE;
6624 /* sv1 is the UTF-8 one,
6625 * if is equal it must be downgrade-able */
6626 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6632 /* sv2 is the UTF-8 one,
6633 * if is equal it must be downgrade-able */
6634 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6640 /* Downgrade not possible - cannot be eq */
6648 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6650 SvREFCNT_dec(svrecode);
6660 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6661 string in C<sv1> is less than, equal to, or greater than the string in
6662 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6663 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6669 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6673 const char *pv1, *pv2;
6676 SV *svrecode = NULL;
6683 pv1 = SvPV_const(sv1, cur1);
6690 pv2 = SvPV_const(sv2, cur2);
6692 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6693 /* Differing utf8ness.
6694 * Do not UTF8size the comparands as a side-effect. */
6697 svrecode = newSVpvn(pv2, cur2);
6698 sv_recode_to_utf8(svrecode, PL_encoding);
6699 pv2 = SvPV_const(svrecode, cur2);
6702 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6707 svrecode = newSVpvn(pv1, cur1);
6708 sv_recode_to_utf8(svrecode, PL_encoding);
6709 pv1 = SvPV_const(svrecode, cur1);
6712 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6718 cmp = cur2 ? -1 : 0;
6722 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6725 cmp = retval < 0 ? -1 : 1;
6726 } else if (cur1 == cur2) {
6729 cmp = cur1 < cur2 ? -1 : 1;
6733 SvREFCNT_dec(svrecode);
6741 =for apidoc sv_cmp_locale
6743 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6744 'use bytes' aware, handles get magic, and will coerce its args to strings
6745 if necessary. See also C<sv_cmp>.
6751 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6754 #ifdef USE_LOCALE_COLLATE
6760 if (PL_collation_standard)
6764 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6766 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6768 if (!pv1 || !len1) {
6779 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6782 return retval < 0 ? -1 : 1;
6785 * When the result of collation is equality, that doesn't mean
6786 * that there are no differences -- some locales exclude some
6787 * characters from consideration. So to avoid false equalities,
6788 * we use the raw string as a tiebreaker.
6794 #endif /* USE_LOCALE_COLLATE */
6796 return sv_cmp(sv1, sv2);
6800 #ifdef USE_LOCALE_COLLATE
6803 =for apidoc sv_collxfrm
6805 Add Collate Transform magic to an SV if it doesn't already have it.
6807 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6808 scalar data of the variable, but transformed to such a format that a normal
6809 memory comparison can be used to compare the data according to the locale
6816 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6821 PERL_ARGS_ASSERT_SV_COLLXFRM;
6823 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6824 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6830 Safefree(mg->mg_ptr);
6831 s = SvPV_const(sv, len);
6832 if ((xf = mem_collxfrm(s, len, &xlen))) {
6834 #ifdef PERL_OLD_COPY_ON_WRITE
6836 sv_force_normal_flags(sv, 0);
6838 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6852 if (mg && mg->mg_ptr) {
6854 return mg->mg_ptr + sizeof(PL_collation_ix);
6862 #endif /* USE_LOCALE_COLLATE */
6867 Get a line from the filehandle and store it into the SV, optionally
6868 appending to the currently-stored string.
6874 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6879 register STDCHAR rslast;
6880 register STDCHAR *bp;
6885 PERL_ARGS_ASSERT_SV_GETS;
6887 if (SvTHINKFIRST(sv))
6888 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6889 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6891 However, perlbench says it's slower, because the existing swipe code
6892 is faster than copy on write.
6893 Swings and roundabouts. */
6894 SvUPGRADE(sv, SVt_PV);
6899 if (PerlIO_isutf8(fp)) {
6901 sv_utf8_upgrade_nomg(sv);
6902 sv_pos_u2b(sv,&append,0);
6904 } else if (SvUTF8(sv)) {
6905 SV * const tsv = newSV(0);
6906 sv_gets(tsv, fp, 0);
6907 sv_utf8_upgrade_nomg(tsv);
6908 SvCUR_set(sv,append);
6911 goto return_string_or_null;
6916 if (PerlIO_isutf8(fp))
6919 if (IN_PERL_COMPILETIME) {
6920 /* we always read code in line mode */
6924 else if (RsSNARF(PL_rs)) {
6925 /* If it is a regular disk file use size from stat() as estimate
6926 of amount we are going to read -- may result in mallocing
6927 more memory than we really need if the layers below reduce
6928 the size we read (e.g. CRLF or a gzip layer).
6931 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6932 const Off_t offset = PerlIO_tell(fp);
6933 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6934 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6940 else if (RsRECORD(PL_rs)) {
6948 /* Grab the size of the record we're getting */
6949 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6950 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6953 /* VMS wants read instead of fread, because fread doesn't respect */
6954 /* RMS record boundaries. This is not necessarily a good thing to be */
6955 /* doing, but we've got no other real choice - except avoid stdio
6956 as implementation - perhaps write a :vms layer ?
6958 fd = PerlIO_fileno(fp);
6959 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6960 bytesread = PerlIO_read(fp, buffer, recsize);
6963 bytesread = PerlLIO_read(fd, buffer, recsize);
6966 bytesread = PerlIO_read(fp, buffer, recsize);
6970 SvCUR_set(sv, bytesread + append);
6971 buffer[bytesread] = '\0';
6972 goto return_string_or_null;
6974 else if (RsPARA(PL_rs)) {
6980 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6981 if (PerlIO_isutf8(fp)) {
6982 rsptr = SvPVutf8(PL_rs, rslen);
6985 if (SvUTF8(PL_rs)) {
6986 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6987 Perl_croak(aTHX_ "Wide character in $/");
6990 rsptr = SvPV_const(PL_rs, rslen);
6994 rslast = rslen ? rsptr[rslen - 1] : '\0';
6996 if (rspara) { /* have to do this both before and after */
6997 do { /* to make sure file boundaries work right */
7000 i = PerlIO_getc(fp);
7004 PerlIO_ungetc(fp,i);
7010 /* See if we know enough about I/O mechanism to cheat it ! */
7012 /* This used to be #ifdef test - it is made run-time test for ease
7013 of abstracting out stdio interface. One call should be cheap
7014 enough here - and may even be a macro allowing compile
7018 if (PerlIO_fast_gets(fp)) {
7021 * We're going to steal some values from the stdio struct
7022 * and put EVERYTHING in the innermost loop into registers.
7024 register STDCHAR *ptr;
7028 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7029 /* An ungetc()d char is handled separately from the regular
7030 * buffer, so we getc() it back out and stuff it in the buffer.
7032 i = PerlIO_getc(fp);
7033 if (i == EOF) return 0;
7034 *(--((*fp)->_ptr)) = (unsigned char) i;
7038 /* Here is some breathtakingly efficient cheating */
7040 cnt = PerlIO_get_cnt(fp); /* get count into register */
7041 /* make sure we have the room */
7042 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7043 /* Not room for all of it
7044 if we are looking for a separator and room for some
7046 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7047 /* just process what we have room for */
7048 shortbuffered = cnt - SvLEN(sv) + append + 1;
7049 cnt -= shortbuffered;
7053 /* remember that cnt can be negative */
7054 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7059 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7060 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7061 DEBUG_P(PerlIO_printf(Perl_debug_log,
7062 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7063 DEBUG_P(PerlIO_printf(Perl_debug_log,
7064 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7065 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7066 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7071 while (cnt > 0) { /* this | eat */
7073 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7074 goto thats_all_folks; /* screams | sed :-) */
7078 Copy(ptr, bp, cnt, char); /* this | eat */
7079 bp += cnt; /* screams | dust */
7080 ptr += cnt; /* louder | sed :-) */
7085 if (shortbuffered) { /* oh well, must extend */
7086 cnt = shortbuffered;
7088 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7090 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7091 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7095 DEBUG_P(PerlIO_printf(Perl_debug_log,
7096 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7097 PTR2UV(ptr),(long)cnt));
7098 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7100 DEBUG_P(PerlIO_printf(Perl_debug_log,
7101 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7102 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7103 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7105 /* This used to call 'filbuf' in stdio form, but as that behaves like
7106 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7107 another abstraction. */
7108 i = PerlIO_getc(fp); /* get more characters */
7110 DEBUG_P(PerlIO_printf(Perl_debug_log,
7111 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7112 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7113 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7115 cnt = PerlIO_get_cnt(fp);
7116 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7117 DEBUG_P(PerlIO_printf(Perl_debug_log,
7118 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7120 if (i == EOF) /* all done for ever? */
7121 goto thats_really_all_folks;
7123 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7125 SvGROW(sv, bpx + cnt + 2);
7126 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7128 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7130 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7131 goto thats_all_folks;
7135 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7136 memNE((char*)bp - rslen, rsptr, rslen))
7137 goto screamer; /* go back to the fray */
7138 thats_really_all_folks:
7140 cnt += shortbuffered;
7141 DEBUG_P(PerlIO_printf(Perl_debug_log,
7142 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7143 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7144 DEBUG_P(PerlIO_printf(Perl_debug_log,
7145 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7146 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7147 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7149 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7150 DEBUG_P(PerlIO_printf(Perl_debug_log,
7151 "Screamer: done, len=%ld, string=|%.*s|\n",
7152 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7156 /*The big, slow, and stupid way. */
7157 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7158 STDCHAR *buf = NULL;
7159 Newx(buf, 8192, STDCHAR);
7167 register const STDCHAR * const bpe = buf + sizeof(buf);
7169 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7170 ; /* keep reading */
7174 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7175 /* Accomodate broken VAXC compiler, which applies U8 cast to
7176 * both args of ?: operator, causing EOF to change into 255
7179 i = (U8)buf[cnt - 1];
7185 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7187 sv_catpvn(sv, (char *) buf, cnt);
7189 sv_setpvn(sv, (char *) buf, cnt);
7191 if (i != EOF && /* joy */
7193 SvCUR(sv) < rslen ||
7194 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7198 * If we're reading from a TTY and we get a short read,
7199 * indicating that the user hit his EOF character, we need
7200 * to notice it now, because if we try to read from the TTY
7201 * again, the EOF condition will disappear.
7203 * The comparison of cnt to sizeof(buf) is an optimization
7204 * that prevents unnecessary calls to feof().
7208 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7212 #ifdef USE_HEAP_INSTEAD_OF_STACK
7217 if (rspara) { /* have to do this both before and after */
7218 while (i != EOF) { /* to make sure file boundaries work right */
7219 i = PerlIO_getc(fp);
7221 PerlIO_ungetc(fp,i);
7227 return_string_or_null:
7228 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7234 Auto-increment of the value in the SV, doing string to numeric conversion
7235 if necessary. Handles 'get' magic.
7241 Perl_sv_inc(pTHX_ register SV *const sv)
7250 if (SvTHINKFIRST(sv)) {
7252 sv_force_normal_flags(sv, 0);
7253 if (SvREADONLY(sv)) {
7254 if (IN_PERL_RUNTIME)
7255 Perl_croak(aTHX_ "%s", PL_no_modify);
7259 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7261 i = PTR2IV(SvRV(sv));
7266 flags = SvFLAGS(sv);
7267 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7268 /* It's (privately or publicly) a float, but not tested as an
7269 integer, so test it to see. */
7271 flags = SvFLAGS(sv);
7273 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7274 /* It's publicly an integer, or privately an integer-not-float */
7275 #ifdef PERL_PRESERVE_IVUV
7279 if (SvUVX(sv) == UV_MAX)
7280 sv_setnv(sv, UV_MAX_P1);
7282 (void)SvIOK_only_UV(sv);
7283 SvUV_set(sv, SvUVX(sv) + 1);
7285 if (SvIVX(sv) == IV_MAX)
7286 sv_setuv(sv, (UV)IV_MAX + 1);
7288 (void)SvIOK_only(sv);
7289 SvIV_set(sv, SvIVX(sv) + 1);
7294 if (flags & SVp_NOK) {
7295 const NV was = SvNVX(sv);
7296 if (NV_OVERFLOWS_INTEGERS_AT &&
7297 was >= NV_OVERFLOWS_INTEGERS_AT) {
7298 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7299 "Lost precision when incrementing %" NVff " by 1",
7302 (void)SvNOK_only(sv);
7303 SvNV_set(sv, was + 1.0);
7307 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7308 if ((flags & SVTYPEMASK) < SVt_PVIV)
7309 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7310 (void)SvIOK_only(sv);
7315 while (isALPHA(*d)) d++;
7316 while (isDIGIT(*d)) d++;
7317 if (d < SvEND(sv)) {
7318 #ifdef PERL_PRESERVE_IVUV
7319 /* Got to punt this as an integer if needs be, but we don't issue
7320 warnings. Probably ought to make the sv_iv_please() that does
7321 the conversion if possible, and silently. */
7322 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7323 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7324 /* Need to try really hard to see if it's an integer.
7325 9.22337203685478e+18 is an integer.
7326 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7327 so $a="9.22337203685478e+18"; $a+0; $a++
7328 needs to be the same as $a="9.22337203685478e+18"; $a++
7335 /* sv_2iv *should* have made this an NV */
7336 if (flags & SVp_NOK) {
7337 (void)SvNOK_only(sv);
7338 SvNV_set(sv, SvNVX(sv) + 1.0);
7341 /* I don't think we can get here. Maybe I should assert this
7342 And if we do get here I suspect that sv_setnv will croak. NWC
7344 #if defined(USE_LONG_DOUBLE)
7345 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",
7346 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7348 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7349 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7352 #endif /* PERL_PRESERVE_IVUV */
7353 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7357 while (d >= SvPVX_const(sv)) {
7365 /* MKS: The original code here died if letters weren't consecutive.
7366 * at least it didn't have to worry about non-C locales. The
7367 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7368 * arranged in order (although not consecutively) and that only
7369 * [A-Za-z] are accepted by isALPHA in the C locale.
7371 if (*d != 'z' && *d != 'Z') {
7372 do { ++*d; } while (!isALPHA(*d));
7375 *(d--) -= 'z' - 'a';
7380 *(d--) -= 'z' - 'a' + 1;
7384 /* oh,oh, the number grew */
7385 SvGROW(sv, SvCUR(sv) + 2);
7386 SvCUR_set(sv, SvCUR(sv) + 1);
7387 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7398 Auto-decrement of the value in the SV, doing string to numeric conversion
7399 if necessary. Handles 'get' magic.
7405 Perl_sv_dec(pTHX_ register SV *const sv)
7413 if (SvTHINKFIRST(sv)) {
7415 sv_force_normal_flags(sv, 0);
7416 if (SvREADONLY(sv)) {
7417 if (IN_PERL_RUNTIME)
7418 Perl_croak(aTHX_ "%s", PL_no_modify);
7422 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7424 i = PTR2IV(SvRV(sv));
7429 /* Unlike sv_inc we don't have to worry about string-never-numbers
7430 and keeping them magic. But we mustn't warn on punting */
7431 flags = SvFLAGS(sv);
7432 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7433 /* It's publicly an integer, or privately an integer-not-float */
7434 #ifdef PERL_PRESERVE_IVUV
7438 if (SvUVX(sv) == 0) {
7439 (void)SvIOK_only(sv);
7443 (void)SvIOK_only_UV(sv);
7444 SvUV_set(sv, SvUVX(sv) - 1);
7447 if (SvIVX(sv) == IV_MIN) {
7448 sv_setnv(sv, (NV)IV_MIN);
7452 (void)SvIOK_only(sv);
7453 SvIV_set(sv, SvIVX(sv) - 1);
7458 if (flags & SVp_NOK) {
7461 const NV was = SvNVX(sv);
7462 if (NV_OVERFLOWS_INTEGERS_AT &&
7463 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7464 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7465 "Lost precision when decrementing %" NVff " by 1",
7468 (void)SvNOK_only(sv);
7469 SvNV_set(sv, was - 1.0);
7473 if (!(flags & SVp_POK)) {
7474 if ((flags & SVTYPEMASK) < SVt_PVIV)
7475 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7477 (void)SvIOK_only(sv);
7480 #ifdef PERL_PRESERVE_IVUV
7482 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7483 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7484 /* Need to try really hard to see if it's an integer.
7485 9.22337203685478e+18 is an integer.
7486 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7487 so $a="9.22337203685478e+18"; $a+0; $a--
7488 needs to be the same as $a="9.22337203685478e+18"; $a--
7495 /* sv_2iv *should* have made this an NV */
7496 if (flags & SVp_NOK) {
7497 (void)SvNOK_only(sv);
7498 SvNV_set(sv, SvNVX(sv) - 1.0);
7501 /* I don't think we can get here. Maybe I should assert this
7502 And if we do get here I suspect that sv_setnv will croak. NWC
7504 #if defined(USE_LONG_DOUBLE)
7505 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",
7506 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7508 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7509 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7513 #endif /* PERL_PRESERVE_IVUV */
7514 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7517 /* this define is used to eliminate a chunk of duplicated but shared logic
7518 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7519 * used anywhere but here - yves
7521 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7524 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7528 =for apidoc sv_mortalcopy
7530 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7531 The new SV is marked as mortal. It will be destroyed "soon", either by an
7532 explicit call to FREETMPS, or by an implicit call at places such as
7533 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7538 /* Make a string that will exist for the duration of the expression
7539 * evaluation. Actually, it may have to last longer than that, but
7540 * hopefully we won't free it until it has been assigned to a
7541 * permanent location. */
7544 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7550 sv_setsv(sv,oldstr);
7551 PUSH_EXTEND_MORTAL__SV_C(sv);
7557 =for apidoc sv_newmortal
7559 Creates a new null SV which is mortal. The reference count of the SV is
7560 set to 1. It will be destroyed "soon", either by an explicit call to
7561 FREETMPS, or by an implicit call at places such as statement boundaries.
7562 See also C<sv_mortalcopy> and C<sv_2mortal>.
7568 Perl_sv_newmortal(pTHX)
7574 SvFLAGS(sv) = SVs_TEMP;
7575 PUSH_EXTEND_MORTAL__SV_C(sv);
7581 =for apidoc newSVpvn_flags
7583 Creates a new SV and copies a string into it. The reference count for the
7584 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7585 string. You are responsible for ensuring that the source string is at least
7586 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7587 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7588 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7589 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7590 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7592 #define newSVpvn_utf8(s, len, u) \
7593 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7599 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7604 /* All the flags we don't support must be zero.
7605 And we're new code so I'm going to assert this from the start. */
7606 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7608 sv_setpvn(sv,s,len);
7610 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7611 * and do what it does outselves here.
7612 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7613 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7614 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7615 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7618 SvFLAGS(sv) |= flags;
7620 if(flags & SVs_TEMP){
7621 PUSH_EXTEND_MORTAL__SV_C(sv);
7628 =for apidoc sv_2mortal
7630 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7631 by an explicit call to FREETMPS, or by an implicit call at places such as
7632 statement boundaries. SvTEMP() is turned on which means that the SV's
7633 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7634 and C<sv_mortalcopy>.
7640 Perl_sv_2mortal(pTHX_ register SV *const sv)
7645 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7647 PUSH_EXTEND_MORTAL__SV_C(sv);
7655 Creates a new SV and copies a string into it. The reference count for the
7656 SV is set to 1. If C<len> is zero, Perl will compute the length using
7657 strlen(). For efficiency, consider using C<newSVpvn> instead.
7663 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7669 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7674 =for apidoc newSVpvn
7676 Creates a new SV and copies a string into it. The reference count for the
7677 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7678 string. You are responsible for ensuring that the source string is at least
7679 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7685 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7691 sv_setpvn(sv,s,len);
7696 =for apidoc newSVhek
7698 Creates a new SV from the hash key structure. It will generate scalars that
7699 point to the shared string table where possible. Returns a new (undefined)
7700 SV if the hek is NULL.
7706 Perl_newSVhek(pTHX_ const HEK *const hek)
7716 if (HEK_LEN(hek) == HEf_SVKEY) {
7717 return newSVsv(*(SV**)HEK_KEY(hek));
7719 const int flags = HEK_FLAGS(hek);
7720 if (flags & HVhek_WASUTF8) {
7722 Andreas would like keys he put in as utf8 to come back as utf8
7724 STRLEN utf8_len = HEK_LEN(hek);
7725 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7726 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7729 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7731 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7732 /* We don't have a pointer to the hv, so we have to replicate the
7733 flag into every HEK. This hv is using custom a hasing
7734 algorithm. Hence we can't return a shared string scalar, as
7735 that would contain the (wrong) hash value, and might get passed
7736 into an hv routine with a regular hash.
7737 Similarly, a hash that isn't using shared hash keys has to have
7738 the flag in every key so that we know not to try to call
7739 share_hek_kek on it. */
7741 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7746 /* This will be overwhelminly the most common case. */
7748 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7749 more efficient than sharepvn(). */
7753 sv_upgrade(sv, SVt_PV);
7754 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7755 SvCUR_set(sv, HEK_LEN(hek));
7768 =for apidoc newSVpvn_share
7770 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7771 table. If the string does not already exist in the table, it is created
7772 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7773 value is used; otherwise the hash is computed. The string's hash can be later
7774 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7775 that as the string table is used for shared hash keys these strings will have
7776 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7782 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7786 bool is_utf8 = FALSE;
7787 const char *const orig_src = src;
7790 STRLEN tmplen = -len;
7792 /* See the note in hv.c:hv_fetch() --jhi */
7793 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7797 PERL_HASH(hash, src, len);
7799 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7800 changes here, update it there too. */
7801 sv_upgrade(sv, SVt_PV);
7802 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7810 if (src != orig_src)
7816 #if defined(PERL_IMPLICIT_CONTEXT)
7818 /* pTHX_ magic can't cope with varargs, so this is a no-context
7819 * version of the main function, (which may itself be aliased to us).
7820 * Don't access this version directly.
7824 Perl_newSVpvf_nocontext(const char *const pat, ...)
7830 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7832 va_start(args, pat);
7833 sv = vnewSVpvf(pat, &args);
7840 =for apidoc newSVpvf
7842 Creates a new SV and initializes it with the string formatted like
7849 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7854 PERL_ARGS_ASSERT_NEWSVPVF;
7856 va_start(args, pat);
7857 sv = vnewSVpvf(pat, &args);
7862 /* backend for newSVpvf() and newSVpvf_nocontext() */
7865 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7870 PERL_ARGS_ASSERT_VNEWSVPVF;
7873 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7880 Creates a new SV and copies a floating point value into it.
7881 The reference count for the SV is set to 1.
7887 Perl_newSVnv(pTHX_ const NV n)
7900 Creates a new SV and copies an integer into it. The reference count for the
7907 Perl_newSViv(pTHX_ const IV i)
7920 Creates a new SV and copies an unsigned integer into it.
7921 The reference count for the SV is set to 1.
7927 Perl_newSVuv(pTHX_ const UV u)
7938 =for apidoc newSV_type
7940 Creates a new SV, of the type specified. The reference count for the new SV
7947 Perl_newSV_type(pTHX_ const svtype type)
7952 sv_upgrade(sv, type);
7957 =for apidoc newRV_noinc
7959 Creates an RV wrapper for an SV. The reference count for the original
7960 SV is B<not> incremented.
7966 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7969 register SV *sv = newSV_type(SVt_IV);
7971 PERL_ARGS_ASSERT_NEWRV_NOINC;
7974 SvRV_set(sv, tmpRef);
7979 /* newRV_inc is the official function name to use now.
7980 * newRV_inc is in fact #defined to newRV in sv.h
7984 Perl_newRV(pTHX_ SV *const sv)
7988 PERL_ARGS_ASSERT_NEWRV;
7990 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7996 Creates a new SV which is an exact duplicate of the original SV.
8003 Perl_newSVsv(pTHX_ register SV *const old)
8010 if (SvTYPE(old) == SVTYPEMASK) {
8011 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8015 /* SV_GMAGIC is the default for sv_setv()
8016 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8017 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8018 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8023 =for apidoc sv_reset
8025 Underlying implementation for the C<reset> Perl function.
8026 Note that the perl-level function is vaguely deprecated.
8032 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8035 char todo[PERL_UCHAR_MAX+1];
8037 PERL_ARGS_ASSERT_SV_RESET;
8042 if (!*s) { /* reset ?? searches */
8043 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8045 const U32 count = mg->mg_len / sizeof(PMOP**);
8046 PMOP **pmp = (PMOP**) mg->mg_ptr;
8047 PMOP *const *const end = pmp + count;
8051 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8053 (*pmp)->op_pmflags &= ~PMf_USED;
8061 /* reset variables */
8063 if (!HvARRAY(stash))
8066 Zero(todo, 256, char);
8069 I32 i = (unsigned char)*s;
8073 max = (unsigned char)*s++;
8074 for ( ; i <= max; i++) {
8077 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8079 for (entry = HvARRAY(stash)[i];
8081 entry = HeNEXT(entry))
8086 if (!todo[(U8)*HeKEY(entry)])
8088 gv = MUTABLE_GV(HeVAL(entry));
8091 if (SvTHINKFIRST(sv)) {
8092 if (!SvREADONLY(sv) && SvROK(sv))
8094 /* XXX Is this continue a bug? Why should THINKFIRST
8095 exempt us from resetting arrays and hashes? */
8099 if (SvTYPE(sv) >= SVt_PV) {
8101 if (SvPVX_const(sv) != NULL)
8109 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8111 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8114 # if defined(USE_ENVIRON_ARRAY)
8117 # endif /* USE_ENVIRON_ARRAY */
8128 Using various gambits, try to get an IO from an SV: the IO slot if its a
8129 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8130 named after the PV if we're a string.
8136 Perl_sv_2io(pTHX_ SV *const sv)
8141 PERL_ARGS_ASSERT_SV_2IO;
8143 switch (SvTYPE(sv)) {
8145 io = MUTABLE_IO(sv);
8148 if (isGV_with_GP(sv)) {
8149 gv = MUTABLE_GV(sv);
8152 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8158 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8160 return sv_2io(SvRV(sv));
8161 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8167 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8176 Using various gambits, try to get a CV from an SV; in addition, try if
8177 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8178 The flags in C<lref> are passed to gv_fetchsv.
8184 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8190 PERL_ARGS_ASSERT_SV_2CV;
8197 switch (SvTYPE(sv)) {
8201 return MUTABLE_CV(sv);
8208 if (isGV_with_GP(sv)) {
8209 gv = MUTABLE_GV(sv);
8218 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8220 tryAMAGICunDEREF(to_cv);
8223 if (SvTYPE(sv) == SVt_PVCV) {
8224 cv = MUTABLE_CV(sv);
8229 else if(isGV_with_GP(sv))
8230 gv = MUTABLE_GV(sv);
8232 Perl_croak(aTHX_ "Not a subroutine reference");
8234 else if (isGV_with_GP(sv)) {
8236 gv = MUTABLE_GV(sv);
8239 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8245 /* Some flags to gv_fetchsv mean don't really create the GV */
8246 if (!isGV_with_GP(gv)) {
8252 if (lref && !GvCVu(gv)) {
8256 gv_efullname3(tmpsv, gv, NULL);
8257 /* XXX this is probably not what they think they're getting.
8258 * It has the same effect as "sub name;", i.e. just a forward
8260 newSUB(start_subparse(FALSE, 0),
8261 newSVOP(OP_CONST, 0, tmpsv),
8265 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8266 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8275 Returns true if the SV has a true value by Perl's rules.
8276 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8277 instead use an in-line version.
8283 Perl_sv_true(pTHX_ register SV *const sv)
8288 register const XPV* const tXpv = (XPV*)SvANY(sv);
8290 (tXpv->xpv_cur > 1 ||
8291 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8298 return SvIVX(sv) != 0;
8301 return SvNVX(sv) != 0.0;
8303 return sv_2bool(sv);
8309 =for apidoc sv_pvn_force
8311 Get a sensible string out of the SV somehow.
8312 A private implementation of the C<SvPV_force> macro for compilers which
8313 can't cope with complex macro expressions. Always use the macro instead.
8315 =for apidoc sv_pvn_force_flags
8317 Get a sensible string out of the SV somehow.
8318 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8319 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8320 implemented in terms of this function.
8321 You normally want to use the various wrapper macros instead: see
8322 C<SvPV_force> and C<SvPV_force_nomg>
8328 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8332 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8334 if (SvTHINKFIRST(sv) && !SvROK(sv))
8335 sv_force_normal_flags(sv, 0);
8345 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8346 const char * const ref = sv_reftype(sv,0);
8348 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8349 ref, OP_NAME(PL_op));
8351 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8353 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8354 || isGV_with_GP(sv))
8355 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8357 s = sv_2pv_flags(sv, &len, flags);
8361 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8364 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8365 SvGROW(sv, len + 1);
8366 Move(s,SvPVX(sv),len,char);
8368 SvPVX(sv)[len] = '\0';
8371 SvPOK_on(sv); /* validate pointer */
8373 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8374 PTR2UV(sv),SvPVX_const(sv)));
8377 return SvPVX_mutable(sv);
8381 =for apidoc sv_pvbyten_force
8383 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8389 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8391 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8393 sv_pvn_force(sv,lp);
8394 sv_utf8_downgrade(sv,0);
8400 =for apidoc sv_pvutf8n_force
8402 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8408 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8410 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8412 sv_pvn_force(sv,lp);
8413 sv_utf8_upgrade(sv);
8419 =for apidoc sv_reftype
8421 Returns a string describing what the SV is a reference to.
8427 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8429 PERL_ARGS_ASSERT_SV_REFTYPE;
8431 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8432 inside return suggests a const propagation bug in g++. */
8433 if (ob && SvOBJECT(sv)) {
8434 char * const name = HvNAME_get(SvSTASH(sv));
8435 return name ? name : (char *) "__ANON__";
8438 switch (SvTYPE(sv)) {
8453 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8454 /* tied lvalues should appear to be
8455 * scalars for backwards compatitbility */
8456 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8457 ? "SCALAR" : "LVALUE");
8458 case SVt_PVAV: return "ARRAY";
8459 case SVt_PVHV: return "HASH";
8460 case SVt_PVCV: return "CODE";
8461 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8462 ? "GLOB" : "SCALAR");
8463 case SVt_PVFM: return "FORMAT";
8464 case SVt_PVIO: return "IO";
8465 case SVt_BIND: return "BIND";
8466 case SVt_REGEXP: return "REGEXP";
8467 default: return "UNKNOWN";
8473 =for apidoc sv_isobject
8475 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8476 object. If the SV is not an RV, or if the object is not blessed, then this
8483 Perl_sv_isobject(pTHX_ SV *sv)
8499 Returns a boolean indicating whether the SV is blessed into the specified
8500 class. This does not check for subtypes; use C<sv_derived_from> to verify
8501 an inheritance relationship.
8507 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8511 PERL_ARGS_ASSERT_SV_ISA;
8521 hvname = HvNAME_get(SvSTASH(sv));
8525 return strEQ(hvname, name);
8531 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8532 it will be upgraded to one. If C<classname> is non-null then the new SV will
8533 be blessed in the specified package. The new SV is returned and its
8534 reference count is 1.
8540 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8545 PERL_ARGS_ASSERT_NEWSVRV;
8549 SV_CHECK_THINKFIRST_COW_DROP(rv);
8550 (void)SvAMAGIC_off(rv);
8552 if (SvTYPE(rv) >= SVt_PVMG) {
8553 const U32 refcnt = SvREFCNT(rv);
8557 SvREFCNT(rv) = refcnt;
8559 sv_upgrade(rv, SVt_IV);
8560 } else if (SvROK(rv)) {
8561 SvREFCNT_dec(SvRV(rv));
8563 prepare_SV_for_RV(rv);
8571 HV* const stash = gv_stashpv(classname, GV_ADD);
8572 (void)sv_bless(rv, stash);
8578 =for apidoc sv_setref_pv
8580 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8581 argument will be upgraded to an RV. That RV will be modified to point to
8582 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8583 into the SV. The C<classname> argument indicates the package for the
8584 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8585 will have a reference count of 1, and the RV will be returned.
8587 Do not use with other Perl types such as HV, AV, SV, CV, because those
8588 objects will become corrupted by the pointer copy process.
8590 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8596 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8600 PERL_ARGS_ASSERT_SV_SETREF_PV;
8603 sv_setsv(rv, &PL_sv_undef);
8607 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8612 =for apidoc sv_setref_iv
8614 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8615 argument will be upgraded to an RV. That RV will be modified to point to
8616 the new SV. The C<classname> argument indicates the package for the
8617 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8618 will have a reference count of 1, and the RV will be returned.
8624 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8626 PERL_ARGS_ASSERT_SV_SETREF_IV;
8628 sv_setiv(newSVrv(rv,classname), iv);
8633 =for apidoc sv_setref_uv
8635 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8636 argument will be upgraded to an RV. That RV will be modified to point to
8637 the new SV. The C<classname> argument indicates the package for the
8638 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8639 will have a reference count of 1, and the RV will be returned.
8645 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8647 PERL_ARGS_ASSERT_SV_SETREF_UV;
8649 sv_setuv(newSVrv(rv,classname), uv);
8654 =for apidoc sv_setref_nv
8656 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8657 argument will be upgraded to an RV. That RV will be modified to point to
8658 the new SV. The C<classname> argument indicates the package for the
8659 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8660 will have a reference count of 1, and the RV will be returned.
8666 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8668 PERL_ARGS_ASSERT_SV_SETREF_NV;
8670 sv_setnv(newSVrv(rv,classname), nv);
8675 =for apidoc sv_setref_pvn
8677 Copies a string into a new SV, optionally blessing the SV. The length of the
8678 string must be specified with C<n>. The C<rv> argument will be upgraded to
8679 an RV. That RV will be modified to point to the new SV. The C<classname>
8680 argument indicates the package for the blessing. Set C<classname> to
8681 C<NULL> to avoid the blessing. The new SV will have a reference count
8682 of 1, and the RV will be returned.
8684 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8690 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8691 const char *const pv, const STRLEN n)
8693 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8695 sv_setpvn(newSVrv(rv,classname), pv, n);
8700 =for apidoc sv_bless
8702 Blesses an SV into a specified package. The SV must be an RV. The package
8703 must be designated by its stash (see C<gv_stashpv()>). The reference count
8704 of the SV is unaffected.
8710 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8715 PERL_ARGS_ASSERT_SV_BLESS;
8718 Perl_croak(aTHX_ "Can't bless non-reference value");
8720 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8721 if (SvIsCOW(tmpRef))
8722 sv_force_normal_flags(tmpRef, 0);
8723 if (SvREADONLY(tmpRef))
8724 Perl_croak(aTHX_ "%s", PL_no_modify);
8725 if (SvOBJECT(tmpRef)) {
8726 if (SvTYPE(tmpRef) != SVt_PVIO)
8728 SvREFCNT_dec(SvSTASH(tmpRef));
8731 SvOBJECT_on(tmpRef);
8732 if (SvTYPE(tmpRef) != SVt_PVIO)
8734 SvUPGRADE(tmpRef, SVt_PVMG);
8735 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8740 (void)SvAMAGIC_off(sv);
8742 if(SvSMAGICAL(tmpRef))
8743 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8751 /* Downgrades a PVGV to a PVMG.
8755 S_sv_unglob(pTHX_ SV *const sv)
8760 SV * const temp = sv_newmortal();
8762 PERL_ARGS_ASSERT_SV_UNGLOB;
8764 assert(SvTYPE(sv) == SVt_PVGV);
8766 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8769 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8770 && HvNAME_get(stash))
8771 mro_method_changed_in(stash);
8772 gp_free(MUTABLE_GV(sv));
8775 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8779 if (GvNAME_HEK(sv)) {
8780 unshare_hek(GvNAME_HEK(sv));
8782 isGV_with_GP_off(sv);
8784 /* need to keep SvANY(sv) in the right arena */
8785 xpvmg = new_XPVMG();
8786 StructCopy(SvANY(sv), xpvmg, XPVMG);
8787 del_XPVGV(SvANY(sv));
8790 SvFLAGS(sv) &= ~SVTYPEMASK;
8791 SvFLAGS(sv) |= SVt_PVMG;
8793 /* Intentionally not calling any local SET magic, as this isn't so much a
8794 set operation as merely an internal storage change. */
8795 sv_setsv_flags(sv, temp, 0);
8799 =for apidoc sv_unref_flags
8801 Unsets the RV status of the SV, and decrements the reference count of
8802 whatever was being referenced by the RV. This can almost be thought of
8803 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8804 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8805 (otherwise the decrementing is conditional on the reference count being
8806 different from one or the reference being a readonly SV).
8813 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8815 SV* const target = SvRV(ref);
8817 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8819 if (SvWEAKREF(ref)) {
8820 sv_del_backref(target, ref);
8822 SvRV_set(ref, NULL);
8825 SvRV_set(ref, NULL);
8827 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8828 assigned to as BEGIN {$a = \"Foo"} will fail. */
8829 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8830 SvREFCNT_dec(target);
8831 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8832 sv_2mortal(target); /* Schedule for freeing later */
8836 =for apidoc sv_untaint
8838 Untaint an SV. Use C<SvTAINTED_off> instead.
8843 Perl_sv_untaint(pTHX_ SV *const sv)
8845 PERL_ARGS_ASSERT_SV_UNTAINT;
8847 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8848 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8855 =for apidoc sv_tainted
8857 Test an SV for taintedness. Use C<SvTAINTED> instead.
8862 Perl_sv_tainted(pTHX_ SV *const sv)
8864 PERL_ARGS_ASSERT_SV_TAINTED;
8866 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8867 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8868 if (mg && (mg->mg_len & 1) )
8875 =for apidoc sv_setpviv
8877 Copies an integer into the given SV, also updating its string value.
8878 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8884 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8886 char buf[TYPE_CHARS(UV)];
8888 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8890 PERL_ARGS_ASSERT_SV_SETPVIV;
8892 sv_setpvn(sv, ptr, ebuf - ptr);
8896 =for apidoc sv_setpviv_mg
8898 Like C<sv_setpviv>, but also handles 'set' magic.
8904 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8906 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8912 #if defined(PERL_IMPLICIT_CONTEXT)
8914 /* pTHX_ magic can't cope with varargs, so this is a no-context
8915 * version of the main function, (which may itself be aliased to us).
8916 * Don't access this version directly.
8920 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8925 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8927 va_start(args, pat);
8928 sv_vsetpvf(sv, pat, &args);
8932 /* pTHX_ magic can't cope with varargs, so this is a no-context
8933 * version of the main function, (which may itself be aliased to us).
8934 * Don't access this version directly.
8938 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8943 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8945 va_start(args, pat);
8946 sv_vsetpvf_mg(sv, pat, &args);
8952 =for apidoc sv_setpvf
8954 Works like C<sv_catpvf> but copies the text into the SV instead of
8955 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8961 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8965 PERL_ARGS_ASSERT_SV_SETPVF;
8967 va_start(args, pat);
8968 sv_vsetpvf(sv, pat, &args);
8973 =for apidoc sv_vsetpvf
8975 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8976 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8978 Usually used via its frontend C<sv_setpvf>.
8984 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8986 PERL_ARGS_ASSERT_SV_VSETPVF;
8988 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8992 =for apidoc sv_setpvf_mg
8994 Like C<sv_setpvf>, but also handles 'set' magic.
9000 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9004 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9006 va_start(args, pat);
9007 sv_vsetpvf_mg(sv, pat, &args);
9012 =for apidoc sv_vsetpvf_mg
9014 Like C<sv_vsetpvf>, but also handles 'set' magic.
9016 Usually used via its frontend C<sv_setpvf_mg>.
9022 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9024 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9026 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9030 #if defined(PERL_IMPLICIT_CONTEXT)
9032 /* pTHX_ magic can't cope with varargs, so this is a no-context
9033 * version of the main function, (which may itself be aliased to us).
9034 * Don't access this version directly.
9038 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9043 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9045 va_start(args, pat);
9046 sv_vcatpvf(sv, pat, &args);
9050 /* pTHX_ magic can't cope with varargs, so this is a no-context
9051 * version of the main function, (which may itself be aliased to us).
9052 * Don't access this version directly.
9056 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9061 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9063 va_start(args, pat);
9064 sv_vcatpvf_mg(sv, pat, &args);
9070 =for apidoc sv_catpvf
9072 Processes its arguments like C<sprintf> and appends the formatted
9073 output to an SV. If the appended data contains "wide" characters
9074 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9075 and characters >255 formatted with %c), the original SV might get
9076 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9077 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9078 valid UTF-8; if the original SV was bytes, the pattern should be too.
9083 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9087 PERL_ARGS_ASSERT_SV_CATPVF;
9089 va_start(args, pat);
9090 sv_vcatpvf(sv, pat, &args);
9095 =for apidoc sv_vcatpvf
9097 Processes its arguments like C<vsprintf> and appends the formatted output
9098 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9100 Usually used via its frontend C<sv_catpvf>.
9106 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9108 PERL_ARGS_ASSERT_SV_VCATPVF;
9110 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9114 =for apidoc sv_catpvf_mg
9116 Like C<sv_catpvf>, but also handles 'set' magic.
9122 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9126 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9128 va_start(args, pat);
9129 sv_vcatpvf_mg(sv, pat, &args);
9134 =for apidoc sv_vcatpvf_mg
9136 Like C<sv_vcatpvf>, but also handles 'set' magic.
9138 Usually used via its frontend C<sv_catpvf_mg>.
9144 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9146 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9148 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9153 =for apidoc sv_vsetpvfn
9155 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9158 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9164 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9165 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9167 PERL_ARGS_ASSERT_SV_VSETPVFN;
9170 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9175 * Warn of missing argument to sprintf, and then return a defined value
9176 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9178 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9180 S_vcatpvfn_missing_argument(pTHX) {
9181 if (ckWARN(WARN_MISSING)) {
9182 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9183 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9190 S_expect_number(pTHX_ char **const pattern)
9195 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9197 switch (**pattern) {
9198 case '1': case '2': case '3':
9199 case '4': case '5': case '6':
9200 case '7': case '8': case '9':
9201 var = *(*pattern)++ - '0';
9202 while (isDIGIT(**pattern)) {
9203 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9205 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9213 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9215 const int neg = nv < 0;
9218 PERL_ARGS_ASSERT_F0CONVERT;
9226 if (uv & 1 && uv == nv)
9227 uv--; /* Round to even */
9229 const unsigned dig = uv % 10;
9242 =for apidoc sv_vcatpvfn
9244 Processes its arguments like C<vsprintf> and appends the formatted output
9245 to an SV. Uses an array of SVs if the C style variable argument list is
9246 missing (NULL). When running with taint checks enabled, indicates via
9247 C<maybe_tainted> if results are untrustworthy (often due to the use of
9250 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9256 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9257 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9258 vec_utf8 = DO_UTF8(vecsv);
9260 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9263 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9264 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9272 static const char nullstr[] = "(null)";
9274 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9275 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9277 /* Times 4: a decimal digit takes more than 3 binary digits.
9278 * NV_DIG: mantissa takes than many decimal digits.
9279 * Plus 32: Playing safe. */
9280 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9281 /* large enough for "%#.#f" --chip */
9282 /* what about long double NVs? --jhi */
9284 PERL_ARGS_ASSERT_SV_VCATPVFN;
9285 PERL_UNUSED_ARG(maybe_tainted);
9287 /* no matter what, this is a string now */
9288 (void)SvPV_force(sv, origlen);
9290 /* special-case "", "%s", and "%-p" (SVf - see below) */
9293 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9295 const char * const s = va_arg(*args, char*);
9296 sv_catpv(sv, s ? s : nullstr);
9298 else if (svix < svmax) {
9299 sv_catsv(sv, *svargs);
9303 if (args && patlen == 3 && pat[0] == '%' &&
9304 pat[1] == '-' && pat[2] == 'p') {
9305 argsv = MUTABLE_SV(va_arg(*args, void*));
9306 sv_catsv(sv, argsv);
9310 #ifndef USE_LONG_DOUBLE
9311 /* special-case "%.<number>[gf]" */
9312 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9313 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9314 unsigned digits = 0;
9318 while (*pp >= '0' && *pp <= '9')
9319 digits = 10 * digits + (*pp++ - '0');
9320 if (pp - pat == (int)patlen - 1) {
9328 /* Add check for digits != 0 because it seems that some
9329 gconverts are buggy in this case, and we don't yet have
9330 a Configure test for this. */
9331 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9332 /* 0, point, slack */
9333 Gconvert(nv, (int)digits, 0, ebuf);
9335 if (*ebuf) /* May return an empty string for digits==0 */
9338 } else if (!digits) {
9341 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9342 sv_catpvn(sv, p, l);
9348 #endif /* !USE_LONG_DOUBLE */
9350 if (!args && svix < svmax && DO_UTF8(*svargs))
9353 patend = (char*)pat + patlen;
9354 for (p = (char*)pat; p < patend; p = q) {
9357 bool vectorize = FALSE;
9358 bool vectorarg = FALSE;
9359 bool vec_utf8 = FALSE;
9365 bool has_precis = FALSE;
9367 const I32 osvix = svix;
9368 bool is_utf8 = FALSE; /* is this item utf8? */
9369 #ifdef HAS_LDBL_SPRINTF_BUG
9370 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9371 with sfio - Allen <allens@cpan.org> */
9372 bool fix_ldbl_sprintf_bug = FALSE;
9376 U8 utf8buf[UTF8_MAXBYTES+1];
9377 STRLEN esignlen = 0;
9379 const char *eptr = NULL;
9380 const char *fmtstart;
9383 const U8 *vecstr = NULL;
9390 /* we need a long double target in case HAS_LONG_DOUBLE but
9393 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9401 const char *dotstr = ".";
9402 STRLEN dotstrlen = 1;
9403 I32 efix = 0; /* explicit format parameter index */
9404 I32 ewix = 0; /* explicit width index */
9405 I32 epix = 0; /* explicit precision index */
9406 I32 evix = 0; /* explicit vector index */
9407 bool asterisk = FALSE;
9409 /* echo everything up to the next format specification */
9410 for (q = p; q < patend && *q != '%'; ++q) ;
9412 if (has_utf8 && !pat_utf8)
9413 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9415 sv_catpvn(sv, p, q - p);
9424 We allow format specification elements in this order:
9425 \d+\$ explicit format parameter index
9427 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9428 0 flag (as above): repeated to allow "v02"
9429 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9430 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9432 [%bcdefginopsuxDFOUX] format (mandatory)
9437 As of perl5.9.3, printf format checking is on by default.
9438 Internally, perl uses %p formats to provide an escape to
9439 some extended formatting. This block deals with those
9440 extensions: if it does not match, (char*)q is reset and
9441 the normal format processing code is used.
9443 Currently defined extensions are:
9444 %p include pointer address (standard)
9445 %-p (SVf) include an SV (previously %_)
9446 %-<num>p include an SV with precision <num>
9447 %<num>p reserved for future extensions
9449 Robin Barker 2005-07-14
9451 %1p (VDf) removed. RMB 2007-10-19
9458 n = expect_number(&q);
9465 argsv = MUTABLE_SV(va_arg(*args, void*));
9466 eptr = SvPV_const(argsv, elen);
9472 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9473 "internal %%<num>p might conflict with future printf extensions");
9479 if ( (width = expect_number(&q)) ) {
9494 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9523 if ( (ewix = expect_number(&q)) )
9532 if ((vectorarg = asterisk)) {
9545 width = expect_number(&q);
9551 vecsv = va_arg(*args, SV*);
9553 vecsv = (evix > 0 && evix <= svmax)
9554 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
9556 vecsv = svix < svmax
9557 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9559 dotstr = SvPV_const(vecsv, dotstrlen);
9560 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9561 bad with tied or overloaded values that return UTF8. */
9564 else if (has_utf8) {
9565 vecsv = sv_mortalcopy(vecsv);
9566 sv_utf8_upgrade(vecsv);
9567 dotstr = SvPV_const(vecsv, dotstrlen);
9574 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9575 vecsv = svargs[efix ? efix-1 : svix++];
9576 vecstr = (U8*)SvPV_const(vecsv,veclen);
9577 vec_utf8 = DO_UTF8(vecsv);
9579 /* if this is a version object, we need to convert
9580 * back into v-string notation and then let the
9581 * vectorize happen normally
9583 if (sv_derived_from(vecsv, "version")) {
9584 char *version = savesvpv(vecsv);
9585 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9586 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9587 "vector argument not supported with alpha versions");
9590 vecsv = sv_newmortal();
9591 scan_vstring(version, version + veclen, vecsv);
9592 vecstr = (U8*)SvPV_const(vecsv, veclen);
9593 vec_utf8 = DO_UTF8(vecsv);
9605 i = va_arg(*args, int);
9607 i = (ewix ? ewix <= svmax : svix < svmax) ?
9608 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9610 width = (i < 0) ? -i : i;
9620 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9622 /* XXX: todo, support specified precision parameter */
9626 i = va_arg(*args, int);
9628 i = (ewix ? ewix <= svmax : svix < svmax)
9629 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9631 has_precis = !(i < 0);
9636 precis = precis * 10 + (*q++ - '0');
9645 case 'I': /* Ix, I32x, and I64x */
9647 if (q[1] == '6' && q[2] == '4') {
9653 if (q[1] == '3' && q[2] == '2') {
9663 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9674 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9675 if (*(q + 1) == 'l') { /* lld, llf */
9701 if (!vectorize && !args) {
9703 const I32 i = efix-1;
9704 argsv = (i >= 0 && i < svmax)
9705 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
9707 argsv = (svix >= 0 && svix < svmax)
9708 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9719 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9721 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9723 eptr = (char*)utf8buf;
9724 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9738 eptr = va_arg(*args, char*);
9740 elen = strlen(eptr);
9742 eptr = (char *)nullstr;
9743 elen = sizeof nullstr - 1;
9747 eptr = SvPV_const(argsv, elen);
9748 if (DO_UTF8(argsv)) {
9749 STRLEN old_precis = precis;
9750 if (has_precis && precis < elen) {
9751 STRLEN ulen = sv_len_utf8(argsv);
9752 I32 p = precis > ulen ? ulen : precis;
9753 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9756 if (width) { /* fudge width (can't fudge elen) */
9757 if (has_precis && precis < elen)
9758 width += precis - old_precis;
9760 width += elen - sv_len_utf8(argsv);
9767 if (has_precis && precis < elen)
9774 if (alt || vectorize)
9776 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9797 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9806 esignbuf[esignlen++] = plus;
9810 case 'h': iv = (short)va_arg(*args, int); break;
9811 case 'l': iv = va_arg(*args, long); break;
9812 case 'V': iv = va_arg(*args, IV); break;
9813 default: iv = va_arg(*args, int); break;
9816 iv = va_arg(*args, Quad_t); break;
9823 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9825 case 'h': iv = (short)tiv; break;
9826 case 'l': iv = (long)tiv; break;
9828 default: iv = tiv; break;
9831 iv = (Quad_t)tiv; break;
9837 if ( !vectorize ) /* we already set uv above */
9842 esignbuf[esignlen++] = plus;
9846 esignbuf[esignlen++] = '-';
9890 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9901 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9902 case 'l': uv = va_arg(*args, unsigned long); break;
9903 case 'V': uv = va_arg(*args, UV); break;
9904 default: uv = va_arg(*args, unsigned); break;
9907 uv = va_arg(*args, Uquad_t); break;
9914 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9916 case 'h': uv = (unsigned short)tuv; break;
9917 case 'l': uv = (unsigned long)tuv; break;
9919 default: uv = tuv; break;
9922 uv = (Uquad_t)tuv; break;
9931 char *ptr = ebuf + sizeof ebuf;
9932 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9938 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9944 esignbuf[esignlen++] = '0';
9945 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9953 if (alt && *ptr != '0')
9962 esignbuf[esignlen++] = '0';
9963 esignbuf[esignlen++] = c;
9966 default: /* it had better be ten or less */
9970 } while (uv /= base);
9973 elen = (ebuf + sizeof ebuf) - ptr;
9977 zeros = precis - elen;
9978 else if (precis == 0 && elen == 1 && *eptr == '0'
9979 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9982 /* a precision nullifies the 0 flag. */
9989 /* FLOATING POINT */
9992 c = 'f'; /* maybe %F isn't supported here */
10000 /* This is evil, but floating point is even more evil */
10002 /* for SV-style calling, we can only get NV
10003 for C-style calling, we assume %f is double;
10004 for simplicity we allow any of %Lf, %llf, %qf for long double
10008 #if defined(USE_LONG_DOUBLE)
10012 /* [perl #20339] - we should accept and ignore %lf rather than die */
10016 #if defined(USE_LONG_DOUBLE)
10017 intsize = args ? 0 : 'q';
10021 #if defined(HAS_LONG_DOUBLE)
10030 /* now we need (long double) if intsize == 'q', else (double) */
10032 #if LONG_DOUBLESIZE > DOUBLESIZE
10034 va_arg(*args, long double) :
10035 va_arg(*args, double)
10037 va_arg(*args, double)
10042 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10043 else. frexp() has some unspecified behaviour for those three */
10044 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10046 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10047 will cast our (long double) to (double) */
10048 (void)Perl_frexp(nv, &i);
10049 if (i == PERL_INT_MIN)
10050 Perl_die(aTHX_ "panic: frexp");
10052 need = BIT_DIGITS(i);
10054 need += has_precis ? precis : 6; /* known default */
10059 #ifdef HAS_LDBL_SPRINTF_BUG
10060 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10061 with sfio - Allen <allens@cpan.org> */
10064 # define MY_DBL_MAX DBL_MAX
10065 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10066 # if DOUBLESIZE >= 8
10067 # define MY_DBL_MAX 1.7976931348623157E+308L
10069 # define MY_DBL_MAX 3.40282347E+38L
10073 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10074 # define MY_DBL_MAX_BUG 1L
10076 # define MY_DBL_MAX_BUG MY_DBL_MAX
10080 # define MY_DBL_MIN DBL_MIN
10081 # else /* XXX guessing! -Allen */
10082 # if DOUBLESIZE >= 8
10083 # define MY_DBL_MIN 2.2250738585072014E-308L
10085 # define MY_DBL_MIN 1.17549435E-38L
10089 if ((intsize == 'q') && (c == 'f') &&
10090 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10091 (need < DBL_DIG)) {
10092 /* it's going to be short enough that
10093 * long double precision is not needed */
10095 if ((nv <= 0L) && (nv >= -0L))
10096 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10098 /* would use Perl_fp_class as a double-check but not
10099 * functional on IRIX - see perl.h comments */
10101 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10102 /* It's within the range that a double can represent */
10103 #if defined(DBL_MAX) && !defined(DBL_MIN)
10104 if ((nv >= ((long double)1/DBL_MAX)) ||
10105 (nv <= (-(long double)1/DBL_MAX)))
10107 fix_ldbl_sprintf_bug = TRUE;
10110 if (fix_ldbl_sprintf_bug == TRUE) {
10120 # undef MY_DBL_MAX_BUG
10123 #endif /* HAS_LDBL_SPRINTF_BUG */
10125 need += 20; /* fudge factor */
10126 if (PL_efloatsize < need) {
10127 Safefree(PL_efloatbuf);
10128 PL_efloatsize = need + 20; /* more fudge */
10129 Newx(PL_efloatbuf, PL_efloatsize, char);
10130 PL_efloatbuf[0] = '\0';
10133 if ( !(width || left || plus || alt) && fill != '0'
10134 && has_precis && intsize != 'q' ) { /* Shortcuts */
10135 /* See earlier comment about buggy Gconvert when digits,
10137 if ( c == 'g' && precis) {
10138 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10139 /* May return an empty string for digits==0 */
10140 if (*PL_efloatbuf) {
10141 elen = strlen(PL_efloatbuf);
10142 goto float_converted;
10144 } else if ( c == 'f' && !precis) {
10145 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10150 char *ptr = ebuf + sizeof ebuf;
10153 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10154 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10155 if (intsize == 'q') {
10156 /* Copy the one or more characters in a long double
10157 * format before the 'base' ([efgEFG]) character to
10158 * the format string. */
10159 static char const prifldbl[] = PERL_PRIfldbl;
10160 char const *p = prifldbl + sizeof(prifldbl) - 3;
10161 while (p >= prifldbl) { *--ptr = *p--; }
10166 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10171 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10183 /* No taint. Otherwise we are in the strange situation
10184 * where printf() taints but print($float) doesn't.
10186 #if defined(HAS_LONG_DOUBLE)
10187 elen = ((intsize == 'q')
10188 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10189 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10191 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10195 eptr = PL_efloatbuf;
10203 i = SvCUR(sv) - origlen;
10206 case 'h': *(va_arg(*args, short*)) = i; break;
10207 default: *(va_arg(*args, int*)) = i; break;
10208 case 'l': *(va_arg(*args, long*)) = i; break;
10209 case 'V': *(va_arg(*args, IV*)) = i; break;
10212 *(va_arg(*args, Quad_t*)) = i; break;
10219 sv_setuv_mg(argsv, (UV)i);
10220 continue; /* not "break" */
10227 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10228 && ckWARN(WARN_PRINTF))
10230 SV * const msg = sv_newmortal();
10231 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10232 (PL_op->op_type == OP_PRTF) ? "" : "s");
10233 if (fmtstart < patend) {
10234 const char * const fmtend = q < patend ? q : patend;
10236 sv_catpvs(msg, "\"%");
10237 for (f = fmtstart; f < fmtend; f++) {
10239 sv_catpvn(msg, f, 1);
10241 Perl_sv_catpvf(aTHX_ msg,
10242 "\\%03"UVof, (UV)*f & 0xFF);
10245 sv_catpvs(msg, "\"");
10247 sv_catpvs(msg, "end of string");
10249 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10252 /* output mangled stuff ... */
10258 /* ... right here, because formatting flags should not apply */
10259 SvGROW(sv, SvCUR(sv) + elen + 1);
10261 Copy(eptr, p, elen, char);
10264 SvCUR_set(sv, p - SvPVX_const(sv));
10266 continue; /* not "break" */
10269 if (is_utf8 != has_utf8) {
10272 sv_utf8_upgrade(sv);
10275 const STRLEN old_elen = elen;
10276 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10277 sv_utf8_upgrade(nsv);
10278 eptr = SvPVX_const(nsv);
10281 if (width) { /* fudge width (can't fudge elen) */
10282 width += elen - old_elen;
10288 have = esignlen + zeros + elen;
10290 Perl_croak_nocontext("%s", PL_memory_wrap);
10292 need = (have > width ? have : width);
10295 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10296 Perl_croak_nocontext("%s", PL_memory_wrap);
10297 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10299 if (esignlen && fill == '0') {
10301 for (i = 0; i < (int)esignlen; i++)
10302 *p++ = esignbuf[i];
10304 if (gap && !left) {
10305 memset(p, fill, gap);
10308 if (esignlen && fill != '0') {
10310 for (i = 0; i < (int)esignlen; i++)
10311 *p++ = esignbuf[i];
10315 for (i = zeros; i; i--)
10319 Copy(eptr, p, elen, char);
10323 memset(p, ' ', gap);
10328 Copy(dotstr, p, dotstrlen, char);
10332 vectorize = FALSE; /* done iterating over vecstr */
10339 SvCUR_set(sv, p - SvPVX_const(sv));
10347 /* =========================================================================
10349 =head1 Cloning an interpreter
10351 All the macros and functions in this section are for the private use of
10352 the main function, perl_clone().
10354 The foo_dup() functions make an exact copy of an existing foo thingy.
10355 During the course of a cloning, a hash table is used to map old addresses
10356 to new addresses. The table is created and manipulated with the
10357 ptr_table_* functions.
10361 * =========================================================================*/
10364 #if defined(USE_ITHREADS)
10366 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10367 #ifndef GpREFCNT_inc
10368 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10372 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10373 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10374 If this changes, please unmerge ss_dup.
10375 Likewise, sv_dup_inc_multiple() relies on this fact. */
10376 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10377 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10378 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10379 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10380 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10381 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10382 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10383 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10384 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10385 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10386 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10387 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10388 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10389 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10391 /* clone a parser */
10394 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10398 PERL_ARGS_ASSERT_PARSER_DUP;
10403 /* look for it in the table first */
10404 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10408 /* create anew and remember what it is */
10409 Newxz(parser, 1, yy_parser);
10410 ptr_table_store(PL_ptr_table, proto, parser);
10412 parser->yyerrstatus = 0;
10413 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10415 /* XXX these not yet duped */
10416 parser->old_parser = NULL;
10417 parser->stack = NULL;
10419 parser->stack_size = 0;
10420 /* XXX parser->stack->state = 0; */
10422 /* XXX eventually, just Copy() most of the parser struct ? */
10424 parser->lex_brackets = proto->lex_brackets;
10425 parser->lex_casemods = proto->lex_casemods;
10426 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10427 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10428 parser->lex_casestack = savepvn(proto->lex_casestack,
10429 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10430 parser->lex_defer = proto->lex_defer;
10431 parser->lex_dojoin = proto->lex_dojoin;
10432 parser->lex_expect = proto->lex_expect;
10433 parser->lex_formbrack = proto->lex_formbrack;
10434 parser->lex_inpat = proto->lex_inpat;
10435 parser->lex_inwhat = proto->lex_inwhat;
10436 parser->lex_op = proto->lex_op;
10437 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10438 parser->lex_starts = proto->lex_starts;
10439 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10440 parser->multi_close = proto->multi_close;
10441 parser->multi_open = proto->multi_open;
10442 parser->multi_start = proto->multi_start;
10443 parser->multi_end = proto->multi_end;
10444 parser->pending_ident = proto->pending_ident;
10445 parser->preambled = proto->preambled;
10446 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10447 parser->linestr = sv_dup_inc(proto->linestr, param);
10448 parser->expect = proto->expect;
10449 parser->copline = proto->copline;
10450 parser->last_lop_op = proto->last_lop_op;
10451 parser->lex_state = proto->lex_state;
10452 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10453 /* rsfp_filters entries have fake IoDIRP() */
10454 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10455 parser->in_my = proto->in_my;
10456 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10457 parser->error_count = proto->error_count;
10460 parser->linestr = sv_dup_inc(proto->linestr, param);
10463 char * const ols = SvPVX(proto->linestr);
10464 char * const ls = SvPVX(parser->linestr);
10466 parser->bufptr = ls + (proto->bufptr >= ols ?
10467 proto->bufptr - ols : 0);
10468 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10469 proto->oldbufptr - ols : 0);
10470 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10471 proto->oldoldbufptr - ols : 0);
10472 parser->linestart = ls + (proto->linestart >= ols ?
10473 proto->linestart - ols : 0);
10474 parser->last_uni = ls + (proto->last_uni >= ols ?
10475 proto->last_uni - ols : 0);
10476 parser->last_lop = ls + (proto->last_lop >= ols ?
10477 proto->last_lop - ols : 0);
10479 parser->bufend = ls + SvCUR(parser->linestr);
10482 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10486 parser->endwhite = proto->endwhite;
10487 parser->faketokens = proto->faketokens;
10488 parser->lasttoke = proto->lasttoke;
10489 parser->nextwhite = proto->nextwhite;
10490 parser->realtokenstart = proto->realtokenstart;
10491 parser->skipwhite = proto->skipwhite;
10492 parser->thisclose = proto->thisclose;
10493 parser->thismad = proto->thismad;
10494 parser->thisopen = proto->thisopen;
10495 parser->thisstuff = proto->thisstuff;
10496 parser->thistoken = proto->thistoken;
10497 parser->thiswhite = proto->thiswhite;
10499 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10500 parser->curforce = proto->curforce;
10502 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10503 Copy(proto->nexttype, parser->nexttype, 5, I32);
10504 parser->nexttoke = proto->nexttoke;
10507 /* XXX should clone saved_curcop here, but we aren't passed
10508 * proto_perl; so do it in perl_clone_using instead */
10514 /* duplicate a file handle */
10517 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10521 PERL_ARGS_ASSERT_FP_DUP;
10522 PERL_UNUSED_ARG(type);
10525 return (PerlIO*)NULL;
10527 /* look for it in the table first */
10528 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10532 /* create anew and remember what it is */
10533 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10534 ptr_table_store(PL_ptr_table, fp, ret);
10538 /* duplicate a directory handle */
10541 Perl_dirp_dup(pTHX_ DIR *const dp)
10543 PERL_UNUSED_CONTEXT;
10550 /* duplicate a typeglob */
10553 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10557 PERL_ARGS_ASSERT_GP_DUP;
10561 /* look for it in the table first */
10562 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10566 /* create anew and remember what it is */
10568 ptr_table_store(PL_ptr_table, gp, ret);
10571 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10572 on Newxz() to do this for us. */
10573 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10574 ret->gp_io = io_dup_inc(gp->gp_io, param);
10575 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10576 ret->gp_av = av_dup_inc(gp->gp_av, param);
10577 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10578 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10579 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10580 ret->gp_cvgen = gp->gp_cvgen;
10581 ret->gp_line = gp->gp_line;
10582 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10586 /* duplicate a chain of magic */
10589 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10591 MAGIC *mgret = NULL;
10592 MAGIC **mgprev_p = &mgret;
10594 PERL_ARGS_ASSERT_MG_DUP;
10596 for (; mg; mg = mg->mg_moremagic) {
10598 Newx(nmg, 1, MAGIC);
10600 mgprev_p = &(nmg->mg_moremagic);
10602 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10603 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10604 from the original commit adding Perl_mg_dup() - revision 4538.
10605 Similarly there is the annotation "XXX random ptr?" next to the
10606 assignment to nmg->mg_ptr. */
10609 /* FIXME for plugins
10610 if (nmg->mg_type == PERL_MAGIC_qr) {
10611 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10615 if(nmg->mg_type == PERL_MAGIC_backref) {
10616 /* The backref AV has its reference count deliberately bumped by
10619 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10622 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10623 ? sv_dup_inc(nmg->mg_obj, param)
10624 : sv_dup(nmg->mg_obj, param);
10627 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10628 if (nmg->mg_len > 0) {
10629 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10630 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10631 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10633 AMT * const namtp = (AMT*)nmg->mg_ptr;
10634 sv_dup_inc_multiple((SV**)(namtp->table),
10635 (SV**)(namtp->table), NofAMmeth, param);
10638 else if (nmg->mg_len == HEf_SVKEY)
10639 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10641 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10642 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10648 #endif /* USE_ITHREADS */
10650 /* create a new pointer-mapping table */
10653 Perl_ptr_table_new(pTHX)
10656 PERL_UNUSED_CONTEXT;
10658 Newx(tbl, 1, PTR_TBL_t);
10659 tbl->tbl_max = 511;
10660 tbl->tbl_items = 0;
10661 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10665 #define PTR_TABLE_HASH(ptr) \
10666 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10669 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10670 following define) and at call to new_body_inline made below in
10671 Perl_ptr_table_store()
10674 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10676 /* map an existing pointer using a table */
10678 STATIC PTR_TBL_ENT_t *
10679 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10681 PTR_TBL_ENT_t *tblent;
10682 const UV hash = PTR_TABLE_HASH(sv);
10684 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10686 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10687 for (; tblent; tblent = tblent->next) {
10688 if (tblent->oldval == sv)
10695 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10697 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10699 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10700 PERL_UNUSED_CONTEXT;
10702 return tblent ? tblent->newval : NULL;
10705 /* add a new entry to a pointer-mapping table */
10708 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10710 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10712 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10713 PERL_UNUSED_CONTEXT;
10716 tblent->newval = newsv;
10718 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10720 new_body_inline(tblent, PTE_SVSLOT);
10722 tblent->oldval = oldsv;
10723 tblent->newval = newsv;
10724 tblent->next = tbl->tbl_ary[entry];
10725 tbl->tbl_ary[entry] = tblent;
10727 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10728 ptr_table_split(tbl);
10732 /* double the hash bucket size of an existing ptr table */
10735 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10737 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10738 const UV oldsize = tbl->tbl_max + 1;
10739 UV newsize = oldsize * 2;
10742 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10743 PERL_UNUSED_CONTEXT;
10745 Renew(ary, newsize, PTR_TBL_ENT_t*);
10746 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10747 tbl->tbl_max = --newsize;
10748 tbl->tbl_ary = ary;
10749 for (i=0; i < oldsize; i++, ary++) {
10750 PTR_TBL_ENT_t **curentp, **entp, *ent;
10753 curentp = ary + oldsize;
10754 for (entp = ary, ent = *ary; ent; ent = *entp) {
10755 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10757 ent->next = *curentp;
10767 /* remove all the entries from a ptr table */
10770 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10772 if (tbl && tbl->tbl_items) {
10773 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10774 UV riter = tbl->tbl_max;
10777 PTR_TBL_ENT_t *entry = array[riter];
10780 PTR_TBL_ENT_t * const oentry = entry;
10781 entry = entry->next;
10786 tbl->tbl_items = 0;
10790 /* clear and free a ptr table */
10793 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10798 ptr_table_clear(tbl);
10799 Safefree(tbl->tbl_ary);
10803 #if defined(USE_ITHREADS)
10806 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10808 PERL_ARGS_ASSERT_RVPV_DUP;
10811 SvRV_set(dstr, SvWEAKREF(sstr)
10812 ? sv_dup(SvRV_const(sstr), param)
10813 : sv_dup_inc(SvRV_const(sstr), param));
10816 else if (SvPVX_const(sstr)) {
10817 /* Has something there */
10819 /* Normal PV - clone whole allocated space */
10820 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10821 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10822 /* Not that normal - actually sstr is copy on write.
10823 But we are a true, independant SV, so: */
10824 SvREADONLY_off(dstr);
10829 /* Special case - not normally malloced for some reason */
10830 if (isGV_with_GP(sstr)) {
10831 /* Don't need to do anything here. */
10833 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10834 /* A "shared" PV - clone it as "shared" PV */
10836 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10840 /* Some other special case - random pointer */
10841 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10846 /* Copy the NULL */
10847 SvPV_set(dstr, NULL);
10851 /* duplicate a list of SVs. source and dest may point to the same memory. */
10853 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10854 SSize_t items, CLONE_PARAMS *const param)
10856 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10858 while (items-- > 0) {
10859 *dest++ = sv_dup_inc(*source++, param);
10865 /* duplicate an SV of any type (including AV, HV etc) */
10868 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10873 PERL_ARGS_ASSERT_SV_DUP;
10877 if (SvTYPE(sstr) == SVTYPEMASK) {
10878 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10883 /* look for it in the table first */
10884 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10888 if(param->flags & CLONEf_JOIN_IN) {
10889 /** We are joining here so we don't want do clone
10890 something that is bad **/
10891 if (SvTYPE(sstr) == SVt_PVHV) {
10892 const HEK * const hvname = HvNAME_HEK(sstr);
10894 /** don't clone stashes if they already exist **/
10895 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10899 /* create anew and remember what it is */
10902 #ifdef DEBUG_LEAKING_SCALARS
10903 dstr->sv_debug_optype = sstr->sv_debug_optype;
10904 dstr->sv_debug_line = sstr->sv_debug_line;
10905 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10906 dstr->sv_debug_cloned = 1;
10907 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10910 ptr_table_store(PL_ptr_table, sstr, dstr);
10913 SvFLAGS(dstr) = SvFLAGS(sstr);
10914 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10915 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10918 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10919 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10920 (void*)PL_watch_pvx, SvPVX_const(sstr));
10923 /* don't clone objects whose class has asked us not to */
10924 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10929 switch (SvTYPE(sstr)) {
10931 SvANY(dstr) = NULL;
10934 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10936 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10938 SvIV_set(dstr, SvIVX(sstr));
10942 SvANY(dstr) = new_XNV();
10943 SvNV_set(dstr, SvNVX(sstr));
10945 /* case SVt_BIND: */
10948 /* These are all the types that need complex bodies allocating. */
10950 const svtype sv_type = SvTYPE(sstr);
10951 const struct body_details *const sv_type_details
10952 = bodies_by_type + sv_type;
10956 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10971 assert(sv_type_details->body_size);
10972 if (sv_type_details->arena) {
10973 new_body_inline(new_body, sv_type);
10975 = (void*)((char*)new_body - sv_type_details->offset);
10977 new_body = new_NOARENA(sv_type_details);
10981 SvANY(dstr) = new_body;
10984 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10985 ((char*)SvANY(dstr)) + sv_type_details->offset,
10986 sv_type_details->copy, char);
10988 Copy(((char*)SvANY(sstr)),
10989 ((char*)SvANY(dstr)),
10990 sv_type_details->body_size + sv_type_details->offset, char);
10993 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10994 && !isGV_with_GP(dstr))
10995 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10997 /* The Copy above means that all the source (unduplicated) pointers
10998 are now in the destination. We can check the flags and the
10999 pointers in either, but it's possible that there's less cache
11000 missing by always going for the destination.
11001 FIXME - instrument and check that assumption */
11002 if (sv_type >= SVt_PVMG) {
11003 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11004 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11005 } else if (SvMAGIC(dstr))
11006 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11008 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11011 /* The cast silences a GCC warning about unhandled types. */
11012 switch ((int)sv_type) {
11022 /* FIXME for plugins */
11023 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11026 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11027 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11028 LvTARG(dstr) = dstr;
11029 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11030 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11032 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11034 if(isGV_with_GP(sstr)) {
11035 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11036 /* Don't call sv_add_backref here as it's going to be
11037 created as part of the magic cloning of the symbol
11039 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11040 at the point of this comment. */
11041 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11042 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11043 (void)GpREFCNT_inc(GvGP(dstr));
11045 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11048 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11049 if (IoOFP(dstr) == IoIFP(sstr))
11050 IoOFP(dstr) = IoIFP(dstr);
11052 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11053 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11054 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11055 /* I have no idea why fake dirp (rsfps)
11056 should be treated differently but otherwise
11057 we end up with leaks -- sky*/
11058 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11059 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11060 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11062 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11063 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11064 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11065 if (IoDIRP(dstr)) {
11066 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11069 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11072 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11073 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11074 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11077 /* avoid cloning an empty array */
11078 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11079 SV **dst_ary, **src_ary;
11080 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11082 src_ary = AvARRAY((const AV *)sstr);
11083 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11084 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11085 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11086 AvALLOC((const AV *)dstr) = dst_ary;
11087 if (AvREAL((const AV *)sstr)) {
11088 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11092 while (items-- > 0)
11093 *dst_ary++ = sv_dup(*src_ary++, param);
11095 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11096 while (items-- > 0) {
11097 *dst_ary++ = &PL_sv_undef;
11101 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11102 AvALLOC((const AV *)dstr) = (SV**)NULL;
11103 AvMAX( (const AV *)dstr) = -1;
11104 AvFILLp((const AV *)dstr) = -1;
11108 if (HvARRAY((const HV *)sstr)) {
11110 const bool sharekeys = !!HvSHAREKEYS(sstr);
11111 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11112 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11114 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11115 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11117 HvARRAY(dstr) = (HE**)darray;
11118 while (i <= sxhv->xhv_max) {
11119 const HE * const source = HvARRAY(sstr)[i];
11120 HvARRAY(dstr)[i] = source
11121 ? he_dup(source, sharekeys, param) : 0;
11126 const struct xpvhv_aux * const saux = HvAUX(sstr);
11127 struct xpvhv_aux * const daux = HvAUX(dstr);
11128 /* This flag isn't copied. */
11129 /* SvOOK_on(hv) attacks the IV flags. */
11130 SvFLAGS(dstr) |= SVf_OOK;
11132 hvname = saux->xhv_name;
11133 daux->xhv_name = hek_dup(hvname, param);
11135 daux->xhv_riter = saux->xhv_riter;
11136 daux->xhv_eiter = saux->xhv_eiter
11137 ? he_dup(saux->xhv_eiter,
11138 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11139 /* backref array needs refcnt=2; see sv_add_backref */
11140 daux->xhv_backreferences =
11141 saux->xhv_backreferences
11142 ? MUTABLE_AV(SvREFCNT_inc(
11143 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11146 daux->xhv_mro_meta = saux->xhv_mro_meta
11147 ? mro_meta_dup(saux->xhv_mro_meta, param)
11150 /* Record stashes for possible cloning in Perl_clone(). */
11152 av_push(param->stashes, dstr);
11156 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11159 if (!(param->flags & CLONEf_COPY_STACKS)) {
11163 /* NOTE: not refcounted */
11164 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11166 if (!CvISXSUB(dstr))
11167 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11169 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11170 CvXSUBANY(dstr).any_ptr =
11171 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11173 /* don't dup if copying back - CvGV isn't refcounted, so the
11174 * duped GV may never be freed. A bit of a hack! DAPM */
11175 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11176 NULL : gv_dup(CvGV(dstr), param) ;
11177 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11179 CvWEAKOUTSIDE(sstr)
11180 ? cv_dup( CvOUTSIDE(dstr), param)
11181 : cv_dup_inc(CvOUTSIDE(dstr), param);
11182 if (!CvISXSUB(dstr))
11183 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11189 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11195 /* duplicate a context */
11198 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11200 PERL_CONTEXT *ncxs;
11202 PERL_ARGS_ASSERT_CX_DUP;
11205 return (PERL_CONTEXT*)NULL;
11207 /* look for it in the table first */
11208 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11212 /* create anew and remember what it is */
11213 Newx(ncxs, max + 1, PERL_CONTEXT);
11214 ptr_table_store(PL_ptr_table, cxs, ncxs);
11215 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11218 PERL_CONTEXT * const ncx = &ncxs[ix];
11219 if (CxTYPE(ncx) == CXt_SUBST) {
11220 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11223 switch (CxTYPE(ncx)) {
11225 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11226 ? cv_dup_inc(ncx->blk_sub.cv, param)
11227 : cv_dup(ncx->blk_sub.cv,param));
11228 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11229 ? av_dup_inc(ncx->blk_sub.argarray,
11232 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11234 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11235 ncx->blk_sub.oldcomppad);
11238 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11240 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11242 case CXt_LOOP_LAZYSV:
11243 ncx->blk_loop.state_u.lazysv.end
11244 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11245 /* We are taking advantage of av_dup_inc and sv_dup_inc
11246 actually being the same function, and order equivalance of
11248 We can assert the later [but only at run time :-(] */
11249 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11250 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11252 ncx->blk_loop.state_u.ary.ary
11253 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11254 case CXt_LOOP_LAZYIV:
11255 case CXt_LOOP_PLAIN:
11256 if (CxPADLOOP(ncx)) {
11257 ncx->blk_loop.oldcomppad
11258 = (PAD*)ptr_table_fetch(PL_ptr_table,
11259 ncx->blk_loop.oldcomppad);
11261 ncx->blk_loop.oldcomppad
11262 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11267 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11268 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11269 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11282 /* duplicate a stack info structure */
11285 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11289 PERL_ARGS_ASSERT_SI_DUP;
11292 return (PERL_SI*)NULL;
11294 /* look for it in the table first */
11295 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11299 /* create anew and remember what it is */
11300 Newxz(nsi, 1, PERL_SI);
11301 ptr_table_store(PL_ptr_table, si, nsi);
11303 nsi->si_stack = av_dup_inc(si->si_stack, param);
11304 nsi->si_cxix = si->si_cxix;
11305 nsi->si_cxmax = si->si_cxmax;
11306 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11307 nsi->si_type = si->si_type;
11308 nsi->si_prev = si_dup(si->si_prev, param);
11309 nsi->si_next = si_dup(si->si_next, param);
11310 nsi->si_markoff = si->si_markoff;
11315 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11316 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11317 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11318 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11319 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11320 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11321 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11322 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11323 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11324 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11325 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11326 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11327 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11328 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11331 #define pv_dup_inc(p) SAVEPV(p)
11332 #define pv_dup(p) SAVEPV(p)
11333 #define svp_dup_inc(p,pp) any_dup(p,pp)
11335 /* map any object to the new equivent - either something in the
11336 * ptr table, or something in the interpreter structure
11340 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11344 PERL_ARGS_ASSERT_ANY_DUP;
11347 return (void*)NULL;
11349 /* look for it in the table first */
11350 ret = ptr_table_fetch(PL_ptr_table, v);
11354 /* see if it is part of the interpreter structure */
11355 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11356 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11364 /* duplicate the save stack */
11367 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11370 ANY * const ss = proto_perl->Isavestack;
11371 const I32 max = proto_perl->Isavestack_max;
11372 I32 ix = proto_perl->Isavestack_ix;
11385 void (*dptr) (void*);
11386 void (*dxptr) (pTHX_ void*);
11388 PERL_ARGS_ASSERT_SS_DUP;
11390 Newxz(nss, max, ANY);
11393 const I32 type = POPINT(ss,ix);
11394 TOPINT(nss,ix) = type;
11396 case SAVEt_HELEM: /* hash element */
11397 sv = (const SV *)POPPTR(ss,ix);
11398 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11400 case SAVEt_ITEM: /* normal string */
11401 case SAVEt_SV: /* scalar reference */
11402 sv = (const SV *)POPPTR(ss,ix);
11403 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11406 case SAVEt_MORTALIZESV:
11407 sv = (const SV *)POPPTR(ss,ix);
11408 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11410 case SAVEt_SHARED_PVREF: /* char* in shared space */
11411 c = (char*)POPPTR(ss,ix);
11412 TOPPTR(nss,ix) = savesharedpv(c);
11413 ptr = POPPTR(ss,ix);
11414 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11416 case SAVEt_GENERIC_SVREF: /* generic sv */
11417 case SAVEt_SVREF: /* scalar reference */
11418 sv = (const SV *)POPPTR(ss,ix);
11419 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11420 ptr = POPPTR(ss,ix);
11421 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11423 case SAVEt_HV: /* hash reference */
11424 case SAVEt_AV: /* array reference */
11425 sv = (const SV *) POPPTR(ss,ix);
11426 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11428 case SAVEt_COMPPAD:
11430 sv = (const SV *) POPPTR(ss,ix);
11431 TOPPTR(nss,ix) = sv_dup(sv, param);
11433 case SAVEt_INT: /* int reference */
11434 ptr = POPPTR(ss,ix);
11435 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11436 intval = (int)POPINT(ss,ix);
11437 TOPINT(nss,ix) = intval;
11439 case SAVEt_LONG: /* long reference */
11440 ptr = POPPTR(ss,ix);
11441 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11443 case SAVEt_CLEARSV:
11444 longval = (long)POPLONG(ss,ix);
11445 TOPLONG(nss,ix) = longval;
11447 case SAVEt_I32: /* I32 reference */
11448 case SAVEt_I16: /* I16 reference */
11449 case SAVEt_I8: /* I8 reference */
11450 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11451 ptr = POPPTR(ss,ix);
11452 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11454 TOPINT(nss,ix) = i;
11456 case SAVEt_IV: /* IV reference */
11457 ptr = POPPTR(ss,ix);
11458 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11460 TOPIV(nss,ix) = iv;
11462 case SAVEt_HPTR: /* HV* reference */
11463 case SAVEt_APTR: /* AV* reference */
11464 case SAVEt_SPTR: /* SV* reference */
11465 ptr = POPPTR(ss,ix);
11466 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11467 sv = (const SV *)POPPTR(ss,ix);
11468 TOPPTR(nss,ix) = sv_dup(sv, param);
11470 case SAVEt_VPTR: /* random* reference */
11471 ptr = POPPTR(ss,ix);
11472 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11473 ptr = POPPTR(ss,ix);
11474 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11476 case SAVEt_GENERIC_PVREF: /* generic char* */
11477 case SAVEt_PPTR: /* char* reference */
11478 ptr = POPPTR(ss,ix);
11479 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11480 c = (char*)POPPTR(ss,ix);
11481 TOPPTR(nss,ix) = pv_dup(c);
11483 case SAVEt_GP: /* scalar reference */
11484 gp = (GP*)POPPTR(ss,ix);
11485 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11486 (void)GpREFCNT_inc(gp);
11487 gv = (const GV *)POPPTR(ss,ix);
11488 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11491 ptr = POPPTR(ss,ix);
11492 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11493 /* these are assumed to be refcounted properly */
11495 switch (((OP*)ptr)->op_type) {
11497 case OP_LEAVESUBLV:
11501 case OP_LEAVEWRITE:
11502 TOPPTR(nss,ix) = ptr;
11505 (void) OpREFCNT_inc(o);
11509 TOPPTR(nss,ix) = NULL;
11514 TOPPTR(nss,ix) = NULL;
11517 hv = (const HV *)POPPTR(ss,ix);
11518 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11520 TOPINT(nss,ix) = i;
11523 c = (char*)POPPTR(ss,ix);
11524 TOPPTR(nss,ix) = pv_dup_inc(c);
11526 case SAVEt_STACK_POS: /* Position on Perl stack */
11528 TOPINT(nss,ix) = i;
11530 case SAVEt_DESTRUCTOR:
11531 ptr = POPPTR(ss,ix);
11532 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11533 dptr = POPDPTR(ss,ix);
11534 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11535 any_dup(FPTR2DPTR(void *, dptr),
11538 case SAVEt_DESTRUCTOR_X:
11539 ptr = POPPTR(ss,ix);
11540 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11541 dxptr = POPDXPTR(ss,ix);
11542 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11543 any_dup(FPTR2DPTR(void *, dxptr),
11546 case SAVEt_REGCONTEXT:
11549 TOPINT(nss,ix) = i;
11552 case SAVEt_AELEM: /* array element */
11553 sv = (const SV *)POPPTR(ss,ix);
11554 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11556 TOPINT(nss,ix) = i;
11557 av = (const AV *)POPPTR(ss,ix);
11558 TOPPTR(nss,ix) = av_dup_inc(av, param);
11561 ptr = POPPTR(ss,ix);
11562 TOPPTR(nss,ix) = ptr;
11565 ptr = POPPTR(ss,ix);
11568 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11569 HINTS_REFCNT_UNLOCK;
11571 TOPPTR(nss,ix) = ptr;
11573 TOPINT(nss,ix) = i;
11574 if (i & HINT_LOCALIZE_HH) {
11575 hv = (const HV *)POPPTR(ss,ix);
11576 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11579 case SAVEt_PADSV_AND_MORTALIZE:
11580 longval = (long)POPLONG(ss,ix);
11581 TOPLONG(nss,ix) = longval;
11582 ptr = POPPTR(ss,ix);
11583 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11584 sv = (const SV *)POPPTR(ss,ix);
11585 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11588 ptr = POPPTR(ss,ix);
11589 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11590 longval = (long)POPBOOL(ss,ix);
11591 TOPBOOL(nss,ix) = (bool)longval;
11593 case SAVEt_SET_SVFLAGS:
11595 TOPINT(nss,ix) = i;
11597 TOPINT(nss,ix) = i;
11598 sv = (const SV *)POPPTR(ss,ix);
11599 TOPPTR(nss,ix) = sv_dup(sv, param);
11601 case SAVEt_RE_STATE:
11603 const struct re_save_state *const old_state
11604 = (struct re_save_state *)
11605 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11606 struct re_save_state *const new_state
11607 = (struct re_save_state *)
11608 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11610 Copy(old_state, new_state, 1, struct re_save_state);
11611 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11613 new_state->re_state_bostr
11614 = pv_dup(old_state->re_state_bostr);
11615 new_state->re_state_reginput
11616 = pv_dup(old_state->re_state_reginput);
11617 new_state->re_state_regeol
11618 = pv_dup(old_state->re_state_regeol);
11619 new_state->re_state_regoffs
11620 = (regexp_paren_pair*)
11621 any_dup(old_state->re_state_regoffs, proto_perl);
11622 new_state->re_state_reglastparen
11623 = (U32*) any_dup(old_state->re_state_reglastparen,
11625 new_state->re_state_reglastcloseparen
11626 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11628 /* XXX This just has to be broken. The old save_re_context
11629 code did SAVEGENERICPV(PL_reg_start_tmp);
11630 PL_reg_start_tmp is char **.
11631 Look above to what the dup code does for
11632 SAVEt_GENERIC_PVREF
11633 It can never have worked.
11634 So this is merely a faithful copy of the exiting bug: */
11635 new_state->re_state_reg_start_tmp
11636 = (char **) pv_dup((char *)
11637 old_state->re_state_reg_start_tmp);
11638 /* I assume that it only ever "worked" because no-one called
11639 (pseudo)fork while the regexp engine had re-entered itself.
11641 #ifdef PERL_OLD_COPY_ON_WRITE
11642 new_state->re_state_nrs
11643 = sv_dup(old_state->re_state_nrs, param);
11645 new_state->re_state_reg_magic
11646 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11648 new_state->re_state_reg_oldcurpm
11649 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11651 new_state->re_state_reg_curpm
11652 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11654 new_state->re_state_reg_oldsaved
11655 = pv_dup(old_state->re_state_reg_oldsaved);
11656 new_state->re_state_reg_poscache
11657 = pv_dup(old_state->re_state_reg_poscache);
11658 new_state->re_state_reg_starttry
11659 = pv_dup(old_state->re_state_reg_starttry);
11662 case SAVEt_COMPILE_WARNINGS:
11663 ptr = POPPTR(ss,ix);
11664 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11667 ptr = POPPTR(ss,ix);
11668 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11672 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11680 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11681 * flag to the result. This is done for each stash before cloning starts,
11682 * so we know which stashes want their objects cloned */
11685 do_mark_cloneable_stash(pTHX_ SV *const sv)
11687 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11689 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11690 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11691 if (cloner && GvCV(cloner)) {
11698 mXPUSHs(newSVhek(hvname));
11700 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11707 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11715 =for apidoc perl_clone
11717 Create and return a new interpreter by cloning the current one.
11719 perl_clone takes these flags as parameters:
11721 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11722 without it we only clone the data and zero the stacks,
11723 with it we copy the stacks and the new perl interpreter is
11724 ready to run at the exact same point as the previous one.
11725 The pseudo-fork code uses COPY_STACKS while the
11726 threads->create doesn't.
11728 CLONEf_KEEP_PTR_TABLE
11729 perl_clone keeps a ptr_table with the pointer of the old
11730 variable as a key and the new variable as a value,
11731 this allows it to check if something has been cloned and not
11732 clone it again but rather just use the value and increase the
11733 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11734 the ptr_table using the function
11735 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11736 reason to keep it around is if you want to dup some of your own
11737 variable who are outside the graph perl scans, example of this
11738 code is in threads.xs create
11741 This is a win32 thing, it is ignored on unix, it tells perls
11742 win32host code (which is c++) to clone itself, this is needed on
11743 win32 if you want to run two threads at the same time,
11744 if you just want to do some stuff in a separate perl interpreter
11745 and then throw it away and return to the original one,
11746 you don't need to do anything.
11751 /* XXX the above needs expanding by someone who actually understands it ! */
11752 EXTERN_C PerlInterpreter *
11753 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11756 perl_clone(PerlInterpreter *proto_perl, UV flags)
11759 #ifdef PERL_IMPLICIT_SYS
11761 PERL_ARGS_ASSERT_PERL_CLONE;
11763 /* perlhost.h so we need to call into it
11764 to clone the host, CPerlHost should have a c interface, sky */
11766 if (flags & CLONEf_CLONE_HOST) {
11767 return perl_clone_host(proto_perl,flags);
11769 return perl_clone_using(proto_perl, flags,
11771 proto_perl->IMemShared,
11772 proto_perl->IMemParse,
11774 proto_perl->IStdIO,
11778 proto_perl->IProc);
11782 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11783 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11784 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11785 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11786 struct IPerlDir* ipD, struct IPerlSock* ipS,
11787 struct IPerlProc* ipP)
11789 /* XXX many of the string copies here can be optimized if they're
11790 * constants; they need to be allocated as common memory and just
11791 * their pointers copied. */
11794 CLONE_PARAMS clone_params;
11795 CLONE_PARAMS* const param = &clone_params;
11797 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11799 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11801 /* for each stash, determine whether its objects should be cloned */
11802 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11803 PERL_SET_THX(my_perl);
11806 PoisonNew(my_perl, 1, PerlInterpreter);
11811 PL_scopestack_name = 0;
11813 PL_savestack_ix = 0;
11814 PL_savestack_max = -1;
11815 PL_sig_pending = 0;
11817 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11818 # else /* !DEBUGGING */
11819 Zero(my_perl, 1, PerlInterpreter);
11820 # endif /* DEBUGGING */
11822 /* host pointers */
11824 PL_MemShared = ipMS;
11825 PL_MemParse = ipMP;
11832 #else /* !PERL_IMPLICIT_SYS */
11834 CLONE_PARAMS clone_params;
11835 CLONE_PARAMS* param = &clone_params;
11836 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11838 PERL_ARGS_ASSERT_PERL_CLONE;
11840 /* for each stash, determine whether its objects should be cloned */
11841 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11842 PERL_SET_THX(my_perl);
11845 PoisonNew(my_perl, 1, PerlInterpreter);
11850 PL_scopestack_name = 0;
11852 PL_savestack_ix = 0;
11853 PL_savestack_max = -1;
11854 PL_sig_pending = 0;
11856 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11857 # else /* !DEBUGGING */
11858 Zero(my_perl, 1, PerlInterpreter);
11859 # endif /* DEBUGGING */
11860 #endif /* PERL_IMPLICIT_SYS */
11861 param->flags = flags;
11862 param->proto_perl = proto_perl;
11864 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11866 PL_body_arenas = NULL;
11867 Zero(&PL_body_roots, 1, PL_body_roots);
11869 PL_nice_chunk = NULL;
11870 PL_nice_chunk_size = 0;
11872 PL_sv_objcount = 0;
11874 PL_sv_arenaroot = NULL;
11876 PL_debug = proto_perl->Idebug;
11878 PL_hash_seed = proto_perl->Ihash_seed;
11879 PL_rehash_seed = proto_perl->Irehash_seed;
11881 #ifdef USE_REENTRANT_API
11882 /* XXX: things like -Dm will segfault here in perlio, but doing
11883 * PERL_SET_CONTEXT(proto_perl);
11884 * breaks too many other things
11886 Perl_reentrant_init(aTHX);
11889 /* create SV map for pointer relocation */
11890 PL_ptr_table = ptr_table_new();
11892 /* initialize these special pointers as early as possible */
11893 SvANY(&PL_sv_undef) = NULL;
11894 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11895 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11896 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11898 SvANY(&PL_sv_no) = new_XPVNV();
11899 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11900 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11901 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11902 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11903 SvCUR_set(&PL_sv_no, 0);
11904 SvLEN_set(&PL_sv_no, 1);
11905 SvIV_set(&PL_sv_no, 0);
11906 SvNV_set(&PL_sv_no, 0);
11907 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11909 SvANY(&PL_sv_yes) = new_XPVNV();
11910 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11911 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11912 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11913 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11914 SvCUR_set(&PL_sv_yes, 1);
11915 SvLEN_set(&PL_sv_yes, 2);
11916 SvIV_set(&PL_sv_yes, 1);
11917 SvNV_set(&PL_sv_yes, 1);
11918 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11920 /* create (a non-shared!) shared string table */
11921 PL_strtab = newHV();
11922 HvSHAREKEYS_off(PL_strtab);
11923 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11924 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11926 PL_compiling = proto_perl->Icompiling;
11928 /* These two PVs will be free'd special way so must set them same way op.c does */
11929 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11930 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11932 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11933 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11935 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11936 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11937 if (PL_compiling.cop_hints_hash) {
11939 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11940 HINTS_REFCNT_UNLOCK;
11942 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11943 #ifdef PERL_DEBUG_READONLY_OPS
11948 /* pseudo environmental stuff */
11949 PL_origargc = proto_perl->Iorigargc;
11950 PL_origargv = proto_perl->Iorigargv;
11952 param->stashes = newAV(); /* Setup array of objects to call clone on */
11954 /* Set tainting stuff before PerlIO_debug can possibly get called */
11955 PL_tainting = proto_perl->Itainting;
11956 PL_taint_warn = proto_perl->Itaint_warn;
11958 #ifdef PERLIO_LAYERS
11959 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11960 PerlIO_clone(aTHX_ proto_perl, param);
11963 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11964 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11965 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11966 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11967 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11968 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11971 PL_minus_c = proto_perl->Iminus_c;
11972 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11973 PL_localpatches = proto_perl->Ilocalpatches;
11974 PL_splitstr = proto_perl->Isplitstr;
11975 PL_minus_n = proto_perl->Iminus_n;
11976 PL_minus_p = proto_perl->Iminus_p;
11977 PL_minus_l = proto_perl->Iminus_l;
11978 PL_minus_a = proto_perl->Iminus_a;
11979 PL_minus_E = proto_perl->Iminus_E;
11980 PL_minus_F = proto_perl->Iminus_F;
11981 PL_doswitches = proto_perl->Idoswitches;
11982 PL_dowarn = proto_perl->Idowarn;
11983 PL_doextract = proto_perl->Idoextract;
11984 PL_sawampersand = proto_perl->Isawampersand;
11985 PL_unsafe = proto_perl->Iunsafe;
11986 PL_inplace = SAVEPV(proto_perl->Iinplace);
11987 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11988 PL_perldb = proto_perl->Iperldb;
11989 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11990 PL_exit_flags = proto_perl->Iexit_flags;
11992 /* magical thingies */
11993 /* XXX time(&PL_basetime) when asked for? */
11994 PL_basetime = proto_perl->Ibasetime;
11995 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11997 PL_maxsysfd = proto_perl->Imaxsysfd;
11998 PL_statusvalue = proto_perl->Istatusvalue;
12000 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12002 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12004 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12006 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12007 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12008 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12011 /* RE engine related */
12012 Zero(&PL_reg_state, 1, struct re_save_state);
12013 PL_reginterp_cnt = 0;
12014 PL_regmatch_slab = NULL;
12016 /* Clone the regex array */
12017 /* ORANGE FIXME for plugins, probably in the SV dup code.
12018 newSViv(PTR2IV(CALLREGDUPE(
12019 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12021 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12022 PL_regex_pad = AvARRAY(PL_regex_padav);
12024 /* shortcuts to various I/O objects */
12025 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
12026 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12027 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12028 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12029 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12030 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12031 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12033 /* shortcuts to regexp stuff */
12034 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12036 /* shortcuts to misc objects */
12037 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12039 /* shortcuts to debugging objects */
12040 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12041 PL_DBline = gv_dup(proto_perl->IDBline, param);
12042 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12043 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12044 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12045 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12046 PL_dbargs = av_dup(proto_perl->Idbargs, param);
12048 /* symbol tables */
12049 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12050 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12051 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12052 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12053 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12055 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12056 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12057 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12058 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12059 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12060 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12061 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12062 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12064 PL_sub_generation = proto_perl->Isub_generation;
12065 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12067 /* funky return mechanisms */
12068 PL_forkprocess = proto_perl->Iforkprocess;
12070 /* subprocess state */
12071 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12073 /* internal state */
12074 PL_maxo = proto_perl->Imaxo;
12075 if (proto_perl->Iop_mask)
12076 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12079 /* PL_asserting = proto_perl->Iasserting; */
12081 /* current interpreter roots */
12082 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12084 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12086 PL_main_start = proto_perl->Imain_start;
12087 PL_eval_root = proto_perl->Ieval_root;
12088 PL_eval_start = proto_perl->Ieval_start;
12090 /* runtime control stuff */
12091 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12093 PL_filemode = proto_perl->Ifilemode;
12094 PL_lastfd = proto_perl->Ilastfd;
12095 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12098 PL_gensym = proto_perl->Igensym;
12099 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12100 PL_laststatval = proto_perl->Ilaststatval;
12101 PL_laststype = proto_perl->Ilaststype;
12104 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12106 /* interpreter atexit processing */
12107 PL_exitlistlen = proto_perl->Iexitlistlen;
12108 if (PL_exitlistlen) {
12109 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12110 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12113 PL_exitlist = (PerlExitListEntry*)NULL;
12115 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12116 if (PL_my_cxt_size) {
12117 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12118 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12119 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12120 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12121 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12125 PL_my_cxt_list = (void**)NULL;
12126 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12127 PL_my_cxt_keys = (const char**)NULL;
12130 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12131 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12132 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12134 PL_profiledata = NULL;
12136 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12138 PAD_CLONE_VARS(proto_perl, param);
12140 #ifdef HAVE_INTERP_INTERN
12141 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12144 /* more statics moved here */
12145 PL_generation = proto_perl->Igeneration;
12146 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12148 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12149 PL_in_clean_all = proto_perl->Iin_clean_all;
12151 PL_uid = proto_perl->Iuid;
12152 PL_euid = proto_perl->Ieuid;
12153 PL_gid = proto_perl->Igid;
12154 PL_egid = proto_perl->Iegid;
12155 PL_nomemok = proto_perl->Inomemok;
12156 PL_an = proto_perl->Ian;
12157 PL_evalseq = proto_perl->Ievalseq;
12158 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12159 PL_origalen = proto_perl->Iorigalen;
12160 #ifdef PERL_USES_PL_PIDSTATUS
12161 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12163 PL_osname = SAVEPV(proto_perl->Iosname);
12164 PL_sighandlerp = proto_perl->Isighandlerp;
12166 PL_runops = proto_perl->Irunops;
12168 PL_parser = parser_dup(proto_perl->Iparser, param);
12170 /* XXX this only works if the saved cop has already been cloned */
12171 if (proto_perl->Iparser) {
12172 PL_parser->saved_curcop = (COP*)any_dup(
12173 proto_perl->Iparser->saved_curcop,
12177 PL_subline = proto_perl->Isubline;
12178 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12181 PL_cryptseen = proto_perl->Icryptseen;
12184 PL_hints = proto_perl->Ihints;
12186 PL_amagic_generation = proto_perl->Iamagic_generation;
12188 #ifdef USE_LOCALE_COLLATE
12189 PL_collation_ix = proto_perl->Icollation_ix;
12190 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12191 PL_collation_standard = proto_perl->Icollation_standard;
12192 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12193 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12194 #endif /* USE_LOCALE_COLLATE */
12196 #ifdef USE_LOCALE_NUMERIC
12197 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12198 PL_numeric_standard = proto_perl->Inumeric_standard;
12199 PL_numeric_local = proto_perl->Inumeric_local;
12200 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12201 #endif /* !USE_LOCALE_NUMERIC */
12203 /* utf8 character classes */
12204 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12205 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12206 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12207 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12208 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12209 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12210 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12211 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12212 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12213 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12214 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12215 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12216 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12217 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12218 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12219 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12220 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12221 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12222 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12224 /* Did the locale setup indicate UTF-8? */
12225 PL_utf8locale = proto_perl->Iutf8locale;
12226 /* Unicode features (see perlrun/-C) */
12227 PL_unicode = proto_perl->Iunicode;
12229 /* Pre-5.8 signals control */
12230 PL_signals = proto_perl->Isignals;
12232 /* times() ticks per second */
12233 PL_clocktick = proto_perl->Iclocktick;
12235 /* Recursion stopper for PerlIO_find_layer */
12236 PL_in_load_module = proto_perl->Iin_load_module;
12238 /* sort() routine */
12239 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12241 /* Not really needed/useful since the reenrant_retint is "volatile",
12242 * but do it for consistency's sake. */
12243 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12245 /* Hooks to shared SVs and locks. */
12246 PL_sharehook = proto_perl->Isharehook;
12247 PL_lockhook = proto_perl->Ilockhook;
12248 PL_unlockhook = proto_perl->Iunlockhook;
12249 PL_threadhook = proto_perl->Ithreadhook;
12250 PL_destroyhook = proto_perl->Idestroyhook;
12252 #ifdef THREADS_HAVE_PIDS
12253 PL_ppid = proto_perl->Ippid;
12257 PL_last_swash_hv = NULL; /* reinits on demand */
12258 PL_last_swash_klen = 0;
12259 PL_last_swash_key[0]= '\0';
12260 PL_last_swash_tmps = (U8*)NULL;
12261 PL_last_swash_slen = 0;
12263 PL_glob_index = proto_perl->Iglob_index;
12264 PL_srand_called = proto_perl->Isrand_called;
12266 if (proto_perl->Ipsig_pend) {
12267 Newxz(PL_psig_pend, SIG_SIZE, int);
12270 PL_psig_pend = (int*)NULL;
12273 if (proto_perl->Ipsig_name) {
12274 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12275 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12277 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12280 PL_psig_ptr = (SV**)NULL;
12281 PL_psig_name = (SV**)NULL;
12284 /* intrpvar.h stuff */
12286 if (flags & CLONEf_COPY_STACKS) {
12287 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12288 PL_tmps_ix = proto_perl->Itmps_ix;
12289 PL_tmps_max = proto_perl->Itmps_max;
12290 PL_tmps_floor = proto_perl->Itmps_floor;
12291 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12292 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12293 PL_tmps_ix+1, param);
12295 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12296 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12297 Newxz(PL_markstack, i, I32);
12298 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12299 - proto_perl->Imarkstack);
12300 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12301 - proto_perl->Imarkstack);
12302 Copy(proto_perl->Imarkstack, PL_markstack,
12303 PL_markstack_ptr - PL_markstack + 1, I32);
12305 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12306 * NOTE: unlike the others! */
12307 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12308 PL_scopestack_max = proto_perl->Iscopestack_max;
12309 Newxz(PL_scopestack, PL_scopestack_max, I32);
12310 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12313 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12314 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12316 /* NOTE: si_dup() looks at PL_markstack */
12317 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12319 /* PL_curstack = PL_curstackinfo->si_stack; */
12320 PL_curstack = av_dup(proto_perl->Icurstack, param);
12321 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12323 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12324 PL_stack_base = AvARRAY(PL_curstack);
12325 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12326 - proto_perl->Istack_base);
12327 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12329 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12330 * NOTE: unlike the others! */
12331 PL_savestack_ix = proto_perl->Isavestack_ix;
12332 PL_savestack_max = proto_perl->Isavestack_max;
12333 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12334 PL_savestack = ss_dup(proto_perl, param);
12338 ENTER; /* perl_destruct() wants to LEAVE; */
12340 /* although we're not duplicating the tmps stack, we should still
12341 * add entries for any SVs on the tmps stack that got cloned by a
12342 * non-refcount means (eg a temp in @_); otherwise they will be
12345 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12346 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12347 proto_perl->Itmps_stack[i]));
12348 if (nsv && !SvREFCNT(nsv)) {
12349 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12354 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12355 PL_top_env = &PL_start_env;
12357 PL_op = proto_perl->Iop;
12360 PL_Xpv = (XPV*)NULL;
12361 my_perl->Ina = proto_perl->Ina;
12363 PL_statbuf = proto_perl->Istatbuf;
12364 PL_statcache = proto_perl->Istatcache;
12365 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12366 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12368 PL_timesbuf = proto_perl->Itimesbuf;
12371 PL_tainted = proto_perl->Itainted;
12372 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12373 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12374 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12375 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12376 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12377 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12378 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12379 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12381 PL_restartop = proto_perl->Irestartop;
12382 PL_in_eval = proto_perl->Iin_eval;
12383 PL_delaymagic = proto_perl->Idelaymagic;
12384 PL_dirty = proto_perl->Idirty;
12385 PL_localizing = proto_perl->Ilocalizing;
12387 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12388 PL_hv_fetch_ent_mh = NULL;
12389 PL_modcount = proto_perl->Imodcount;
12390 PL_lastgotoprobe = NULL;
12391 PL_dumpindent = proto_perl->Idumpindent;
12393 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12394 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12395 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12396 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12397 PL_efloatbuf = NULL; /* reinits on demand */
12398 PL_efloatsize = 0; /* reinits on demand */
12402 PL_screamfirst = NULL;
12403 PL_screamnext = NULL;
12404 PL_maxscream = -1; /* reinits on demand */
12405 PL_lastscream = NULL;
12408 PL_regdummy = proto_perl->Iregdummy;
12409 PL_colorset = 0; /* reinits PL_colors[] */
12410 /*PL_colors[6] = {0,0,0,0,0,0};*/
12414 /* Pluggable optimizer */
12415 PL_peepp = proto_perl->Ipeepp;
12416 /* op_free() hook */
12417 PL_opfreehook = proto_perl->Iopfreehook;
12419 PL_stashcache = newHV();
12421 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12422 proto_perl->Iwatchaddr);
12423 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12424 if (PL_debug && PL_watchaddr) {
12425 PerlIO_printf(Perl_debug_log,
12426 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12427 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12428 PTR2UV(PL_watchok));
12431 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12433 /* Call the ->CLONE method, if it exists, for each of the stashes
12434 identified by sv_dup() above.
12436 while(av_len(param->stashes) != -1) {
12437 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12438 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12439 if (cloner && GvCV(cloner)) {
12444 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12446 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12452 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12453 ptr_table_free(PL_ptr_table);
12454 PL_ptr_table = NULL;
12458 SvREFCNT_dec(param->stashes);
12460 /* orphaned? eg threads->new inside BEGIN or use */
12461 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12462 SvREFCNT_inc_simple_void(PL_compcv);
12463 SAVEFREESV(PL_compcv);
12469 #endif /* USE_ITHREADS */
12472 =head1 Unicode Support
12474 =for apidoc sv_recode_to_utf8
12476 The encoding is assumed to be an Encode object, on entry the PV
12477 of the sv is assumed to be octets in that encoding, and the sv
12478 will be converted into Unicode (and UTF-8).
12480 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12481 is not a reference, nothing is done to the sv. If the encoding is not
12482 an C<Encode::XS> Encoding object, bad things will happen.
12483 (See F<lib/encoding.pm> and L<Encode>).
12485 The PV of the sv is returned.
12490 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12494 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12496 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12510 Passing sv_yes is wrong - it needs to be or'ed set of constants
12511 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12512 remove converted chars from source.
12514 Both will default the value - let them.
12516 XPUSHs(&PL_sv_yes);
12519 call_method("decode", G_SCALAR);
12523 s = SvPV_const(uni, len);
12524 if (s != SvPVX_const(sv)) {
12525 SvGROW(sv, len + 1);
12526 Move(s, SvPVX(sv), len + 1, char);
12527 SvCUR_set(sv, len);
12534 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12538 =for apidoc sv_cat_decode
12540 The encoding is assumed to be an Encode object, the PV of the ssv is
12541 assumed to be octets in that encoding and decoding the input starts
12542 from the position which (PV + *offset) pointed to. The dsv will be
12543 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12544 when the string tstr appears in decoding output or the input ends on
12545 the PV of the ssv. The value which the offset points will be modified
12546 to the last input position on the ssv.
12548 Returns TRUE if the terminator was found, else returns FALSE.
12553 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12554 SV *ssv, int *offset, char *tstr, int tlen)
12559 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12561 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12572 offsv = newSViv(*offset);
12574 mXPUSHp(tstr, tlen);
12576 call_method("cat_decode", G_SCALAR);
12578 ret = SvTRUE(TOPs);
12579 *offset = SvIV(offsv);
12585 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12590 /* ---------------------------------------------------------------------
12592 * support functions for report_uninit()
12595 /* the maxiumum size of array or hash where we will scan looking
12596 * for the undefined element that triggered the warning */
12598 #define FUV_MAX_SEARCH_SIZE 1000
12600 /* Look for an entry in the hash whose value has the same SV as val;
12601 * If so, return a mortal copy of the key. */
12604 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12607 register HE **array;
12610 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12612 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12613 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12616 array = HvARRAY(hv);
12618 for (i=HvMAX(hv); i>0; i--) {
12619 register HE *entry;
12620 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12621 if (HeVAL(entry) != val)
12623 if ( HeVAL(entry) == &PL_sv_undef ||
12624 HeVAL(entry) == &PL_sv_placeholder)
12628 if (HeKLEN(entry) == HEf_SVKEY)
12629 return sv_mortalcopy(HeKEY_sv(entry));
12630 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12636 /* Look for an entry in the array whose value has the same SV as val;
12637 * If so, return the index, otherwise return -1. */
12640 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12644 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12646 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12647 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12650 if (val != &PL_sv_undef) {
12651 SV ** const svp = AvARRAY(av);
12654 for (i=AvFILLp(av); i>=0; i--)
12661 /* S_varname(): return the name of a variable, optionally with a subscript.
12662 * If gv is non-zero, use the name of that global, along with gvtype (one
12663 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12664 * targ. Depending on the value of the subscript_type flag, return:
12667 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12668 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12669 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12670 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12673 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12674 const SV *const keyname, I32 aindex, int subscript_type)
12677 SV * const name = sv_newmortal();
12680 buffer[0] = gvtype;
12683 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12685 gv_fullname4(name, gv, buffer, 0);
12687 if ((unsigned int)SvPVX(name)[1] <= 26) {
12689 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12691 /* Swap the 1 unprintable control character for the 2 byte pretty
12692 version - ie substr($name, 1, 1) = $buffer; */
12693 sv_insert(name, 1, 1, buffer, 2);
12697 CV * const cv = find_runcv(NULL);
12701 if (!cv || !CvPADLIST(cv))
12703 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12704 sv = *av_fetch(av, targ, FALSE);
12705 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12708 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12709 SV * const sv = newSV(0);
12710 *SvPVX(name) = '$';
12711 Perl_sv_catpvf(aTHX_ name, "{%s}",
12712 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12715 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12716 *SvPVX(name) = '$';
12717 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12719 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12720 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12721 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12729 =for apidoc find_uninit_var
12731 Find the name of the undefined variable (if any) that caused the operator o
12732 to issue a "Use of uninitialized value" warning.
12733 If match is true, only return a name if it's value matches uninit_sv.
12734 So roughly speaking, if a unary operator (such as OP_COS) generates a
12735 warning, then following the direct child of the op may yield an
12736 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12737 other hand, with OP_ADD there are two branches to follow, so we only print
12738 the variable name if we get an exact match.
12740 The name is returned as a mortal SV.
12742 Assumes that PL_op is the op that originally triggered the error, and that
12743 PL_comppad/PL_curpad points to the currently executing pad.
12749 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12755 const OP *o, *o2, *kid;
12757 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12758 uninit_sv == &PL_sv_placeholder)))
12761 switch (obase->op_type) {
12768 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12769 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12772 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12774 if (pad) { /* @lex, %lex */
12775 sv = PAD_SVl(obase->op_targ);
12779 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12780 /* @global, %global */
12781 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12784 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12786 else /* @{expr}, %{expr} */
12787 return find_uninit_var(cUNOPx(obase)->op_first,
12791 /* attempt to find a match within the aggregate */
12793 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12795 subscript_type = FUV_SUBSCRIPT_HASH;
12798 index = find_array_subscript((const AV *)sv, uninit_sv);
12800 subscript_type = FUV_SUBSCRIPT_ARRAY;
12803 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12806 return varname(gv, hash ? '%' : '@', obase->op_targ,
12807 keysv, index, subscript_type);
12811 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12813 return varname(NULL, '$', obase->op_targ,
12814 NULL, 0, FUV_SUBSCRIPT_NONE);
12817 gv = cGVOPx_gv(obase);
12818 if (!gv || (match && GvSV(gv) != uninit_sv))
12820 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12823 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12826 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12827 if (!av || SvRMAGICAL(av))
12829 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12830 if (!svp || *svp != uninit_sv)
12833 return varname(NULL, '$', obase->op_targ,
12834 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12837 gv = cGVOPx_gv(obase);
12842 AV *const av = GvAV(gv);
12843 if (!av || SvRMAGICAL(av))
12845 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12846 if (!svp || *svp != uninit_sv)
12849 return varname(gv, '$', 0,
12850 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12855 o = cUNOPx(obase)->op_first;
12856 if (!o || o->op_type != OP_NULL ||
12857 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12859 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12863 if (PL_op == obase)
12864 /* $a[uninit_expr] or $h{uninit_expr} */
12865 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12868 o = cBINOPx(obase)->op_first;
12869 kid = cBINOPx(obase)->op_last;
12871 /* get the av or hv, and optionally the gv */
12873 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12874 sv = PAD_SV(o->op_targ);
12876 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12877 && cUNOPo->op_first->op_type == OP_GV)
12879 gv = cGVOPx_gv(cUNOPo->op_first);
12883 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12888 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12889 /* index is constant */
12893 if (obase->op_type == OP_HELEM) {
12894 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12895 if (!he || HeVAL(he) != uninit_sv)
12899 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12900 if (!svp || *svp != uninit_sv)
12904 if (obase->op_type == OP_HELEM)
12905 return varname(gv, '%', o->op_targ,
12906 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12908 return varname(gv, '@', o->op_targ, NULL,
12909 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12912 /* index is an expression;
12913 * attempt to find a match within the aggregate */
12914 if (obase->op_type == OP_HELEM) {
12915 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12917 return varname(gv, '%', o->op_targ,
12918 keysv, 0, FUV_SUBSCRIPT_HASH);
12922 = find_array_subscript((const AV *)sv, uninit_sv);
12924 return varname(gv, '@', o->op_targ,
12925 NULL, index, FUV_SUBSCRIPT_ARRAY);
12930 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12932 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12937 /* only examine RHS */
12938 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12941 o = cUNOPx(obase)->op_first;
12942 if (o->op_type == OP_PUSHMARK)
12945 if (!o->op_sibling) {
12946 /* one-arg version of open is highly magical */
12948 if (o->op_type == OP_GV) { /* open FOO; */
12950 if (match && GvSV(gv) != uninit_sv)
12952 return varname(gv, '$', 0,
12953 NULL, 0, FUV_SUBSCRIPT_NONE);
12955 /* other possibilities not handled are:
12956 * open $x; or open my $x; should return '${*$x}'
12957 * open expr; should return '$'.expr ideally
12963 /* ops where $_ may be an implicit arg */
12967 if ( !(obase->op_flags & OPf_STACKED)) {
12968 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12969 ? PAD_SVl(obase->op_targ)
12972 sv = sv_newmortal();
12973 sv_setpvs(sv, "$_");
12982 match = 1; /* print etc can return undef on defined args */
12983 /* skip filehandle as it can't produce 'undef' warning */
12984 o = cUNOPx(obase)->op_first;
12985 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12986 o = o->op_sibling->op_sibling;
12990 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12992 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12994 /* the following ops are capable of returning PL_sv_undef even for
12995 * defined arg(s) */
13014 case OP_GETPEERNAME:
13062 case OP_SMARTMATCH:
13071 /* XXX tmp hack: these two may call an XS sub, and currently
13072 XS subs don't have a SUB entry on the context stack, so CV and
13073 pad determination goes wrong, and BAD things happen. So, just
13074 don't try to determine the value under those circumstances.
13075 Need a better fix at dome point. DAPM 11/2007 */
13081 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13082 if (gv && GvSV(gv) == uninit_sv)
13083 return newSVpvs_flags("$.", SVs_TEMP);
13088 /* def-ness of rval pos() is independent of the def-ness of its arg */
13089 if ( !(obase->op_flags & OPf_MOD))
13094 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13095 return newSVpvs_flags("${$/}", SVs_TEMP);
13100 if (!(obase->op_flags & OPf_KIDS))
13102 o = cUNOPx(obase)->op_first;
13108 /* if all except one arg are constant, or have no side-effects,
13109 * or are optimized away, then it's unambiguous */
13111 for (kid=o; kid; kid = kid->op_sibling) {
13113 const OPCODE type = kid->op_type;
13114 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13115 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13116 || (type == OP_PUSHMARK)
13120 if (o2) { /* more than one found */
13127 return find_uninit_var(o2, uninit_sv, match);
13129 /* scan all args */
13131 sv = find_uninit_var(o, uninit_sv, 1);
13143 =for apidoc report_uninit
13145 Print appropriate "Use of uninitialized variable" warning
13151 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13155 SV* varname = NULL;
13157 varname = find_uninit_var(PL_op, uninit_sv,0);
13159 sv_insert(varname, 0, 0, " ", 1);
13161 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13162 varname ? SvPV_nolen_const(varname) : "",
13163 " in ", OP_DESC(PL_op));
13166 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13172 * c-indentation-style: bsd
13173 * c-basic-offset: 4
13174 * indent-tabs-mode: t
13177 * ex: set ts=8 sts=4 sw=4 noet: