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
693 Arena types 2 & 3 are chained by body-type off an array of
694 arena-root pointers, which is indexed by svtype. Some of the
695 larger/less used body types are malloced singly, since a large
696 unused block of them is wasteful. Also, several svtypes dont have
697 bodies; the data fits into the sv-head itself. The arena-root
698 pointer thus has a few unused root-pointers (which may be hijacked
699 later for arena types 4,5)
701 3 differs from 2 as an optimization; some body types have several
702 unused fields in the front of the structure (which are kept in-place
703 for consistency). These bodies can be allocated in smaller chunks,
704 because the leading fields arent accessed. Pointers to such bodies
705 are decremented to point at the unused 'ghost' memory, knowing that
706 the pointers are used with offsets to the real memory.
708 HE, HEK arenas are managed separately, with separate code, but may
709 be merge-able later..
712 /* get_arena(size): this creates custom-sized arenas
713 TBD: export properly for hv.c: S_more_he().
716 Perl_get_arena(pTHX_ const size_t arena_size, const svtype bodytype)
719 struct arena_desc* adesc;
720 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
723 /* shouldnt need this
724 if (!arena_size) arena_size = PERL_ARENA_SIZE;
727 /* may need new arena-set to hold new arena */
728 if (!aroot || aroot->curr >= aroot->set_size) {
729 struct arena_set *newroot;
730 Newxz(newroot, 1, struct arena_set);
731 newroot->set_size = ARENAS_PER_SET;
732 newroot->next = aroot;
734 PL_body_arenas = (void *) newroot;
735 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
738 /* ok, now have arena-set with at least 1 empty/available arena-desc */
739 curr = aroot->curr++;
740 adesc = &(aroot->set[curr]);
741 assert(!adesc->arena);
743 Newx(adesc->arena, arena_size, char);
744 adesc->size = arena_size;
745 adesc->utype = bodytype;
746 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
747 curr, (void*)adesc->arena, (UV)arena_size));
753 /* return a thing to the free list */
755 #define del_body(thing, root) \
757 void ** const thing_copy = (void **)thing;\
758 *thing_copy = *root; \
759 *root = (void*)thing_copy; \
764 =head1 SV-Body Allocation
766 Allocation of SV-bodies is similar to SV-heads, differing as follows;
767 the allocation mechanism is used for many body types, so is somewhat
768 more complicated, it uses arena-sets, and has no need for still-live
771 At the outermost level, (new|del)_X*V macros return bodies of the
772 appropriate type. These macros call either (new|del)_body_type or
773 (new|del)_body_allocated macro pairs, depending on specifics of the
774 type. Most body types use the former pair, the latter pair is used to
775 allocate body types with "ghost fields".
777 "ghost fields" are fields that are unused in certain types, and
778 consequently don't need to actually exist. They are declared because
779 they're part of a "base type", which allows use of functions as
780 methods. The simplest examples are AVs and HVs, 2 aggregate types
781 which don't use the fields which support SCALAR semantics.
783 For these types, the arenas are carved up into appropriately sized
784 chunks, we thus avoid wasted memory for those unaccessed members.
785 When bodies are allocated, we adjust the pointer back in memory by the
786 size of the part not allocated, so it's as if we allocated the full
787 structure. (But things will all go boom if you write to the part that
788 is "not there", because you'll be overwriting the last members of the
789 preceding structure in memory.)
791 We calculate the correction using the STRUCT_OFFSET macro on the first
792 member present. If the allocated structure is smaller (no initial NV
793 actually allocated) then the net effect is to subtract the size of the NV
794 from the pointer, to return a new pointer as if an initial NV were actually
795 allocated. (We were using structures named *_allocated for this, but
796 this turned out to be a subtle bug, because a structure without an NV
797 could have a lower alignment constraint, but the compiler is allowed to
798 optimised accesses based on the alignment constraint of the actual pointer
799 to the full structure, for example, using a single 64 bit load instruction
800 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
802 This is the same trick as was used for NV and IV bodies. Ironically it
803 doesn't need to be used for NV bodies any more, because NV is now at
804 the start of the structure. IV bodies don't need it either, because
805 they are no longer allocated.
807 In turn, the new_body_* allocators call S_new_body(), which invokes
808 new_body_inline macro, which takes a lock, and takes a body off the
809 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
810 necessary to refresh an empty list. Then the lock is released, and
811 the body is returned.
813 S_more_bodies calls get_arena(), and carves it up into an array of N
814 bodies, which it strings into a linked list. It looks up arena-size
815 and body-size from the body_details table described below, thus
816 supporting the multiple body-types.
818 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
819 the (new|del)_X*V macros are mapped directly to malloc/free.
825 For each sv-type, struct body_details bodies_by_type[] carries
826 parameters which control these aspects of SV handling:
828 Arena_size determines whether arenas are used for this body type, and if
829 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
830 zero, forcing individual mallocs and frees.
832 Body_size determines how big a body is, and therefore how many fit into
833 each arena. Offset carries the body-pointer adjustment needed for
834 "ghost fields", and is used in *_allocated macros.
836 But its main purpose is to parameterize info needed in
837 Perl_sv_upgrade(). The info here dramatically simplifies the function
838 vs the implementation in 5.8.8, making it table-driven. All fields
839 are used for this, except for arena_size.
841 For the sv-types that have no bodies, arenas are not used, so those
842 PL_body_roots[sv_type] are unused, and can be overloaded. In
843 something of a special case, SVt_NULL is borrowed for HE arenas;
844 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
845 bodies_by_type[SVt_NULL] slot is not used, as the table is not
850 struct body_details {
851 U8 body_size; /* Size to allocate */
852 U8 copy; /* Size of structure to copy (may be shorter) */
854 unsigned int type : 4; /* We have space for a sanity check. */
855 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
856 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
857 unsigned int arena : 1; /* Allocated from an arena */
858 size_t arena_size; /* Size of arena to allocate */
866 /* With -DPURFIY we allocate everything directly, and don't use arenas.
867 This seems a rather elegant way to simplify some of the code below. */
868 #define HASARENA FALSE
870 #define HASARENA TRUE
872 #define NOARENA FALSE
874 /* Size the arenas to exactly fit a given number of bodies. A count
875 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
876 simplifying the default. If count > 0, the arena is sized to fit
877 only that many bodies, allowing arenas to be used for large, rare
878 bodies (XPVFM, XPVIO) without undue waste. The arena size is
879 limited by PERL_ARENA_SIZE, so we can safely oversize the
882 #define FIT_ARENA0(body_size) \
883 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
884 #define FIT_ARENAn(count,body_size) \
885 ( count * body_size <= PERL_ARENA_SIZE) \
886 ? count * body_size \
887 : FIT_ARENA0 (body_size)
888 #define FIT_ARENA(count,body_size) \
890 ? FIT_ARENAn (count, body_size) \
891 : FIT_ARENA0 (body_size)
893 /* Calculate the length to copy. Specifically work out the length less any
894 final padding the compiler needed to add. See the comment in sv_upgrade
895 for why copying the padding proved to be a bug. */
897 #define copy_length(type, last_member) \
898 STRUCT_OFFSET(type, last_member) \
899 + sizeof (((type*)SvANY((const SV *)0))->last_member)
901 static const struct body_details bodies_by_type[] = {
902 { sizeof(HE), 0, 0, SVt_NULL,
903 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
905 /* The bind placeholder pretends to be an RV for now.
906 Also it's marked as "can't upgrade" to stop anyone using it before it's
908 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
910 /* IVs are in the head, so the allocation size is 0. */
912 sizeof(IV), /* This is used to copy out the IV body. */
913 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
914 NOARENA /* IVS don't need an arena */, 0
917 /* 8 bytes on most ILP32 with IEEE doubles */
918 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
919 FIT_ARENA(0, sizeof(NV)) },
921 /* 8 bytes on most ILP32 with IEEE doubles */
922 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
923 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
924 + STRUCT_OFFSET(XPV, xpv_cur),
925 SVt_PV, FALSE, NONV, HASARENA,
926 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
929 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
930 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
931 + STRUCT_OFFSET(XPVIV, xpv_cur),
932 SVt_PVIV, FALSE, NONV, HASARENA,
933 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
936 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
937 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
940 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
941 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
944 { sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
945 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
946 + STRUCT_OFFSET(regexp, xpv_cur),
947 SVt_REGEXP, FALSE, NONV, HASARENA,
948 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
952 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
953 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
956 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
957 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
959 { sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill),
960 copy_length(XPVAV, xmg_stash) - STRUCT_OFFSET(XPVAV, xav_fill),
961 + STRUCT_OFFSET(XPVAV, xav_fill),
962 SVt_PVAV, TRUE, NONV, HASARENA,
963 FIT_ARENA(0, sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill)) },
965 { sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill),
966 copy_length(XPVHV, xmg_stash) - STRUCT_OFFSET(XPVHV, xhv_fill),
967 + STRUCT_OFFSET(XPVHV, xhv_fill),
968 SVt_PVHV, TRUE, NONV, HASARENA,
969 FIT_ARENA(0, sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill)) },
972 { sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
973 sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
974 + STRUCT_OFFSET(XPVCV, xpv_cur),
975 SVt_PVCV, TRUE, NONV, HASARENA,
976 FIT_ARENA(0, sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur)) },
978 { sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
979 sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
980 + STRUCT_OFFSET(XPVFM, xpv_cur),
981 SVt_PVFM, TRUE, NONV, NOARENA,
982 FIT_ARENA(20, sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur)) },
984 /* XPVIO is 84 bytes, fits 48x */
985 { sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
986 sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
987 + STRUCT_OFFSET(XPVIO, xpv_cur),
988 SVt_PVIO, TRUE, NONV, HASARENA,
989 FIT_ARENA(24, sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur)) },
992 #define new_body_type(sv_type) \
993 (void *)((char *)S_new_body(aTHX_ sv_type))
995 #define del_body_type(p, sv_type) \
996 del_body(p, &PL_body_roots[sv_type])
999 #define new_body_allocated(sv_type) \
1000 (void *)((char *)S_new_body(aTHX_ sv_type) \
1001 - bodies_by_type[sv_type].offset)
1003 #define del_body_allocated(p, sv_type) \
1004 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1007 #define my_safemalloc(s) (void*)safemalloc(s)
1008 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1009 #define my_safefree(p) safefree((char*)p)
1013 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1014 #define del_XNV(p) my_safefree(p)
1016 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1017 #define del_XPVNV(p) my_safefree(p)
1019 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1020 #define del_XPVAV(p) my_safefree(p)
1022 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1023 #define del_XPVHV(p) my_safefree(p)
1025 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1026 #define del_XPVMG(p) my_safefree(p)
1028 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1029 #define del_XPVGV(p) my_safefree(p)
1033 #define new_XNV() new_body_type(SVt_NV)
1034 #define del_XNV(p) del_body_type(p, SVt_NV)
1036 #define new_XPVNV() new_body_type(SVt_PVNV)
1037 #define del_XPVNV(p) del_body_type(p, SVt_PVNV)
1039 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1040 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1042 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1043 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1045 #define new_XPVMG() new_body_type(SVt_PVMG)
1046 #define del_XPVMG(p) del_body_type(p, SVt_PVMG)
1048 #define new_XPVGV() new_body_type(SVt_PVGV)
1049 #define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1053 /* no arena for you! */
1055 #define new_NOARENA(details) \
1056 my_safemalloc((details)->body_size + (details)->offset)
1057 #define new_NOARENAZ(details) \
1058 my_safecalloc((details)->body_size + (details)->offset)
1061 S_more_bodies (pTHX_ const svtype sv_type)
1064 void ** const root = &PL_body_roots[sv_type];
1065 const struct body_details * const bdp = &bodies_by_type[sv_type];
1066 const size_t body_size = bdp->body_size;
1069 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1070 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1071 static bool done_sanity_check;
1073 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1074 * variables like done_sanity_check. */
1075 if (!done_sanity_check) {
1076 unsigned int i = SVt_LAST;
1078 done_sanity_check = TRUE;
1081 assert (bodies_by_type[i].type == i);
1085 assert(bdp->arena_size);
1087 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1089 end = start + arena_size - 2 * body_size;
1091 /* computed count doesnt reflect the 1st slot reservation */
1092 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1093 DEBUG_m(PerlIO_printf(Perl_debug_log,
1094 "arena %p end %p arena-size %d (from %d) type %d "
1096 (void*)start, (void*)end, (int)arena_size,
1097 (int)bdp->arena_size, sv_type, (int)body_size,
1098 (int)arena_size / (int)body_size));
1100 DEBUG_m(PerlIO_printf(Perl_debug_log,
1101 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1102 (void*)start, (void*)end,
1103 (int)bdp->arena_size, sv_type, (int)body_size,
1104 (int)bdp->arena_size / (int)body_size));
1106 *root = (void *)start;
1108 while (start <= end) {
1109 char * const next = start + body_size;
1110 *(void**) start = (void *)next;
1113 *(void **)start = 0;
1118 /* grab a new thing from the free list, allocating more if necessary.
1119 The inline version is used for speed in hot routines, and the
1120 function using it serves the rest (unless PURIFY).
1122 #define new_body_inline(xpv, sv_type) \
1124 void ** const r3wt = &PL_body_roots[sv_type]; \
1125 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1126 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1127 *(r3wt) = *(void**)(xpv); \
1133 S_new_body(pTHX_ const svtype sv_type)
1137 new_body_inline(xpv, sv_type);
1143 static const struct body_details fake_rv =
1144 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1147 =for apidoc sv_upgrade
1149 Upgrade an SV to a more complex form. Generally adds a new body type to the
1150 SV, then copies across as much information as possible from the old body.
1151 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1157 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1162 const svtype old_type = SvTYPE(sv);
1163 const struct body_details *new_type_details;
1164 const struct body_details *old_type_details
1165 = bodies_by_type + old_type;
1166 SV *referant = NULL;
1168 PERL_ARGS_ASSERT_SV_UPGRADE;
1170 if (old_type == new_type)
1173 /* This clause was purposefully added ahead of the early return above to
1174 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1175 inference by Nick I-S that it would fix other troublesome cases. See
1176 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1178 Given that shared hash key scalars are no longer PVIV, but PV, there is
1179 no longer need to unshare so as to free up the IVX slot for its proper
1180 purpose. So it's safe to move the early return earlier. */
1182 if (new_type != SVt_PV && SvIsCOW(sv)) {
1183 sv_force_normal_flags(sv, 0);
1186 old_body = SvANY(sv);
1188 /* Copying structures onto other structures that have been neatly zeroed
1189 has a subtle gotcha. Consider XPVMG
1191 +------+------+------+------+------+-------+-------+
1192 | NV | CUR | LEN | IV | MAGIC | STASH |
1193 +------+------+------+------+------+-------+-------+
1194 0 4 8 12 16 20 24 28
1196 where NVs are aligned to 8 bytes, so that sizeof that structure is
1197 actually 32 bytes long, with 4 bytes of padding at the end:
1199 +------+------+------+------+------+-------+-------+------+
1200 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1201 +------+------+------+------+------+-------+-------+------+
1202 0 4 8 12 16 20 24 28 32
1204 so what happens if you allocate memory for this structure:
1206 +------+------+------+------+------+-------+-------+------+------+...
1207 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1208 +------+------+------+------+------+-------+-------+------+------+...
1209 0 4 8 12 16 20 24 28 32 36
1211 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1212 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1213 started out as zero once, but it's quite possible that it isn't. So now,
1214 rather than a nicely zeroed GP, you have it pointing somewhere random.
1217 (In fact, GP ends up pointing at a previous GP structure, because the
1218 principle cause of the padding in XPVMG getting garbage is a copy of
1219 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1220 this happens to be moot because XPVGV has been re-ordered, with GP
1221 no longer after STASH)
1223 So we are careful and work out the size of used parts of all the
1231 referant = SvRV(sv);
1232 old_type_details = &fake_rv;
1233 if (new_type == SVt_NV)
1234 new_type = SVt_PVNV;
1236 if (new_type < SVt_PVIV) {
1237 new_type = (new_type == SVt_NV)
1238 ? SVt_PVNV : SVt_PVIV;
1243 if (new_type < SVt_PVNV) {
1244 new_type = SVt_PVNV;
1248 assert(new_type > SVt_PV);
1249 assert(SVt_IV < SVt_PV);
1250 assert(SVt_NV < SVt_PV);
1257 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1258 there's no way that it can be safely upgraded, because perl.c
1259 expects to Safefree(SvANY(PL_mess_sv)) */
1260 assert(sv != PL_mess_sv);
1261 /* This flag bit is used to mean other things in other scalar types.
1262 Given that it only has meaning inside the pad, it shouldn't be set
1263 on anything that can get upgraded. */
1264 assert(!SvPAD_TYPED(sv));
1267 if (old_type_details->cant_upgrade)
1268 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1269 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1272 if (old_type > new_type)
1273 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1274 (int)old_type, (int)new_type);
1276 new_type_details = bodies_by_type + new_type;
1278 SvFLAGS(sv) &= ~SVTYPEMASK;
1279 SvFLAGS(sv) |= new_type;
1281 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1282 the return statements above will have triggered. */
1283 assert (new_type != SVt_NULL);
1286 assert(old_type == SVt_NULL);
1287 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1291 assert(old_type == SVt_NULL);
1292 SvANY(sv) = new_XNV();
1297 assert(new_type_details->body_size);
1300 assert(new_type_details->arena);
1301 assert(new_type_details->arena_size);
1302 /* This points to the start of the allocated area. */
1303 new_body_inline(new_body, new_type);
1304 Zero(new_body, new_type_details->body_size, char);
1305 new_body = ((char *)new_body) - new_type_details->offset;
1307 /* We always allocated the full length item with PURIFY. To do this
1308 we fake things so that arena is false for all 16 types.. */
1309 new_body = new_NOARENAZ(new_type_details);
1311 SvANY(sv) = new_body;
1312 if (new_type == SVt_PVAV) {
1316 if (old_type_details->body_size) {
1319 /* It will have been zeroed when the new body was allocated.
1320 Lets not write to it, in case it confuses a write-back
1326 #ifndef NODEFAULT_SHAREKEYS
1327 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1329 HvMAX(sv) = 7; /* (start with 8 buckets) */
1330 if (old_type_details->body_size) {
1333 /* It will have been zeroed when the new body was allocated.
1334 Lets not write to it, in case it confuses a write-back
1339 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1340 The target created by newSVrv also is, and it can have magic.
1341 However, it never has SvPVX set.
1343 if (old_type == SVt_IV) {
1345 } else if (old_type >= SVt_PV) {
1346 assert(SvPVX_const(sv) == 0);
1349 if (old_type >= SVt_PVMG) {
1350 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1351 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1353 sv->sv_u.svu_array = NULL; /* or svu_hash */
1359 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1360 sv_force_normal_flags(sv) is called. */
1363 /* XXX Is this still needed? Was it ever needed? Surely as there is
1364 no route from NV to PVIV, NOK can never be true */
1365 assert(!SvNOKp(sv));
1376 assert(new_type_details->body_size);
1377 /* We always allocated the full length item with PURIFY. To do this
1378 we fake things so that arena is false for all 16 types.. */
1379 if(new_type_details->arena) {
1380 /* This points to the start of the allocated area. */
1381 new_body_inline(new_body, new_type);
1382 Zero(new_body, new_type_details->body_size, char);
1383 new_body = ((char *)new_body) - new_type_details->offset;
1385 new_body = new_NOARENAZ(new_type_details);
1387 SvANY(sv) = new_body;
1389 if (old_type_details->copy) {
1390 /* There is now the potential for an upgrade from something without
1391 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1392 int offset = old_type_details->offset;
1393 int length = old_type_details->copy;
1395 if (new_type_details->offset > old_type_details->offset) {
1396 const int difference
1397 = new_type_details->offset - old_type_details->offset;
1398 offset += difference;
1399 length -= difference;
1401 assert (length >= 0);
1403 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1407 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1408 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1409 * correct 0.0 for us. Otherwise, if the old body didn't have an
1410 * NV slot, but the new one does, then we need to initialise the
1411 * freshly created NV slot with whatever the correct bit pattern is
1413 if (old_type_details->zero_nv && !new_type_details->zero_nv
1414 && !isGV_with_GP(sv))
1418 if (new_type == SVt_PVIO) {
1419 IO * const io = MUTABLE_IO(sv);
1420 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1423 /* Clear the stashcache because a new IO could overrule a package
1425 hv_clear(PL_stashcache);
1427 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1428 IoPAGE_LEN(sv) = 60;
1430 if (old_type < SVt_PV) {
1431 /* referant will be NULL unless the old type was SVt_IV emulating
1433 sv->sv_u.svu_rv = referant;
1437 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1438 (unsigned long)new_type);
1441 if (old_type > SVt_IV) {
1443 my_safefree(old_body);
1445 /* Note that there is an assumption that all bodies of types that
1446 can be upgraded came from arenas. Only the more complex non-
1447 upgradable types are allowed to be directly malloc()ed. */
1448 assert(old_type_details->arena);
1449 del_body((void*)((char*)old_body + old_type_details->offset),
1450 &PL_body_roots[old_type]);
1456 =for apidoc sv_backoff
1458 Remove any string offset. You should normally use the C<SvOOK_off> macro
1465 Perl_sv_backoff(pTHX_ register SV *const sv)
1468 const char * const s = SvPVX_const(sv);
1470 PERL_ARGS_ASSERT_SV_BACKOFF;
1471 PERL_UNUSED_CONTEXT;
1474 assert(SvTYPE(sv) != SVt_PVHV);
1475 assert(SvTYPE(sv) != SVt_PVAV);
1477 SvOOK_offset(sv, delta);
1479 SvLEN_set(sv, SvLEN(sv) + delta);
1480 SvPV_set(sv, SvPVX(sv) - delta);
1481 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1482 SvFLAGS(sv) &= ~SVf_OOK;
1489 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1490 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1491 Use the C<SvGROW> wrapper instead.
1497 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1501 PERL_ARGS_ASSERT_SV_GROW;
1503 if (PL_madskills && newlen >= 0x100000) {
1504 PerlIO_printf(Perl_debug_log,
1505 "Allocation too large: %"UVxf"\n", (UV)newlen);
1507 #ifdef HAS_64K_LIMIT
1508 if (newlen >= 0x10000) {
1509 PerlIO_printf(Perl_debug_log,
1510 "Allocation too large: %"UVxf"\n", (UV)newlen);
1513 #endif /* HAS_64K_LIMIT */
1516 if (SvTYPE(sv) < SVt_PV) {
1517 sv_upgrade(sv, SVt_PV);
1518 s = SvPVX_mutable(sv);
1520 else if (SvOOK(sv)) { /* pv is offset? */
1522 s = SvPVX_mutable(sv);
1523 if (newlen > SvLEN(sv))
1524 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1525 #ifdef HAS_64K_LIMIT
1526 if (newlen >= 0x10000)
1531 s = SvPVX_mutable(sv);
1533 if (newlen > SvLEN(sv)) { /* need more room? */
1534 #ifndef Perl_safesysmalloc_size
1535 newlen = PERL_STRLEN_ROUNDUP(newlen);
1537 if (SvLEN(sv) && s) {
1538 s = (char*)saferealloc(s, newlen);
1541 s = (char*)safemalloc(newlen);
1542 if (SvPVX_const(sv) && SvCUR(sv)) {
1543 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1547 #ifdef Perl_safesysmalloc_size
1548 /* Do this here, do it once, do it right, and then we will never get
1549 called back into sv_grow() unless there really is some growing
1551 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1553 SvLEN_set(sv, newlen);
1560 =for apidoc sv_setiv
1562 Copies an integer into the given SV, upgrading first if necessary.
1563 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1569 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1573 PERL_ARGS_ASSERT_SV_SETIV;
1575 SV_CHECK_THINKFIRST_COW_DROP(sv);
1576 switch (SvTYPE(sv)) {
1579 sv_upgrade(sv, SVt_IV);
1582 sv_upgrade(sv, SVt_PVIV);
1586 if (!isGV_with_GP(sv))
1593 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1597 (void)SvIOK_only(sv); /* validate number */
1603 =for apidoc sv_setiv_mg
1605 Like C<sv_setiv>, but also handles 'set' magic.
1611 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1613 PERL_ARGS_ASSERT_SV_SETIV_MG;
1620 =for apidoc sv_setuv
1622 Copies an unsigned integer into the given SV, upgrading first if necessary.
1623 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1629 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1631 PERL_ARGS_ASSERT_SV_SETUV;
1633 /* With these two if statements:
1634 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1637 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1639 If you wish to remove them, please benchmark to see what the effect is
1641 if (u <= (UV)IV_MAX) {
1642 sv_setiv(sv, (IV)u);
1651 =for apidoc sv_setuv_mg
1653 Like C<sv_setuv>, but also handles 'set' magic.
1659 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1661 PERL_ARGS_ASSERT_SV_SETUV_MG;
1668 =for apidoc sv_setnv
1670 Copies a double into the given SV, upgrading first if necessary.
1671 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1677 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1681 PERL_ARGS_ASSERT_SV_SETNV;
1683 SV_CHECK_THINKFIRST_COW_DROP(sv);
1684 switch (SvTYPE(sv)) {
1687 sv_upgrade(sv, SVt_NV);
1691 sv_upgrade(sv, SVt_PVNV);
1695 if (!isGV_with_GP(sv))
1702 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1707 (void)SvNOK_only(sv); /* validate number */
1712 =for apidoc sv_setnv_mg
1714 Like C<sv_setnv>, but also handles 'set' magic.
1720 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1722 PERL_ARGS_ASSERT_SV_SETNV_MG;
1728 /* Print an "isn't numeric" warning, using a cleaned-up,
1729 * printable version of the offending string
1733 S_not_a_number(pTHX_ SV *const sv)
1740 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1743 dsv = newSVpvs_flags("", SVs_TEMP);
1744 pv = sv_uni_display(dsv, sv, 10, 0);
1747 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1748 /* each *s can expand to 4 chars + "...\0",
1749 i.e. need room for 8 chars */
1751 const char *s = SvPVX_const(sv);
1752 const char * const end = s + SvCUR(sv);
1753 for ( ; s < end && d < limit; s++ ) {
1755 if (ch & 128 && !isPRINT_LC(ch)) {
1764 else if (ch == '\r') {
1768 else if (ch == '\f') {
1772 else if (ch == '\\') {
1776 else if (ch == '\0') {
1780 else if (isPRINT_LC(ch))
1797 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1798 "Argument \"%s\" isn't numeric in %s", pv,
1801 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1802 "Argument \"%s\" isn't numeric", pv);
1806 =for apidoc looks_like_number
1808 Test if the content of an SV looks like a number (or is a number).
1809 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1810 non-numeric warning), even if your atof() doesn't grok them.
1816 Perl_looks_like_number(pTHX_ SV *const sv)
1818 register const char *sbegin;
1821 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1824 sbegin = SvPVX_const(sv);
1827 else if (SvPOKp(sv))
1828 sbegin = SvPV_const(sv, len);
1830 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1831 return grok_number(sbegin, len, NULL);
1835 S_glob_2number(pTHX_ GV * const gv)
1837 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1838 SV *const buffer = sv_newmortal();
1840 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1842 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1845 gv_efullname3(buffer, gv, "*");
1846 SvFLAGS(gv) |= wasfake;
1848 /* We know that all GVs stringify to something that is not-a-number,
1849 so no need to test that. */
1850 if (ckWARN(WARN_NUMERIC))
1851 not_a_number(buffer);
1852 /* We just want something true to return, so that S_sv_2iuv_common
1853 can tail call us and return true. */
1857 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1858 until proven guilty, assume that things are not that bad... */
1863 As 64 bit platforms often have an NV that doesn't preserve all bits of
1864 an IV (an assumption perl has been based on to date) it becomes necessary
1865 to remove the assumption that the NV always carries enough precision to
1866 recreate the IV whenever needed, and that the NV is the canonical form.
1867 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1868 precision as a side effect of conversion (which would lead to insanity
1869 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1870 1) to distinguish between IV/UV/NV slots that have cached a valid
1871 conversion where precision was lost and IV/UV/NV slots that have a
1872 valid conversion which has lost no precision
1873 2) to ensure that if a numeric conversion to one form is requested that
1874 would lose precision, the precise conversion (or differently
1875 imprecise conversion) is also performed and cached, to prevent
1876 requests for different numeric formats on the same SV causing
1877 lossy conversion chains. (lossless conversion chains are perfectly
1882 SvIOKp is true if the IV slot contains a valid value
1883 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1884 SvNOKp is true if the NV slot contains a valid value
1885 SvNOK is true only if the NV value is accurate
1888 while converting from PV to NV, check to see if converting that NV to an
1889 IV(or UV) would lose accuracy over a direct conversion from PV to
1890 IV(or UV). If it would, cache both conversions, return NV, but mark
1891 SV as IOK NOKp (ie not NOK).
1893 While converting from PV to IV, check to see if converting that IV to an
1894 NV would lose accuracy over a direct conversion from PV to NV. If it
1895 would, cache both conversions, flag similarly.
1897 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1898 correctly because if IV & NV were set NV *always* overruled.
1899 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1900 changes - now IV and NV together means that the two are interchangeable:
1901 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1903 The benefit of this is that operations such as pp_add know that if
1904 SvIOK is true for both left and right operands, then integer addition
1905 can be used instead of floating point (for cases where the result won't
1906 overflow). Before, floating point was always used, which could lead to
1907 loss of precision compared with integer addition.
1909 * making IV and NV equal status should make maths accurate on 64 bit
1911 * may speed up maths somewhat if pp_add and friends start to use
1912 integers when possible instead of fp. (Hopefully the overhead in
1913 looking for SvIOK and checking for overflow will not outweigh the
1914 fp to integer speedup)
1915 * will slow down integer operations (callers of SvIV) on "inaccurate"
1916 values, as the change from SvIOK to SvIOKp will cause a call into
1917 sv_2iv each time rather than a macro access direct to the IV slot
1918 * should speed up number->string conversion on integers as IV is
1919 favoured when IV and NV are equally accurate
1921 ####################################################################
1922 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1923 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1924 On the other hand, SvUOK is true iff UV.
1925 ####################################################################
1927 Your mileage will vary depending your CPU's relative fp to integer
1931 #ifndef NV_PRESERVES_UV
1932 # define IS_NUMBER_UNDERFLOW_IV 1
1933 # define IS_NUMBER_UNDERFLOW_UV 2
1934 # define IS_NUMBER_IV_AND_UV 2
1935 # define IS_NUMBER_OVERFLOW_IV 4
1936 # define IS_NUMBER_OVERFLOW_UV 5
1938 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1940 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1942 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1950 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1952 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));
1953 if (SvNVX(sv) < (NV)IV_MIN) {
1954 (void)SvIOKp_on(sv);
1956 SvIV_set(sv, IV_MIN);
1957 return IS_NUMBER_UNDERFLOW_IV;
1959 if (SvNVX(sv) > (NV)UV_MAX) {
1960 (void)SvIOKp_on(sv);
1963 SvUV_set(sv, UV_MAX);
1964 return IS_NUMBER_OVERFLOW_UV;
1966 (void)SvIOKp_on(sv);
1968 /* Can't use strtol etc to convert this string. (See truth table in
1970 if (SvNVX(sv) <= (UV)IV_MAX) {
1971 SvIV_set(sv, I_V(SvNVX(sv)));
1972 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1973 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1975 /* Integer is imprecise. NOK, IOKp */
1977 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1980 SvUV_set(sv, U_V(SvNVX(sv)));
1981 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1982 if (SvUVX(sv) == UV_MAX) {
1983 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1984 possibly be preserved by NV. Hence, it must be overflow.
1986 return IS_NUMBER_OVERFLOW_UV;
1988 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1990 /* Integer is imprecise. NOK, IOKp */
1992 return IS_NUMBER_OVERFLOW_IV;
1994 #endif /* !NV_PRESERVES_UV*/
1997 S_sv_2iuv_common(pTHX_ SV *const sv)
2001 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2004 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2005 * without also getting a cached IV/UV from it at the same time
2006 * (ie PV->NV conversion should detect loss of accuracy and cache
2007 * IV or UV at same time to avoid this. */
2008 /* IV-over-UV optimisation - choose to cache IV if possible */
2010 if (SvTYPE(sv) == SVt_NV)
2011 sv_upgrade(sv, SVt_PVNV);
2013 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2014 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2015 certainly cast into the IV range at IV_MAX, whereas the correct
2016 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2018 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2019 if (Perl_isnan(SvNVX(sv))) {
2025 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2026 SvIV_set(sv, I_V(SvNVX(sv)));
2027 if (SvNVX(sv) == (NV) SvIVX(sv)
2028 #ifndef NV_PRESERVES_UV
2029 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2030 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2031 /* Don't flag it as "accurately an integer" if the number
2032 came from a (by definition imprecise) NV operation, and
2033 we're outside the range of NV integer precision */
2037 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2039 /* scalar has trailing garbage, eg "42a" */
2041 DEBUG_c(PerlIO_printf(Perl_debug_log,
2042 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2048 /* IV not precise. No need to convert from PV, as NV
2049 conversion would already have cached IV if it detected
2050 that PV->IV would be better than PV->NV->IV
2051 flags already correct - don't set public IOK. */
2052 DEBUG_c(PerlIO_printf(Perl_debug_log,
2053 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2058 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2059 but the cast (NV)IV_MIN rounds to a the value less (more
2060 negative) than IV_MIN which happens to be equal to SvNVX ??
2061 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2062 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2063 (NV)UVX == NVX are both true, but the values differ. :-(
2064 Hopefully for 2s complement IV_MIN is something like
2065 0x8000000000000000 which will be exact. NWC */
2068 SvUV_set(sv, U_V(SvNVX(sv)));
2070 (SvNVX(sv) == (NV) SvUVX(sv))
2071 #ifndef NV_PRESERVES_UV
2072 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2073 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2074 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2075 /* Don't flag it as "accurately an integer" if the number
2076 came from a (by definition imprecise) NV operation, and
2077 we're outside the range of NV integer precision */
2083 DEBUG_c(PerlIO_printf(Perl_debug_log,
2084 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2090 else if (SvPOKp(sv) && SvLEN(sv)) {
2092 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2093 /* We want to avoid a possible problem when we cache an IV/ a UV which
2094 may be later translated to an NV, and the resulting NV is not
2095 the same as the direct translation of the initial string
2096 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2097 be careful to ensure that the value with the .456 is around if the
2098 NV value is requested in the future).
2100 This means that if we cache such an IV/a UV, we need to cache the
2101 NV as well. Moreover, we trade speed for space, and do not
2102 cache the NV if we are sure it's not needed.
2105 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2106 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2107 == IS_NUMBER_IN_UV) {
2108 /* It's definitely an integer, only upgrade to PVIV */
2109 if (SvTYPE(sv) < SVt_PVIV)
2110 sv_upgrade(sv, SVt_PVIV);
2112 } else if (SvTYPE(sv) < SVt_PVNV)
2113 sv_upgrade(sv, SVt_PVNV);
2115 /* If NVs preserve UVs then we only use the UV value if we know that
2116 we aren't going to call atof() below. If NVs don't preserve UVs
2117 then the value returned may have more precision than atof() will
2118 return, even though value isn't perfectly accurate. */
2119 if ((numtype & (IS_NUMBER_IN_UV
2120 #ifdef NV_PRESERVES_UV
2123 )) == IS_NUMBER_IN_UV) {
2124 /* This won't turn off the public IOK flag if it was set above */
2125 (void)SvIOKp_on(sv);
2127 if (!(numtype & IS_NUMBER_NEG)) {
2129 if (value <= (UV)IV_MAX) {
2130 SvIV_set(sv, (IV)value);
2132 /* it didn't overflow, and it was positive. */
2133 SvUV_set(sv, value);
2137 /* 2s complement assumption */
2138 if (value <= (UV)IV_MIN) {
2139 SvIV_set(sv, -(IV)value);
2141 /* Too negative for an IV. This is a double upgrade, but
2142 I'm assuming it will be rare. */
2143 if (SvTYPE(sv) < SVt_PVNV)
2144 sv_upgrade(sv, SVt_PVNV);
2148 SvNV_set(sv, -(NV)value);
2149 SvIV_set(sv, IV_MIN);
2153 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2154 will be in the previous block to set the IV slot, and the next
2155 block to set the NV slot. So no else here. */
2157 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2158 != IS_NUMBER_IN_UV) {
2159 /* It wasn't an (integer that doesn't overflow the UV). */
2160 SvNV_set(sv, Atof(SvPVX_const(sv)));
2162 if (! numtype && ckWARN(WARN_NUMERIC))
2165 #if defined(USE_LONG_DOUBLE)
2166 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2167 PTR2UV(sv), SvNVX(sv)));
2169 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2170 PTR2UV(sv), SvNVX(sv)));
2173 #ifdef NV_PRESERVES_UV
2174 (void)SvIOKp_on(sv);
2176 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2177 SvIV_set(sv, I_V(SvNVX(sv)));
2178 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2181 NOOP; /* Integer is imprecise. NOK, IOKp */
2183 /* UV will not work better than IV */
2185 if (SvNVX(sv) > (NV)UV_MAX) {
2187 /* Integer is inaccurate. NOK, IOKp, is UV */
2188 SvUV_set(sv, UV_MAX);
2190 SvUV_set(sv, U_V(SvNVX(sv)));
2191 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2192 NV preservse UV so can do correct comparison. */
2193 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2196 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2201 #else /* NV_PRESERVES_UV */
2202 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2203 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2204 /* The IV/UV slot will have been set from value returned by
2205 grok_number above. The NV slot has just been set using
2208 assert (SvIOKp(sv));
2210 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2211 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2212 /* Small enough to preserve all bits. */
2213 (void)SvIOKp_on(sv);
2215 SvIV_set(sv, I_V(SvNVX(sv)));
2216 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2218 /* Assumption: first non-preserved integer is < IV_MAX,
2219 this NV is in the preserved range, therefore: */
2220 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2222 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);
2226 0 0 already failed to read UV.
2227 0 1 already failed to read UV.
2228 1 0 you won't get here in this case. IV/UV
2229 slot set, public IOK, Atof() unneeded.
2230 1 1 already read UV.
2231 so there's no point in sv_2iuv_non_preserve() attempting
2232 to use atol, strtol, strtoul etc. */
2234 sv_2iuv_non_preserve (sv, numtype);
2236 sv_2iuv_non_preserve (sv);
2240 #endif /* NV_PRESERVES_UV */
2241 /* It might be more code efficient to go through the entire logic above
2242 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2243 gets complex and potentially buggy, so more programmer efficient
2244 to do it this way, by turning off the public flags: */
2246 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2250 if (isGV_with_GP(sv))
2251 return glob_2number(MUTABLE_GV(sv));
2253 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2254 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2257 if (SvTYPE(sv) < SVt_IV)
2258 /* Typically the caller expects that sv_any is not NULL now. */
2259 sv_upgrade(sv, SVt_IV);
2260 /* Return 0 from the caller. */
2267 =for apidoc sv_2iv_flags
2269 Return the integer value of an SV, doing any necessary string
2270 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2271 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2277 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2282 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2283 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2284 cache IVs just in case. In practice it seems that they never
2285 actually anywhere accessible by user Perl code, let alone get used
2286 in anything other than a string context. */
2287 if (flags & SV_GMAGIC)
2292 return I_V(SvNVX(sv));
2294 if (SvPOKp(sv) && SvLEN(sv)) {
2297 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2299 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2300 == IS_NUMBER_IN_UV) {
2301 /* It's definitely an integer */
2302 if (numtype & IS_NUMBER_NEG) {
2303 if (value < (UV)IV_MIN)
2306 if (value < (UV)IV_MAX)
2311 if (ckWARN(WARN_NUMERIC))
2314 return I_V(Atof(SvPVX_const(sv)));
2319 assert(SvTYPE(sv) >= SVt_PVMG);
2320 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2321 } else if (SvTHINKFIRST(sv)) {
2325 SV * const tmpstr=AMG_CALLun(sv,numer);
2326 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2327 return SvIV(tmpstr);
2330 return PTR2IV(SvRV(sv));
2333 sv_force_normal_flags(sv, 0);
2335 if (SvREADONLY(sv) && !SvOK(sv)) {
2336 if (ckWARN(WARN_UNINITIALIZED))
2342 if (S_sv_2iuv_common(aTHX_ sv))
2345 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2346 PTR2UV(sv),SvIVX(sv)));
2347 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2351 =for apidoc sv_2uv_flags
2353 Return the unsigned integer value of an SV, doing any necessary string
2354 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2355 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2361 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2366 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2367 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2368 cache IVs just in case. */
2369 if (flags & SV_GMAGIC)
2374 return U_V(SvNVX(sv));
2375 if (SvPOKp(sv) && SvLEN(sv)) {
2378 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2380 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2381 == IS_NUMBER_IN_UV) {
2382 /* It's definitely an integer */
2383 if (!(numtype & IS_NUMBER_NEG))
2387 if (ckWARN(WARN_NUMERIC))
2390 return U_V(Atof(SvPVX_const(sv)));
2395 assert(SvTYPE(sv) >= SVt_PVMG);
2396 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2397 } else if (SvTHINKFIRST(sv)) {
2401 SV *const tmpstr = AMG_CALLun(sv,numer);
2402 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2403 return SvUV(tmpstr);
2406 return PTR2UV(SvRV(sv));
2409 sv_force_normal_flags(sv, 0);
2411 if (SvREADONLY(sv) && !SvOK(sv)) {
2412 if (ckWARN(WARN_UNINITIALIZED))
2418 if (S_sv_2iuv_common(aTHX_ sv))
2422 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2423 PTR2UV(sv),SvUVX(sv)));
2424 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2430 Return the num value of an SV, doing any necessary string or integer
2431 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2438 Perl_sv_2nv(pTHX_ register SV *const sv)
2443 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2444 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2445 cache IVs just in case. */
2449 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2450 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2451 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2453 return Atof(SvPVX_const(sv));
2457 return (NV)SvUVX(sv);
2459 return (NV)SvIVX(sv);
2464 assert(SvTYPE(sv) >= SVt_PVMG);
2465 /* This falls through to the report_uninit near the end of the
2467 } else if (SvTHINKFIRST(sv)) {
2471 SV *const tmpstr = AMG_CALLun(sv,numer);
2472 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2473 return SvNV(tmpstr);
2476 return PTR2NV(SvRV(sv));
2479 sv_force_normal_flags(sv, 0);
2481 if (SvREADONLY(sv) && !SvOK(sv)) {
2482 if (ckWARN(WARN_UNINITIALIZED))
2487 if (SvTYPE(sv) < SVt_NV) {
2488 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2489 sv_upgrade(sv, SVt_NV);
2490 #ifdef USE_LONG_DOUBLE
2492 STORE_NUMERIC_LOCAL_SET_STANDARD();
2493 PerlIO_printf(Perl_debug_log,
2494 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2495 PTR2UV(sv), SvNVX(sv));
2496 RESTORE_NUMERIC_LOCAL();
2500 STORE_NUMERIC_LOCAL_SET_STANDARD();
2501 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2502 PTR2UV(sv), SvNVX(sv));
2503 RESTORE_NUMERIC_LOCAL();
2507 else if (SvTYPE(sv) < SVt_PVNV)
2508 sv_upgrade(sv, SVt_PVNV);
2513 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2514 #ifdef NV_PRESERVES_UV
2520 /* Only set the public NV OK flag if this NV preserves the IV */
2521 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2523 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2524 : (SvIVX(sv) == I_V(SvNVX(sv))))
2530 else if (SvPOKp(sv) && SvLEN(sv)) {
2532 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2533 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2535 #ifdef NV_PRESERVES_UV
2536 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2537 == IS_NUMBER_IN_UV) {
2538 /* It's definitely an integer */
2539 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2541 SvNV_set(sv, Atof(SvPVX_const(sv)));
2547 SvNV_set(sv, Atof(SvPVX_const(sv)));
2548 /* Only set the public NV OK flag if this NV preserves the value in
2549 the PV at least as well as an IV/UV would.
2550 Not sure how to do this 100% reliably. */
2551 /* if that shift count is out of range then Configure's test is
2552 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2554 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2555 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2556 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2557 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2558 /* Can't use strtol etc to convert this string, so don't try.
2559 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2562 /* value has been set. It may not be precise. */
2563 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2564 /* 2s complement assumption for (UV)IV_MIN */
2565 SvNOK_on(sv); /* Integer is too negative. */
2570 if (numtype & IS_NUMBER_NEG) {
2571 SvIV_set(sv, -(IV)value);
2572 } else if (value <= (UV)IV_MAX) {
2573 SvIV_set(sv, (IV)value);
2575 SvUV_set(sv, value);
2579 if (numtype & IS_NUMBER_NOT_INT) {
2580 /* I believe that even if the original PV had decimals,
2581 they are lost beyond the limit of the FP precision.
2582 However, neither is canonical, so both only get p
2583 flags. NWC, 2000/11/25 */
2584 /* Both already have p flags, so do nothing */
2586 const NV nv = SvNVX(sv);
2587 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2588 if (SvIVX(sv) == I_V(nv)) {
2591 /* It had no "." so it must be integer. */
2595 /* between IV_MAX and NV(UV_MAX).
2596 Could be slightly > UV_MAX */
2598 if (numtype & IS_NUMBER_NOT_INT) {
2599 /* UV and NV both imprecise. */
2601 const UV nv_as_uv = U_V(nv);
2603 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2612 /* It might be more code efficient to go through the entire logic above
2613 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2614 gets complex and potentially buggy, so more programmer efficient
2615 to do it this way, by turning off the public flags: */
2617 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2618 #endif /* NV_PRESERVES_UV */
2621 if (isGV_with_GP(sv)) {
2622 glob_2number(MUTABLE_GV(sv));
2626 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2628 assert (SvTYPE(sv) >= SVt_NV);
2629 /* Typically the caller expects that sv_any is not NULL now. */
2630 /* XXX Ilya implies that this is a bug in callers that assume this
2631 and ideally should be fixed. */
2634 #if defined(USE_LONG_DOUBLE)
2636 STORE_NUMERIC_LOCAL_SET_STANDARD();
2637 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2638 PTR2UV(sv), SvNVX(sv));
2639 RESTORE_NUMERIC_LOCAL();
2643 STORE_NUMERIC_LOCAL_SET_STANDARD();
2644 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2645 PTR2UV(sv), SvNVX(sv));
2646 RESTORE_NUMERIC_LOCAL();
2655 Return an SV with the numeric value of the source SV, doing any necessary
2656 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2657 access this function.
2663 Perl_sv_2num(pTHX_ register SV *const sv)
2665 PERL_ARGS_ASSERT_SV_2NUM;
2670 SV * const tmpsv = AMG_CALLun(sv,numer);
2671 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2672 return sv_2num(tmpsv);
2674 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2677 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2678 * UV as a string towards the end of buf, and return pointers to start and
2681 * We assume that buf is at least TYPE_CHARS(UV) long.
2685 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2687 char *ptr = buf + TYPE_CHARS(UV);
2688 char * const ebuf = ptr;
2691 PERL_ARGS_ASSERT_UIV_2BUF;
2703 *--ptr = '0' + (char)(uv % 10);
2712 =for apidoc sv_2pv_flags
2714 Returns a pointer to the string value of an SV, and sets *lp to its length.
2715 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2717 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2718 usually end up here too.
2724 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2734 if (SvGMAGICAL(sv)) {
2735 if (flags & SV_GMAGIC)
2740 if (flags & SV_MUTABLE_RETURN)
2741 return SvPVX_mutable(sv);
2742 if (flags & SV_CONST_RETURN)
2743 return (char *)SvPVX_const(sv);
2746 if (SvIOKp(sv) || SvNOKp(sv)) {
2747 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2752 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2753 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2755 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2762 #ifdef FIXNEGATIVEZERO
2763 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2769 SvUPGRADE(sv, SVt_PV);
2772 s = SvGROW_mutable(sv, len + 1);
2775 return (char*)memcpy(s, tbuf, len + 1);
2781 assert(SvTYPE(sv) >= SVt_PVMG);
2782 /* This falls through to the report_uninit near the end of the
2784 } else if (SvTHINKFIRST(sv)) {
2788 SV *const tmpstr = AMG_CALLun(sv,string);
2789 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2791 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2795 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2796 if (flags & SV_CONST_RETURN) {
2797 pv = (char *) SvPVX_const(tmpstr);
2799 pv = (flags & SV_MUTABLE_RETURN)
2800 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2803 *lp = SvCUR(tmpstr);
2805 pv = sv_2pv_flags(tmpstr, lp, flags);
2818 SV *const referent = SvRV(sv);
2822 retval = buffer = savepvn("NULLREF", len);
2823 } else if (SvTYPE(referent) == SVt_REGEXP) {
2824 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2829 /* If the regex is UTF-8 we want the containing scalar to
2830 have an UTF-8 flag too */
2836 if ((seen_evals = RX_SEEN_EVALS(re)))
2837 PL_reginterp_cnt += seen_evals;
2840 *lp = RX_WRAPLEN(re);
2842 return RX_WRAPPED(re);
2844 const char *const typestr = sv_reftype(referent, 0);
2845 const STRLEN typelen = strlen(typestr);
2846 UV addr = PTR2UV(referent);
2847 const char *stashname = NULL;
2848 STRLEN stashnamelen = 0; /* hush, gcc */
2849 const char *buffer_end;
2851 if (SvOBJECT(referent)) {
2852 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2855 stashname = HEK_KEY(name);
2856 stashnamelen = HEK_LEN(name);
2858 if (HEK_UTF8(name)) {
2864 stashname = "__ANON__";
2867 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2868 + 2 * sizeof(UV) + 2 /* )\0 */;
2870 len = typelen + 3 /* (0x */
2871 + 2 * sizeof(UV) + 2 /* )\0 */;
2874 Newx(buffer, len, char);
2875 buffer_end = retval = buffer + len;
2877 /* Working backwards */
2881 *--retval = PL_hexdigit[addr & 15];
2882 } while (addr >>= 4);
2888 memcpy(retval, typestr, typelen);
2892 retval -= stashnamelen;
2893 memcpy(retval, stashname, stashnamelen);
2895 /* retval may not neccesarily have reached the start of the
2897 assert (retval >= buffer);
2899 len = buffer_end - retval - 1; /* -1 for that \0 */
2907 if (SvREADONLY(sv) && !SvOK(sv)) {
2910 if (flags & SV_UNDEF_RETURNS_NULL)
2912 if (ckWARN(WARN_UNINITIALIZED))
2917 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2918 /* I'm assuming that if both IV and NV are equally valid then
2919 converting the IV is going to be more efficient */
2920 const U32 isUIOK = SvIsUV(sv);
2921 char buf[TYPE_CHARS(UV)];
2925 if (SvTYPE(sv) < SVt_PVIV)
2926 sv_upgrade(sv, SVt_PVIV);
2927 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2929 /* inlined from sv_setpvn */
2930 s = SvGROW_mutable(sv, len + 1);
2931 Move(ptr, s, len, char);
2935 else if (SvNOKp(sv)) {
2937 if (SvTYPE(sv) < SVt_PVNV)
2938 sv_upgrade(sv, SVt_PVNV);
2939 /* The +20 is pure guesswork. Configure test needed. --jhi */
2940 s = SvGROW_mutable(sv, NV_DIG + 20);
2941 /* some Xenix systems wipe out errno here */
2943 if (SvNVX(sv) == 0.0)
2944 my_strlcpy(s, "0", SvLEN(sv));
2948 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2951 #ifdef FIXNEGATIVEZERO
2952 if (*s == '-' && s[1] == '0' && !s[2]) {
2964 if (isGV_with_GP(sv)) {
2965 GV *const gv = MUTABLE_GV(sv);
2966 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2967 SV *const buffer = sv_newmortal();
2969 /* FAKE globs can get coerced, so need to turn this off temporarily
2972 gv_efullname3(buffer, gv, "*");
2973 SvFLAGS(gv) |= wasfake;
2975 if (SvPOK(buffer)) {
2977 *lp = SvCUR(buffer);
2979 return SvPVX(buffer);
2990 if (flags & SV_UNDEF_RETURNS_NULL)
2992 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2994 if (SvTYPE(sv) < SVt_PV)
2995 /* Typically the caller expects that sv_any is not NULL now. */
2996 sv_upgrade(sv, SVt_PV);
3000 const STRLEN len = s - SvPVX_const(sv);
3006 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3007 PTR2UV(sv),SvPVX_const(sv)));
3008 if (flags & SV_CONST_RETURN)
3009 return (char *)SvPVX_const(sv);
3010 if (flags & SV_MUTABLE_RETURN)
3011 return SvPVX_mutable(sv);
3016 =for apidoc sv_copypv
3018 Copies a stringified representation of the source SV into the
3019 destination SV. Automatically performs any necessary mg_get and
3020 coercion of numeric values into strings. Guaranteed to preserve
3021 UTF8 flag even from overloaded objects. Similar in nature to
3022 sv_2pv[_flags] but operates directly on an SV instead of just the
3023 string. Mostly uses sv_2pv_flags to do its work, except when that
3024 would lose the UTF-8'ness of the PV.
3030 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3033 const char * const s = SvPV_const(ssv,len);
3035 PERL_ARGS_ASSERT_SV_COPYPV;
3037 sv_setpvn(dsv,s,len);
3045 =for apidoc sv_2pvbyte
3047 Return a pointer to the byte-encoded representation of the SV, and set *lp
3048 to its length. May cause the SV to be downgraded from UTF-8 as a
3051 Usually accessed via the C<SvPVbyte> macro.
3057 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3059 PERL_ARGS_ASSERT_SV_2PVBYTE;
3061 sv_utf8_downgrade(sv,0);
3062 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3066 =for apidoc sv_2pvutf8
3068 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3069 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3071 Usually accessed via the C<SvPVutf8> macro.
3077 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3079 PERL_ARGS_ASSERT_SV_2PVUTF8;
3081 sv_utf8_upgrade(sv);
3082 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3087 =for apidoc sv_2bool
3089 This function is only called on magical items, and is only used by
3090 sv_true() or its macro equivalent.
3096 Perl_sv_2bool(pTHX_ register SV *const sv)
3100 PERL_ARGS_ASSERT_SV_2BOOL;
3108 SV * const tmpsv = AMG_CALLun(sv,bool_);
3109 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3110 return cBOOL(SvTRUE(tmpsv));
3112 return SvRV(sv) != 0;
3115 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3117 (*sv->sv_u.svu_pv > '0' ||
3118 Xpvtmp->xpv_cur > 1 ||
3119 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3126 return SvIVX(sv) != 0;
3129 return SvNVX(sv) != 0.0;
3131 if (isGV_with_GP(sv))
3141 =for apidoc sv_utf8_upgrade
3143 Converts the PV of an SV to its UTF-8-encoded form.
3144 Forces the SV to string form if it is not already.
3145 Will C<mg_get> on C<sv> if appropriate.
3146 Always sets the SvUTF8 flag to avoid future validity checks even
3147 if the whole string is the same in UTF-8 as not.
3148 Returns the number of bytes in the converted string
3150 This is not as a general purpose byte encoding to Unicode interface:
3151 use the Encode extension for that.
3153 =for apidoc sv_utf8_upgrade_nomg
3155 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3157 =for apidoc sv_utf8_upgrade_flags
3159 Converts the PV of an SV to its UTF-8-encoded form.
3160 Forces the SV to string form if it is not already.
3161 Always sets the SvUTF8 flag to avoid future validity checks even
3162 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3163 will C<mg_get> on C<sv> if appropriate, else not.
3164 Returns the number of bytes in the converted string
3165 C<sv_utf8_upgrade> and
3166 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3168 This is not as a general purpose byte encoding to Unicode interface:
3169 use the Encode extension for that.
3173 The grow version is currently not externally documented. It adds a parameter,
3174 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3175 have free after it upon return. This allows the caller to reserve extra space
3176 that it intends to fill, to avoid extra grows.
3178 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3179 which can be used to tell this function to not first check to see if there are
3180 any characters that are different in UTF-8 (variant characters) which would
3181 force it to allocate a new string to sv, but to assume there are. Typically
3182 this flag is used by a routine that has already parsed the string to find that
3183 there are such characters, and passes this information on so that the work
3184 doesn't have to be repeated.
3186 (One might think that the calling routine could pass in the position of the
3187 first such variant, so it wouldn't have to be found again. But that is not the
3188 case, because typically when the caller is likely to use this flag, it won't be
3189 calling this routine unless it finds something that won't fit into a byte.
3190 Otherwise it tries to not upgrade and just use bytes. But some things that
3191 do fit into a byte are variants in utf8, and the caller may not have been
3192 keeping track of these.)
3194 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3195 isn't guaranteed due to having other routines do the work in some input cases,
3196 or if the input is already flagged as being in utf8.
3198 The speed of this could perhaps be improved for many cases if someone wanted to
3199 write a fast function that counts the number of variant characters in a string,
3200 especially if it could return the position of the first one.
3205 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3209 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3211 if (sv == &PL_sv_undef)
3215 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3216 (void) sv_2pv_flags(sv,&len, flags);
3218 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3222 (void) SvPV_force(sv,len);
3227 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3232 sv_force_normal_flags(sv, 0);
3235 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3236 sv_recode_to_utf8(sv, PL_encoding);
3237 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3241 if (SvCUR(sv) == 0) {
3242 if (extra) SvGROW(sv, extra);
3243 } else { /* Assume Latin-1/EBCDIC */
3244 /* This function could be much more efficient if we
3245 * had a FLAG in SVs to signal if there are any variant
3246 * chars in the PV. Given that there isn't such a flag
3247 * make the loop as fast as possible (although there are certainly ways
3248 * to speed this up, eg. through vectorization) */
3249 U8 * s = (U8 *) SvPVX_const(sv);
3250 U8 * e = (U8 *) SvEND(sv);
3252 STRLEN two_byte_count = 0;
3254 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3256 /* See if really will need to convert to utf8. We mustn't rely on our
3257 * incoming SV being well formed and having a trailing '\0', as certain
3258 * code in pp_formline can send us partially built SVs. */
3262 if (NATIVE_IS_INVARIANT(ch)) continue;
3264 t--; /* t already incremented; re-point to first variant */
3269 /* utf8 conversion not needed because all are invariants. Mark as
3270 * UTF-8 even if no variant - saves scanning loop */
3276 /* Here, the string should be converted to utf8, either because of an
3277 * input flag (two_byte_count = 0), or because a character that
3278 * requires 2 bytes was found (two_byte_count = 1). t points either to
3279 * the beginning of the string (if we didn't examine anything), or to
3280 * the first variant. In either case, everything from s to t - 1 will
3281 * occupy only 1 byte each on output.
3283 * There are two main ways to convert. One is to create a new string
3284 * and go through the input starting from the beginning, appending each
3285 * converted value onto the new string as we go along. It's probably
3286 * best to allocate enough space in the string for the worst possible
3287 * case rather than possibly running out of space and having to
3288 * reallocate and then copy what we've done so far. Since everything
3289 * from s to t - 1 is invariant, the destination can be initialized
3290 * with these using a fast memory copy
3292 * The other way is to figure out exactly how big the string should be
3293 * by parsing the entire input. Then you don't have to make it big
3294 * enough to handle the worst possible case, and more importantly, if
3295 * the string you already have is large enough, you don't have to
3296 * allocate a new string, you can copy the last character in the input
3297 * string to the final position(s) that will be occupied by the
3298 * converted string and go backwards, stopping at t, since everything
3299 * before that is invariant.
3301 * There are advantages and disadvantages to each method.
3303 * In the first method, we can allocate a new string, do the memory
3304 * copy from the s to t - 1, and then proceed through the rest of the
3305 * string byte-by-byte.
3307 * In the second method, we proceed through the rest of the input
3308 * string just calculating how big the converted string will be. Then
3309 * there are two cases:
3310 * 1) if the string has enough extra space to handle the converted
3311 * value. We go backwards through the string, converting until we
3312 * get to the position we are at now, and then stop. If this
3313 * position is far enough along in the string, this method is
3314 * faster than the other method. If the memory copy were the same
3315 * speed as the byte-by-byte loop, that position would be about
3316 * half-way, as at the half-way mark, parsing to the end and back
3317 * is one complete string's parse, the same amount as starting
3318 * over and going all the way through. Actually, it would be
3319 * somewhat less than half-way, as it's faster to just count bytes
3320 * than to also copy, and we don't have the overhead of allocating
3321 * a new string, changing the scalar to use it, and freeing the
3322 * existing one. But if the memory copy is fast, the break-even
3323 * point is somewhere after half way. The counting loop could be
3324 * sped up by vectorization, etc, to move the break-even point
3325 * further towards the beginning.
3326 * 2) if the string doesn't have enough space to handle the converted
3327 * value. A new string will have to be allocated, and one might
3328 * as well, given that, start from the beginning doing the first
3329 * method. We've spent extra time parsing the string and in
3330 * exchange all we've gotten is that we know precisely how big to
3331 * make the new one. Perl is more optimized for time than space,
3332 * so this case is a loser.
3333 * So what I've decided to do is not use the 2nd method unless it is
3334 * guaranteed that a new string won't have to be allocated, assuming
3335 * the worst case. I also decided not to put any more conditions on it
3336 * than this, for now. It seems likely that, since the worst case is
3337 * twice as big as the unknown portion of the string (plus 1), we won't
3338 * be guaranteed enough space, causing us to go to the first method,
3339 * unless the string is short, or the first variant character is near
3340 * the end of it. In either of these cases, it seems best to use the
3341 * 2nd method. The only circumstance I can think of where this would
3342 * be really slower is if the string had once had much more data in it
3343 * than it does now, but there is still a substantial amount in it */
3346 STRLEN invariant_head = t - s;
3347 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3348 if (SvLEN(sv) < size) {
3350 /* Here, have decided to allocate a new string */
3355 Newx(dst, size, U8);
3357 /* If no known invariants at the beginning of the input string,
3358 * set so starts from there. Otherwise, can use memory copy to
3359 * get up to where we are now, and then start from here */
3361 if (invariant_head <= 0) {
3364 Copy(s, dst, invariant_head, char);
3365 d = dst + invariant_head;
3369 const UV uv = NATIVE8_TO_UNI(*t++);
3370 if (UNI_IS_INVARIANT(uv))
3371 *d++ = (U8)UNI_TO_NATIVE(uv);
3373 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3374 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3378 SvPV_free(sv); /* No longer using pre-existing string */
3379 SvPV_set(sv, (char*)dst);
3380 SvCUR_set(sv, d - dst);
3381 SvLEN_set(sv, size);
3384 /* Here, have decided to get the exact size of the string.
3385 * Currently this happens only when we know that there is
3386 * guaranteed enough space to fit the converted string, so
3387 * don't have to worry about growing. If two_byte_count is 0,
3388 * then t points to the first byte of the string which hasn't
3389 * been examined yet. Otherwise two_byte_count is 1, and t
3390 * points to the first byte in the string that will expand to
3391 * two. Depending on this, start examining at t or 1 after t.
3394 U8 *d = t + two_byte_count;
3397 /* Count up the remaining bytes that expand to two */
3400 const U8 chr = *d++;
3401 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3404 /* The string will expand by just the number of bytes that
3405 * occupy two positions. But we are one afterwards because of
3406 * the increment just above. This is the place to put the
3407 * trailing NUL, and to set the length before we decrement */
3409 d += two_byte_count;
3410 SvCUR_set(sv, d - s);
3414 /* Having decremented d, it points to the position to put the
3415 * very last byte of the expanded string. Go backwards through
3416 * the string, copying and expanding as we go, stopping when we
3417 * get to the part that is invariant the rest of the way down */
3421 const U8 ch = NATIVE8_TO_UNI(*e--);
3422 if (UNI_IS_INVARIANT(ch)) {
3423 *d-- = UNI_TO_NATIVE(ch);
3425 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3426 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3433 /* Mark as UTF-8 even if no variant - saves scanning loop */
3439 =for apidoc sv_utf8_downgrade
3441 Attempts to convert the PV of an SV from characters to bytes.
3442 If the PV contains a character that cannot fit
3443 in a byte, this conversion will fail;
3444 in this case, either returns false or, if C<fail_ok> is not
3447 This is not as a general purpose Unicode to byte encoding interface:
3448 use the Encode extension for that.
3454 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3458 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3460 if (SvPOKp(sv) && SvUTF8(sv)) {
3466 sv_force_normal_flags(sv, 0);
3468 s = (U8 *) SvPV(sv, len);
3469 if (!utf8_to_bytes(s, &len)) {
3474 Perl_croak(aTHX_ "Wide character in %s",
3477 Perl_croak(aTHX_ "Wide character");
3488 =for apidoc sv_utf8_encode
3490 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3491 flag off so that it looks like octets again.
3497 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3499 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3502 sv_force_normal_flags(sv, 0);
3504 if (SvREADONLY(sv)) {
3505 Perl_croak(aTHX_ "%s", PL_no_modify);
3507 (void) sv_utf8_upgrade(sv);
3512 =for apidoc sv_utf8_decode
3514 If the PV of the SV is an octet sequence in UTF-8
3515 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3516 so that it looks like a character. If the PV contains only single-byte
3517 characters, the C<SvUTF8> flag stays being off.
3518 Scans PV for validity and returns false if the PV is invalid UTF-8.
3524 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3526 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3532 /* The octets may have got themselves encoded - get them back as
3535 if (!sv_utf8_downgrade(sv, TRUE))
3538 /* it is actually just a matter of turning the utf8 flag on, but
3539 * we want to make sure everything inside is valid utf8 first.
3541 c = (const U8 *) SvPVX_const(sv);
3542 if (!is_utf8_string(c, SvCUR(sv)+1))
3544 e = (const U8 *) SvEND(sv);
3547 if (!UTF8_IS_INVARIANT(ch)) {
3557 =for apidoc sv_setsv
3559 Copies the contents of the source SV C<ssv> into the destination SV
3560 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3561 function if the source SV needs to be reused. Does not handle 'set' magic.
3562 Loosely speaking, it performs a copy-by-value, obliterating any previous
3563 content of the destination.
3565 You probably want to use one of the assortment of wrappers, such as
3566 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3567 C<SvSetMagicSV_nosteal>.
3569 =for apidoc sv_setsv_flags
3571 Copies the contents of the source SV C<ssv> into the destination SV
3572 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3573 function if the source SV needs to be reused. Does not handle 'set' magic.
3574 Loosely speaking, it performs a copy-by-value, obliterating any previous
3575 content of the destination.
3576 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3577 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3578 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3579 and C<sv_setsv_nomg> are implemented in terms of this function.
3581 You probably want to use one of the assortment of wrappers, such as
3582 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3583 C<SvSetMagicSV_nosteal>.
3585 This is the primary function for copying scalars, and most other
3586 copy-ish functions and macros use this underneath.
3592 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3594 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3596 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3598 if (dtype != SVt_PVGV) {
3599 const char * const name = GvNAME(sstr);
3600 const STRLEN len = GvNAMELEN(sstr);
3602 if (dtype >= SVt_PV) {
3608 SvUPGRADE(dstr, SVt_PVGV);
3609 (void)SvOK_off(dstr);
3610 /* FIXME - why are we doing this, then turning it off and on again
3612 isGV_with_GP_on(dstr);
3614 GvSTASH(dstr) = GvSTASH(sstr);
3616 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3617 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3618 SvFAKE_on(dstr); /* can coerce to non-glob */
3621 if(GvGP(MUTABLE_GV(sstr))) {
3622 /* If source has method cache entry, clear it */
3624 SvREFCNT_dec(GvCV(sstr));
3628 /* If source has a real method, then a method is
3630 else if(GvCV((const GV *)sstr)) {
3635 /* If dest already had a real method, that's a change as well */
3636 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3640 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3643 gp_free(MUTABLE_GV(dstr));
3644 isGV_with_GP_off(dstr);
3645 (void)SvOK_off(dstr);
3646 isGV_with_GP_on(dstr);
3647 GvINTRO_off(dstr); /* one-shot flag */
3648 GvGP(dstr) = gp_ref(GvGP(sstr));
3649 if (SvTAINTED(sstr))
3651 if (GvIMPORTED(dstr) != GVf_IMPORTED
3652 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3654 GvIMPORTED_on(dstr);
3657 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3658 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3663 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3665 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3667 const int intro = GvINTRO(dstr);
3670 const U32 stype = SvTYPE(sref);
3672 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3675 GvINTRO_off(dstr); /* one-shot flag */
3676 GvLINE(dstr) = CopLINE(PL_curcop);
3677 GvEGV(dstr) = MUTABLE_GV(dstr);
3682 location = (SV **) &GvCV(dstr);
3683 import_flag = GVf_IMPORTED_CV;
3686 location = (SV **) &GvHV(dstr);
3687 import_flag = GVf_IMPORTED_HV;
3690 location = (SV **) &GvAV(dstr);
3691 import_flag = GVf_IMPORTED_AV;
3694 location = (SV **) &GvIOp(dstr);
3697 location = (SV **) &GvFORM(dstr);
3700 location = &GvSV(dstr);
3701 import_flag = GVf_IMPORTED_SV;
3704 if (stype == SVt_PVCV) {
3705 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3706 if (GvCVGEN(dstr)) {
3707 SvREFCNT_dec(GvCV(dstr));
3709 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3712 SAVEGENERICSV(*location);
3716 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3717 CV* const cv = MUTABLE_CV(*location);
3719 if (!GvCVGEN((const GV *)dstr) &&
3720 (CvROOT(cv) || CvXSUB(cv)))
3722 /* Redefining a sub - warning is mandatory if
3723 it was a const and its value changed. */
3724 if (CvCONST(cv) && CvCONST((const CV *)sref)
3726 == cv_const_sv((const CV *)sref)) {
3728 /* They are 2 constant subroutines generated from
3729 the same constant. This probably means that
3730 they are really the "same" proxy subroutine
3731 instantiated in 2 places. Most likely this is
3732 when a constant is exported twice. Don't warn.
3735 else if (ckWARN(WARN_REDEFINE)
3737 && (!CvCONST((const CV *)sref)
3738 || sv_cmp(cv_const_sv(cv),
3739 cv_const_sv((const CV *)
3741 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3744 ? "Constant subroutine %s::%s redefined"
3745 : "Subroutine %s::%s redefined"),
3746 HvNAME_get(GvSTASH((const GV *)dstr)),
3747 GvENAME(MUTABLE_GV(dstr)));
3751 cv_ckproto_len(cv, (const GV *)dstr,
3752 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3753 SvPOK(sref) ? SvCUR(sref) : 0);
3755 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3756 GvASSUMECV_on(dstr);
3757 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3760 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3761 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3762 GvFLAGS(dstr) |= import_flag;
3764 if (stype == SVt_PVAV && strEQ(GvNAME((GV*)dstr), "ISA")) {
3765 sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3766 mro_isa_changed_in(GvSTASH(dstr));
3771 if (SvTAINTED(sstr))
3777 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3780 register U32 sflags;
3782 register svtype stype;
3784 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3789 if (SvIS_FREED(dstr)) {
3790 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3791 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3793 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3795 sstr = &PL_sv_undef;
3796 if (SvIS_FREED(sstr)) {
3797 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3798 (void*)sstr, (void*)dstr);
3800 stype = SvTYPE(sstr);
3801 dtype = SvTYPE(dstr);
3803 (void)SvAMAGIC_off(dstr);
3806 /* need to nuke the magic */
3810 /* There's a lot of redundancy below but we're going for speed here */
3815 if (dtype != SVt_PVGV) {
3816 (void)SvOK_off(dstr);
3824 sv_upgrade(dstr, SVt_IV);
3828 sv_upgrade(dstr, SVt_PVIV);
3831 goto end_of_first_switch;
3833 (void)SvIOK_only(dstr);
3834 SvIV_set(dstr, SvIVX(sstr));
3837 /* SvTAINTED can only be true if the SV has taint magic, which in
3838 turn means that the SV type is PVMG (or greater). This is the
3839 case statement for SVt_IV, so this cannot be true (whatever gcov
3841 assert(!SvTAINTED(sstr));
3846 if (dtype < SVt_PV && dtype != SVt_IV)
3847 sv_upgrade(dstr, SVt_IV);
3855 sv_upgrade(dstr, SVt_NV);
3859 sv_upgrade(dstr, SVt_PVNV);
3862 goto end_of_first_switch;
3864 SvNV_set(dstr, SvNVX(sstr));
3865 (void)SvNOK_only(dstr);
3866 /* SvTAINTED can only be true if the SV has taint magic, which in
3867 turn means that the SV type is PVMG (or greater). This is the
3868 case statement for SVt_NV, so this cannot be true (whatever gcov
3870 assert(!SvTAINTED(sstr));
3876 #ifdef PERL_OLD_COPY_ON_WRITE
3877 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3878 if (dtype < SVt_PVIV)
3879 sv_upgrade(dstr, SVt_PVIV);
3886 sv_upgrade(dstr, SVt_PV);
3889 if (dtype < SVt_PVIV)
3890 sv_upgrade(dstr, SVt_PVIV);
3893 if (dtype < SVt_PVNV)
3894 sv_upgrade(dstr, SVt_PVNV);
3898 const char * const type = sv_reftype(sstr,0);
3900 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
3902 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3907 if (dtype < SVt_REGEXP)
3908 sv_upgrade(dstr, SVt_REGEXP);
3911 /* case SVt_BIND: */
3914 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3915 glob_assign_glob(dstr, sstr, dtype);
3918 /* SvVALID means that this PVGV is playing at being an FBM. */
3922 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3924 if (SvTYPE(sstr) != stype) {
3925 stype = SvTYPE(sstr);
3926 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3927 glob_assign_glob(dstr, sstr, dtype);
3932 if (stype == SVt_PVLV)
3933 SvUPGRADE(dstr, SVt_PVNV);
3935 SvUPGRADE(dstr, (svtype)stype);
3937 end_of_first_switch:
3939 /* dstr may have been upgraded. */
3940 dtype = SvTYPE(dstr);
3941 sflags = SvFLAGS(sstr);
3943 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3944 /* Assigning to a subroutine sets the prototype. */
3947 const char *const ptr = SvPV_const(sstr, len);
3949 SvGROW(dstr, len + 1);
3950 Copy(ptr, SvPVX(dstr), len + 1, char);
3951 SvCUR_set(dstr, len);
3953 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3957 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3958 const char * const type = sv_reftype(dstr,0);
3960 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
3962 Perl_croak(aTHX_ "Cannot copy to %s", type);
3963 } else if (sflags & SVf_ROK) {
3964 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3965 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3968 if (GvIMPORTED(dstr) != GVf_IMPORTED
3969 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3971 GvIMPORTED_on(dstr);
3976 glob_assign_glob(dstr, sstr, dtype);
3980 if (dtype >= SVt_PV) {
3981 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3982 glob_assign_ref(dstr, sstr);
3985 if (SvPVX_const(dstr)) {
3991 (void)SvOK_off(dstr);
3992 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3993 SvFLAGS(dstr) |= sflags & SVf_ROK;
3994 assert(!(sflags & SVp_NOK));
3995 assert(!(sflags & SVp_IOK));
3996 assert(!(sflags & SVf_NOK));
3997 assert(!(sflags & SVf_IOK));
3999 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4000 if (!(sflags & SVf_OK)) {
4001 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4002 "Undefined value assigned to typeglob");
4005 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4006 if (dstr != (const SV *)gv) {
4008 gp_free(MUTABLE_GV(dstr));
4009 GvGP(dstr) = gp_ref(GvGP(gv));
4013 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4014 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4016 else if (sflags & SVp_POK) {
4020 * Check to see if we can just swipe the string. If so, it's a
4021 * possible small lose on short strings, but a big win on long ones.
4022 * It might even be a win on short strings if SvPVX_const(dstr)
4023 * has to be allocated and SvPVX_const(sstr) has to be freed.
4024 * Likewise if we can set up COW rather than doing an actual copy, we
4025 * drop to the else clause, as the swipe code and the COW setup code
4026 * have much in common.
4029 /* Whichever path we take through the next code, we want this true,
4030 and doing it now facilitates the COW check. */
4031 (void)SvPOK_only(dstr);
4034 /* If we're already COW then this clause is not true, and if COW
4035 is allowed then we drop down to the else and make dest COW
4036 with us. If caller hasn't said that we're allowed to COW
4037 shared hash keys then we don't do the COW setup, even if the
4038 source scalar is a shared hash key scalar. */
4039 (((flags & SV_COW_SHARED_HASH_KEYS)
4040 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4041 : 1 /* If making a COW copy is forbidden then the behaviour we
4042 desire is as if the source SV isn't actually already
4043 COW, even if it is. So we act as if the source flags
4044 are not COW, rather than actually testing them. */
4046 #ifndef PERL_OLD_COPY_ON_WRITE
4047 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4048 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4049 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4050 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4051 but in turn, it's somewhat dead code, never expected to go
4052 live, but more kept as a placeholder on how to do it better
4053 in a newer implementation. */
4054 /* If we are COW and dstr is a suitable target then we drop down
4055 into the else and make dest a COW of us. */
4056 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4061 (sflags & SVs_TEMP) && /* slated for free anyway? */
4062 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4063 (!(flags & SV_NOSTEAL)) &&
4064 /* and we're allowed to steal temps */
4065 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4066 SvLEN(sstr)) /* and really is a string */
4067 #ifdef PERL_OLD_COPY_ON_WRITE
4068 && ((flags & SV_COW_SHARED_HASH_KEYS)
4069 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4070 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4071 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4075 /* Failed the swipe test, and it's not a shared hash key either.
4076 Have to copy the string. */
4077 STRLEN len = SvCUR(sstr);
4078 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4079 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4080 SvCUR_set(dstr, len);
4081 *SvEND(dstr) = '\0';
4083 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4085 /* Either it's a shared hash key, or it's suitable for
4086 copy-on-write or we can swipe the string. */
4088 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4092 #ifdef PERL_OLD_COPY_ON_WRITE
4094 if ((sflags & (SVf_FAKE | SVf_READONLY))
4095 != (SVf_FAKE | SVf_READONLY)) {
4096 SvREADONLY_on(sstr);
4098 /* Make the source SV into a loop of 1.
4099 (about to become 2) */
4100 SV_COW_NEXT_SV_SET(sstr, sstr);
4104 /* Initial code is common. */
4105 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4110 /* making another shared SV. */
4111 STRLEN cur = SvCUR(sstr);
4112 STRLEN len = SvLEN(sstr);
4113 #ifdef PERL_OLD_COPY_ON_WRITE
4115 assert (SvTYPE(dstr) >= SVt_PVIV);
4116 /* SvIsCOW_normal */
4117 /* splice us in between source and next-after-source. */
4118 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4119 SV_COW_NEXT_SV_SET(sstr, dstr);
4120 SvPV_set(dstr, SvPVX_mutable(sstr));
4124 /* SvIsCOW_shared_hash */
4125 DEBUG_C(PerlIO_printf(Perl_debug_log,
4126 "Copy on write: Sharing hash\n"));
4128 assert (SvTYPE(dstr) >= SVt_PV);
4130 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4132 SvLEN_set(dstr, len);
4133 SvCUR_set(dstr, cur);
4134 SvREADONLY_on(dstr);
4138 { /* Passes the swipe test. */
4139 SvPV_set(dstr, SvPVX_mutable(sstr));
4140 SvLEN_set(dstr, SvLEN(sstr));
4141 SvCUR_set(dstr, SvCUR(sstr));
4144 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4145 SvPV_set(sstr, NULL);
4151 if (sflags & SVp_NOK) {
4152 SvNV_set(dstr, SvNVX(sstr));
4154 if (sflags & SVp_IOK) {
4155 SvIV_set(dstr, SvIVX(sstr));
4156 /* Must do this otherwise some other overloaded use of 0x80000000
4157 gets confused. I guess SVpbm_VALID */
4158 if (sflags & SVf_IVisUV)
4161 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4163 const MAGIC * const smg = SvVSTRING_mg(sstr);
4165 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4166 smg->mg_ptr, smg->mg_len);
4167 SvRMAGICAL_on(dstr);
4171 else if (sflags & (SVp_IOK|SVp_NOK)) {
4172 (void)SvOK_off(dstr);
4173 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4174 if (sflags & SVp_IOK) {
4175 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4176 SvIV_set(dstr, SvIVX(sstr));
4178 if (sflags & SVp_NOK) {
4179 SvNV_set(dstr, SvNVX(sstr));
4183 if (isGV_with_GP(sstr)) {
4184 /* This stringification rule for globs is spread in 3 places.
4185 This feels bad. FIXME. */
4186 const U32 wasfake = sflags & SVf_FAKE;
4188 /* FAKE globs can get coerced, so need to turn this off
4189 temporarily if it is on. */
4191 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4192 SvFLAGS(sstr) |= wasfake;
4195 (void)SvOK_off(dstr);
4197 if (SvTAINTED(sstr))
4202 =for apidoc sv_setsv_mg
4204 Like C<sv_setsv>, but also handles 'set' magic.
4210 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4212 PERL_ARGS_ASSERT_SV_SETSV_MG;
4214 sv_setsv(dstr,sstr);
4218 #ifdef PERL_OLD_COPY_ON_WRITE
4220 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4222 STRLEN cur = SvCUR(sstr);
4223 STRLEN len = SvLEN(sstr);
4224 register char *new_pv;
4226 PERL_ARGS_ASSERT_SV_SETSV_COW;
4229 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4230 (void*)sstr, (void*)dstr);
4237 if (SvTHINKFIRST(dstr))
4238 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4239 else if (SvPVX_const(dstr))
4240 Safefree(SvPVX_const(dstr));
4244 SvUPGRADE(dstr, SVt_PVIV);
4246 assert (SvPOK(sstr));
4247 assert (SvPOKp(sstr));
4248 assert (!SvIOK(sstr));
4249 assert (!SvIOKp(sstr));
4250 assert (!SvNOK(sstr));
4251 assert (!SvNOKp(sstr));
4253 if (SvIsCOW(sstr)) {
4255 if (SvLEN(sstr) == 0) {
4256 /* source is a COW shared hash key. */
4257 DEBUG_C(PerlIO_printf(Perl_debug_log,
4258 "Fast copy on write: Sharing hash\n"));
4259 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4262 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4264 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4265 SvUPGRADE(sstr, SVt_PVIV);
4266 SvREADONLY_on(sstr);
4268 DEBUG_C(PerlIO_printf(Perl_debug_log,
4269 "Fast copy on write: Converting sstr to COW\n"));
4270 SV_COW_NEXT_SV_SET(dstr, sstr);
4272 SV_COW_NEXT_SV_SET(sstr, dstr);
4273 new_pv = SvPVX_mutable(sstr);
4276 SvPV_set(dstr, new_pv);
4277 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4280 SvLEN_set(dstr, len);
4281 SvCUR_set(dstr, cur);
4290 =for apidoc sv_setpvn
4292 Copies a string into an SV. The C<len> parameter indicates the number of
4293 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4294 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4300 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4303 register char *dptr;
4305 PERL_ARGS_ASSERT_SV_SETPVN;
4307 SV_CHECK_THINKFIRST_COW_DROP(sv);
4313 /* len is STRLEN which is unsigned, need to copy to signed */
4316 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4318 SvUPGRADE(sv, SVt_PV);
4320 dptr = SvGROW(sv, len + 1);
4321 Move(ptr,dptr,len,char);
4324 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4329 =for apidoc sv_setpvn_mg
4331 Like C<sv_setpvn>, but also handles 'set' magic.
4337 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4339 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4341 sv_setpvn(sv,ptr,len);
4346 =for apidoc sv_setpv
4348 Copies a string into an SV. The string must be null-terminated. Does not
4349 handle 'set' magic. See C<sv_setpv_mg>.
4355 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4358 register STRLEN len;
4360 PERL_ARGS_ASSERT_SV_SETPV;
4362 SV_CHECK_THINKFIRST_COW_DROP(sv);
4368 SvUPGRADE(sv, SVt_PV);
4370 SvGROW(sv, len + 1);
4371 Move(ptr,SvPVX(sv),len+1,char);
4373 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4378 =for apidoc sv_setpv_mg
4380 Like C<sv_setpv>, but also handles 'set' magic.
4386 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4388 PERL_ARGS_ASSERT_SV_SETPV_MG;
4395 =for apidoc sv_usepvn_flags
4397 Tells an SV to use C<ptr> to find its string value. Normally the
4398 string is stored inside the SV but sv_usepvn allows the SV to use an
4399 outside string. The C<ptr> should point to memory that was allocated
4400 by C<malloc>. The string length, C<len>, must be supplied. By default
4401 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4402 so that pointer should not be freed or used by the programmer after
4403 giving it to sv_usepvn, and neither should any pointers from "behind"
4404 that pointer (e.g. ptr + 1) be used.
4406 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4407 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4408 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4409 C<len>, and already meets the requirements for storing in C<SvPVX>)
4415 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4420 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4422 SV_CHECK_THINKFIRST_COW_DROP(sv);
4423 SvUPGRADE(sv, SVt_PV);
4426 if (flags & SV_SMAGIC)
4430 if (SvPVX_const(sv))
4434 if (flags & SV_HAS_TRAILING_NUL)
4435 assert(ptr[len] == '\0');
4438 allocate = (flags & SV_HAS_TRAILING_NUL)
4440 #ifdef Perl_safesysmalloc_size
4443 PERL_STRLEN_ROUNDUP(len + 1);
4445 if (flags & SV_HAS_TRAILING_NUL) {
4446 /* It's long enough - do nothing.
4447 Specfically Perl_newCONSTSUB is relying on this. */
4450 /* Force a move to shake out bugs in callers. */
4451 char *new_ptr = (char*)safemalloc(allocate);
4452 Copy(ptr, new_ptr, len, char);
4453 PoisonFree(ptr,len,char);
4457 ptr = (char*) saferealloc (ptr, allocate);
4460 #ifdef Perl_safesysmalloc_size
4461 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4463 SvLEN_set(sv, allocate);
4467 if (!(flags & SV_HAS_TRAILING_NUL)) {
4470 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4472 if (flags & SV_SMAGIC)
4476 #ifdef PERL_OLD_COPY_ON_WRITE
4477 /* Need to do this *after* making the SV normal, as we need the buffer
4478 pointer to remain valid until after we've copied it. If we let go too early,
4479 another thread could invalidate it by unsharing last of the same hash key
4480 (which it can do by means other than releasing copy-on-write Svs)
4481 or by changing the other copy-on-write SVs in the loop. */
4483 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4485 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4487 { /* this SV was SvIsCOW_normal(sv) */
4488 /* we need to find the SV pointing to us. */
4489 SV *current = SV_COW_NEXT_SV(after);
4491 if (current == sv) {
4492 /* The SV we point to points back to us (there were only two of us
4494 Hence other SV is no longer copy on write either. */
4496 SvREADONLY_off(after);
4498 /* We need to follow the pointers around the loop. */
4500 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4503 /* don't loop forever if the structure is bust, and we have
4504 a pointer into a closed loop. */
4505 assert (current != after);
4506 assert (SvPVX_const(current) == pvx);
4508 /* Make the SV before us point to the SV after us. */
4509 SV_COW_NEXT_SV_SET(current, after);
4515 =for apidoc sv_force_normal_flags
4517 Undo various types of fakery on an SV: if the PV is a shared string, make
4518 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4519 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4520 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4521 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4522 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4523 set to some other value.) In addition, the C<flags> parameter gets passed to
4524 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4525 with flags set to 0.
4531 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4535 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4537 #ifdef PERL_OLD_COPY_ON_WRITE
4538 if (SvREADONLY(sv)) {
4540 const char * const pvx = SvPVX_const(sv);
4541 const STRLEN len = SvLEN(sv);
4542 const STRLEN cur = SvCUR(sv);
4543 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4544 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4545 we'll fail an assertion. */
4546 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4549 PerlIO_printf(Perl_debug_log,
4550 "Copy on write: Force normal %ld\n",
4556 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4559 if (flags & SV_COW_DROP_PV) {
4560 /* OK, so we don't need to copy our buffer. */
4563 SvGROW(sv, cur + 1);
4564 Move(pvx,SvPVX(sv),cur,char);
4569 sv_release_COW(sv, pvx, next);
4571 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4577 else if (IN_PERL_RUNTIME)
4578 Perl_croak(aTHX_ "%s", PL_no_modify);
4581 if (SvREADONLY(sv)) {
4583 const char * const pvx = SvPVX_const(sv);
4584 const STRLEN len = SvCUR(sv);
4589 SvGROW(sv, len + 1);
4590 Move(pvx,SvPVX(sv),len,char);
4592 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4594 else if (IN_PERL_RUNTIME)
4595 Perl_croak(aTHX_ "%s", PL_no_modify);
4599 sv_unref_flags(sv, flags);
4600 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4602 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4603 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analagous
4604 to sv_unglob. We only need it here, so inline it. */
4605 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4606 SV *const temp = newSV_type(new_type);
4607 void *const temp_p = SvANY(sv);
4609 if (new_type == SVt_PVMG) {
4610 SvMAGIC_set(temp, SvMAGIC(sv));
4611 SvMAGIC_set(sv, NULL);
4612 SvSTASH_set(temp, SvSTASH(sv));
4613 SvSTASH_set(sv, NULL);
4615 SvCUR_set(temp, SvCUR(sv));
4616 /* Remember that SvPVX is in the head, not the body. */
4618 SvLEN_set(temp, SvLEN(sv));
4619 /* This signals "buffer is owned by someone else" in sv_clear,
4620 which is the least effort way to stop it freeing the buffer.
4622 SvLEN_set(sv, SvLEN(sv)+1);
4624 /* Their buffer is already owned by someone else. */
4625 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4626 SvLEN_set(temp, SvCUR(sv)+1);
4629 /* Now swap the rest of the bodies. */
4631 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4632 SvFLAGS(sv) |= new_type;
4633 SvANY(sv) = SvANY(temp);
4635 SvFLAGS(temp) &= ~(SVTYPEMASK);
4636 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4637 SvANY(temp) = temp_p;
4646 Efficient removal of characters from the beginning of the string buffer.
4647 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4648 the string buffer. The C<ptr> becomes the first character of the adjusted
4649 string. Uses the "OOK hack".
4650 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4651 refer to the same chunk of data.
4657 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4663 const U8 *real_start;
4667 PERL_ARGS_ASSERT_SV_CHOP;
4669 if (!ptr || !SvPOKp(sv))
4671 delta = ptr - SvPVX_const(sv);
4673 /* Nothing to do. */
4676 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4677 nothing uses the value of ptr any more. */
4678 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4679 if (ptr <= SvPVX_const(sv))
4680 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4681 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4682 SV_CHECK_THINKFIRST(sv);
4683 if (delta > max_delta)
4684 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4685 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4686 SvPVX_const(sv) + max_delta);
4689 if (!SvLEN(sv)) { /* make copy of shared string */
4690 const char *pvx = SvPVX_const(sv);
4691 const STRLEN len = SvCUR(sv);
4692 SvGROW(sv, len + 1);
4693 Move(pvx,SvPVX(sv),len,char);
4696 SvFLAGS(sv) |= SVf_OOK;
4699 SvOOK_offset(sv, old_delta);
4701 SvLEN_set(sv, SvLEN(sv) - delta);
4702 SvCUR_set(sv, SvCUR(sv) - delta);
4703 SvPV_set(sv, SvPVX(sv) + delta);
4705 p = (U8 *)SvPVX_const(sv);
4710 real_start = p - delta;
4714 if (delta < 0x100) {
4718 p -= sizeof(STRLEN);
4719 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4723 /* Fill the preceding buffer with sentinals to verify that no-one is
4725 while (p > real_start) {
4733 =for apidoc sv_catpvn
4735 Concatenates the string onto the end of the string which is in the SV. The
4736 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4737 status set, then the bytes appended should be valid UTF-8.
4738 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4740 =for apidoc sv_catpvn_flags
4742 Concatenates the string onto the end of the string which is in the SV. The
4743 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4744 status set, then the bytes appended should be valid UTF-8.
4745 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4746 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4747 in terms of this function.
4753 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4757 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4759 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4761 SvGROW(dsv, dlen + slen + 1);
4763 sstr = SvPVX_const(dsv);
4764 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4765 SvCUR_set(dsv, SvCUR(dsv) + slen);
4767 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4769 if (flags & SV_SMAGIC)
4774 =for apidoc sv_catsv
4776 Concatenates the string from SV C<ssv> onto the end of the string in
4777 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4778 not 'set' magic. See C<sv_catsv_mg>.
4780 =for apidoc sv_catsv_flags
4782 Concatenates the string from SV C<ssv> onto the end of the string in
4783 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4784 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4785 and C<sv_catsv_nomg> are implemented in terms of this function.
4790 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4794 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4798 const char *spv = SvPV_const(ssv, slen);
4800 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4801 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4802 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4803 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4804 dsv->sv_flags doesn't have that bit set.
4805 Andy Dougherty 12 Oct 2001
4807 const I32 sutf8 = DO_UTF8(ssv);
4810 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4812 dutf8 = DO_UTF8(dsv);
4814 if (dutf8 != sutf8) {
4816 /* Not modifying source SV, so taking a temporary copy. */
4817 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4819 sv_utf8_upgrade(csv);
4820 spv = SvPV_const(csv, slen);
4823 /* Leave enough space for the cat that's about to happen */
4824 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4826 sv_catpvn_nomg(dsv, spv, slen);
4829 if (flags & SV_SMAGIC)
4834 =for apidoc sv_catpv
4836 Concatenates the string onto the end of the string which is in the SV.
4837 If the SV has the UTF-8 status set, then the bytes appended should be
4838 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4843 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4846 register STRLEN len;
4850 PERL_ARGS_ASSERT_SV_CATPV;
4854 junk = SvPV_force(sv, tlen);
4856 SvGROW(sv, tlen + len + 1);
4858 ptr = SvPVX_const(sv);
4859 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4860 SvCUR_set(sv, SvCUR(sv) + len);
4861 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4866 =for apidoc sv_catpv_mg
4868 Like C<sv_catpv>, but also handles 'set' magic.
4874 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4876 PERL_ARGS_ASSERT_SV_CATPV_MG;
4885 Creates a new SV. A non-zero C<len> parameter indicates the number of
4886 bytes of preallocated string space the SV should have. An extra byte for a
4887 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4888 space is allocated.) The reference count for the new SV is set to 1.
4890 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4891 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4892 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4893 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4894 modules supporting older perls.
4900 Perl_newSV(pTHX_ const STRLEN len)
4907 sv_upgrade(sv, SVt_PV);
4908 SvGROW(sv, len + 1);
4913 =for apidoc sv_magicext
4915 Adds magic to an SV, upgrading it if necessary. Applies the
4916 supplied vtable and returns a pointer to the magic added.
4918 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4919 In particular, you can add magic to SvREADONLY SVs, and add more than
4920 one instance of the same 'how'.
4922 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4923 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4924 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4925 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4927 (This is now used as a subroutine by C<sv_magic>.)
4932 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4933 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4938 PERL_ARGS_ASSERT_SV_MAGICEXT;
4940 SvUPGRADE(sv, SVt_PVMG);
4941 Newxz(mg, 1, MAGIC);
4942 mg->mg_moremagic = SvMAGIC(sv);
4943 SvMAGIC_set(sv, mg);
4945 /* Sometimes a magic contains a reference loop, where the sv and
4946 object refer to each other. To prevent a reference loop that
4947 would prevent such objects being freed, we look for such loops
4948 and if we find one we avoid incrementing the object refcount.
4950 Note we cannot do this to avoid self-tie loops as intervening RV must
4951 have its REFCNT incremented to keep it in existence.
4954 if (!obj || obj == sv ||
4955 how == PERL_MAGIC_arylen ||
4956 how == PERL_MAGIC_symtab ||
4957 (SvTYPE(obj) == SVt_PVGV &&
4958 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4959 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4960 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4965 mg->mg_obj = SvREFCNT_inc_simple(obj);
4966 mg->mg_flags |= MGf_REFCOUNTED;
4969 /* Normal self-ties simply pass a null object, and instead of
4970 using mg_obj directly, use the SvTIED_obj macro to produce a
4971 new RV as needed. For glob "self-ties", we are tieing the PVIO
4972 with an RV obj pointing to the glob containing the PVIO. In
4973 this case, to avoid a reference loop, we need to weaken the
4977 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4978 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4984 mg->mg_len = namlen;
4987 mg->mg_ptr = savepvn(name, namlen);
4988 else if (namlen == HEf_SVKEY) {
4989 /* Yes, this is casting away const. This is only for the case of
4990 HEf_SVKEY. I think we need to document this abberation of the
4991 constness of the API, rather than making name non-const, as
4992 that change propagating outwards a long way. */
4993 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4995 mg->mg_ptr = (char *) name;
4997 mg->mg_virtual = (MGVTBL *) vtable;
5001 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5006 =for apidoc sv_magic
5008 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5009 then adds a new magic item of type C<how> to the head of the magic list.
5011 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5012 handling of the C<name> and C<namlen> arguments.
5014 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5015 to add more than one instance of the same 'how'.
5021 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5022 const char *const name, const I32 namlen)
5025 const MGVTBL *vtable;
5028 PERL_ARGS_ASSERT_SV_MAGIC;
5030 #ifdef PERL_OLD_COPY_ON_WRITE
5032 sv_force_normal_flags(sv, 0);
5034 if (SvREADONLY(sv)) {
5036 /* its okay to attach magic to shared strings; the subsequent
5037 * upgrade to PVMG will unshare the string */
5038 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5041 && how != PERL_MAGIC_regex_global
5042 && how != PERL_MAGIC_bm
5043 && how != PERL_MAGIC_fm
5044 && how != PERL_MAGIC_sv
5045 && how != PERL_MAGIC_backref
5048 Perl_croak(aTHX_ "%s", PL_no_modify);
5051 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5052 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5053 /* sv_magic() refuses to add a magic of the same 'how' as an
5056 if (how == PERL_MAGIC_taint) {
5058 /* Any scalar which already had taint magic on which someone
5059 (erroneously?) did SvIOK_on() or similar will now be
5060 incorrectly sporting public "OK" flags. */
5061 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5069 vtable = &PL_vtbl_sv;
5071 case PERL_MAGIC_overload:
5072 vtable = &PL_vtbl_amagic;
5074 case PERL_MAGIC_overload_elem:
5075 vtable = &PL_vtbl_amagicelem;
5077 case PERL_MAGIC_overload_table:
5078 vtable = &PL_vtbl_ovrld;
5081 vtable = &PL_vtbl_bm;
5083 case PERL_MAGIC_regdata:
5084 vtable = &PL_vtbl_regdata;
5086 case PERL_MAGIC_regdatum:
5087 vtable = &PL_vtbl_regdatum;
5089 case PERL_MAGIC_env:
5090 vtable = &PL_vtbl_env;
5093 vtable = &PL_vtbl_fm;
5095 case PERL_MAGIC_envelem:
5096 vtable = &PL_vtbl_envelem;
5098 case PERL_MAGIC_regex_global:
5099 vtable = &PL_vtbl_mglob;
5101 case PERL_MAGIC_isa:
5102 vtable = &PL_vtbl_isa;
5104 case PERL_MAGIC_isaelem:
5105 vtable = &PL_vtbl_isaelem;
5107 case PERL_MAGIC_nkeys:
5108 vtable = &PL_vtbl_nkeys;
5110 case PERL_MAGIC_dbfile:
5113 case PERL_MAGIC_dbline:
5114 vtable = &PL_vtbl_dbline;
5116 #ifdef USE_LOCALE_COLLATE
5117 case PERL_MAGIC_collxfrm:
5118 vtable = &PL_vtbl_collxfrm;
5120 #endif /* USE_LOCALE_COLLATE */
5121 case PERL_MAGIC_tied:
5122 vtable = &PL_vtbl_pack;
5124 case PERL_MAGIC_tiedelem:
5125 case PERL_MAGIC_tiedscalar:
5126 vtable = &PL_vtbl_packelem;
5129 vtable = &PL_vtbl_regexp;
5131 case PERL_MAGIC_sig:
5132 vtable = &PL_vtbl_sig;
5134 case PERL_MAGIC_sigelem:
5135 vtable = &PL_vtbl_sigelem;
5137 case PERL_MAGIC_taint:
5138 vtable = &PL_vtbl_taint;
5140 case PERL_MAGIC_uvar:
5141 vtable = &PL_vtbl_uvar;
5143 case PERL_MAGIC_vec:
5144 vtable = &PL_vtbl_vec;
5146 case PERL_MAGIC_arylen_p:
5147 case PERL_MAGIC_rhash:
5148 case PERL_MAGIC_symtab:
5149 case PERL_MAGIC_vstring:
5152 case PERL_MAGIC_utf8:
5153 vtable = &PL_vtbl_utf8;
5155 case PERL_MAGIC_substr:
5156 vtable = &PL_vtbl_substr;
5158 case PERL_MAGIC_defelem:
5159 vtable = &PL_vtbl_defelem;
5161 case PERL_MAGIC_arylen:
5162 vtable = &PL_vtbl_arylen;
5164 case PERL_MAGIC_pos:
5165 vtable = &PL_vtbl_pos;
5167 case PERL_MAGIC_backref:
5168 vtable = &PL_vtbl_backref;
5170 case PERL_MAGIC_hintselem:
5171 vtable = &PL_vtbl_hintselem;
5173 case PERL_MAGIC_hints:
5174 vtable = &PL_vtbl_hints;
5176 case PERL_MAGIC_ext:
5177 /* Reserved for use by extensions not perl internals. */
5178 /* Useful for attaching extension internal data to perl vars. */
5179 /* Note that multiple extensions may clash if magical scalars */
5180 /* etc holding private data from one are passed to another. */
5184 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5187 /* Rest of work is done else where */
5188 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5191 case PERL_MAGIC_taint:
5194 case PERL_MAGIC_ext:
5195 case PERL_MAGIC_dbfile:
5202 =for apidoc sv_unmagic
5204 Removes all magic of type C<type> from an SV.
5210 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5215 PERL_ARGS_ASSERT_SV_UNMAGIC;
5217 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5219 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5220 for (mg = *mgp; mg; mg = *mgp) {
5221 if (mg->mg_type == type) {
5222 const MGVTBL* const vtbl = mg->mg_virtual;
5223 *mgp = mg->mg_moremagic;
5224 if (vtbl && vtbl->svt_free)
5225 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5226 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5228 Safefree(mg->mg_ptr);
5229 else if (mg->mg_len == HEf_SVKEY)
5230 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5231 else if (mg->mg_type == PERL_MAGIC_utf8)
5232 Safefree(mg->mg_ptr);
5234 if (mg->mg_flags & MGf_REFCOUNTED)
5235 SvREFCNT_dec(mg->mg_obj);
5239 mgp = &mg->mg_moremagic;
5242 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5243 mg_magical(sv); /* else fix the flags now */
5247 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5253 =for apidoc sv_rvweaken
5255 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5256 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5257 push a back-reference to this RV onto the array of backreferences
5258 associated with that magic. If the RV is magical, set magic will be
5259 called after the RV is cleared.
5265 Perl_sv_rvweaken(pTHX_ SV *const sv)
5269 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5271 if (!SvOK(sv)) /* let undefs pass */
5274 Perl_croak(aTHX_ "Can't weaken a nonreference");
5275 else if (SvWEAKREF(sv)) {
5276 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5280 Perl_sv_add_backref(aTHX_ tsv, sv);
5286 /* Give tsv backref magic if it hasn't already got it, then push a
5287 * back-reference to sv onto the array associated with the backref magic.
5290 /* A discussion about the backreferences array and its refcount:
5292 * The AV holding the backreferences is pointed to either as the mg_obj of
5293 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5294 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5295 * have the standard magic instead.) The array is created with a refcount
5296 * of 2. This means that if during global destruction the array gets
5297 * picked on first to have its refcount decremented by the random zapper,
5298 * it won't actually be freed, meaning it's still theere for when its
5299 * parent gets freed.
5300 * When the parent SV is freed, in the case of magic, the magic is freed,
5301 * Perl_magic_killbackrefs is called which decrements one refcount, then
5302 * mg_obj is freed which kills the second count.
5303 * In the vase of a HV being freed, one ref is removed by
5304 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5309 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5314 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5316 if (SvTYPE(tsv) == SVt_PVHV) {
5317 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5321 /* There is no AV in the offical place - try a fixup. */
5322 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5325 /* Aha. They've got it stowed in magic. Bring it back. */
5326 av = MUTABLE_AV(mg->mg_obj);
5327 /* Stop mg_free decreasing the refernce count. */
5329 /* Stop mg_free even calling the destructor, given that
5330 there's no AV to free up. */
5332 sv_unmagic(tsv, PERL_MAGIC_backref);
5336 SvREFCNT_inc_simple_void(av); /* see discussion above */
5341 const MAGIC *const mg
5342 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5344 av = MUTABLE_AV(mg->mg_obj);
5348 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5349 /* av now has a refcnt of 2; see discussion above */
5352 if (AvFILLp(av) >= AvMAX(av)) {
5353 av_extend(av, AvFILLp(av)+1);
5355 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5358 /* delete a back-reference to ourselves from the backref magic associated
5359 * with the SV we point to.
5363 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5370 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5372 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5373 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5374 /* We mustn't attempt to "fix up" the hash here by moving the
5375 backreference array back to the hv_aux structure, as that is stored
5376 in the main HvARRAY(), and hfreentries assumes that no-one
5377 reallocates HvARRAY() while it is running. */
5380 const MAGIC *const mg
5381 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5383 av = MUTABLE_AV(mg->mg_obj);
5387 Perl_croak(aTHX_ "panic: del_backref");
5389 assert(!SvIS_FREED(av));
5392 /* We shouldn't be in here more than once, but for paranoia reasons lets
5394 for (i = AvFILLp(av); i >= 0; i--) {
5396 const SSize_t fill = AvFILLp(av);
5398 /* We weren't the last entry.
5399 An unordered list has this property that you can take the
5400 last element off the end to fill the hole, and it's still
5401 an unordered list :-)
5406 AvFILLp(av) = fill - 1;
5412 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5414 SV **svp = AvARRAY(av);
5416 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5417 PERL_UNUSED_ARG(sv);
5419 assert(!svp || !SvIS_FREED(av));
5421 SV *const *const last = svp + AvFILLp(av);
5423 while (svp <= last) {
5425 SV *const referrer = *svp;
5426 if (SvWEAKREF(referrer)) {
5427 /* XXX Should we check that it hasn't changed? */
5428 SvRV_set(referrer, 0);
5430 SvWEAKREF_off(referrer);
5431 SvSETMAGIC(referrer);
5432 } else if (SvTYPE(referrer) == SVt_PVGV ||
5433 SvTYPE(referrer) == SVt_PVLV) {
5434 /* You lookin' at me? */
5435 assert(GvSTASH(referrer));
5436 assert(GvSTASH(referrer) == (const HV *)sv);
5437 GvSTASH(referrer) = 0;
5440 "panic: magic_killbackrefs (flags=%"UVxf")",
5441 (UV)SvFLAGS(referrer));
5449 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5454 =for apidoc sv_insert
5456 Inserts a string at the specified offset/length within the SV. Similar to
5457 the Perl substr() function. Handles get magic.
5459 =for apidoc sv_insert_flags
5461 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5467 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5472 register char *midend;
5473 register char *bigend;
5477 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5480 Perl_croak(aTHX_ "Can't modify non-existent substring");
5481 SvPV_force_flags(bigstr, curlen, flags);
5482 (void)SvPOK_only_UTF8(bigstr);
5483 if (offset + len > curlen) {
5484 SvGROW(bigstr, offset+len+1);
5485 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5486 SvCUR_set(bigstr, offset+len);
5490 i = littlelen - len;
5491 if (i > 0) { /* string might grow */
5492 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5493 mid = big + offset + len;
5494 midend = bigend = big + SvCUR(bigstr);
5497 while (midend > mid) /* shove everything down */
5498 *--bigend = *--midend;
5499 Move(little,big+offset,littlelen,char);
5500 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5505 Move(little,SvPVX(bigstr)+offset,len,char);
5510 big = SvPVX(bigstr);
5513 bigend = big + SvCUR(bigstr);
5515 if (midend > bigend)
5516 Perl_croak(aTHX_ "panic: sv_insert");
5518 if (mid - big > bigend - midend) { /* faster to shorten from end */
5520 Move(little, mid, littlelen,char);
5523 i = bigend - midend;
5525 Move(midend, mid, i,char);
5529 SvCUR_set(bigstr, mid - big);
5531 else if ((i = mid - big)) { /* faster from front */
5532 midend -= littlelen;
5534 Move(big, midend - i, i, char);
5535 sv_chop(bigstr,midend-i);
5537 Move(little, mid, littlelen,char);
5539 else if (littlelen) {
5540 midend -= littlelen;
5541 sv_chop(bigstr,midend);
5542 Move(little,midend,littlelen,char);
5545 sv_chop(bigstr,midend);
5551 =for apidoc sv_replace
5553 Make the first argument a copy of the second, then delete the original.
5554 The target SV physically takes over ownership of the body of the source SV
5555 and inherits its flags; however, the target keeps any magic it owns,
5556 and any magic in the source is discarded.
5557 Note that this is a rather specialist SV copying operation; most of the
5558 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5564 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5567 const U32 refcnt = SvREFCNT(sv);
5569 PERL_ARGS_ASSERT_SV_REPLACE;
5571 SV_CHECK_THINKFIRST_COW_DROP(sv);
5572 if (SvREFCNT(nsv) != 1) {
5573 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5574 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5576 if (SvMAGICAL(sv)) {
5580 sv_upgrade(nsv, SVt_PVMG);
5581 SvMAGIC_set(nsv, SvMAGIC(sv));
5582 SvFLAGS(nsv) |= SvMAGICAL(sv);
5584 SvMAGIC_set(sv, NULL);
5588 assert(!SvREFCNT(sv));
5589 #ifdef DEBUG_LEAKING_SCALARS
5590 sv->sv_flags = nsv->sv_flags;
5591 sv->sv_any = nsv->sv_any;
5592 sv->sv_refcnt = nsv->sv_refcnt;
5593 sv->sv_u = nsv->sv_u;
5595 StructCopy(nsv,sv,SV);
5597 if(SvTYPE(sv) == SVt_IV) {
5599 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5603 #ifdef PERL_OLD_COPY_ON_WRITE
5604 if (SvIsCOW_normal(nsv)) {
5605 /* We need to follow the pointers around the loop to make the
5606 previous SV point to sv, rather than nsv. */
5609 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5612 assert(SvPVX_const(current) == SvPVX_const(nsv));
5614 /* Make the SV before us point to the SV after us. */
5616 PerlIO_printf(Perl_debug_log, "previous is\n");
5618 PerlIO_printf(Perl_debug_log,
5619 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5620 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5622 SV_COW_NEXT_SV_SET(current, sv);
5625 SvREFCNT(sv) = refcnt;
5626 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5632 =for apidoc sv_clear
5634 Clear an SV: call any destructors, free up any memory used by the body,
5635 and free the body itself. The SV's head is I<not> freed, although
5636 its type is set to all 1's so that it won't inadvertently be assumed
5637 to be live during global destruction etc.
5638 This function should only be called when REFCNT is zero. Most of the time
5639 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5646 Perl_sv_clear(pTHX_ register SV *const sv)
5649 const U32 type = SvTYPE(sv);
5650 const struct body_details *const sv_type_details
5651 = bodies_by_type + type;
5654 PERL_ARGS_ASSERT_SV_CLEAR;
5655 assert(SvREFCNT(sv) == 0);
5656 assert(SvTYPE(sv) != SVTYPEMASK);
5658 if (type <= SVt_IV) {
5659 /* See the comment in sv.h about the collusion between this early
5660 return and the overloading of the NULL slots in the size table. */
5663 SvFLAGS(sv) &= SVf_BREAK;
5664 SvFLAGS(sv) |= SVTYPEMASK;
5669 if (PL_defstash && /* Still have a symbol table? */
5676 stash = SvSTASH(sv);
5677 destructor = StashHANDLER(stash,DESTROY);
5679 /* A constant subroutine can have no side effects, so
5680 don't bother calling it. */
5681 && !CvCONST(destructor)
5682 /* Don't bother calling an empty destructor */
5683 && (CvISXSUB(destructor)
5684 || (CvSTART(destructor)
5685 && (CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))))
5687 SV* const tmpref = newRV(sv);
5688 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5690 PUSHSTACKi(PERLSI_DESTROY);
5695 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5701 if(SvREFCNT(tmpref) < 2) {
5702 /* tmpref is not kept alive! */
5704 SvRV_set(tmpref, NULL);
5707 SvREFCNT_dec(tmpref);
5709 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5713 if (PL_in_clean_objs)
5714 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5716 /* DESTROY gave object new lease on life */
5722 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5723 SvOBJECT_off(sv); /* Curse the object. */
5724 if (type != SVt_PVIO)
5725 --PL_sv_objcount; /* XXX Might want something more general */
5728 if (type >= SVt_PVMG) {
5729 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5730 SvREFCNT_dec(SvOURSTASH(sv));
5731 } else if (SvMAGIC(sv))
5733 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5734 SvREFCNT_dec(SvSTASH(sv));
5737 /* case SVt_BIND: */
5740 IoIFP(sv) != PerlIO_stdin() &&
5741 IoIFP(sv) != PerlIO_stdout() &&
5742 IoIFP(sv) != PerlIO_stderr())
5744 io_close(MUTABLE_IO(sv), FALSE);
5746 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5747 PerlDir_close(IoDIRP(sv));
5748 IoDIRP(sv) = (DIR*)NULL;
5749 Safefree(IoTOP_NAME(sv));
5750 Safefree(IoFMT_NAME(sv));
5751 Safefree(IoBOTTOM_NAME(sv));
5754 /* FIXME for plugins */
5755 pregfree2((REGEXP*) sv);
5759 cv_undef(MUTABLE_CV(sv));
5762 if (PL_last_swash_hv == (const HV *)sv) {
5763 PL_last_swash_hv = NULL;
5765 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5766 hv_undef(MUTABLE_HV(sv));
5769 if (PL_comppad == MUTABLE_AV(sv)) {
5773 av_undef(MUTABLE_AV(sv));
5776 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5777 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5778 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5779 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5781 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5782 SvREFCNT_dec(LvTARG(sv));
5784 if (isGV_with_GP(sv)) {
5785 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5786 && HvNAME_get(stash))
5787 mro_method_changed_in(stash);
5788 gp_free(MUTABLE_GV(sv));
5790 unshare_hek(GvNAME_HEK(sv));
5791 /* If we're in a stash, we don't own a reference to it. However it does
5792 have a back reference to us, which needs to be cleared. */
5793 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5794 sv_del_backref(MUTABLE_SV(stash), sv);
5796 /* FIXME. There are probably more unreferenced pointers to SVs in the
5797 interpreter struct that we should check and tidy in a similar
5799 if ((const GV *)sv == PL_last_in_gv)
5800 PL_last_in_gv = NULL;
5806 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5809 SvOOK_offset(sv, offset);
5810 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5811 /* Don't even bother with turning off the OOK flag. */
5816 SV * const target = SvRV(sv);
5818 sv_del_backref(target, sv);
5820 SvREFCNT_dec(target);
5823 #ifdef PERL_OLD_COPY_ON_WRITE
5824 else if (SvPVX_const(sv)) {
5827 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5831 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5833 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5837 } else if (SvLEN(sv)) {
5838 Safefree(SvPVX_const(sv));
5842 else if (SvPVX_const(sv) && SvLEN(sv))
5843 Safefree(SvPVX_mutable(sv));
5844 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5845 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5854 SvFLAGS(sv) &= SVf_BREAK;
5855 SvFLAGS(sv) |= SVTYPEMASK;
5857 if (sv_type_details->arena) {
5858 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5859 &PL_body_roots[type]);
5861 else if (sv_type_details->body_size) {
5862 my_safefree(SvANY(sv));
5867 =for apidoc sv_newref
5869 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5876 Perl_sv_newref(pTHX_ SV *const sv)
5878 PERL_UNUSED_CONTEXT;
5887 Decrement an SV's reference count, and if it drops to zero, call
5888 C<sv_clear> to invoke destructors and free up any memory used by
5889 the body; finally, deallocate the SV's head itself.
5890 Normally called via a wrapper macro C<SvREFCNT_dec>.
5896 Perl_sv_free(pTHX_ SV *const sv)
5901 if (SvREFCNT(sv) == 0) {
5902 if (SvFLAGS(sv) & SVf_BREAK)
5903 /* this SV's refcnt has been artificially decremented to
5904 * trigger cleanup */
5906 if (PL_in_clean_all) /* All is fair */
5908 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5909 /* make sure SvREFCNT(sv)==0 happens very seldom */
5910 SvREFCNT(sv) = (~(U32)0)/2;
5913 if (ckWARN_d(WARN_INTERNAL)) {
5914 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5915 Perl_dump_sv_child(aTHX_ sv);
5917 #ifdef DEBUG_LEAKING_SCALARS
5920 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5921 if (PL_warnhook == PERL_WARNHOOK_FATAL
5922 || ckDEAD(packWARN(WARN_INTERNAL))) {
5923 /* Don't let Perl_warner cause us to escape our fate: */
5927 /* This may not return: */
5928 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5929 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5930 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5933 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5938 if (--(SvREFCNT(sv)) > 0)
5940 Perl_sv_free2(aTHX_ sv);
5944 Perl_sv_free2(pTHX_ SV *const sv)
5948 PERL_ARGS_ASSERT_SV_FREE2;
5952 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5953 "Attempt to free temp prematurely: SV 0x%"UVxf
5954 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5958 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5959 /* make sure SvREFCNT(sv)==0 happens very seldom */
5960 SvREFCNT(sv) = (~(U32)0)/2;
5971 Returns the length of the string in the SV. Handles magic and type
5972 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5978 Perl_sv_len(pTHX_ register SV *const sv)
5986 len = mg_length(sv);
5988 (void)SvPV_const(sv, len);
5993 =for apidoc sv_len_utf8
5995 Returns the number of characters in the string in an SV, counting wide
5996 UTF-8 bytes as a single character. Handles magic and type coercion.
6002 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6003 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6004 * (Note that the mg_len is not the length of the mg_ptr field.
6005 * This allows the cache to store the character length of the string without
6006 * needing to malloc() extra storage to attach to the mg_ptr.)
6011 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6017 return mg_length(sv);
6021 const U8 *s = (U8*)SvPV_const(sv, len);
6025 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6027 if (mg && mg->mg_len != -1) {
6029 if (PL_utf8cache < 0) {
6030 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6032 /* Need to turn the assertions off otherwise we may
6033 recurse infinitely while printing error messages.
6035 SAVEI8(PL_utf8cache);
6037 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6038 " real %"UVuf" for %"SVf,
6039 (UV) ulen, (UV) real, SVfARG(sv));
6044 ulen = Perl_utf8_length(aTHX_ s, s + len);
6045 if (!SvREADONLY(sv)) {
6046 if (!mg && (SvTYPE(sv) < SVt_PVMG ||
6047 !(mg = mg_find(sv, PERL_MAGIC_utf8)))) {
6048 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6049 &PL_vtbl_utf8, 0, 0);
6053 /* For now, treat "overflowed" as "still unknown".
6055 if (ulen != (STRLEN) mg->mg_len)
6061 return Perl_utf8_length(aTHX_ s, s + len);
6065 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6068 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6071 const U8 *s = start;
6073 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6075 while (s < send && uoffset--)
6078 /* This is the existing behaviour. Possibly it should be a croak, as
6079 it's actually a bounds error */
6085 /* Given the length of the string in both bytes and UTF-8 characters, decide
6086 whether to walk forwards or backwards to find the byte corresponding to
6087 the passed in UTF-8 offset. */
6089 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6090 const STRLEN uoffset, const STRLEN uend)
6092 STRLEN backw = uend - uoffset;
6094 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6096 if (uoffset < 2 * backw) {
6097 /* The assumption is that going forwards is twice the speed of going
6098 forward (that's where the 2 * backw comes from).
6099 (The real figure of course depends on the UTF-8 data.) */
6100 return sv_pos_u2b_forwards(start, send, uoffset);
6105 while (UTF8_IS_CONTINUATION(*send))
6108 return send - start;
6111 /* For the string representation of the given scalar, find the byte
6112 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6113 give another position in the string, *before* the sought offset, which
6114 (which is always true, as 0, 0 is a valid pair of positions), which should
6115 help reduce the amount of linear searching.
6116 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6117 will be used to reduce the amount of linear searching. The cache will be
6118 created if necessary, and the found value offered to it for update. */
6120 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6121 const U8 *const send, const STRLEN uoffset,
6122 STRLEN uoffset0, STRLEN boffset0)
6124 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6127 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6129 assert (uoffset >= uoffset0);
6133 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6134 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6135 if ((*mgp)->mg_ptr) {
6136 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6137 if (cache[0] == uoffset) {
6138 /* An exact match. */
6141 if (cache[2] == uoffset) {
6142 /* An exact match. */
6146 if (cache[0] < uoffset) {
6147 /* The cache already knows part of the way. */
6148 if (cache[0] > uoffset0) {
6149 /* The cache knows more than the passed in pair */
6150 uoffset0 = cache[0];
6151 boffset0 = cache[1];
6153 if ((*mgp)->mg_len != -1) {
6154 /* And we know the end too. */
6156 + sv_pos_u2b_midway(start + boffset0, send,
6158 (*mgp)->mg_len - uoffset0);
6161 + sv_pos_u2b_forwards(start + boffset0,
6162 send, uoffset - uoffset0);
6165 else if (cache[2] < uoffset) {
6166 /* We're between the two cache entries. */
6167 if (cache[2] > uoffset0) {
6168 /* and the cache knows more than the passed in pair */
6169 uoffset0 = cache[2];
6170 boffset0 = cache[3];
6174 + sv_pos_u2b_midway(start + boffset0,
6177 cache[0] - uoffset0);
6180 + sv_pos_u2b_midway(start + boffset0,
6183 cache[2] - uoffset0);
6187 else if ((*mgp)->mg_len != -1) {
6188 /* If we can take advantage of a passed in offset, do so. */
6189 /* In fact, offset0 is either 0, or less than offset, so don't
6190 need to worry about the other possibility. */
6192 + sv_pos_u2b_midway(start + boffset0, send,
6194 (*mgp)->mg_len - uoffset0);
6199 if (!found || PL_utf8cache < 0) {
6200 const STRLEN real_boffset
6201 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6202 send, uoffset - uoffset0);
6204 if (found && PL_utf8cache < 0) {
6205 if (real_boffset != boffset) {
6206 /* Need to turn the assertions off otherwise we may recurse
6207 infinitely while printing error messages. */
6208 SAVEI8(PL_utf8cache);
6210 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6211 " real %"UVuf" for %"SVf,
6212 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6215 boffset = real_boffset;
6219 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6225 =for apidoc sv_pos_u2b_flags
6227 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6228 the start of the string, to a count of the equivalent number of bytes; if
6229 lenp is non-zero, it does the same to lenp, but this time starting from
6230 the offset, rather than from the start of the string. Handles type coercion.
6231 I<flags> is passed to C<SvPV_flags>, and usually should be
6232 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6238 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6239 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6240 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6245 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6252 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6254 start = (U8*)SvPV_flags(sv, len, flags);
6256 const U8 * const send = start + len;
6258 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6261 /* Convert the relative offset to absolute. */
6262 const STRLEN uoffset2 = uoffset + *lenp;
6263 const STRLEN boffset2
6264 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6265 uoffset, boffset) - boffset;
6279 =for apidoc sv_pos_u2b
6281 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6282 the start of the string, to a count of the equivalent number of bytes; if
6283 lenp is non-zero, it does the same to lenp, but this time starting from
6284 the offset, rather than from the start of the string. Handles magic and
6287 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6294 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6295 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6296 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6300 /* This function is subject to size and sign problems */
6303 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6305 PERL_ARGS_ASSERT_SV_POS_U2B;
6308 STRLEN ulen = (STRLEN)*lenp;
6309 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6310 SV_GMAGIC|SV_CONST_RETURN);
6313 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6314 SV_GMAGIC|SV_CONST_RETURN);
6318 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6319 byte length pairing. The (byte) length of the total SV is passed in too,
6320 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6321 may not have updated SvCUR, so we can't rely on reading it directly.
6323 The proffered utf8/byte length pairing isn't used if the cache already has
6324 two pairs, and swapping either for the proffered pair would increase the
6325 RMS of the intervals between known byte offsets.
6327 The cache itself consists of 4 STRLEN values
6328 0: larger UTF-8 offset
6329 1: corresponding byte offset
6330 2: smaller UTF-8 offset
6331 3: corresponding byte offset
6333 Unused cache pairs have the value 0, 0.
6334 Keeping the cache "backwards" means that the invariant of
6335 cache[0] >= cache[2] is maintained even with empty slots, which means that
6336 the code that uses it doesn't need to worry if only 1 entry has actually
6337 been set to non-zero. It also makes the "position beyond the end of the
6338 cache" logic much simpler, as the first slot is always the one to start
6342 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6343 const STRLEN utf8, const STRLEN blen)
6347 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6352 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6353 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6354 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6356 (*mgp)->mg_len = -1;
6360 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6361 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6362 (*mgp)->mg_ptr = (char *) cache;
6366 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6367 /* SvPOKp() because it's possible that sv has string overloading, and
6368 therefore is a reference, hence SvPVX() is actually a pointer.
6369 This cures the (very real) symptoms of RT 69422, but I'm not actually
6370 sure whether we should even be caching the results of UTF-8
6371 operations on overloading, given that nothing stops overloading
6372 returning a different value every time it's called. */
6373 const U8 *start = (const U8 *) SvPVX_const(sv);
6374 const STRLEN realutf8 = utf8_length(start, start + byte);
6376 if (realutf8 != utf8) {
6377 /* Need to turn the assertions off otherwise we may recurse
6378 infinitely while printing error messages. */
6379 SAVEI8(PL_utf8cache);
6381 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6382 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6386 /* Cache is held with the later position first, to simplify the code
6387 that deals with unbounded ends. */
6389 ASSERT_UTF8_CACHE(cache);
6390 if (cache[1] == 0) {
6391 /* Cache is totally empty */
6394 } else if (cache[3] == 0) {
6395 if (byte > cache[1]) {
6396 /* New one is larger, so goes first. */
6397 cache[2] = cache[0];
6398 cache[3] = cache[1];
6406 #define THREEWAY_SQUARE(a,b,c,d) \
6407 ((float)((d) - (c))) * ((float)((d) - (c))) \
6408 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6409 + ((float)((b) - (a))) * ((float)((b) - (a)))
6411 /* Cache has 2 slots in use, and we know three potential pairs.
6412 Keep the two that give the lowest RMS distance. Do the
6413 calcualation in bytes simply because we always know the byte
6414 length. squareroot has the same ordering as the positive value,
6415 so don't bother with the actual square root. */
6416 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6417 if (byte > cache[1]) {
6418 /* New position is after the existing pair of pairs. */
6419 const float keep_earlier
6420 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6421 const float keep_later
6422 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6424 if (keep_later < keep_earlier) {
6425 if (keep_later < existing) {
6426 cache[2] = cache[0];
6427 cache[3] = cache[1];
6433 if (keep_earlier < existing) {
6439 else if (byte > cache[3]) {
6440 /* New position is between the existing pair of pairs. */
6441 const float keep_earlier
6442 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6443 const float keep_later
6444 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6446 if (keep_later < keep_earlier) {
6447 if (keep_later < existing) {
6453 if (keep_earlier < existing) {
6460 /* New position is before the existing pair of pairs. */
6461 const float keep_earlier
6462 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6463 const float keep_later
6464 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6466 if (keep_later < keep_earlier) {
6467 if (keep_later < existing) {
6473 if (keep_earlier < existing) {
6474 cache[0] = cache[2];
6475 cache[1] = cache[3];
6482 ASSERT_UTF8_CACHE(cache);
6485 /* We already know all of the way, now we may be able to walk back. The same
6486 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6487 backward is half the speed of walking forward. */
6489 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6490 const U8 *end, STRLEN endu)
6492 const STRLEN forw = target - s;
6493 STRLEN backw = end - target;
6495 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6497 if (forw < 2 * backw) {
6498 return utf8_length(s, target);
6501 while (end > target) {
6503 while (UTF8_IS_CONTINUATION(*end)) {
6512 =for apidoc sv_pos_b2u
6514 Converts the value pointed to by offsetp from a count of bytes from the
6515 start of the string, to a count of the equivalent number of UTF-8 chars.
6516 Handles magic and type coercion.
6522 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6523 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6528 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6531 const STRLEN byte = *offsetp;
6532 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6538 PERL_ARGS_ASSERT_SV_POS_B2U;
6543 s = (const U8*)SvPV_const(sv, blen);
6546 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6552 && SvTYPE(sv) >= SVt_PVMG
6553 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6556 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6557 if (cache[1] == byte) {
6558 /* An exact match. */
6559 *offsetp = cache[0];
6562 if (cache[3] == byte) {
6563 /* An exact match. */
6564 *offsetp = cache[2];
6568 if (cache[1] < byte) {
6569 /* We already know part of the way. */
6570 if (mg->mg_len != -1) {
6571 /* Actually, we know the end too. */
6573 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6574 s + blen, mg->mg_len - cache[0]);
6576 len = cache[0] + utf8_length(s + cache[1], send);
6579 else if (cache[3] < byte) {
6580 /* We're between the two cached pairs, so we do the calculation
6581 offset by the byte/utf-8 positions for the earlier pair,
6582 then add the utf-8 characters from the string start to
6584 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6585 s + cache[1], cache[0] - cache[2])
6589 else { /* cache[3] > byte */
6590 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6594 ASSERT_UTF8_CACHE(cache);
6596 } else if (mg->mg_len != -1) {
6597 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6601 if (!found || PL_utf8cache < 0) {
6602 const STRLEN real_len = utf8_length(s, send);
6604 if (found && PL_utf8cache < 0) {
6605 if (len != real_len) {
6606 /* Need to turn the assertions off otherwise we may recurse
6607 infinitely while printing error messages. */
6608 SAVEI8(PL_utf8cache);
6610 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6611 " real %"UVuf" for %"SVf,
6612 (UV) len, (UV) real_len, SVfARG(sv));
6620 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6626 Returns a boolean indicating whether the strings in the two SVs are
6627 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6628 coerce its args to strings if necessary.
6634 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6643 SV* svrecode = NULL;
6650 /* if pv1 and pv2 are the same, second SvPV_const call may
6651 * invalidate pv1, so we may need to make a copy */
6652 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6653 pv1 = SvPV_const(sv1, cur1);
6654 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6656 pv1 = SvPV_const(sv1, cur1);
6664 pv2 = SvPV_const(sv2, cur2);
6666 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6667 /* Differing utf8ness.
6668 * Do not UTF8size the comparands as a side-effect. */
6671 svrecode = newSVpvn(pv2, cur2);
6672 sv_recode_to_utf8(svrecode, PL_encoding);
6673 pv2 = SvPV_const(svrecode, cur2);
6676 svrecode = newSVpvn(pv1, cur1);
6677 sv_recode_to_utf8(svrecode, PL_encoding);
6678 pv1 = SvPV_const(svrecode, cur1);
6680 /* Now both are in UTF-8. */
6682 SvREFCNT_dec(svrecode);
6687 bool is_utf8 = TRUE;
6690 /* sv1 is the UTF-8 one,
6691 * if is equal it must be downgrade-able */
6692 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6698 /* sv2 is the UTF-8 one,
6699 * if is equal it must be downgrade-able */
6700 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6706 /* Downgrade not possible - cannot be eq */
6714 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6716 SvREFCNT_dec(svrecode);
6726 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6727 string in C<sv1> is less than, equal to, or greater than the string in
6728 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6729 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6735 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6739 const char *pv1, *pv2;
6742 SV *svrecode = NULL;
6749 pv1 = SvPV_const(sv1, cur1);
6756 pv2 = SvPV_const(sv2, cur2);
6758 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6759 /* Differing utf8ness.
6760 * Do not UTF8size the comparands as a side-effect. */
6763 svrecode = newSVpvn(pv2, cur2);
6764 sv_recode_to_utf8(svrecode, PL_encoding);
6765 pv2 = SvPV_const(svrecode, cur2);
6768 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6773 svrecode = newSVpvn(pv1, cur1);
6774 sv_recode_to_utf8(svrecode, PL_encoding);
6775 pv1 = SvPV_const(svrecode, cur1);
6778 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6784 cmp = cur2 ? -1 : 0;
6788 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6791 cmp = retval < 0 ? -1 : 1;
6792 } else if (cur1 == cur2) {
6795 cmp = cur1 < cur2 ? -1 : 1;
6799 SvREFCNT_dec(svrecode);
6807 =for apidoc sv_cmp_locale
6809 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6810 'use bytes' aware, handles get magic, and will coerce its args to strings
6811 if necessary. See also C<sv_cmp>.
6817 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6820 #ifdef USE_LOCALE_COLLATE
6826 if (PL_collation_standard)
6830 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6832 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6834 if (!pv1 || !len1) {
6845 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6848 return retval < 0 ? -1 : 1;
6851 * When the result of collation is equality, that doesn't mean
6852 * that there are no differences -- some locales exclude some
6853 * characters from consideration. So to avoid false equalities,
6854 * we use the raw string as a tiebreaker.
6860 #endif /* USE_LOCALE_COLLATE */
6862 return sv_cmp(sv1, sv2);
6866 #ifdef USE_LOCALE_COLLATE
6869 =for apidoc sv_collxfrm
6871 Add Collate Transform magic to an SV if it doesn't already have it.
6873 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6874 scalar data of the variable, but transformed to such a format that a normal
6875 memory comparison can be used to compare the data according to the locale
6882 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6887 PERL_ARGS_ASSERT_SV_COLLXFRM;
6889 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6890 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6896 Safefree(mg->mg_ptr);
6897 s = SvPV_const(sv, len);
6898 if ((xf = mem_collxfrm(s, len, &xlen))) {
6900 #ifdef PERL_OLD_COPY_ON_WRITE
6902 sv_force_normal_flags(sv, 0);
6904 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6918 if (mg && mg->mg_ptr) {
6920 return mg->mg_ptr + sizeof(PL_collation_ix);
6928 #endif /* USE_LOCALE_COLLATE */
6933 Get a line from the filehandle and store it into the SV, optionally
6934 appending to the currently-stored string.
6940 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6945 register STDCHAR rslast;
6946 register STDCHAR *bp;
6951 PERL_ARGS_ASSERT_SV_GETS;
6953 if (SvTHINKFIRST(sv))
6954 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6955 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6957 However, perlbench says it's slower, because the existing swipe code
6958 is faster than copy on write.
6959 Swings and roundabouts. */
6960 SvUPGRADE(sv, SVt_PV);
6965 if (PerlIO_isutf8(fp)) {
6967 sv_utf8_upgrade_nomg(sv);
6968 sv_pos_u2b(sv,&append,0);
6970 } else if (SvUTF8(sv)) {
6971 SV * const tsv = newSV(0);
6972 sv_gets(tsv, fp, 0);
6973 sv_utf8_upgrade_nomg(tsv);
6974 SvCUR_set(sv,append);
6977 goto return_string_or_null;
6982 if (PerlIO_isutf8(fp))
6985 if (IN_PERL_COMPILETIME) {
6986 /* we always read code in line mode */
6990 else if (RsSNARF(PL_rs)) {
6991 /* If it is a regular disk file use size from stat() as estimate
6992 of amount we are going to read -- may result in mallocing
6993 more memory than we really need if the layers below reduce
6994 the size we read (e.g. CRLF or a gzip layer).
6997 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6998 const Off_t offset = PerlIO_tell(fp);
6999 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7000 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7006 else if (RsRECORD(PL_rs)) {
7014 /* Grab the size of the record we're getting */
7015 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7016 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7019 /* VMS wants read instead of fread, because fread doesn't respect */
7020 /* RMS record boundaries. This is not necessarily a good thing to be */
7021 /* doing, but we've got no other real choice - except avoid stdio
7022 as implementation - perhaps write a :vms layer ?
7024 fd = PerlIO_fileno(fp);
7025 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
7026 bytesread = PerlIO_read(fp, buffer, recsize);
7029 bytesread = PerlLIO_read(fd, buffer, recsize);
7032 bytesread = PerlIO_read(fp, buffer, recsize);
7036 SvCUR_set(sv, bytesread + append);
7037 buffer[bytesread] = '\0';
7038 goto return_string_or_null;
7040 else if (RsPARA(PL_rs)) {
7046 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7047 if (PerlIO_isutf8(fp)) {
7048 rsptr = SvPVutf8(PL_rs, rslen);
7051 if (SvUTF8(PL_rs)) {
7052 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7053 Perl_croak(aTHX_ "Wide character in $/");
7056 rsptr = SvPV_const(PL_rs, rslen);
7060 rslast = rslen ? rsptr[rslen - 1] : '\0';
7062 if (rspara) { /* have to do this both before and after */
7063 do { /* to make sure file boundaries work right */
7066 i = PerlIO_getc(fp);
7070 PerlIO_ungetc(fp,i);
7076 /* See if we know enough about I/O mechanism to cheat it ! */
7078 /* This used to be #ifdef test - it is made run-time test for ease
7079 of abstracting out stdio interface. One call should be cheap
7080 enough here - and may even be a macro allowing compile
7084 if (PerlIO_fast_gets(fp)) {
7087 * We're going to steal some values from the stdio struct
7088 * and put EVERYTHING in the innermost loop into registers.
7090 register STDCHAR *ptr;
7094 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7095 /* An ungetc()d char is handled separately from the regular
7096 * buffer, so we getc() it back out and stuff it in the buffer.
7098 i = PerlIO_getc(fp);
7099 if (i == EOF) return 0;
7100 *(--((*fp)->_ptr)) = (unsigned char) i;
7104 /* Here is some breathtakingly efficient cheating */
7106 cnt = PerlIO_get_cnt(fp); /* get count into register */
7107 /* make sure we have the room */
7108 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7109 /* Not room for all of it
7110 if we are looking for a separator and room for some
7112 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7113 /* just process what we have room for */
7114 shortbuffered = cnt - SvLEN(sv) + append + 1;
7115 cnt -= shortbuffered;
7119 /* remember that cnt can be negative */
7120 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7125 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7126 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7127 DEBUG_P(PerlIO_printf(Perl_debug_log,
7128 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7129 DEBUG_P(PerlIO_printf(Perl_debug_log,
7130 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7131 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7132 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7137 while (cnt > 0) { /* this | eat */
7139 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7140 goto thats_all_folks; /* screams | sed :-) */
7144 Copy(ptr, bp, cnt, char); /* this | eat */
7145 bp += cnt; /* screams | dust */
7146 ptr += cnt; /* louder | sed :-) */
7151 if (shortbuffered) { /* oh well, must extend */
7152 cnt = shortbuffered;
7154 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7156 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7157 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7161 DEBUG_P(PerlIO_printf(Perl_debug_log,
7162 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7163 PTR2UV(ptr),(long)cnt));
7164 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7166 DEBUG_P(PerlIO_printf(Perl_debug_log,
7167 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7168 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7169 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7171 /* This used to call 'filbuf' in stdio form, but as that behaves like
7172 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7173 another abstraction. */
7174 i = PerlIO_getc(fp); /* get more characters */
7176 DEBUG_P(PerlIO_printf(Perl_debug_log,
7177 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7178 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7179 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7181 cnt = PerlIO_get_cnt(fp);
7182 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7183 DEBUG_P(PerlIO_printf(Perl_debug_log,
7184 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7186 if (i == EOF) /* all done for ever? */
7187 goto thats_really_all_folks;
7189 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7191 SvGROW(sv, bpx + cnt + 2);
7192 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7194 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7196 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7197 goto thats_all_folks;
7201 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7202 memNE((char*)bp - rslen, rsptr, rslen))
7203 goto screamer; /* go back to the fray */
7204 thats_really_all_folks:
7206 cnt += shortbuffered;
7207 DEBUG_P(PerlIO_printf(Perl_debug_log,
7208 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7209 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7210 DEBUG_P(PerlIO_printf(Perl_debug_log,
7211 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7212 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7213 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7215 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7216 DEBUG_P(PerlIO_printf(Perl_debug_log,
7217 "Screamer: done, len=%ld, string=|%.*s|\n",
7218 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7222 /*The big, slow, and stupid way. */
7223 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7224 STDCHAR *buf = NULL;
7225 Newx(buf, 8192, STDCHAR);
7233 register const STDCHAR * const bpe = buf + sizeof(buf);
7235 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7236 ; /* keep reading */
7240 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7241 /* Accomodate broken VAXC compiler, which applies U8 cast to
7242 * both args of ?: operator, causing EOF to change into 255
7245 i = (U8)buf[cnt - 1];
7251 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7253 sv_catpvn(sv, (char *) buf, cnt);
7255 sv_setpvn(sv, (char *) buf, cnt);
7257 if (i != EOF && /* joy */
7259 SvCUR(sv) < rslen ||
7260 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7264 * If we're reading from a TTY and we get a short read,
7265 * indicating that the user hit his EOF character, we need
7266 * to notice it now, because if we try to read from the TTY
7267 * again, the EOF condition will disappear.
7269 * The comparison of cnt to sizeof(buf) is an optimization
7270 * that prevents unnecessary calls to feof().
7274 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7278 #ifdef USE_HEAP_INSTEAD_OF_STACK
7283 if (rspara) { /* have to do this both before and after */
7284 while (i != EOF) { /* to make sure file boundaries work right */
7285 i = PerlIO_getc(fp);
7287 PerlIO_ungetc(fp,i);
7293 return_string_or_null:
7294 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7300 Auto-increment of the value in the SV, doing string to numeric conversion
7301 if necessary. Handles 'get' magic.
7307 Perl_sv_inc(pTHX_ register SV *const sv)
7316 if (SvTHINKFIRST(sv)) {
7318 sv_force_normal_flags(sv, 0);
7319 if (SvREADONLY(sv)) {
7320 if (IN_PERL_RUNTIME)
7321 Perl_croak(aTHX_ "%s", PL_no_modify);
7325 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7327 i = PTR2IV(SvRV(sv));
7332 flags = SvFLAGS(sv);
7333 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7334 /* It's (privately or publicly) a float, but not tested as an
7335 integer, so test it to see. */
7337 flags = SvFLAGS(sv);
7339 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7340 /* It's publicly an integer, or privately an integer-not-float */
7341 #ifdef PERL_PRESERVE_IVUV
7345 if (SvUVX(sv) == UV_MAX)
7346 sv_setnv(sv, UV_MAX_P1);
7348 (void)SvIOK_only_UV(sv);
7349 SvUV_set(sv, SvUVX(sv) + 1);
7351 if (SvIVX(sv) == IV_MAX)
7352 sv_setuv(sv, (UV)IV_MAX + 1);
7354 (void)SvIOK_only(sv);
7355 SvIV_set(sv, SvIVX(sv) + 1);
7360 if (flags & SVp_NOK) {
7361 const NV was = SvNVX(sv);
7362 if (NV_OVERFLOWS_INTEGERS_AT &&
7363 was >= NV_OVERFLOWS_INTEGERS_AT) {
7364 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7365 "Lost precision when incrementing %" NVff " by 1",
7368 (void)SvNOK_only(sv);
7369 SvNV_set(sv, was + 1.0);
7373 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7374 if ((flags & SVTYPEMASK) < SVt_PVIV)
7375 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7376 (void)SvIOK_only(sv);
7381 while (isALPHA(*d)) d++;
7382 while (isDIGIT(*d)) d++;
7383 if (d < SvEND(sv)) {
7384 #ifdef PERL_PRESERVE_IVUV
7385 /* Got to punt this as an integer if needs be, but we don't issue
7386 warnings. Probably ought to make the sv_iv_please() that does
7387 the conversion if possible, and silently. */
7388 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7389 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7390 /* Need to try really hard to see if it's an integer.
7391 9.22337203685478e+18 is an integer.
7392 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7393 so $a="9.22337203685478e+18"; $a+0; $a++
7394 needs to be the same as $a="9.22337203685478e+18"; $a++
7401 /* sv_2iv *should* have made this an NV */
7402 if (flags & SVp_NOK) {
7403 (void)SvNOK_only(sv);
7404 SvNV_set(sv, SvNVX(sv) + 1.0);
7407 /* I don't think we can get here. Maybe I should assert this
7408 And if we do get here I suspect that sv_setnv will croak. NWC
7410 #if defined(USE_LONG_DOUBLE)
7411 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",
7412 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7414 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7415 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7418 #endif /* PERL_PRESERVE_IVUV */
7419 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7423 while (d >= SvPVX_const(sv)) {
7431 /* MKS: The original code here died if letters weren't consecutive.
7432 * at least it didn't have to worry about non-C locales. The
7433 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7434 * arranged in order (although not consecutively) and that only
7435 * [A-Za-z] are accepted by isALPHA in the C locale.
7437 if (*d != 'z' && *d != 'Z') {
7438 do { ++*d; } while (!isALPHA(*d));
7441 *(d--) -= 'z' - 'a';
7446 *(d--) -= 'z' - 'a' + 1;
7450 /* oh,oh, the number grew */
7451 SvGROW(sv, SvCUR(sv) + 2);
7452 SvCUR_set(sv, SvCUR(sv) + 1);
7453 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7464 Auto-decrement of the value in the SV, doing string to numeric conversion
7465 if necessary. Handles 'get' magic.
7471 Perl_sv_dec(pTHX_ register SV *const sv)
7479 if (SvTHINKFIRST(sv)) {
7481 sv_force_normal_flags(sv, 0);
7482 if (SvREADONLY(sv)) {
7483 if (IN_PERL_RUNTIME)
7484 Perl_croak(aTHX_ "%s", PL_no_modify);
7488 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7490 i = PTR2IV(SvRV(sv));
7495 /* Unlike sv_inc we don't have to worry about string-never-numbers
7496 and keeping them magic. But we mustn't warn on punting */
7497 flags = SvFLAGS(sv);
7498 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7499 /* It's publicly an integer, or privately an integer-not-float */
7500 #ifdef PERL_PRESERVE_IVUV
7504 if (SvUVX(sv) == 0) {
7505 (void)SvIOK_only(sv);
7509 (void)SvIOK_only_UV(sv);
7510 SvUV_set(sv, SvUVX(sv) - 1);
7513 if (SvIVX(sv) == IV_MIN) {
7514 sv_setnv(sv, (NV)IV_MIN);
7518 (void)SvIOK_only(sv);
7519 SvIV_set(sv, SvIVX(sv) - 1);
7524 if (flags & SVp_NOK) {
7527 const NV was = SvNVX(sv);
7528 if (NV_OVERFLOWS_INTEGERS_AT &&
7529 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7530 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7531 "Lost precision when decrementing %" NVff " by 1",
7534 (void)SvNOK_only(sv);
7535 SvNV_set(sv, was - 1.0);
7539 if (!(flags & SVp_POK)) {
7540 if ((flags & SVTYPEMASK) < SVt_PVIV)
7541 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7543 (void)SvIOK_only(sv);
7546 #ifdef PERL_PRESERVE_IVUV
7548 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7549 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7550 /* Need to try really hard to see if it's an integer.
7551 9.22337203685478e+18 is an integer.
7552 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7553 so $a="9.22337203685478e+18"; $a+0; $a--
7554 needs to be the same as $a="9.22337203685478e+18"; $a--
7561 /* sv_2iv *should* have made this an NV */
7562 if (flags & SVp_NOK) {
7563 (void)SvNOK_only(sv);
7564 SvNV_set(sv, SvNVX(sv) - 1.0);
7567 /* I don't think we can get here. Maybe I should assert this
7568 And if we do get here I suspect that sv_setnv will croak. NWC
7570 #if defined(USE_LONG_DOUBLE)
7571 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",
7572 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7574 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7575 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7579 #endif /* PERL_PRESERVE_IVUV */
7580 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7583 /* this define is used to eliminate a chunk of duplicated but shared logic
7584 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7585 * used anywhere but here - yves
7587 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7590 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7594 =for apidoc sv_mortalcopy
7596 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7597 The new SV is marked as mortal. It will be destroyed "soon", either by an
7598 explicit call to FREETMPS, or by an implicit call at places such as
7599 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7604 /* Make a string that will exist for the duration of the expression
7605 * evaluation. Actually, it may have to last longer than that, but
7606 * hopefully we won't free it until it has been assigned to a
7607 * permanent location. */
7610 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7616 sv_setsv(sv,oldstr);
7617 PUSH_EXTEND_MORTAL__SV_C(sv);
7623 =for apidoc sv_newmortal
7625 Creates a new null SV which is mortal. The reference count of the SV is
7626 set to 1. It will be destroyed "soon", either by an explicit call to
7627 FREETMPS, or by an implicit call at places such as statement boundaries.
7628 See also C<sv_mortalcopy> and C<sv_2mortal>.
7634 Perl_sv_newmortal(pTHX)
7640 SvFLAGS(sv) = SVs_TEMP;
7641 PUSH_EXTEND_MORTAL__SV_C(sv);
7647 =for apidoc newSVpvn_flags
7649 Creates a new SV and copies a string into it. The reference count for the
7650 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7651 string. You are responsible for ensuring that the source string is at least
7652 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7653 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7654 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7655 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
7656 C<SVf_UTF8> flag will be set on the new SV.
7657 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7659 #define newSVpvn_utf8(s, len, u) \
7660 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7666 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7671 /* All the flags we don't support must be zero.
7672 And we're new code so I'm going to assert this from the start. */
7673 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7675 sv_setpvn(sv,s,len);
7677 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7678 * and do what it does outselves here.
7679 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7680 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7681 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7682 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7685 SvFLAGS(sv) |= flags;
7687 if(flags & SVs_TEMP){
7688 PUSH_EXTEND_MORTAL__SV_C(sv);
7695 =for apidoc sv_2mortal
7697 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7698 by an explicit call to FREETMPS, or by an implicit call at places such as
7699 statement boundaries. SvTEMP() is turned on which means that the SV's
7700 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7701 and C<sv_mortalcopy>.
7707 Perl_sv_2mortal(pTHX_ register SV *const sv)
7712 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7714 PUSH_EXTEND_MORTAL__SV_C(sv);
7722 Creates a new SV and copies a string into it. The reference count for the
7723 SV is set to 1. If C<len> is zero, Perl will compute the length using
7724 strlen(). For efficiency, consider using C<newSVpvn> instead.
7730 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7736 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7741 =for apidoc newSVpvn
7743 Creates a new SV and copies a string into it. The reference count for the
7744 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7745 string. You are responsible for ensuring that the source string is at least
7746 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7752 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7758 sv_setpvn(sv,s,len);
7763 =for apidoc newSVhek
7765 Creates a new SV from the hash key structure. It will generate scalars that
7766 point to the shared string table where possible. Returns a new (undefined)
7767 SV if the hek is NULL.
7773 Perl_newSVhek(pTHX_ const HEK *const hek)
7783 if (HEK_LEN(hek) == HEf_SVKEY) {
7784 return newSVsv(*(SV**)HEK_KEY(hek));
7786 const int flags = HEK_FLAGS(hek);
7787 if (flags & HVhek_WASUTF8) {
7789 Andreas would like keys he put in as utf8 to come back as utf8
7791 STRLEN utf8_len = HEK_LEN(hek);
7792 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7793 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7796 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7798 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7799 /* We don't have a pointer to the hv, so we have to replicate the
7800 flag into every HEK. This hv is using custom a hasing
7801 algorithm. Hence we can't return a shared string scalar, as
7802 that would contain the (wrong) hash value, and might get passed
7803 into an hv routine with a regular hash.
7804 Similarly, a hash that isn't using shared hash keys has to have
7805 the flag in every key so that we know not to try to call
7806 share_hek_kek on it. */
7808 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7813 /* This will be overwhelminly the most common case. */
7815 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7816 more efficient than sharepvn(). */
7820 sv_upgrade(sv, SVt_PV);
7821 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7822 SvCUR_set(sv, HEK_LEN(hek));
7835 =for apidoc newSVpvn_share
7837 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7838 table. If the string does not already exist in the table, it is created
7839 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7840 value is used; otherwise the hash is computed. The string's hash can be later
7841 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7842 that as the string table is used for shared hash keys these strings will have
7843 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7849 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7853 bool is_utf8 = FALSE;
7854 const char *const orig_src = src;
7857 STRLEN tmplen = -len;
7859 /* See the note in hv.c:hv_fetch() --jhi */
7860 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7864 PERL_HASH(hash, src, len);
7866 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7867 changes here, update it there too. */
7868 sv_upgrade(sv, SVt_PV);
7869 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7877 if (src != orig_src)
7883 #if defined(PERL_IMPLICIT_CONTEXT)
7885 /* pTHX_ magic can't cope with varargs, so this is a no-context
7886 * version of the main function, (which may itself be aliased to us).
7887 * Don't access this version directly.
7891 Perl_newSVpvf_nocontext(const char *const pat, ...)
7897 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7899 va_start(args, pat);
7900 sv = vnewSVpvf(pat, &args);
7907 =for apidoc newSVpvf
7909 Creates a new SV and initializes it with the string formatted like
7916 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7921 PERL_ARGS_ASSERT_NEWSVPVF;
7923 va_start(args, pat);
7924 sv = vnewSVpvf(pat, &args);
7929 /* backend for newSVpvf() and newSVpvf_nocontext() */
7932 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7937 PERL_ARGS_ASSERT_VNEWSVPVF;
7940 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7947 Creates a new SV and copies a floating point value into it.
7948 The reference count for the SV is set to 1.
7954 Perl_newSVnv(pTHX_ const NV n)
7967 Creates a new SV and copies an integer into it. The reference count for the
7974 Perl_newSViv(pTHX_ const IV i)
7987 Creates a new SV and copies an unsigned integer into it.
7988 The reference count for the SV is set to 1.
7994 Perl_newSVuv(pTHX_ const UV u)
8005 =for apidoc newSV_type
8007 Creates a new SV, of the type specified. The reference count for the new SV
8014 Perl_newSV_type(pTHX_ const svtype type)
8019 sv_upgrade(sv, type);
8024 =for apidoc newRV_noinc
8026 Creates an RV wrapper for an SV. The reference count for the original
8027 SV is B<not> incremented.
8033 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8036 register SV *sv = newSV_type(SVt_IV);
8038 PERL_ARGS_ASSERT_NEWRV_NOINC;
8041 SvRV_set(sv, tmpRef);
8046 /* newRV_inc is the official function name to use now.
8047 * newRV_inc is in fact #defined to newRV in sv.h
8051 Perl_newRV(pTHX_ SV *const sv)
8055 PERL_ARGS_ASSERT_NEWRV;
8057 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8063 Creates a new SV which is an exact duplicate of the original SV.
8070 Perl_newSVsv(pTHX_ register SV *const old)
8077 if (SvTYPE(old) == SVTYPEMASK) {
8078 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8082 /* SV_GMAGIC is the default for sv_setv()
8083 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8084 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8085 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8090 =for apidoc sv_reset
8092 Underlying implementation for the C<reset> Perl function.
8093 Note that the perl-level function is vaguely deprecated.
8099 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8102 char todo[PERL_UCHAR_MAX+1];
8104 PERL_ARGS_ASSERT_SV_RESET;
8109 if (!*s) { /* reset ?? searches */
8110 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8112 const U32 count = mg->mg_len / sizeof(PMOP**);
8113 PMOP **pmp = (PMOP**) mg->mg_ptr;
8114 PMOP *const *const end = pmp + count;
8118 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8120 (*pmp)->op_pmflags &= ~PMf_USED;
8128 /* reset variables */
8130 if (!HvARRAY(stash))
8133 Zero(todo, 256, char);
8136 I32 i = (unsigned char)*s;
8140 max = (unsigned char)*s++;
8141 for ( ; i <= max; i++) {
8144 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8146 for (entry = HvARRAY(stash)[i];
8148 entry = HeNEXT(entry))
8153 if (!todo[(U8)*HeKEY(entry)])
8155 gv = MUTABLE_GV(HeVAL(entry));
8158 if (SvTHINKFIRST(sv)) {
8159 if (!SvREADONLY(sv) && SvROK(sv))
8161 /* XXX Is this continue a bug? Why should THINKFIRST
8162 exempt us from resetting arrays and hashes? */
8166 if (SvTYPE(sv) >= SVt_PV) {
8168 if (SvPVX_const(sv) != NULL)
8176 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8178 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8181 # if defined(USE_ENVIRON_ARRAY)
8184 # endif /* USE_ENVIRON_ARRAY */
8195 Using various gambits, try to get an IO from an SV: the IO slot if its a
8196 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8197 named after the PV if we're a string.
8203 Perl_sv_2io(pTHX_ SV *const sv)
8208 PERL_ARGS_ASSERT_SV_2IO;
8210 switch (SvTYPE(sv)) {
8212 io = MUTABLE_IO(sv);
8215 if (isGV_with_GP(sv)) {
8216 gv = MUTABLE_GV(sv);
8219 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8225 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8227 return sv_2io(SvRV(sv));
8228 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8234 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8243 Using various gambits, try to get a CV from an SV; in addition, try if
8244 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8245 The flags in C<lref> are passed to gv_fetchsv.
8251 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8257 PERL_ARGS_ASSERT_SV_2CV;
8264 switch (SvTYPE(sv)) {
8268 return MUTABLE_CV(sv);
8275 if (isGV_with_GP(sv)) {
8276 gv = MUTABLE_GV(sv);
8285 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8287 tryAMAGICunDEREF(to_cv);
8290 if (SvTYPE(sv) == SVt_PVCV) {
8291 cv = MUTABLE_CV(sv);
8296 else if(isGV_with_GP(sv))
8297 gv = MUTABLE_GV(sv);
8299 Perl_croak(aTHX_ "Not a subroutine reference");
8301 else if (isGV_with_GP(sv)) {
8303 gv = MUTABLE_GV(sv);
8306 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8312 /* Some flags to gv_fetchsv mean don't really create the GV */
8313 if (!isGV_with_GP(gv)) {
8319 if (lref && !GvCVu(gv)) {
8323 gv_efullname3(tmpsv, gv, NULL);
8324 /* XXX this is probably not what they think they're getting.
8325 * It has the same effect as "sub name;", i.e. just a forward
8327 newSUB(start_subparse(FALSE, 0),
8328 newSVOP(OP_CONST, 0, tmpsv),
8332 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8333 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8342 Returns true if the SV has a true value by Perl's rules.
8343 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8344 instead use an in-line version.
8350 Perl_sv_true(pTHX_ register SV *const sv)
8355 register const XPV* const tXpv = (XPV*)SvANY(sv);
8357 (tXpv->xpv_cur > 1 ||
8358 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8365 return SvIVX(sv) != 0;
8368 return SvNVX(sv) != 0.0;
8370 return sv_2bool(sv);
8376 =for apidoc sv_pvn_force
8378 Get a sensible string out of the SV somehow.
8379 A private implementation of the C<SvPV_force> macro for compilers which
8380 can't cope with complex macro expressions. Always use the macro instead.
8382 =for apidoc sv_pvn_force_flags
8384 Get a sensible string out of the SV somehow.
8385 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8386 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8387 implemented in terms of this function.
8388 You normally want to use the various wrapper macros instead: see
8389 C<SvPV_force> and C<SvPV_force_nomg>
8395 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8399 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8401 if (SvTHINKFIRST(sv) && !SvROK(sv))
8402 sv_force_normal_flags(sv, 0);
8412 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8413 const char * const ref = sv_reftype(sv,0);
8415 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8416 ref, OP_DESC(PL_op));
8418 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8420 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8421 || isGV_with_GP(sv))
8422 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8424 s = sv_2pv_flags(sv, &len, flags);
8428 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8431 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8432 SvGROW(sv, len + 1);
8433 Move(s,SvPVX(sv),len,char);
8435 SvPVX(sv)[len] = '\0';
8438 SvPOK_on(sv); /* validate pointer */
8440 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8441 PTR2UV(sv),SvPVX_const(sv)));
8444 return SvPVX_mutable(sv);
8448 =for apidoc sv_pvbyten_force
8450 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8456 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8458 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8460 sv_pvn_force(sv,lp);
8461 sv_utf8_downgrade(sv,0);
8467 =for apidoc sv_pvutf8n_force
8469 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8475 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8477 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8479 sv_pvn_force(sv,lp);
8480 sv_utf8_upgrade(sv);
8486 =for apidoc sv_reftype
8488 Returns a string describing what the SV is a reference to.
8494 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8496 PERL_ARGS_ASSERT_SV_REFTYPE;
8498 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8499 inside return suggests a const propagation bug in g++. */
8500 if (ob && SvOBJECT(sv)) {
8501 char * const name = HvNAME_get(SvSTASH(sv));
8502 return name ? name : (char *) "__ANON__";
8505 switch (SvTYPE(sv)) {
8520 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8521 /* tied lvalues should appear to be
8522 * scalars for backwards compatitbility */
8523 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8524 ? "SCALAR" : "LVALUE");
8525 case SVt_PVAV: return "ARRAY";
8526 case SVt_PVHV: return "HASH";
8527 case SVt_PVCV: return "CODE";
8528 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8529 ? "GLOB" : "SCALAR");
8530 case SVt_PVFM: return "FORMAT";
8531 case SVt_PVIO: return "IO";
8532 case SVt_BIND: return "BIND";
8533 case SVt_REGEXP: return "REGEXP";
8534 default: return "UNKNOWN";
8540 =for apidoc sv_isobject
8542 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8543 object. If the SV is not an RV, or if the object is not blessed, then this
8550 Perl_sv_isobject(pTHX_ SV *sv)
8566 Returns a boolean indicating whether the SV is blessed into the specified
8567 class. This does not check for subtypes; use C<sv_derived_from> to verify
8568 an inheritance relationship.
8574 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8578 PERL_ARGS_ASSERT_SV_ISA;
8588 hvname = HvNAME_get(SvSTASH(sv));
8592 return strEQ(hvname, name);
8598 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8599 it will be upgraded to one. If C<classname> is non-null then the new SV will
8600 be blessed in the specified package. The new SV is returned and its
8601 reference count is 1.
8607 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8612 PERL_ARGS_ASSERT_NEWSVRV;
8616 SV_CHECK_THINKFIRST_COW_DROP(rv);
8617 (void)SvAMAGIC_off(rv);
8619 if (SvTYPE(rv) >= SVt_PVMG) {
8620 const U32 refcnt = SvREFCNT(rv);
8624 SvREFCNT(rv) = refcnt;
8626 sv_upgrade(rv, SVt_IV);
8627 } else if (SvROK(rv)) {
8628 SvREFCNT_dec(SvRV(rv));
8630 prepare_SV_for_RV(rv);
8638 HV* const stash = gv_stashpv(classname, GV_ADD);
8639 (void)sv_bless(rv, stash);
8645 =for apidoc sv_setref_pv
8647 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8648 argument will be upgraded to an RV. That RV will be modified to point to
8649 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8650 into the SV. The C<classname> argument indicates the package for the
8651 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8652 will have a reference count of 1, and the RV will be returned.
8654 Do not use with other Perl types such as HV, AV, SV, CV, because those
8655 objects will become corrupted by the pointer copy process.
8657 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8663 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8667 PERL_ARGS_ASSERT_SV_SETREF_PV;
8670 sv_setsv(rv, &PL_sv_undef);
8674 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8679 =for apidoc sv_setref_iv
8681 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8682 argument will be upgraded to an RV. That RV will be modified to point to
8683 the new SV. The C<classname> argument indicates the package for the
8684 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8685 will have a reference count of 1, and the RV will be returned.
8691 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8693 PERL_ARGS_ASSERT_SV_SETREF_IV;
8695 sv_setiv(newSVrv(rv,classname), iv);
8700 =for apidoc sv_setref_uv
8702 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8703 argument will be upgraded to an RV. That RV will be modified to point to
8704 the new SV. The C<classname> argument indicates the package for the
8705 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8706 will have a reference count of 1, and the RV will be returned.
8712 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8714 PERL_ARGS_ASSERT_SV_SETREF_UV;
8716 sv_setuv(newSVrv(rv,classname), uv);
8721 =for apidoc sv_setref_nv
8723 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8724 argument will be upgraded to an RV. That RV will be modified to point to
8725 the new SV. The C<classname> argument indicates the package for the
8726 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8727 will have a reference count of 1, and the RV will be returned.
8733 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8735 PERL_ARGS_ASSERT_SV_SETREF_NV;
8737 sv_setnv(newSVrv(rv,classname), nv);
8742 =for apidoc sv_setref_pvn
8744 Copies a string into a new SV, optionally blessing the SV. The length of the
8745 string must be specified with C<n>. The C<rv> argument will be upgraded to
8746 an RV. That RV will be modified to point to the new SV. The C<classname>
8747 argument indicates the package for the blessing. Set C<classname> to
8748 C<NULL> to avoid the blessing. The new SV will have a reference count
8749 of 1, and the RV will be returned.
8751 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8757 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8758 const char *const pv, const STRLEN n)
8760 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8762 sv_setpvn(newSVrv(rv,classname), pv, n);
8767 =for apidoc sv_bless
8769 Blesses an SV into a specified package. The SV must be an RV. The package
8770 must be designated by its stash (see C<gv_stashpv()>). The reference count
8771 of the SV is unaffected.
8777 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8782 PERL_ARGS_ASSERT_SV_BLESS;
8785 Perl_croak(aTHX_ "Can't bless non-reference value");
8787 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8788 if (SvIsCOW(tmpRef))
8789 sv_force_normal_flags(tmpRef, 0);
8790 if (SvREADONLY(tmpRef))
8791 Perl_croak(aTHX_ "%s", PL_no_modify);
8792 if (SvOBJECT(tmpRef)) {
8793 if (SvTYPE(tmpRef) != SVt_PVIO)
8795 SvREFCNT_dec(SvSTASH(tmpRef));
8798 SvOBJECT_on(tmpRef);
8799 if (SvTYPE(tmpRef) != SVt_PVIO)
8801 SvUPGRADE(tmpRef, SVt_PVMG);
8802 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8807 (void)SvAMAGIC_off(sv);
8809 if(SvSMAGICAL(tmpRef))
8810 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8818 /* Downgrades a PVGV to a PVMG.
8822 S_sv_unglob(pTHX_ SV *const sv)
8827 SV * const temp = sv_newmortal();
8829 PERL_ARGS_ASSERT_SV_UNGLOB;
8831 assert(SvTYPE(sv) == SVt_PVGV);
8833 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8836 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8837 && HvNAME_get(stash))
8838 mro_method_changed_in(stash);
8839 gp_free(MUTABLE_GV(sv));
8842 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8846 if (GvNAME_HEK(sv)) {
8847 unshare_hek(GvNAME_HEK(sv));
8849 isGV_with_GP_off(sv);
8851 /* need to keep SvANY(sv) in the right arena */
8852 xpvmg = new_XPVMG();
8853 StructCopy(SvANY(sv), xpvmg, XPVMG);
8854 del_XPVGV(SvANY(sv));
8857 SvFLAGS(sv) &= ~SVTYPEMASK;
8858 SvFLAGS(sv) |= SVt_PVMG;
8860 /* Intentionally not calling any local SET magic, as this isn't so much a
8861 set operation as merely an internal storage change. */
8862 sv_setsv_flags(sv, temp, 0);
8866 =for apidoc sv_unref_flags
8868 Unsets the RV status of the SV, and decrements the reference count of
8869 whatever was being referenced by the RV. This can almost be thought of
8870 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8871 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8872 (otherwise the decrementing is conditional on the reference count being
8873 different from one or the reference being a readonly SV).
8880 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8882 SV* const target = SvRV(ref);
8884 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8886 if (SvWEAKREF(ref)) {
8887 sv_del_backref(target, ref);
8889 SvRV_set(ref, NULL);
8892 SvRV_set(ref, NULL);
8894 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8895 assigned to as BEGIN {$a = \"Foo"} will fail. */
8896 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8897 SvREFCNT_dec(target);
8898 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8899 sv_2mortal(target); /* Schedule for freeing later */
8903 =for apidoc sv_untaint
8905 Untaint an SV. Use C<SvTAINTED_off> instead.
8910 Perl_sv_untaint(pTHX_ SV *const sv)
8912 PERL_ARGS_ASSERT_SV_UNTAINT;
8914 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8915 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8922 =for apidoc sv_tainted
8924 Test an SV for taintedness. Use C<SvTAINTED> instead.
8929 Perl_sv_tainted(pTHX_ SV *const sv)
8931 PERL_ARGS_ASSERT_SV_TAINTED;
8933 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8934 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8935 if (mg && (mg->mg_len & 1) )
8942 =for apidoc sv_setpviv
8944 Copies an integer into the given SV, also updating its string value.
8945 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8951 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8953 char buf[TYPE_CHARS(UV)];
8955 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8957 PERL_ARGS_ASSERT_SV_SETPVIV;
8959 sv_setpvn(sv, ptr, ebuf - ptr);
8963 =for apidoc sv_setpviv_mg
8965 Like C<sv_setpviv>, but also handles 'set' magic.
8971 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8973 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8979 #if defined(PERL_IMPLICIT_CONTEXT)
8981 /* pTHX_ magic can't cope with varargs, so this is a no-context
8982 * version of the main function, (which may itself be aliased to us).
8983 * Don't access this version directly.
8987 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8992 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8994 va_start(args, pat);
8995 sv_vsetpvf(sv, pat, &args);
8999 /* pTHX_ magic can't cope with varargs, so this is a no-context
9000 * version of the main function, (which may itself be aliased to us).
9001 * Don't access this version directly.
9005 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9010 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9012 va_start(args, pat);
9013 sv_vsetpvf_mg(sv, pat, &args);
9019 =for apidoc sv_setpvf
9021 Works like C<sv_catpvf> but copies the text into the SV instead of
9022 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9028 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9032 PERL_ARGS_ASSERT_SV_SETPVF;
9034 va_start(args, pat);
9035 sv_vsetpvf(sv, pat, &args);
9040 =for apidoc sv_vsetpvf
9042 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9043 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9045 Usually used via its frontend C<sv_setpvf>.
9051 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9053 PERL_ARGS_ASSERT_SV_VSETPVF;
9055 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9059 =for apidoc sv_setpvf_mg
9061 Like C<sv_setpvf>, but also handles 'set' magic.
9067 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9071 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9073 va_start(args, pat);
9074 sv_vsetpvf_mg(sv, pat, &args);
9079 =for apidoc sv_vsetpvf_mg
9081 Like C<sv_vsetpvf>, but also handles 'set' magic.
9083 Usually used via its frontend C<sv_setpvf_mg>.
9089 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9091 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9093 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9097 #if defined(PERL_IMPLICIT_CONTEXT)
9099 /* pTHX_ magic can't cope with varargs, so this is a no-context
9100 * version of the main function, (which may itself be aliased to us).
9101 * Don't access this version directly.
9105 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9110 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9112 va_start(args, pat);
9113 sv_vcatpvf(sv, pat, &args);
9117 /* pTHX_ magic can't cope with varargs, so this is a no-context
9118 * version of the main function, (which may itself be aliased to us).
9119 * Don't access this version directly.
9123 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9128 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9130 va_start(args, pat);
9131 sv_vcatpvf_mg(sv, pat, &args);
9137 =for apidoc sv_catpvf
9139 Processes its arguments like C<sprintf> and appends the formatted
9140 output to an SV. If the appended data contains "wide" characters
9141 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9142 and characters >255 formatted with %c), the original SV might get
9143 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9144 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9145 valid UTF-8; if the original SV was bytes, the pattern should be too.
9150 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9154 PERL_ARGS_ASSERT_SV_CATPVF;
9156 va_start(args, pat);
9157 sv_vcatpvf(sv, pat, &args);
9162 =for apidoc sv_vcatpvf
9164 Processes its arguments like C<vsprintf> and appends the formatted output
9165 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9167 Usually used via its frontend C<sv_catpvf>.
9173 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9175 PERL_ARGS_ASSERT_SV_VCATPVF;
9177 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9181 =for apidoc sv_catpvf_mg
9183 Like C<sv_catpvf>, but also handles 'set' magic.
9189 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9193 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9195 va_start(args, pat);
9196 sv_vcatpvf_mg(sv, pat, &args);
9201 =for apidoc sv_vcatpvf_mg
9203 Like C<sv_vcatpvf>, but also handles 'set' magic.
9205 Usually used via its frontend C<sv_catpvf_mg>.
9211 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9213 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9215 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9220 =for apidoc sv_vsetpvfn
9222 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9225 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9231 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9232 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9234 PERL_ARGS_ASSERT_SV_VSETPVFN;
9237 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9242 * Warn of missing argument to sprintf, and then return a defined value
9243 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9245 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9247 S_vcatpvfn_missing_argument(pTHX) {
9248 if (ckWARN(WARN_MISSING)) {
9249 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9250 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9257 S_expect_number(pTHX_ char **const pattern)
9262 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9264 switch (**pattern) {
9265 case '1': case '2': case '3':
9266 case '4': case '5': case '6':
9267 case '7': case '8': case '9':
9268 var = *(*pattern)++ - '0';
9269 while (isDIGIT(**pattern)) {
9270 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9272 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9280 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9282 const int neg = nv < 0;
9285 PERL_ARGS_ASSERT_F0CONVERT;
9293 if (uv & 1 && uv == nv)
9294 uv--; /* Round to even */
9296 const unsigned dig = uv % 10;
9309 =for apidoc sv_vcatpvfn
9311 Processes its arguments like C<vsprintf> and appends the formatted output
9312 to an SV. Uses an array of SVs if the C style variable argument list is
9313 missing (NULL). When running with taint checks enabled, indicates via
9314 C<maybe_tainted> if results are untrustworthy (often due to the use of
9317 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9323 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9324 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9325 vec_utf8 = DO_UTF8(vecsv);
9327 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9330 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9331 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9339 static const char nullstr[] = "(null)";
9341 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9342 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9344 /* Times 4: a decimal digit takes more than 3 binary digits.
9345 * NV_DIG: mantissa takes than many decimal digits.
9346 * Plus 32: Playing safe. */
9347 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9348 /* large enough for "%#.#f" --chip */
9349 /* what about long double NVs? --jhi */
9351 PERL_ARGS_ASSERT_SV_VCATPVFN;
9352 PERL_UNUSED_ARG(maybe_tainted);
9354 /* no matter what, this is a string now */
9355 (void)SvPV_force(sv, origlen);
9357 /* special-case "", "%s", and "%-p" (SVf - see below) */
9360 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9362 const char * const s = va_arg(*args, char*);
9363 sv_catpv(sv, s ? s : nullstr);
9365 else if (svix < svmax) {
9366 sv_catsv(sv, *svargs);
9369 S_vcatpvfn_missing_argument(aTHX);
9372 if (args && patlen == 3 && pat[0] == '%' &&
9373 pat[1] == '-' && pat[2] == 'p') {
9374 argsv = MUTABLE_SV(va_arg(*args, void*));
9375 sv_catsv(sv, argsv);
9379 #ifndef USE_LONG_DOUBLE
9380 /* special-case "%.<number>[gf]" */
9381 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9382 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9383 unsigned digits = 0;
9387 while (*pp >= '0' && *pp <= '9')
9388 digits = 10 * digits + (*pp++ - '0');
9389 if (pp - pat == (int)patlen - 1) {
9395 S_vcatpvfn_missing_argument(aTHX);
9399 /* Add check for digits != 0 because it seems that some
9400 gconverts are buggy in this case, and we don't yet have
9401 a Configure test for this. */
9402 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9403 /* 0, point, slack */
9404 Gconvert(nv, (int)digits, 0, ebuf);
9406 if (*ebuf) /* May return an empty string for digits==0 */
9409 } else if (!digits) {
9412 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9413 sv_catpvn(sv, p, l);
9419 #endif /* !USE_LONG_DOUBLE */
9421 if (!args && svix < svmax && DO_UTF8(*svargs))
9424 patend = (char*)pat + patlen;
9425 for (p = (char*)pat; p < patend; p = q) {
9428 bool vectorize = FALSE;
9429 bool vectorarg = FALSE;
9430 bool vec_utf8 = FALSE;
9436 bool has_precis = FALSE;
9438 const I32 osvix = svix;
9439 bool is_utf8 = FALSE; /* is this item utf8? */
9440 #ifdef HAS_LDBL_SPRINTF_BUG
9441 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9442 with sfio - Allen <allens@cpan.org> */
9443 bool fix_ldbl_sprintf_bug = FALSE;
9447 U8 utf8buf[UTF8_MAXBYTES+1];
9448 STRLEN esignlen = 0;
9450 const char *eptr = NULL;
9451 const char *fmtstart;
9454 const U8 *vecstr = NULL;
9461 /* we need a long double target in case HAS_LONG_DOUBLE but
9464 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9472 const char *dotstr = ".";
9473 STRLEN dotstrlen = 1;
9474 I32 efix = 0; /* explicit format parameter index */
9475 I32 ewix = 0; /* explicit width index */
9476 I32 epix = 0; /* explicit precision index */
9477 I32 evix = 0; /* explicit vector index */
9478 bool asterisk = FALSE;
9480 /* echo everything up to the next format specification */
9481 for (q = p; q < patend && *q != '%'; ++q) ;
9483 if (has_utf8 && !pat_utf8)
9484 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9486 sv_catpvn(sv, p, q - p);
9495 We allow format specification elements in this order:
9496 \d+\$ explicit format parameter index
9498 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9499 0 flag (as above): repeated to allow "v02"
9500 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9501 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9503 [%bcdefginopsuxDFOUX] format (mandatory)
9508 As of perl5.9.3, printf format checking is on by default.
9509 Internally, perl uses %p formats to provide an escape to
9510 some extended formatting. This block deals with those
9511 extensions: if it does not match, (char*)q is reset and
9512 the normal format processing code is used.
9514 Currently defined extensions are:
9515 %p include pointer address (standard)
9516 %-p (SVf) include an SV (previously %_)
9517 %-<num>p include an SV with precision <num>
9518 %<num>p reserved for future extensions
9520 Robin Barker 2005-07-14
9522 %1p (VDf) removed. RMB 2007-10-19
9529 n = expect_number(&q);
9536 argsv = MUTABLE_SV(va_arg(*args, void*));
9537 eptr = SvPV_const(argsv, elen);
9543 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9544 "internal %%<num>p might conflict with future printf extensions");
9550 if ( (width = expect_number(&q)) ) {
9565 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9594 if ( (ewix = expect_number(&q)) )
9603 if ((vectorarg = asterisk)) {
9616 width = expect_number(&q);
9622 vecsv = va_arg(*args, SV*);
9624 vecsv = (evix > 0 && evix <= svmax)
9625 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
9627 vecsv = svix < svmax
9628 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9630 dotstr = SvPV_const(vecsv, dotstrlen);
9631 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9632 bad with tied or overloaded values that return UTF8. */
9635 else if (has_utf8) {
9636 vecsv = sv_mortalcopy(vecsv);
9637 sv_utf8_upgrade(vecsv);
9638 dotstr = SvPV_const(vecsv, dotstrlen);
9645 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9646 vecsv = svargs[efix ? efix-1 : svix++];
9647 vecstr = (U8*)SvPV_const(vecsv,veclen);
9648 vec_utf8 = DO_UTF8(vecsv);
9650 /* if this is a version object, we need to convert
9651 * back into v-string notation and then let the
9652 * vectorize happen normally
9654 if (sv_derived_from(vecsv, "version")) {
9655 char *version = savesvpv(vecsv);
9656 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9657 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9658 "vector argument not supported with alpha versions");
9661 vecsv = sv_newmortal();
9662 scan_vstring(version, version + veclen, vecsv);
9663 vecstr = (U8*)SvPV_const(vecsv, veclen);
9664 vec_utf8 = DO_UTF8(vecsv);
9676 i = va_arg(*args, int);
9678 i = (ewix ? ewix <= svmax : svix < svmax) ?
9679 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9681 width = (i < 0) ? -i : i;
9691 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9693 /* XXX: todo, support specified precision parameter */
9697 i = va_arg(*args, int);
9699 i = (ewix ? ewix <= svmax : svix < svmax)
9700 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9702 has_precis = !(i < 0);
9707 precis = precis * 10 + (*q++ - '0');
9716 case 'I': /* Ix, I32x, and I64x */
9718 if (q[1] == '6' && q[2] == '4') {
9724 if (q[1] == '3' && q[2] == '2') {
9734 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9745 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9746 if (*(q + 1) == 'l') { /* lld, llf */
9772 if (!vectorize && !args) {
9774 const I32 i = efix-1;
9775 argsv = (i >= 0 && i < svmax)
9776 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
9778 argsv = (svix >= 0 && svix < svmax)
9779 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9790 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9792 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9794 eptr = (char*)utf8buf;
9795 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9809 eptr = va_arg(*args, char*);
9811 elen = strlen(eptr);
9813 eptr = (char *)nullstr;
9814 elen = sizeof nullstr - 1;
9818 eptr = SvPV_const(argsv, elen);
9819 if (DO_UTF8(argsv)) {
9820 STRLEN old_precis = precis;
9821 if (has_precis && precis < elen) {
9822 STRLEN ulen = sv_len_utf8(argsv);
9823 I32 p = precis > ulen ? ulen : precis;
9824 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9827 if (width) { /* fudge width (can't fudge elen) */
9828 if (has_precis && precis < elen)
9829 width += precis - old_precis;
9831 width += elen - sv_len_utf8(argsv);
9838 if (has_precis && precis < elen)
9845 if (alt || vectorize)
9847 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9868 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9877 esignbuf[esignlen++] = plus;
9881 case 'h': iv = (short)va_arg(*args, int); break;
9882 case 'l': iv = va_arg(*args, long); break;
9883 case 'V': iv = va_arg(*args, IV); break;
9884 default: iv = va_arg(*args, int); break;
9887 iv = va_arg(*args, Quad_t); break;
9894 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9896 case 'h': iv = (short)tiv; break;
9897 case 'l': iv = (long)tiv; break;
9899 default: iv = tiv; break;
9902 iv = (Quad_t)tiv; break;
9908 if ( !vectorize ) /* we already set uv above */
9913 esignbuf[esignlen++] = plus;
9917 esignbuf[esignlen++] = '-';
9961 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9972 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9973 case 'l': uv = va_arg(*args, unsigned long); break;
9974 case 'V': uv = va_arg(*args, UV); break;
9975 default: uv = va_arg(*args, unsigned); break;
9978 uv = va_arg(*args, Uquad_t); break;
9985 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9987 case 'h': uv = (unsigned short)tuv; break;
9988 case 'l': uv = (unsigned long)tuv; break;
9990 default: uv = tuv; break;
9993 uv = (Uquad_t)tuv; break;
10002 char *ptr = ebuf + sizeof ebuf;
10003 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10009 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10013 } while (uv >>= 4);
10015 esignbuf[esignlen++] = '0';
10016 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10022 *--ptr = '0' + dig;
10023 } while (uv >>= 3);
10024 if (alt && *ptr != '0')
10030 *--ptr = '0' + dig;
10031 } while (uv >>= 1);
10033 esignbuf[esignlen++] = '0';
10034 esignbuf[esignlen++] = c;
10037 default: /* it had better be ten or less */
10040 *--ptr = '0' + dig;
10041 } while (uv /= base);
10044 elen = (ebuf + sizeof ebuf) - ptr;
10048 zeros = precis - elen;
10049 else if (precis == 0 && elen == 1 && *eptr == '0'
10050 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10053 /* a precision nullifies the 0 flag. */
10060 /* FLOATING POINT */
10063 c = 'f'; /* maybe %F isn't supported here */
10065 case 'e': case 'E':
10067 case 'g': case 'G':
10071 /* This is evil, but floating point is even more evil */
10073 /* for SV-style calling, we can only get NV
10074 for C-style calling, we assume %f is double;
10075 for simplicity we allow any of %Lf, %llf, %qf for long double
10079 #if defined(USE_LONG_DOUBLE)
10083 /* [perl #20339] - we should accept and ignore %lf rather than die */
10087 #if defined(USE_LONG_DOUBLE)
10088 intsize = args ? 0 : 'q';
10092 #if defined(HAS_LONG_DOUBLE)
10101 /* now we need (long double) if intsize == 'q', else (double) */
10103 #if LONG_DOUBLESIZE > DOUBLESIZE
10105 va_arg(*args, long double) :
10106 va_arg(*args, double)
10108 va_arg(*args, double)
10113 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10114 else. frexp() has some unspecified behaviour for those three */
10115 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10117 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10118 will cast our (long double) to (double) */
10119 (void)Perl_frexp(nv, &i);
10120 if (i == PERL_INT_MIN)
10121 Perl_die(aTHX_ "panic: frexp");
10123 need = BIT_DIGITS(i);
10125 need += has_precis ? precis : 6; /* known default */
10130 #ifdef HAS_LDBL_SPRINTF_BUG
10131 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10132 with sfio - Allen <allens@cpan.org> */
10135 # define MY_DBL_MAX DBL_MAX
10136 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10137 # if DOUBLESIZE >= 8
10138 # define MY_DBL_MAX 1.7976931348623157E+308L
10140 # define MY_DBL_MAX 3.40282347E+38L
10144 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10145 # define MY_DBL_MAX_BUG 1L
10147 # define MY_DBL_MAX_BUG MY_DBL_MAX
10151 # define MY_DBL_MIN DBL_MIN
10152 # else /* XXX guessing! -Allen */
10153 # if DOUBLESIZE >= 8
10154 # define MY_DBL_MIN 2.2250738585072014E-308L
10156 # define MY_DBL_MIN 1.17549435E-38L
10160 if ((intsize == 'q') && (c == 'f') &&
10161 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10162 (need < DBL_DIG)) {
10163 /* it's going to be short enough that
10164 * long double precision is not needed */
10166 if ((nv <= 0L) && (nv >= -0L))
10167 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10169 /* would use Perl_fp_class as a double-check but not
10170 * functional on IRIX - see perl.h comments */
10172 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10173 /* It's within the range that a double can represent */
10174 #if defined(DBL_MAX) && !defined(DBL_MIN)
10175 if ((nv >= ((long double)1/DBL_MAX)) ||
10176 (nv <= (-(long double)1/DBL_MAX)))
10178 fix_ldbl_sprintf_bug = TRUE;
10181 if (fix_ldbl_sprintf_bug == TRUE) {
10191 # undef MY_DBL_MAX_BUG
10194 #endif /* HAS_LDBL_SPRINTF_BUG */
10196 need += 20; /* fudge factor */
10197 if (PL_efloatsize < need) {
10198 Safefree(PL_efloatbuf);
10199 PL_efloatsize = need + 20; /* more fudge */
10200 Newx(PL_efloatbuf, PL_efloatsize, char);
10201 PL_efloatbuf[0] = '\0';
10204 if ( !(width || left || plus || alt) && fill != '0'
10205 && has_precis && intsize != 'q' ) { /* Shortcuts */
10206 /* See earlier comment about buggy Gconvert when digits,
10208 if ( c == 'g' && precis) {
10209 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10210 /* May return an empty string for digits==0 */
10211 if (*PL_efloatbuf) {
10212 elen = strlen(PL_efloatbuf);
10213 goto float_converted;
10215 } else if ( c == 'f' && !precis) {
10216 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10221 char *ptr = ebuf + sizeof ebuf;
10224 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10225 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10226 if (intsize == 'q') {
10227 /* Copy the one or more characters in a long double
10228 * format before the 'base' ([efgEFG]) character to
10229 * the format string. */
10230 static char const prifldbl[] = PERL_PRIfldbl;
10231 char const *p = prifldbl + sizeof(prifldbl) - 3;
10232 while (p >= prifldbl) { *--ptr = *p--; }
10237 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10242 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10254 /* No taint. Otherwise we are in the strange situation
10255 * where printf() taints but print($float) doesn't.
10257 #if defined(HAS_LONG_DOUBLE)
10258 elen = ((intsize == 'q')
10259 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10260 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10262 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10266 eptr = PL_efloatbuf;
10274 i = SvCUR(sv) - origlen;
10277 case 'h': *(va_arg(*args, short*)) = i; break;
10278 default: *(va_arg(*args, int*)) = i; break;
10279 case 'l': *(va_arg(*args, long*)) = i; break;
10280 case 'V': *(va_arg(*args, IV*)) = i; break;
10283 *(va_arg(*args, Quad_t*)) = i; break;
10290 sv_setuv_mg(argsv, (UV)i);
10291 continue; /* not "break" */
10298 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10299 && ckWARN(WARN_PRINTF))
10301 SV * const msg = sv_newmortal();
10302 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10303 (PL_op->op_type == OP_PRTF) ? "" : "s");
10304 if (fmtstart < patend) {
10305 const char * const fmtend = q < patend ? q : patend;
10307 sv_catpvs(msg, "\"%");
10308 for (f = fmtstart; f < fmtend; f++) {
10310 sv_catpvn(msg, f, 1);
10312 Perl_sv_catpvf(aTHX_ msg,
10313 "\\%03"UVof, (UV)*f & 0xFF);
10316 sv_catpvs(msg, "\"");
10318 sv_catpvs(msg, "end of string");
10320 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10323 /* output mangled stuff ... */
10329 /* ... right here, because formatting flags should not apply */
10330 SvGROW(sv, SvCUR(sv) + elen + 1);
10332 Copy(eptr, p, elen, char);
10335 SvCUR_set(sv, p - SvPVX_const(sv));
10337 continue; /* not "break" */
10340 if (is_utf8 != has_utf8) {
10343 sv_utf8_upgrade(sv);
10346 const STRLEN old_elen = elen;
10347 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10348 sv_utf8_upgrade(nsv);
10349 eptr = SvPVX_const(nsv);
10352 if (width) { /* fudge width (can't fudge elen) */
10353 width += elen - old_elen;
10359 have = esignlen + zeros + elen;
10361 Perl_croak_nocontext("%s", PL_memory_wrap);
10363 need = (have > width ? have : width);
10366 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10367 Perl_croak_nocontext("%s", PL_memory_wrap);
10368 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10370 if (esignlen && fill == '0') {
10372 for (i = 0; i < (int)esignlen; i++)
10373 *p++ = esignbuf[i];
10375 if (gap && !left) {
10376 memset(p, fill, gap);
10379 if (esignlen && fill != '0') {
10381 for (i = 0; i < (int)esignlen; i++)
10382 *p++ = esignbuf[i];
10386 for (i = zeros; i; i--)
10390 Copy(eptr, p, elen, char);
10394 memset(p, ' ', gap);
10399 Copy(dotstr, p, dotstrlen, char);
10403 vectorize = FALSE; /* done iterating over vecstr */
10410 SvCUR_set(sv, p - SvPVX_const(sv));
10419 /* =========================================================================
10421 =head1 Cloning an interpreter
10423 All the macros and functions in this section are for the private use of
10424 the main function, perl_clone().
10426 The foo_dup() functions make an exact copy of an existing foo thingy.
10427 During the course of a cloning, a hash table is used to map old addresses
10428 to new addresses. The table is created and manipulated with the
10429 ptr_table_* functions.
10433 * =========================================================================*/
10436 #if defined(USE_ITHREADS)
10438 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10439 #ifndef GpREFCNT_inc
10440 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10444 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10445 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10446 If this changes, please unmerge ss_dup.
10447 Likewise, sv_dup_inc_multiple() relies on this fact. */
10448 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10449 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10450 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10451 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10452 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10453 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10454 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10455 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10456 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10457 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10458 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10459 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10460 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10461 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10463 /* clone a parser */
10466 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10470 PERL_ARGS_ASSERT_PARSER_DUP;
10475 /* look for it in the table first */
10476 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10480 /* create anew and remember what it is */
10481 Newxz(parser, 1, yy_parser);
10482 ptr_table_store(PL_ptr_table, proto, parser);
10484 parser->yyerrstatus = 0;
10485 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10487 /* XXX these not yet duped */
10488 parser->old_parser = NULL;
10489 parser->stack = NULL;
10491 parser->stack_size = 0;
10492 /* XXX parser->stack->state = 0; */
10494 /* XXX eventually, just Copy() most of the parser struct ? */
10496 parser->lex_brackets = proto->lex_brackets;
10497 parser->lex_casemods = proto->lex_casemods;
10498 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10499 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10500 parser->lex_casestack = savepvn(proto->lex_casestack,
10501 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10502 parser->lex_defer = proto->lex_defer;
10503 parser->lex_dojoin = proto->lex_dojoin;
10504 parser->lex_expect = proto->lex_expect;
10505 parser->lex_formbrack = proto->lex_formbrack;
10506 parser->lex_inpat = proto->lex_inpat;
10507 parser->lex_inwhat = proto->lex_inwhat;
10508 parser->lex_op = proto->lex_op;
10509 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10510 parser->lex_starts = proto->lex_starts;
10511 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10512 parser->multi_close = proto->multi_close;
10513 parser->multi_open = proto->multi_open;
10514 parser->multi_start = proto->multi_start;
10515 parser->multi_end = proto->multi_end;
10516 parser->pending_ident = proto->pending_ident;
10517 parser->preambled = proto->preambled;
10518 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10519 parser->linestr = sv_dup_inc(proto->linestr, param);
10520 parser->expect = proto->expect;
10521 parser->copline = proto->copline;
10522 parser->last_lop_op = proto->last_lop_op;
10523 parser->lex_state = proto->lex_state;
10524 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10525 /* rsfp_filters entries have fake IoDIRP() */
10526 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10527 parser->in_my = proto->in_my;
10528 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10529 parser->error_count = proto->error_count;
10532 parser->linestr = sv_dup_inc(proto->linestr, param);
10535 char * const ols = SvPVX(proto->linestr);
10536 char * const ls = SvPVX(parser->linestr);
10538 parser->bufptr = ls + (proto->bufptr >= ols ?
10539 proto->bufptr - ols : 0);
10540 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10541 proto->oldbufptr - ols : 0);
10542 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10543 proto->oldoldbufptr - ols : 0);
10544 parser->linestart = ls + (proto->linestart >= ols ?
10545 proto->linestart - ols : 0);
10546 parser->last_uni = ls + (proto->last_uni >= ols ?
10547 proto->last_uni - ols : 0);
10548 parser->last_lop = ls + (proto->last_lop >= ols ?
10549 proto->last_lop - ols : 0);
10551 parser->bufend = ls + SvCUR(parser->linestr);
10554 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10558 parser->endwhite = proto->endwhite;
10559 parser->faketokens = proto->faketokens;
10560 parser->lasttoke = proto->lasttoke;
10561 parser->nextwhite = proto->nextwhite;
10562 parser->realtokenstart = proto->realtokenstart;
10563 parser->skipwhite = proto->skipwhite;
10564 parser->thisclose = proto->thisclose;
10565 parser->thismad = proto->thismad;
10566 parser->thisopen = proto->thisopen;
10567 parser->thisstuff = proto->thisstuff;
10568 parser->thistoken = proto->thistoken;
10569 parser->thiswhite = proto->thiswhite;
10571 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10572 parser->curforce = proto->curforce;
10574 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10575 Copy(proto->nexttype, parser->nexttype, 5, I32);
10576 parser->nexttoke = proto->nexttoke;
10579 /* XXX should clone saved_curcop here, but we aren't passed
10580 * proto_perl; so do it in perl_clone_using instead */
10586 /* duplicate a file handle */
10589 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10593 PERL_ARGS_ASSERT_FP_DUP;
10594 PERL_UNUSED_ARG(type);
10597 return (PerlIO*)NULL;
10599 /* look for it in the table first */
10600 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10604 /* create anew and remember what it is */
10605 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10606 ptr_table_store(PL_ptr_table, fp, ret);
10610 /* duplicate a directory handle */
10613 Perl_dirp_dup(pTHX_ DIR *const dp)
10615 PERL_UNUSED_CONTEXT;
10622 /* duplicate a typeglob */
10625 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10629 PERL_ARGS_ASSERT_GP_DUP;
10633 /* look for it in the table first */
10634 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10638 /* create anew and remember what it is */
10640 ptr_table_store(PL_ptr_table, gp, ret);
10643 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10644 on Newxz() to do this for us. */
10645 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10646 ret->gp_io = io_dup_inc(gp->gp_io, param);
10647 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10648 ret->gp_av = av_dup_inc(gp->gp_av, param);
10649 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10650 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10651 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10652 ret->gp_cvgen = gp->gp_cvgen;
10653 ret->gp_line = gp->gp_line;
10654 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10658 /* duplicate a chain of magic */
10661 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10663 MAGIC *mgret = NULL;
10664 MAGIC **mgprev_p = &mgret;
10666 PERL_ARGS_ASSERT_MG_DUP;
10668 for (; mg; mg = mg->mg_moremagic) {
10670 Newx(nmg, 1, MAGIC);
10672 mgprev_p = &(nmg->mg_moremagic);
10674 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10675 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10676 from the original commit adding Perl_mg_dup() - revision 4538.
10677 Similarly there is the annotation "XXX random ptr?" next to the
10678 assignment to nmg->mg_ptr. */
10681 /* FIXME for plugins
10682 if (nmg->mg_type == PERL_MAGIC_qr) {
10683 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10687 if(nmg->mg_type == PERL_MAGIC_backref) {
10688 /* The backref AV has its reference count deliberately bumped by
10691 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10694 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10695 ? sv_dup_inc(nmg->mg_obj, param)
10696 : sv_dup(nmg->mg_obj, param);
10699 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10700 if (nmg->mg_len > 0) {
10701 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10702 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10703 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10705 AMT * const namtp = (AMT*)nmg->mg_ptr;
10706 sv_dup_inc_multiple((SV**)(namtp->table),
10707 (SV**)(namtp->table), NofAMmeth, param);
10710 else if (nmg->mg_len == HEf_SVKEY)
10711 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10713 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10714 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10720 #endif /* USE_ITHREADS */
10722 struct ptr_tbl_arena {
10723 struct ptr_tbl_arena *next;
10724 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
10727 /* create a new pointer-mapping table */
10730 Perl_ptr_table_new(pTHX)
10733 PERL_UNUSED_CONTEXT;
10735 Newx(tbl, 1, PTR_TBL_t);
10736 tbl->tbl_max = 511;
10737 tbl->tbl_items = 0;
10738 tbl->tbl_arena = NULL;
10739 tbl->tbl_arena_next = NULL;
10740 tbl->tbl_arena_end = NULL;
10741 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10745 #define PTR_TABLE_HASH(ptr) \
10746 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10748 /* map an existing pointer using a table */
10750 STATIC PTR_TBL_ENT_t *
10751 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10753 PTR_TBL_ENT_t *tblent;
10754 const UV hash = PTR_TABLE_HASH(sv);
10756 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10758 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10759 for (; tblent; tblent = tblent->next) {
10760 if (tblent->oldval == sv)
10767 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10769 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10771 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10772 PERL_UNUSED_CONTEXT;
10774 return tblent ? tblent->newval : NULL;
10777 /* add a new entry to a pointer-mapping table */
10780 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10782 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10784 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10785 PERL_UNUSED_CONTEXT;
10788 tblent->newval = newsv;
10790 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10792 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
10793 struct ptr_tbl_arena *new_arena;
10795 Newx(new_arena, 1, struct ptr_tbl_arena);
10796 new_arena->next = tbl->tbl_arena;
10797 tbl->tbl_arena = new_arena;
10798 tbl->tbl_arena_next = new_arena->array;
10799 tbl->tbl_arena_end = new_arena->array
10800 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
10803 tblent = tbl->tbl_arena_next++;
10805 tblent->oldval = oldsv;
10806 tblent->newval = newsv;
10807 tblent->next = tbl->tbl_ary[entry];
10808 tbl->tbl_ary[entry] = tblent;
10810 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10811 ptr_table_split(tbl);
10815 /* double the hash bucket size of an existing ptr table */
10818 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10820 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10821 const UV oldsize = tbl->tbl_max + 1;
10822 UV newsize = oldsize * 2;
10825 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10826 PERL_UNUSED_CONTEXT;
10828 Renew(ary, newsize, PTR_TBL_ENT_t*);
10829 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10830 tbl->tbl_max = --newsize;
10831 tbl->tbl_ary = ary;
10832 for (i=0; i < oldsize; i++, ary++) {
10833 PTR_TBL_ENT_t **curentp, **entp, *ent;
10836 curentp = ary + oldsize;
10837 for (entp = ary, ent = *ary; ent; ent = *entp) {
10838 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10840 ent->next = *curentp;
10850 /* remove all the entries from a ptr table */
10851 /* Deprecated - will be removed post 5.14 */
10854 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10856 if (tbl && tbl->tbl_items) {
10857 struct ptr_tbl_arena *arena = tbl->tbl_arena;
10859 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
10862 struct ptr_tbl_arena *next = arena->next;
10868 tbl->tbl_items = 0;
10869 tbl->tbl_arena = NULL;
10870 tbl->tbl_arena_next = NULL;
10871 tbl->tbl_arena_end = NULL;
10875 /* clear and free a ptr table */
10878 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10880 struct ptr_tbl_arena *arena;
10886 arena = tbl->tbl_arena;
10889 struct ptr_tbl_arena *next = arena->next;
10895 Safefree(tbl->tbl_ary);
10899 #if defined(USE_ITHREADS)
10902 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10904 PERL_ARGS_ASSERT_RVPV_DUP;
10907 SvRV_set(dstr, SvWEAKREF(sstr)
10908 ? sv_dup(SvRV_const(sstr), param)
10909 : sv_dup_inc(SvRV_const(sstr), param));
10912 else if (SvPVX_const(sstr)) {
10913 /* Has something there */
10915 /* Normal PV - clone whole allocated space */
10916 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10917 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10918 /* Not that normal - actually sstr is copy on write.
10919 But we are a true, independant SV, so: */
10920 SvREADONLY_off(dstr);
10925 /* Special case - not normally malloced for some reason */
10926 if (isGV_with_GP(sstr)) {
10927 /* Don't need to do anything here. */
10929 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10930 /* A "shared" PV - clone it as "shared" PV */
10932 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10936 /* Some other special case - random pointer */
10937 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10942 /* Copy the NULL */
10943 SvPV_set(dstr, NULL);
10947 /* duplicate a list of SVs. source and dest may point to the same memory. */
10949 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10950 SSize_t items, CLONE_PARAMS *const param)
10952 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10954 while (items-- > 0) {
10955 *dest++ = sv_dup_inc(*source++, param);
10961 /* duplicate an SV of any type (including AV, HV etc) */
10964 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10969 PERL_ARGS_ASSERT_SV_DUP;
10973 if (SvTYPE(sstr) == SVTYPEMASK) {
10974 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10979 /* look for it in the table first */
10980 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10984 if(param->flags & CLONEf_JOIN_IN) {
10985 /** We are joining here so we don't want do clone
10986 something that is bad **/
10987 if (SvTYPE(sstr) == SVt_PVHV) {
10988 const HEK * const hvname = HvNAME_HEK(sstr);
10990 /** don't clone stashes if they already exist **/
10991 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10995 /* create anew and remember what it is */
10998 #ifdef DEBUG_LEAKING_SCALARS
10999 dstr->sv_debug_optype = sstr->sv_debug_optype;
11000 dstr->sv_debug_line = sstr->sv_debug_line;
11001 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11002 dstr->sv_debug_cloned = 1;
11003 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11006 ptr_table_store(PL_ptr_table, sstr, dstr);
11009 SvFLAGS(dstr) = SvFLAGS(sstr);
11010 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11011 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11014 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11015 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11016 (void*)PL_watch_pvx, SvPVX_const(sstr));
11019 /* don't clone objects whose class has asked us not to */
11020 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11025 switch (SvTYPE(sstr)) {
11027 SvANY(dstr) = NULL;
11030 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11032 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11034 SvIV_set(dstr, SvIVX(sstr));
11038 SvANY(dstr) = new_XNV();
11039 SvNV_set(dstr, SvNVX(sstr));
11041 /* case SVt_BIND: */
11044 /* These are all the types that need complex bodies allocating. */
11046 const svtype sv_type = SvTYPE(sstr);
11047 const struct body_details *const sv_type_details
11048 = bodies_by_type + sv_type;
11052 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11067 assert(sv_type_details->body_size);
11068 if (sv_type_details->arena) {
11069 new_body_inline(new_body, sv_type);
11071 = (void*)((char*)new_body - sv_type_details->offset);
11073 new_body = new_NOARENA(sv_type_details);
11077 SvANY(dstr) = new_body;
11080 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11081 ((char*)SvANY(dstr)) + sv_type_details->offset,
11082 sv_type_details->copy, char);
11084 Copy(((char*)SvANY(sstr)),
11085 ((char*)SvANY(dstr)),
11086 sv_type_details->body_size + sv_type_details->offset, char);
11089 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11090 && !isGV_with_GP(dstr))
11091 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11093 /* The Copy above means that all the source (unduplicated) pointers
11094 are now in the destination. We can check the flags and the
11095 pointers in either, but it's possible that there's less cache
11096 missing by always going for the destination.
11097 FIXME - instrument and check that assumption */
11098 if (sv_type >= SVt_PVMG) {
11099 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11100 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11101 } else if (SvMAGIC(dstr))
11102 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11104 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11107 /* The cast silences a GCC warning about unhandled types. */
11108 switch ((int)sv_type) {
11118 /* FIXME for plugins */
11119 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11122 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11123 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11124 LvTARG(dstr) = dstr;
11125 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11126 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11128 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11130 if(isGV_with_GP(sstr)) {
11131 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11132 /* Don't call sv_add_backref here as it's going to be
11133 created as part of the magic cloning of the symbol
11134 table--unless this is during a join and the stash
11135 is not actually being cloned. */
11136 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11137 at the point of this comment. */
11138 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11139 if(param->flags & CLONEf_JOIN_IN) {
11140 const HEK * const hvname
11141 = HvNAME_HEK(GvSTASH(dstr));
11143 && GvSTASH(dstr) == gv_stashpvn(
11144 HEK_KEY(hvname), HEK_LEN(hvname), 0
11147 Perl_sv_add_backref(
11148 aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr
11151 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11152 (void)GpREFCNT_inc(GvGP(dstr));
11154 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11157 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11158 if (IoOFP(dstr) == IoIFP(sstr))
11159 IoOFP(dstr) = IoIFP(dstr);
11161 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11162 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11163 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11164 /* I have no idea why fake dirp (rsfps)
11165 should be treated differently but otherwise
11166 we end up with leaks -- sky*/
11167 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11168 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11169 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11171 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11172 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11173 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11174 if (IoDIRP(dstr)) {
11175 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11178 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11181 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11182 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11183 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11186 /* avoid cloning an empty array */
11187 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11188 SV **dst_ary, **src_ary;
11189 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11191 src_ary = AvARRAY((const AV *)sstr);
11192 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11193 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11194 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11195 AvALLOC((const AV *)dstr) = dst_ary;
11196 if (AvREAL((const AV *)sstr)) {
11197 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11201 while (items-- > 0)
11202 *dst_ary++ = sv_dup(*src_ary++, param);
11203 if (!(param->flags & CLONEf_COPY_STACKS)
11206 av_reify(MUTABLE_AV(dstr)); /* #41138 */
11209 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11210 while (items-- > 0) {
11211 *dst_ary++ = &PL_sv_undef;
11215 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11216 AvALLOC((const AV *)dstr) = (SV**)NULL;
11217 AvMAX( (const AV *)dstr) = -1;
11218 AvFILLp((const AV *)dstr) = -1;
11222 if (HvARRAY((const HV *)sstr)) {
11224 const bool sharekeys = !!HvSHAREKEYS(sstr);
11225 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11226 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11228 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11229 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11231 HvARRAY(dstr) = (HE**)darray;
11232 while (i <= sxhv->xhv_max) {
11233 const HE * const source = HvARRAY(sstr)[i];
11234 HvARRAY(dstr)[i] = source
11235 ? he_dup(source, sharekeys, param) : 0;
11240 const struct xpvhv_aux * const saux = HvAUX(sstr);
11241 struct xpvhv_aux * const daux = HvAUX(dstr);
11242 /* This flag isn't copied. */
11243 /* SvOOK_on(hv) attacks the IV flags. */
11244 SvFLAGS(dstr) |= SVf_OOK;
11246 hvname = saux->xhv_name;
11247 daux->xhv_name = hek_dup(hvname, param);
11249 daux->xhv_riter = saux->xhv_riter;
11250 daux->xhv_eiter = saux->xhv_eiter
11251 ? he_dup(saux->xhv_eiter,
11252 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11253 /* backref array needs refcnt=2; see sv_add_backref */
11254 daux->xhv_backreferences =
11255 saux->xhv_backreferences
11256 ? MUTABLE_AV(SvREFCNT_inc(
11257 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11260 daux->xhv_mro_meta = saux->xhv_mro_meta
11261 ? mro_meta_dup(saux->xhv_mro_meta, param)
11264 /* Record stashes for possible cloning in Perl_clone(). */
11266 av_push(param->stashes, dstr);
11270 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11273 if (!(param->flags & CLONEf_COPY_STACKS)) {
11277 /* NOTE: not refcounted */
11278 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11280 if (!CvISXSUB(dstr))
11281 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11283 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11284 CvXSUBANY(dstr).any_ptr =
11285 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11287 /* don't dup if copying back - CvGV isn't refcounted, so the
11288 * duped GV may never be freed. A bit of a hack! DAPM */
11289 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11290 NULL : gv_dup(CvGV(dstr), param) ;
11291 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11293 CvWEAKOUTSIDE(sstr)
11294 ? cv_dup( CvOUTSIDE(dstr), param)
11295 : cv_dup_inc(CvOUTSIDE(dstr), param);
11296 if (!CvISXSUB(dstr))
11297 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11303 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11309 /* duplicate a context */
11312 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11314 PERL_CONTEXT *ncxs;
11316 PERL_ARGS_ASSERT_CX_DUP;
11319 return (PERL_CONTEXT*)NULL;
11321 /* look for it in the table first */
11322 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11326 /* create anew and remember what it is */
11327 Newx(ncxs, max + 1, PERL_CONTEXT);
11328 ptr_table_store(PL_ptr_table, cxs, ncxs);
11329 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11332 PERL_CONTEXT * const ncx = &ncxs[ix];
11333 if (CxTYPE(ncx) == CXt_SUBST) {
11334 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11337 switch (CxTYPE(ncx)) {
11339 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11340 ? cv_dup_inc(ncx->blk_sub.cv, param)
11341 : cv_dup(ncx->blk_sub.cv,param));
11342 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11343 ? av_dup_inc(ncx->blk_sub.argarray,
11346 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11348 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11349 ncx->blk_sub.oldcomppad);
11352 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11354 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11356 case CXt_LOOP_LAZYSV:
11357 ncx->blk_loop.state_u.lazysv.end
11358 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11359 /* We are taking advantage of av_dup_inc and sv_dup_inc
11360 actually being the same function, and order equivalance of
11362 We can assert the later [but only at run time :-(] */
11363 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11364 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11366 ncx->blk_loop.state_u.ary.ary
11367 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11368 case CXt_LOOP_LAZYIV:
11369 case CXt_LOOP_PLAIN:
11370 if (CxPADLOOP(ncx)) {
11371 ncx->blk_loop.oldcomppad
11372 = (PAD*)ptr_table_fetch(PL_ptr_table,
11373 ncx->blk_loop.oldcomppad);
11375 ncx->blk_loop.oldcomppad
11376 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11381 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11382 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11383 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11396 /* duplicate a stack info structure */
11399 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11403 PERL_ARGS_ASSERT_SI_DUP;
11406 return (PERL_SI*)NULL;
11408 /* look for it in the table first */
11409 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11413 /* create anew and remember what it is */
11414 Newxz(nsi, 1, PERL_SI);
11415 ptr_table_store(PL_ptr_table, si, nsi);
11417 nsi->si_stack = av_dup_inc(si->si_stack, param);
11418 nsi->si_cxix = si->si_cxix;
11419 nsi->si_cxmax = si->si_cxmax;
11420 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11421 nsi->si_type = si->si_type;
11422 nsi->si_prev = si_dup(si->si_prev, param);
11423 nsi->si_next = si_dup(si->si_next, param);
11424 nsi->si_markoff = si->si_markoff;
11429 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11430 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11431 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11432 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11433 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11434 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11435 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
11436 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
11437 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11438 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11439 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11440 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11441 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11442 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11443 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11444 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11447 #define pv_dup_inc(p) SAVEPV(p)
11448 #define pv_dup(p) SAVEPV(p)
11449 #define svp_dup_inc(p,pp) any_dup(p,pp)
11451 /* map any object to the new equivent - either something in the
11452 * ptr table, or something in the interpreter structure
11456 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11460 PERL_ARGS_ASSERT_ANY_DUP;
11463 return (void*)NULL;
11465 /* look for it in the table first */
11466 ret = ptr_table_fetch(PL_ptr_table, v);
11470 /* see if it is part of the interpreter structure */
11471 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11472 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11480 /* duplicate the save stack */
11483 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11486 ANY * const ss = proto_perl->Isavestack;
11487 const I32 max = proto_perl->Isavestack_max;
11488 I32 ix = proto_perl->Isavestack_ix;
11501 void (*dptr) (void*);
11502 void (*dxptr) (pTHX_ void*);
11504 PERL_ARGS_ASSERT_SS_DUP;
11506 Newxz(nss, max, ANY);
11509 const UV uv = POPUV(ss,ix);
11510 const U8 type = (U8)uv & SAVE_MASK;
11512 TOPUV(nss,ix) = uv;
11514 case SAVEt_CLEARSV:
11516 case SAVEt_HELEM: /* hash element */
11517 sv = (const SV *)POPPTR(ss,ix);
11518 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11520 case SAVEt_ITEM: /* normal string */
11521 case SAVEt_SV: /* scalar reference */
11522 sv = (const SV *)POPPTR(ss,ix);
11523 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11526 case SAVEt_MORTALIZESV:
11527 sv = (const SV *)POPPTR(ss,ix);
11528 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11530 case SAVEt_SHARED_PVREF: /* char* in shared space */
11531 c = (char*)POPPTR(ss,ix);
11532 TOPPTR(nss,ix) = savesharedpv(c);
11533 ptr = POPPTR(ss,ix);
11534 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11536 case SAVEt_GENERIC_SVREF: /* generic sv */
11537 case SAVEt_SVREF: /* scalar reference */
11538 sv = (const SV *)POPPTR(ss,ix);
11539 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11540 ptr = POPPTR(ss,ix);
11541 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11543 case SAVEt_HV: /* hash reference */
11544 case SAVEt_AV: /* array reference */
11545 sv = (const SV *) POPPTR(ss,ix);
11546 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11548 case SAVEt_COMPPAD:
11550 sv = (const SV *) POPPTR(ss,ix);
11551 TOPPTR(nss,ix) = sv_dup(sv, param);
11553 case SAVEt_INT: /* int reference */
11554 ptr = POPPTR(ss,ix);
11555 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11556 intval = (int)POPINT(ss,ix);
11557 TOPINT(nss,ix) = intval;
11559 case SAVEt_LONG: /* long reference */
11560 ptr = POPPTR(ss,ix);
11561 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11562 longval = (long)POPLONG(ss,ix);
11563 TOPLONG(nss,ix) = longval;
11565 case SAVEt_I32: /* I32 reference */
11566 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11567 ptr = POPPTR(ss,ix);
11568 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11570 TOPINT(nss,ix) = i;
11572 case SAVEt_IV: /* IV reference */
11573 ptr = POPPTR(ss,ix);
11574 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11576 TOPIV(nss,ix) = iv;
11578 case SAVEt_HPTR: /* HV* reference */
11579 case SAVEt_APTR: /* AV* reference */
11580 case SAVEt_SPTR: /* SV* reference */
11581 ptr = POPPTR(ss,ix);
11582 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11583 sv = (const SV *)POPPTR(ss,ix);
11584 TOPPTR(nss,ix) = sv_dup(sv, param);
11586 case SAVEt_VPTR: /* random* reference */
11587 ptr = POPPTR(ss,ix);
11588 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11590 case SAVEt_INT_SMALL:
11591 case SAVEt_I32_SMALL:
11592 case SAVEt_I16: /* I16 reference */
11593 case SAVEt_I8: /* I8 reference */
11595 ptr = POPPTR(ss,ix);
11596 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11598 case SAVEt_GENERIC_PVREF: /* generic char* */
11599 case SAVEt_PPTR: /* char* reference */
11600 ptr = POPPTR(ss,ix);
11601 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11602 c = (char*)POPPTR(ss,ix);
11603 TOPPTR(nss,ix) = pv_dup(c);
11605 case SAVEt_GP: /* scalar reference */
11606 gv = (const GV *)POPPTR(ss,ix);
11607 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11608 gp = (GP*)POPPTR(ss,ix);
11609 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11610 (void)GpREFCNT_inc(gp);
11612 TOPINT(nss,ix) = i;
11615 ptr = POPPTR(ss,ix);
11616 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11617 /* these are assumed to be refcounted properly */
11619 switch (((OP*)ptr)->op_type) {
11621 case OP_LEAVESUBLV:
11625 case OP_LEAVEWRITE:
11626 TOPPTR(nss,ix) = ptr;
11629 (void) OpREFCNT_inc(o);
11633 TOPPTR(nss,ix) = NULL;
11638 TOPPTR(nss,ix) = NULL;
11641 hv = (const HV *)POPPTR(ss,ix);
11642 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11644 TOPINT(nss,ix) = i;
11647 c = (char*)POPPTR(ss,ix);
11648 TOPPTR(nss,ix) = pv_dup_inc(c);
11650 case SAVEt_STACK_POS: /* Position on Perl stack */
11652 TOPINT(nss,ix) = i;
11654 case SAVEt_DESTRUCTOR:
11655 ptr = POPPTR(ss,ix);
11656 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11657 dptr = POPDPTR(ss,ix);
11658 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11659 any_dup(FPTR2DPTR(void *, dptr),
11662 case SAVEt_DESTRUCTOR_X:
11663 ptr = POPPTR(ss,ix);
11664 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11665 dxptr = POPDXPTR(ss,ix);
11666 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11667 any_dup(FPTR2DPTR(void *, dxptr),
11670 case SAVEt_REGCONTEXT:
11672 ix -= uv >> SAVE_TIGHT_SHIFT;
11674 case SAVEt_AELEM: /* array element */
11675 sv = (const SV *)POPPTR(ss,ix);
11676 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11678 TOPINT(nss,ix) = i;
11679 av = (const AV *)POPPTR(ss,ix);
11680 TOPPTR(nss,ix) = av_dup_inc(av, param);
11683 ptr = POPPTR(ss,ix);
11684 TOPPTR(nss,ix) = ptr;
11687 ptr = POPPTR(ss,ix);
11690 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11691 HINTS_REFCNT_UNLOCK;
11693 TOPPTR(nss,ix) = ptr;
11695 TOPINT(nss,ix) = i;
11696 if (i & HINT_LOCALIZE_HH) {
11697 hv = (const HV *)POPPTR(ss,ix);
11698 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11701 case SAVEt_PADSV_AND_MORTALIZE:
11702 longval = (long)POPLONG(ss,ix);
11703 TOPLONG(nss,ix) = longval;
11704 ptr = POPPTR(ss,ix);
11705 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11706 sv = (const SV *)POPPTR(ss,ix);
11707 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11709 case SAVEt_SET_SVFLAGS:
11711 TOPINT(nss,ix) = i;
11713 TOPINT(nss,ix) = i;
11714 sv = (const SV *)POPPTR(ss,ix);
11715 TOPPTR(nss,ix) = sv_dup(sv, param);
11717 case SAVEt_RE_STATE:
11719 const struct re_save_state *const old_state
11720 = (struct re_save_state *)
11721 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11722 struct re_save_state *const new_state
11723 = (struct re_save_state *)
11724 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11726 Copy(old_state, new_state, 1, struct re_save_state);
11727 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11729 new_state->re_state_bostr
11730 = pv_dup(old_state->re_state_bostr);
11731 new_state->re_state_reginput
11732 = pv_dup(old_state->re_state_reginput);
11733 new_state->re_state_regeol
11734 = pv_dup(old_state->re_state_regeol);
11735 new_state->re_state_regoffs
11736 = (regexp_paren_pair*)
11737 any_dup(old_state->re_state_regoffs, proto_perl);
11738 new_state->re_state_reglastparen
11739 = (U32*) any_dup(old_state->re_state_reglastparen,
11741 new_state->re_state_reglastcloseparen
11742 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11744 /* XXX This just has to be broken. The old save_re_context
11745 code did SAVEGENERICPV(PL_reg_start_tmp);
11746 PL_reg_start_tmp is char **.
11747 Look above to what the dup code does for
11748 SAVEt_GENERIC_PVREF
11749 It can never have worked.
11750 So this is merely a faithful copy of the exiting bug: */
11751 new_state->re_state_reg_start_tmp
11752 = (char **) pv_dup((char *)
11753 old_state->re_state_reg_start_tmp);
11754 /* I assume that it only ever "worked" because no-one called
11755 (pseudo)fork while the regexp engine had re-entered itself.
11757 #ifdef PERL_OLD_COPY_ON_WRITE
11758 new_state->re_state_nrs
11759 = sv_dup(old_state->re_state_nrs, param);
11761 new_state->re_state_reg_magic
11762 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11764 new_state->re_state_reg_oldcurpm
11765 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11767 new_state->re_state_reg_curpm
11768 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11770 new_state->re_state_reg_oldsaved
11771 = pv_dup(old_state->re_state_reg_oldsaved);
11772 new_state->re_state_reg_poscache
11773 = pv_dup(old_state->re_state_reg_poscache);
11774 new_state->re_state_reg_starttry
11775 = pv_dup(old_state->re_state_reg_starttry);
11778 case SAVEt_COMPILE_WARNINGS:
11779 ptr = POPPTR(ss,ix);
11780 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11783 ptr = POPPTR(ss,ix);
11784 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11788 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11796 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11797 * flag to the result. This is done for each stash before cloning starts,
11798 * so we know which stashes want their objects cloned */
11801 do_mark_cloneable_stash(pTHX_ SV *const sv)
11803 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11805 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11806 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11807 if (cloner && GvCV(cloner)) {
11814 mXPUSHs(newSVhek(hvname));
11816 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11823 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11831 =for apidoc perl_clone
11833 Create and return a new interpreter by cloning the current one.
11835 perl_clone takes these flags as parameters:
11837 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11838 without it we only clone the data and zero the stacks,
11839 with it we copy the stacks and the new perl interpreter is
11840 ready to run at the exact same point as the previous one.
11841 The pseudo-fork code uses COPY_STACKS while the
11842 threads->create doesn't.
11844 CLONEf_KEEP_PTR_TABLE
11845 perl_clone keeps a ptr_table with the pointer of the old
11846 variable as a key and the new variable as a value,
11847 this allows it to check if something has been cloned and not
11848 clone it again but rather just use the value and increase the
11849 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11850 the ptr_table using the function
11851 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11852 reason to keep it around is if you want to dup some of your own
11853 variable who are outside the graph perl scans, example of this
11854 code is in threads.xs create
11857 This is a win32 thing, it is ignored on unix, it tells perls
11858 win32host code (which is c++) to clone itself, this is needed on
11859 win32 if you want to run two threads at the same time,
11860 if you just want to do some stuff in a separate perl interpreter
11861 and then throw it away and return to the original one,
11862 you don't need to do anything.
11867 /* XXX the above needs expanding by someone who actually understands it ! */
11868 EXTERN_C PerlInterpreter *
11869 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11872 perl_clone(PerlInterpreter *proto_perl, UV flags)
11875 #ifdef PERL_IMPLICIT_SYS
11877 PERL_ARGS_ASSERT_PERL_CLONE;
11879 /* perlhost.h so we need to call into it
11880 to clone the host, CPerlHost should have a c interface, sky */
11882 if (flags & CLONEf_CLONE_HOST) {
11883 return perl_clone_host(proto_perl,flags);
11885 return perl_clone_using(proto_perl, flags,
11887 proto_perl->IMemShared,
11888 proto_perl->IMemParse,
11890 proto_perl->IStdIO,
11894 proto_perl->IProc);
11898 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11899 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11900 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11901 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11902 struct IPerlDir* ipD, struct IPerlSock* ipS,
11903 struct IPerlProc* ipP)
11905 /* XXX many of the string copies here can be optimized if they're
11906 * constants; they need to be allocated as common memory and just
11907 * their pointers copied. */
11910 CLONE_PARAMS clone_params;
11911 CLONE_PARAMS* const param = &clone_params;
11913 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11915 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11916 #else /* !PERL_IMPLICIT_SYS */
11918 CLONE_PARAMS clone_params;
11919 CLONE_PARAMS* param = &clone_params;
11920 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11922 PERL_ARGS_ASSERT_PERL_CLONE;
11923 #endif /* PERL_IMPLICIT_SYS */
11925 /* for each stash, determine whether its objects should be cloned */
11926 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11927 PERL_SET_THX(my_perl);
11930 PoisonNew(my_perl, 1, PerlInterpreter);
11935 PL_scopestack_name = 0;
11937 PL_savestack_ix = 0;
11938 PL_savestack_max = -1;
11939 PL_sig_pending = 0;
11941 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11942 # ifdef DEBUG_LEAKING_SCALARS
11943 PL_sv_serial = (((U32)my_perl >> 2) & 0xfff) * 1000000;
11945 #else /* !DEBUGGING */
11946 Zero(my_perl, 1, PerlInterpreter);
11947 #endif /* DEBUGGING */
11949 #ifdef PERL_IMPLICIT_SYS
11950 /* host pointers */
11952 PL_MemShared = ipMS;
11953 PL_MemParse = ipMP;
11960 #endif /* PERL_IMPLICIT_SYS */
11962 param->flags = flags;
11963 param->proto_perl = proto_perl;
11965 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11967 PL_body_arenas = NULL;
11968 Zero(&PL_body_roots, 1, PL_body_roots);
11970 PL_nice_chunk = NULL;
11971 PL_nice_chunk_size = 0;
11973 PL_sv_objcount = 0;
11975 PL_sv_arenaroot = NULL;
11977 PL_debug = proto_perl->Idebug;
11979 PL_hash_seed = proto_perl->Ihash_seed;
11980 PL_rehash_seed = proto_perl->Irehash_seed;
11982 #ifdef USE_REENTRANT_API
11983 /* XXX: things like -Dm will segfault here in perlio, but doing
11984 * PERL_SET_CONTEXT(proto_perl);
11985 * breaks too many other things
11987 Perl_reentrant_init(aTHX);
11990 /* create SV map for pointer relocation */
11991 PL_ptr_table = ptr_table_new();
11993 /* initialize these special pointers as early as possible */
11994 SvANY(&PL_sv_undef) = NULL;
11995 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11996 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11997 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11999 SvANY(&PL_sv_no) = new_XPVNV();
12000 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12001 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12002 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12003 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12004 SvCUR_set(&PL_sv_no, 0);
12005 SvLEN_set(&PL_sv_no, 1);
12006 SvIV_set(&PL_sv_no, 0);
12007 SvNV_set(&PL_sv_no, 0);
12008 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12010 SvANY(&PL_sv_yes) = new_XPVNV();
12011 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12012 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12013 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12014 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12015 SvCUR_set(&PL_sv_yes, 1);
12016 SvLEN_set(&PL_sv_yes, 2);
12017 SvIV_set(&PL_sv_yes, 1);
12018 SvNV_set(&PL_sv_yes, 1);
12019 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12021 /* dbargs array probably holds garbage */
12024 /* create (a non-shared!) shared string table */
12025 PL_strtab = newHV();
12026 HvSHAREKEYS_off(PL_strtab);
12027 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12028 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12030 PL_compiling = proto_perl->Icompiling;
12032 /* These two PVs will be free'd special way so must set them same way op.c does */
12033 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12034 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12036 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12037 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12039 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12040 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12041 if (PL_compiling.cop_hints_hash) {
12043 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
12044 HINTS_REFCNT_UNLOCK;
12046 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12047 #ifdef PERL_DEBUG_READONLY_OPS
12052 /* pseudo environmental stuff */
12053 PL_origargc = proto_perl->Iorigargc;
12054 PL_origargv = proto_perl->Iorigargv;
12056 param->stashes = newAV(); /* Setup array of objects to call clone on */
12058 /* Set tainting stuff before PerlIO_debug can possibly get called */
12059 PL_tainting = proto_perl->Itainting;
12060 PL_taint_warn = proto_perl->Itaint_warn;
12062 #ifdef PERLIO_LAYERS
12063 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12064 PerlIO_clone(aTHX_ proto_perl, param);
12067 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12068 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12069 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12070 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12071 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12072 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12075 PL_minus_c = proto_perl->Iminus_c;
12076 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12077 PL_localpatches = proto_perl->Ilocalpatches;
12078 PL_splitstr = proto_perl->Isplitstr;
12079 PL_minus_n = proto_perl->Iminus_n;
12080 PL_minus_p = proto_perl->Iminus_p;
12081 PL_minus_l = proto_perl->Iminus_l;
12082 PL_minus_a = proto_perl->Iminus_a;
12083 PL_minus_E = proto_perl->Iminus_E;
12084 PL_minus_F = proto_perl->Iminus_F;
12085 PL_doswitches = proto_perl->Idoswitches;
12086 PL_dowarn = proto_perl->Idowarn;
12087 PL_doextract = proto_perl->Idoextract;
12088 PL_sawampersand = proto_perl->Isawampersand;
12089 PL_unsafe = proto_perl->Iunsafe;
12090 PL_inplace = SAVEPV(proto_perl->Iinplace);
12091 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12092 PL_perldb = proto_perl->Iperldb;
12093 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12094 PL_exit_flags = proto_perl->Iexit_flags;
12096 /* magical thingies */
12097 /* XXX time(&PL_basetime) when asked for? */
12098 PL_basetime = proto_perl->Ibasetime;
12099 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12101 PL_maxsysfd = proto_perl->Imaxsysfd;
12102 PL_statusvalue = proto_perl->Istatusvalue;
12104 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12106 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12108 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12110 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12111 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12112 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12115 /* RE engine related */
12116 Zero(&PL_reg_state, 1, struct re_save_state);
12117 PL_reginterp_cnt = 0;
12118 PL_regmatch_slab = NULL;
12120 /* Clone the regex array */
12121 /* ORANGE FIXME for plugins, probably in the SV dup code.
12122 newSViv(PTR2IV(CALLREGDUPE(
12123 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12125 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12126 PL_regex_pad = AvARRAY(PL_regex_padav);
12128 /* shortcuts to various I/O objects */
12129 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
12130 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12131 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12132 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12133 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12134 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12135 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12137 /* shortcuts to regexp stuff */
12138 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12140 /* shortcuts to misc objects */
12141 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12143 /* shortcuts to debugging objects */
12144 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12145 PL_DBline = gv_dup(proto_perl->IDBline, param);
12146 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12147 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12148 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12149 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12151 /* symbol tables */
12152 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12153 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12154 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12155 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12156 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12158 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12159 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12160 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12161 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12162 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12163 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12164 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12165 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12167 PL_sub_generation = proto_perl->Isub_generation;
12168 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12170 /* funky return mechanisms */
12171 PL_forkprocess = proto_perl->Iforkprocess;
12173 /* subprocess state */
12174 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12176 /* internal state */
12177 PL_maxo = proto_perl->Imaxo;
12178 if (proto_perl->Iop_mask)
12179 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12182 /* PL_asserting = proto_perl->Iasserting; */
12184 /* current interpreter roots */
12185 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12187 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12189 PL_main_start = proto_perl->Imain_start;
12190 PL_eval_root = proto_perl->Ieval_root;
12191 PL_eval_start = proto_perl->Ieval_start;
12193 /* runtime control stuff */
12194 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12196 PL_filemode = proto_perl->Ifilemode;
12197 PL_lastfd = proto_perl->Ilastfd;
12198 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12201 PL_gensym = proto_perl->Igensym;
12202 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12203 PL_laststatval = proto_perl->Ilaststatval;
12204 PL_laststype = proto_perl->Ilaststype;
12207 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12209 /* interpreter atexit processing */
12210 PL_exitlistlen = proto_perl->Iexitlistlen;
12211 if (PL_exitlistlen) {
12212 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12213 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12216 PL_exitlist = (PerlExitListEntry*)NULL;
12218 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12219 if (PL_my_cxt_size) {
12220 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12221 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12222 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12223 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12224 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12228 PL_my_cxt_list = (void**)NULL;
12229 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12230 PL_my_cxt_keys = (const char**)NULL;
12233 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12234 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12235 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12237 PL_profiledata = NULL;
12239 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12241 PAD_CLONE_VARS(proto_perl, param);
12243 #ifdef HAVE_INTERP_INTERN
12244 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12247 /* more statics moved here */
12248 PL_generation = proto_perl->Igeneration;
12249 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12251 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12252 PL_in_clean_all = proto_perl->Iin_clean_all;
12254 PL_uid = proto_perl->Iuid;
12255 PL_euid = proto_perl->Ieuid;
12256 PL_gid = proto_perl->Igid;
12257 PL_egid = proto_perl->Iegid;
12258 PL_nomemok = proto_perl->Inomemok;
12259 PL_an = proto_perl->Ian;
12260 PL_evalseq = proto_perl->Ievalseq;
12261 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12262 PL_origalen = proto_perl->Iorigalen;
12263 #ifdef PERL_USES_PL_PIDSTATUS
12264 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12266 PL_osname = SAVEPV(proto_perl->Iosname);
12267 PL_sighandlerp = proto_perl->Isighandlerp;
12269 PL_runops = proto_perl->Irunops;
12271 PL_parser = parser_dup(proto_perl->Iparser, param);
12273 /* XXX this only works if the saved cop has already been cloned */
12274 if (proto_perl->Iparser) {
12275 PL_parser->saved_curcop = (COP*)any_dup(
12276 proto_perl->Iparser->saved_curcop,
12280 PL_subline = proto_perl->Isubline;
12281 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12284 PL_cryptseen = proto_perl->Icryptseen;
12287 PL_hints = proto_perl->Ihints;
12289 PL_amagic_generation = proto_perl->Iamagic_generation;
12291 #ifdef USE_LOCALE_COLLATE
12292 PL_collation_ix = proto_perl->Icollation_ix;
12293 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12294 PL_collation_standard = proto_perl->Icollation_standard;
12295 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12296 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12297 #endif /* USE_LOCALE_COLLATE */
12299 #ifdef USE_LOCALE_NUMERIC
12300 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12301 PL_numeric_standard = proto_perl->Inumeric_standard;
12302 PL_numeric_local = proto_perl->Inumeric_local;
12303 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12304 #endif /* !USE_LOCALE_NUMERIC */
12306 /* utf8 character classes */
12307 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12308 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12309 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12310 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12311 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12312 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12313 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12314 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12315 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12316 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12317 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12318 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12319 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12320 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
12321 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
12322 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
12323 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
12324 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
12325 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
12326 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
12327 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
12328 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
12329 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
12330 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12331 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12332 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12333 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12334 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12335 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12337 /* Did the locale setup indicate UTF-8? */
12338 PL_utf8locale = proto_perl->Iutf8locale;
12339 /* Unicode features (see perlrun/-C) */
12340 PL_unicode = proto_perl->Iunicode;
12342 /* Pre-5.8 signals control */
12343 PL_signals = proto_perl->Isignals;
12345 /* times() ticks per second */
12346 PL_clocktick = proto_perl->Iclocktick;
12348 /* Recursion stopper for PerlIO_find_layer */
12349 PL_in_load_module = proto_perl->Iin_load_module;
12351 /* sort() routine */
12352 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12354 /* Not really needed/useful since the reenrant_retint is "volatile",
12355 * but do it for consistency's sake. */
12356 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12358 /* Hooks to shared SVs and locks. */
12359 PL_sharehook = proto_perl->Isharehook;
12360 PL_lockhook = proto_perl->Ilockhook;
12361 PL_unlockhook = proto_perl->Iunlockhook;
12362 PL_threadhook = proto_perl->Ithreadhook;
12363 PL_destroyhook = proto_perl->Idestroyhook;
12365 #ifdef THREADS_HAVE_PIDS
12366 PL_ppid = proto_perl->Ippid;
12370 PL_last_swash_hv = NULL; /* reinits on demand */
12371 PL_last_swash_klen = 0;
12372 PL_last_swash_key[0]= '\0';
12373 PL_last_swash_tmps = (U8*)NULL;
12374 PL_last_swash_slen = 0;
12376 PL_glob_index = proto_perl->Iglob_index;
12377 PL_srand_called = proto_perl->Isrand_called;
12379 if (proto_perl->Ipsig_pend) {
12380 Newxz(PL_psig_pend, SIG_SIZE, int);
12383 PL_psig_pend = (int*)NULL;
12386 if (proto_perl->Ipsig_name) {
12387 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12388 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12390 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12393 PL_psig_ptr = (SV**)NULL;
12394 PL_psig_name = (SV**)NULL;
12397 /* intrpvar.h stuff */
12399 if (flags & CLONEf_COPY_STACKS) {
12400 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12401 PL_tmps_ix = proto_perl->Itmps_ix;
12402 PL_tmps_max = proto_perl->Itmps_max;
12403 PL_tmps_floor = proto_perl->Itmps_floor;
12404 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12405 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12406 PL_tmps_ix+1, param);
12408 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12409 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12410 Newxz(PL_markstack, i, I32);
12411 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12412 - proto_perl->Imarkstack);
12413 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12414 - proto_perl->Imarkstack);
12415 Copy(proto_perl->Imarkstack, PL_markstack,
12416 PL_markstack_ptr - PL_markstack + 1, I32);
12418 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12419 * NOTE: unlike the others! */
12420 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12421 PL_scopestack_max = proto_perl->Iscopestack_max;
12422 Newxz(PL_scopestack, PL_scopestack_max, I32);
12423 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12426 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12427 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12429 /* NOTE: si_dup() looks at PL_markstack */
12430 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12432 /* PL_curstack = PL_curstackinfo->si_stack; */
12433 PL_curstack = av_dup(proto_perl->Icurstack, param);
12434 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12436 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12437 PL_stack_base = AvARRAY(PL_curstack);
12438 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12439 - proto_perl->Istack_base);
12440 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12442 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12443 * NOTE: unlike the others! */
12444 PL_savestack_ix = proto_perl->Isavestack_ix;
12445 PL_savestack_max = proto_perl->Isavestack_max;
12446 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12447 PL_savestack = ss_dup(proto_perl, param);
12451 ENTER; /* perl_destruct() wants to LEAVE; */
12453 /* although we're not duplicating the tmps stack, we should still
12454 * add entries for any SVs on the tmps stack that got cloned by a
12455 * non-refcount means (eg a temp in @_); otherwise they will be
12458 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12459 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12460 proto_perl->Itmps_stack[i]));
12461 if (nsv && !SvREFCNT(nsv)) {
12462 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12467 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12468 PL_top_env = &PL_start_env;
12470 PL_op = proto_perl->Iop;
12473 PL_Xpv = (XPV*)NULL;
12474 my_perl->Ina = proto_perl->Ina;
12476 PL_statbuf = proto_perl->Istatbuf;
12477 PL_statcache = proto_perl->Istatcache;
12478 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12479 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12481 PL_timesbuf = proto_perl->Itimesbuf;
12484 PL_tainted = proto_perl->Itainted;
12485 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12486 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12487 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12488 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12489 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12490 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12491 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12492 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12494 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12495 PL_restartop = proto_perl->Irestartop;
12496 PL_in_eval = proto_perl->Iin_eval;
12497 PL_delaymagic = proto_perl->Idelaymagic;
12498 PL_dirty = proto_perl->Idirty;
12499 PL_localizing = proto_perl->Ilocalizing;
12501 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12502 PL_hv_fetch_ent_mh = NULL;
12503 PL_modcount = proto_perl->Imodcount;
12504 PL_lastgotoprobe = NULL;
12505 PL_dumpindent = proto_perl->Idumpindent;
12507 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12508 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12509 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12510 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12511 PL_efloatbuf = NULL; /* reinits on demand */
12512 PL_efloatsize = 0; /* reinits on demand */
12516 PL_screamfirst = NULL;
12517 PL_screamnext = NULL;
12518 PL_maxscream = -1; /* reinits on demand */
12519 PL_lastscream = NULL;
12522 PL_regdummy = proto_perl->Iregdummy;
12523 PL_colorset = 0; /* reinits PL_colors[] */
12524 /*PL_colors[6] = {0,0,0,0,0,0};*/
12528 /* Pluggable optimizer */
12529 PL_peepp = proto_perl->Ipeepp;
12530 /* op_free() hook */
12531 PL_opfreehook = proto_perl->Iopfreehook;
12533 PL_stashcache = newHV();
12535 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12536 proto_perl->Iwatchaddr);
12537 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12538 if (PL_debug && PL_watchaddr) {
12539 PerlIO_printf(Perl_debug_log,
12540 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12541 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12542 PTR2UV(PL_watchok));
12545 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12547 /* Call the ->CLONE method, if it exists, for each of the stashes
12548 identified by sv_dup() above.
12550 while(av_len(param->stashes) != -1) {
12551 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12552 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12553 if (cloner && GvCV(cloner)) {
12558 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12560 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12566 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12567 ptr_table_free(PL_ptr_table);
12568 PL_ptr_table = NULL;
12572 SvREFCNT_dec(param->stashes);
12574 /* orphaned? eg threads->new inside BEGIN or use */
12575 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12576 SvREFCNT_inc_simple_void(PL_compcv);
12577 SAVEFREESV(PL_compcv);
12583 #endif /* USE_ITHREADS */
12586 =head1 Unicode Support
12588 =for apidoc sv_recode_to_utf8
12590 The encoding is assumed to be an Encode object, on entry the PV
12591 of the sv is assumed to be octets in that encoding, and the sv
12592 will be converted into Unicode (and UTF-8).
12594 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12595 is not a reference, nothing is done to the sv. If the encoding is not
12596 an C<Encode::XS> Encoding object, bad things will happen.
12597 (See F<lib/encoding.pm> and L<Encode>).
12599 The PV of the sv is returned.
12604 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12608 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12610 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12624 Passing sv_yes is wrong - it needs to be or'ed set of constants
12625 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12626 remove converted chars from source.
12628 Both will default the value - let them.
12630 XPUSHs(&PL_sv_yes);
12633 call_method("decode", G_SCALAR);
12637 s = SvPV_const(uni, len);
12638 if (s != SvPVX_const(sv)) {
12639 SvGROW(sv, len + 1);
12640 Move(s, SvPVX(sv), len + 1, char);
12641 SvCUR_set(sv, len);
12648 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12652 =for apidoc sv_cat_decode
12654 The encoding is assumed to be an Encode object, the PV of the ssv is
12655 assumed to be octets in that encoding and decoding the input starts
12656 from the position which (PV + *offset) pointed to. The dsv will be
12657 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12658 when the string tstr appears in decoding output or the input ends on
12659 the PV of the ssv. The value which the offset points will be modified
12660 to the last input position on the ssv.
12662 Returns TRUE if the terminator was found, else returns FALSE.
12667 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12668 SV *ssv, int *offset, char *tstr, int tlen)
12673 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12675 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12686 offsv = newSViv(*offset);
12688 mXPUSHp(tstr, tlen);
12690 call_method("cat_decode", G_SCALAR);
12692 ret = SvTRUE(TOPs);
12693 *offset = SvIV(offsv);
12699 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12704 /* ---------------------------------------------------------------------
12706 * support functions for report_uninit()
12709 /* the maxiumum size of array or hash where we will scan looking
12710 * for the undefined element that triggered the warning */
12712 #define FUV_MAX_SEARCH_SIZE 1000
12714 /* Look for an entry in the hash whose value has the same SV as val;
12715 * If so, return a mortal copy of the key. */
12718 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12721 register HE **array;
12724 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12726 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12727 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12730 array = HvARRAY(hv);
12732 for (i=HvMAX(hv); i>0; i--) {
12733 register HE *entry;
12734 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12735 if (HeVAL(entry) != val)
12737 if ( HeVAL(entry) == &PL_sv_undef ||
12738 HeVAL(entry) == &PL_sv_placeholder)
12742 if (HeKLEN(entry) == HEf_SVKEY)
12743 return sv_mortalcopy(HeKEY_sv(entry));
12744 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12750 /* Look for an entry in the array whose value has the same SV as val;
12751 * If so, return the index, otherwise return -1. */
12754 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12758 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12760 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12761 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12764 if (val != &PL_sv_undef) {
12765 SV ** const svp = AvARRAY(av);
12768 for (i=AvFILLp(av); i>=0; i--)
12775 /* S_varname(): return the name of a variable, optionally with a subscript.
12776 * If gv is non-zero, use the name of that global, along with gvtype (one
12777 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12778 * targ. Depending on the value of the subscript_type flag, return:
12781 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12782 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12783 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12784 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12787 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12788 const SV *const keyname, I32 aindex, int subscript_type)
12791 SV * const name = sv_newmortal();
12794 buffer[0] = gvtype;
12797 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12799 gv_fullname4(name, gv, buffer, 0);
12801 if ((unsigned int)SvPVX(name)[1] <= 26) {
12803 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12805 /* Swap the 1 unprintable control character for the 2 byte pretty
12806 version - ie substr($name, 1, 1) = $buffer; */
12807 sv_insert(name, 1, 1, buffer, 2);
12811 CV * const cv = find_runcv(NULL);
12815 if (!cv || !CvPADLIST(cv))
12817 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12818 sv = *av_fetch(av, targ, FALSE);
12819 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12822 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12823 SV * const sv = newSV(0);
12824 *SvPVX(name) = '$';
12825 Perl_sv_catpvf(aTHX_ name, "{%s}",
12826 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12829 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12830 *SvPVX(name) = '$';
12831 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12833 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12834 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12835 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12843 =for apidoc find_uninit_var
12845 Find the name of the undefined variable (if any) that caused the operator o
12846 to issue a "Use of uninitialized value" warning.
12847 If match is true, only return a name if it's value matches uninit_sv.
12848 So roughly speaking, if a unary operator (such as OP_COS) generates a
12849 warning, then following the direct child of the op may yield an
12850 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12851 other hand, with OP_ADD there are two branches to follow, so we only print
12852 the variable name if we get an exact match.
12854 The name is returned as a mortal SV.
12856 Assumes that PL_op is the op that originally triggered the error, and that
12857 PL_comppad/PL_curpad points to the currently executing pad.
12863 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12869 const OP *o, *o2, *kid;
12871 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12872 uninit_sv == &PL_sv_placeholder)))
12875 switch (obase->op_type) {
12882 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12883 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12886 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12888 if (pad) { /* @lex, %lex */
12889 sv = PAD_SVl(obase->op_targ);
12893 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12894 /* @global, %global */
12895 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12898 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12900 else /* @{expr}, %{expr} */
12901 return find_uninit_var(cUNOPx(obase)->op_first,
12905 /* attempt to find a match within the aggregate */
12907 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12909 subscript_type = FUV_SUBSCRIPT_HASH;
12912 index = find_array_subscript((const AV *)sv, uninit_sv);
12914 subscript_type = FUV_SUBSCRIPT_ARRAY;
12917 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12920 return varname(gv, hash ? '%' : '@', obase->op_targ,
12921 keysv, index, subscript_type);
12925 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12927 return varname(NULL, '$', obase->op_targ,
12928 NULL, 0, FUV_SUBSCRIPT_NONE);
12931 gv = cGVOPx_gv(obase);
12932 if (!gv || (match && GvSV(gv) != uninit_sv))
12934 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12937 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12940 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12941 if (!av || SvRMAGICAL(av))
12943 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12944 if (!svp || *svp != uninit_sv)
12947 return varname(NULL, '$', obase->op_targ,
12948 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12951 gv = cGVOPx_gv(obase);
12956 AV *const av = GvAV(gv);
12957 if (!av || SvRMAGICAL(av))
12959 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12960 if (!svp || *svp != uninit_sv)
12963 return varname(gv, '$', 0,
12964 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12969 o = cUNOPx(obase)->op_first;
12970 if (!o || o->op_type != OP_NULL ||
12971 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12973 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12977 if (PL_op == obase)
12978 /* $a[uninit_expr] or $h{uninit_expr} */
12979 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12982 o = cBINOPx(obase)->op_first;
12983 kid = cBINOPx(obase)->op_last;
12985 /* get the av or hv, and optionally the gv */
12987 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12988 sv = PAD_SV(o->op_targ);
12990 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12991 && cUNOPo->op_first->op_type == OP_GV)
12993 gv = cGVOPx_gv(cUNOPo->op_first);
12997 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13002 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13003 /* index is constant */
13007 if (obase->op_type == OP_HELEM) {
13008 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13009 if (!he || HeVAL(he) != uninit_sv)
13013 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13014 if (!svp || *svp != uninit_sv)
13018 if (obase->op_type == OP_HELEM)
13019 return varname(gv, '%', o->op_targ,
13020 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13022 return varname(gv, '@', o->op_targ, NULL,
13023 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13026 /* index is an expression;
13027 * attempt to find a match within the aggregate */
13028 if (obase->op_type == OP_HELEM) {
13029 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13031 return varname(gv, '%', o->op_targ,
13032 keysv, 0, FUV_SUBSCRIPT_HASH);
13036 = find_array_subscript((const AV *)sv, uninit_sv);
13038 return varname(gv, '@', o->op_targ,
13039 NULL, index, FUV_SUBSCRIPT_ARRAY);
13044 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13046 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13051 /* only examine RHS */
13052 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13055 o = cUNOPx(obase)->op_first;
13056 if (o->op_type == OP_PUSHMARK)
13059 if (!o->op_sibling) {
13060 /* one-arg version of open is highly magical */
13062 if (o->op_type == OP_GV) { /* open FOO; */
13064 if (match && GvSV(gv) != uninit_sv)
13066 return varname(gv, '$', 0,
13067 NULL, 0, FUV_SUBSCRIPT_NONE);
13069 /* other possibilities not handled are:
13070 * open $x; or open my $x; should return '${*$x}'
13071 * open expr; should return '$'.expr ideally
13077 /* ops where $_ may be an implicit arg */
13081 if ( !(obase->op_flags & OPf_STACKED)) {
13082 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13083 ? PAD_SVl(obase->op_targ)
13086 sv = sv_newmortal();
13087 sv_setpvs(sv, "$_");
13096 match = 1; /* print etc can return undef on defined args */
13097 /* skip filehandle as it can't produce 'undef' warning */
13098 o = cUNOPx(obase)->op_first;
13099 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13100 o = o->op_sibling->op_sibling;
13104 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13106 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13108 /* the following ops are capable of returning PL_sv_undef even for
13109 * defined arg(s) */
13128 case OP_GETPEERNAME:
13176 case OP_SMARTMATCH:
13185 /* XXX tmp hack: these two may call an XS sub, and currently
13186 XS subs don't have a SUB entry on the context stack, so CV and
13187 pad determination goes wrong, and BAD things happen. So, just
13188 don't try to determine the value under those circumstances.
13189 Need a better fix at dome point. DAPM 11/2007 */
13195 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13196 if (gv && GvSV(gv) == uninit_sv)
13197 return newSVpvs_flags("$.", SVs_TEMP);
13202 /* def-ness of rval pos() is independent of the def-ness of its arg */
13203 if ( !(obase->op_flags & OPf_MOD))
13208 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13209 return newSVpvs_flags("${$/}", SVs_TEMP);
13214 if (!(obase->op_flags & OPf_KIDS))
13216 o = cUNOPx(obase)->op_first;
13222 /* if all except one arg are constant, or have no side-effects,
13223 * or are optimized away, then it's unambiguous */
13225 for (kid=o; kid; kid = kid->op_sibling) {
13227 const OPCODE type = kid->op_type;
13228 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13229 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13230 || (type == OP_PUSHMARK)
13234 if (o2) { /* more than one found */
13241 return find_uninit_var(o2, uninit_sv, match);
13243 /* scan all args */
13245 sv = find_uninit_var(o, uninit_sv, 1);
13257 =for apidoc report_uninit
13259 Print appropriate "Use of uninitialized variable" warning
13265 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13269 SV* varname = NULL;
13271 varname = find_uninit_var(PL_op, uninit_sv,0);
13273 sv_insert(varname, 0, 0, " ", 1);
13275 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13276 varname ? SvPV_nolen_const(varname) : "",
13277 " in ", OP_DESC(PL_op));
13280 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13286 * c-indentation-style: bsd
13287 * c-basic-offset: 4
13288 * indent-tabs-mode: t
13291 * ex: set ts=8 sts=4 sw=4 noet: