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 if (ckWARN_d(WARN_INTERNAL))
357 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
358 "Attempt to free non-arena SV: 0x%"UVxf
359 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
366 #else /* ! DEBUGGING */
368 #define del_SV(p) plant_SV(p)
370 #endif /* DEBUGGING */
374 =head1 SV Manipulation Functions
376 =for apidoc sv_add_arena
378 Given a chunk of memory, link it to the head of the list of arenas,
379 and split it into a list of free SVs.
385 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
388 SV *const sva = MUTABLE_SV(ptr);
392 PERL_ARGS_ASSERT_SV_ADD_ARENA;
394 /* The first SV in an arena isn't an SV. */
395 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
396 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
397 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
399 PL_sv_arenaroot = sva;
400 PL_sv_root = sva + 1;
402 svend = &sva[SvREFCNT(sva) - 1];
405 SvARENA_CHAIN_SET(sv, (sv + 1));
409 /* Must always set typemask because it's always checked in on cleanup
410 when the arenas are walked looking for objects. */
411 SvFLAGS(sv) = SVTYPEMASK;
414 SvARENA_CHAIN_SET(sv, 0);
418 SvFLAGS(sv) = SVTYPEMASK;
421 /* visit(): call the named function for each non-free SV in the arenas
422 * whose flags field matches the flags/mask args. */
425 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
431 PERL_ARGS_ASSERT_VISIT;
433 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
434 register const SV * const svend = &sva[SvREFCNT(sva)];
436 for (sv = sva + 1; sv < svend; ++sv) {
437 if (SvTYPE(sv) != SVTYPEMASK
438 && (sv->sv_flags & mask) == flags
451 /* called by sv_report_used() for each live SV */
454 do_report_used(pTHX_ SV *const sv)
456 if (SvTYPE(sv) != SVTYPEMASK) {
457 PerlIO_printf(Perl_debug_log, "****\n");
464 =for apidoc sv_report_used
466 Dump the contents of all SVs not yet freed. (Debugging aid).
472 Perl_sv_report_used(pTHX)
475 visit(do_report_used, 0, 0);
481 /* called by sv_clean_objs() for each live SV */
484 do_clean_objs(pTHX_ SV *const ref)
489 SV * const target = SvRV(ref);
490 if (SvOBJECT(target)) {
491 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
492 if (SvWEAKREF(ref)) {
493 sv_del_backref(target, ref);
499 SvREFCNT_dec(target);
504 /* XXX Might want to check arrays, etc. */
507 /* called by sv_clean_objs() for each live SV */
509 #ifndef DISABLE_DESTRUCTOR_KLUDGE
511 do_clean_named_objs(pTHX_ SV *const sv)
514 assert(SvTYPE(sv) == SVt_PVGV);
515 assert(isGV_with_GP(sv));
518 #ifdef PERL_DONT_CREATE_GVSV
521 SvOBJECT(GvSV(sv))) ||
522 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
523 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
524 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
525 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
526 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
528 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
529 SvFLAGS(sv) |= SVf_BREAK;
537 =for apidoc sv_clean_objs
539 Attempt to destroy all objects not yet freed
545 Perl_sv_clean_objs(pTHX)
548 PL_in_clean_objs = TRUE;
549 visit(do_clean_objs, SVf_ROK, SVf_ROK);
550 #ifndef DISABLE_DESTRUCTOR_KLUDGE
551 /* some barnacles may yet remain, clinging to typeglobs */
552 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
554 PL_in_clean_objs = FALSE;
557 /* called by sv_clean_all() for each live SV */
560 do_clean_all(pTHX_ SV *const sv)
563 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
564 /* don't clean pid table and strtab */
567 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
568 SvFLAGS(sv) |= SVf_BREAK;
573 =for apidoc sv_clean_all
575 Decrement the refcnt of each remaining SV, possibly triggering a
576 cleanup. This function may have to be called multiple times to free
577 SVs which are in complex self-referential hierarchies.
583 Perl_sv_clean_all(pTHX)
587 PL_in_clean_all = TRUE;
588 cleaned = visit(do_clean_all, 0,0);
589 PL_in_clean_all = FALSE;
594 ARENASETS: a meta-arena implementation which separates arena-info
595 into struct arena_set, which contains an array of struct
596 arena_descs, each holding info for a single arena. By separating
597 the meta-info from the arena, we recover the 1st slot, formerly
598 borrowed for list management. The arena_set is about the size of an
599 arena, avoiding the needless malloc overhead of a naive linked-list.
601 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
602 memory in the last arena-set (1/2 on average). In trade, we get
603 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
604 smaller types). The recovery of the wasted space allows use of
605 small arenas for large, rare body types, by changing array* fields
606 in body_details_by_type[] below.
609 char *arena; /* the raw storage, allocated aligned */
610 size_t size; /* its size ~4k typ */
611 U32 misc; /* type, and in future other things. */
616 /* Get the maximum number of elements in set[] such that struct arena_set
617 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
618 therefore likely to be 1 aligned memory page. */
620 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
621 - 2 * sizeof(int)) / sizeof (struct arena_desc))
624 struct arena_set* next;
625 unsigned int set_size; /* ie ARENAS_PER_SET */
626 unsigned int curr; /* index of next available arena-desc */
627 struct arena_desc set[ARENAS_PER_SET];
631 =for apidoc sv_free_arenas
633 Deallocate the memory used by all arenas. Note that all the individual SV
634 heads and bodies within the arenas must already have been freed.
639 Perl_sv_free_arenas(pTHX)
646 /* Free arenas here, but be careful about fake ones. (We assume
647 contiguity of the fake ones with the corresponding real ones.) */
649 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
650 svanext = MUTABLE_SV(SvANY(sva));
651 while (svanext && SvFAKE(svanext))
652 svanext = MUTABLE_SV(SvANY(svanext));
659 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
662 struct arena_set *current = aroot;
665 assert(aroot->set[i].arena);
666 Safefree(aroot->set[i].arena);
674 i = PERL_ARENA_ROOTS_SIZE;
676 PL_body_roots[i] = 0;
678 Safefree(PL_nice_chunk);
679 PL_nice_chunk = NULL;
680 PL_nice_chunk_size = 0;
686 Here are mid-level routines that manage the allocation of bodies out
687 of the various arenas. There are 5 kinds of arenas:
689 1. SV-head arenas, which are discussed and handled above
690 2. regular body arenas
691 3. arenas for reduced-size bodies
693 5. pte arenas (thread related)
695 Arena types 2 & 3 are chained by body-type off an array of
696 arena-root pointers, which is indexed by svtype. Some of the
697 larger/less used body types are malloced singly, since a large
698 unused block of them is wasteful. Also, several svtypes dont have
699 bodies; the data fits into the sv-head itself. The arena-root
700 pointer thus has a few unused root-pointers (which may be hijacked
701 later for arena types 4,5)
703 3 differs from 2 as an optimization; some body types have several
704 unused fields in the front of the structure (which are kept in-place
705 for consistency). These bodies can be allocated in smaller chunks,
706 because the leading fields arent accessed. Pointers to such bodies
707 are decremented to point at the unused 'ghost' memory, knowing that
708 the pointers are used with offsets to the real memory.
710 HE, HEK arenas are managed separately, with separate code, but may
711 be merge-able later..
713 PTE arenas are not sv-bodies, but they share these mid-level
714 mechanics, so are considered here. The new mid-level mechanics rely
715 on the sv_type of the body being allocated, so we just reserve one
716 of the unused body-slots for PTEs, then use it in those (2) PTE
717 contexts below (line ~10k)
720 /* get_arena(size): this creates custom-sized arenas
721 TBD: export properly for hv.c: S_more_he().
724 Perl_get_arena(pTHX_ const size_t arena_size, const U32 misc)
727 struct arena_desc* adesc;
728 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
731 /* shouldnt need this
732 if (!arena_size) arena_size = PERL_ARENA_SIZE;
735 /* may need new arena-set to hold new arena */
736 if (!aroot || aroot->curr >= aroot->set_size) {
737 struct arena_set *newroot;
738 Newxz(newroot, 1, struct arena_set);
739 newroot->set_size = ARENAS_PER_SET;
740 newroot->next = aroot;
742 PL_body_arenas = (void *) newroot;
743 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
746 /* ok, now have arena-set with at least 1 empty/available arena-desc */
747 curr = aroot->curr++;
748 adesc = &(aroot->set[curr]);
749 assert(!adesc->arena);
751 Newx(adesc->arena, arena_size, char);
752 adesc->size = arena_size;
754 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
755 curr, (void*)adesc->arena, (UV)arena_size));
761 /* return a thing to the free list */
763 #define del_body(thing, root) \
765 void ** const thing_copy = (void **)thing;\
766 *thing_copy = *root; \
767 *root = (void*)thing_copy; \
772 =head1 SV-Body Allocation
774 Allocation of SV-bodies is similar to SV-heads, differing as follows;
775 the allocation mechanism is used for many body types, so is somewhat
776 more complicated, it uses arena-sets, and has no need for still-live
779 At the outermost level, (new|del)_X*V macros return bodies of the
780 appropriate type. These macros call either (new|del)_body_type or
781 (new|del)_body_allocated macro pairs, depending on specifics of the
782 type. Most body types use the former pair, the latter pair is used to
783 allocate body types with "ghost fields".
785 "ghost fields" are fields that are unused in certain types, and
786 consequently don't need to actually exist. They are declared because
787 they're part of a "base type", which allows use of functions as
788 methods. The simplest examples are AVs and HVs, 2 aggregate types
789 which don't use the fields which support SCALAR semantics.
791 For these types, the arenas are carved up into appropriately sized
792 chunks, we thus avoid wasted memory for those unaccessed members.
793 When bodies are allocated, we adjust the pointer back in memory by the
794 size of the part not allocated, so it's as if we allocated the full
795 structure. (But things will all go boom if you write to the part that
796 is "not there", because you'll be overwriting the last members of the
797 preceding structure in memory.)
799 We calculate the correction using the STRUCT_OFFSET macro on the first
800 member present. If the allocated structure is smaller (no initial NV
801 actually allocated) then the net effect is to subtract the size of the NV
802 from the pointer, to return a new pointer as if an initial NV were actually
803 allocated. (We were using structures named *_allocated for this, but
804 this turned out to be a subtle bug, because a structure without an NV
805 could have a lower alignment constraint, but the compiler is allowed to
806 optimised accesses based on the alignment constraint of the actual pointer
807 to the full structure, for example, using a single 64 bit load instruction
808 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
810 This is the same trick as was used for NV and IV bodies. Ironically it
811 doesn't need to be used for NV bodies any more, because NV is now at
812 the start of the structure. IV bodies don't need it either, because
813 they are no longer allocated.
815 In turn, the new_body_* allocators call S_new_body(), which invokes
816 new_body_inline macro, which takes a lock, and takes a body off the
817 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
818 necessary to refresh an empty list. Then the lock is released, and
819 the body is returned.
821 S_more_bodies calls get_arena(), and carves it up into an array of N
822 bodies, which it strings into a linked list. It looks up arena-size
823 and body-size from the body_details table described below, thus
824 supporting the multiple body-types.
826 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
827 the (new|del)_X*V macros are mapped directly to malloc/free.
833 For each sv-type, struct body_details bodies_by_type[] carries
834 parameters which control these aspects of SV handling:
836 Arena_size determines whether arenas are used for this body type, and if
837 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
838 zero, forcing individual mallocs and frees.
840 Body_size determines how big a body is, and therefore how many fit into
841 each arena. Offset carries the body-pointer adjustment needed for
842 "ghost fields", and is used in *_allocated macros.
844 But its main purpose is to parameterize info needed in
845 Perl_sv_upgrade(). The info here dramatically simplifies the function
846 vs the implementation in 5.8.8, making it table-driven. All fields
847 are used for this, except for arena_size.
849 For the sv-types that have no bodies, arenas are not used, so those
850 PL_body_roots[sv_type] are unused, and can be overloaded. In
851 something of a special case, SVt_NULL is borrowed for HE arenas;
852 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
853 bodies_by_type[SVt_NULL] slot is not used, as the table is not
856 PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
857 they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
858 just use the same allocation semantics. At first, PTEs were also
859 overloaded to a non-body sv-type, but this yielded hard-to-find malloc
860 bugs, so was simplified by claiming a new slot. This choice has no
861 consequence at this time.
865 struct body_details {
866 U8 body_size; /* Size to allocate */
867 U8 copy; /* Size of structure to copy (may be shorter) */
869 unsigned int type : 4; /* We have space for a sanity check. */
870 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
871 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
872 unsigned int arena : 1; /* Allocated from an arena */
873 size_t arena_size; /* Size of arena to allocate */
881 /* With -DPURFIY we allocate everything directly, and don't use arenas.
882 This seems a rather elegant way to simplify some of the code below. */
883 #define HASARENA FALSE
885 #define HASARENA TRUE
887 #define NOARENA FALSE
889 /* Size the arenas to exactly fit a given number of bodies. A count
890 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
891 simplifying the default. If count > 0, the arena is sized to fit
892 only that many bodies, allowing arenas to be used for large, rare
893 bodies (XPVFM, XPVIO) without undue waste. The arena size is
894 limited by PERL_ARENA_SIZE, so we can safely oversize the
897 #define FIT_ARENA0(body_size) \
898 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
899 #define FIT_ARENAn(count,body_size) \
900 ( count * body_size <= PERL_ARENA_SIZE) \
901 ? count * body_size \
902 : FIT_ARENA0 (body_size)
903 #define FIT_ARENA(count,body_size) \
905 ? FIT_ARENAn (count, body_size) \
906 : FIT_ARENA0 (body_size)
908 /* Calculate the length to copy. Specifically work out the length less any
909 final padding the compiler needed to add. See the comment in sv_upgrade
910 for why copying the padding proved to be a bug. */
912 #define copy_length(type, last_member) \
913 STRUCT_OFFSET(type, last_member) \
914 + sizeof (((type*)SvANY((const SV *)0))->last_member)
916 static const struct body_details bodies_by_type[] = {
917 { sizeof(HE), 0, 0, SVt_NULL,
918 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
920 /* The bind placeholder pretends to be an RV for now.
921 Also it's marked as "can't upgrade" to stop anyone using it before it's
923 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
925 /* IVs are in the head, so the allocation size is 0.
926 However, the slot is overloaded for PTEs. */
927 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
928 sizeof(IV), /* This is used to copy out the IV body. */
929 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
930 NOARENA /* IVS don't need an arena */,
931 /* But PTEs need to know the size of their arena */
932 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
935 /* 8 bytes on most ILP32 with IEEE doubles */
936 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
937 FIT_ARENA(0, sizeof(NV)) },
939 /* 8 bytes on most ILP32 with IEEE doubles */
940 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
941 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
942 + STRUCT_OFFSET(XPV, xpv_cur),
943 SVt_PV, FALSE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
947 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
948 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
949 + STRUCT_OFFSET(XPVIV, xpv_cur),
950 SVt_PVIV, FALSE, NONV, HASARENA,
951 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
954 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
955 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
958 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
959 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
962 { sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
963 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
964 + STRUCT_OFFSET(regexp, xpv_cur),
965 SVt_REGEXP, FALSE, NONV, HASARENA,
966 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
970 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
971 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
974 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
975 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
977 { sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill),
978 copy_length(XPVAV, xmg_stash) - STRUCT_OFFSET(XPVAV, xav_fill),
979 + STRUCT_OFFSET(XPVAV, xav_fill),
980 SVt_PVAV, TRUE, NONV, HASARENA,
981 FIT_ARENA(0, sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill)) },
983 { sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill),
984 copy_length(XPVHV, xmg_stash) - STRUCT_OFFSET(XPVHV, xhv_fill),
985 + STRUCT_OFFSET(XPVHV, xhv_fill),
986 SVt_PVHV, TRUE, NONV, HASARENA,
987 FIT_ARENA(0, sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill)) },
990 { sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
991 sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
992 + STRUCT_OFFSET(XPVCV, xpv_cur),
993 SVt_PVCV, TRUE, NONV, HASARENA,
994 FIT_ARENA(0, sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur)) },
996 { sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
997 sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
998 + STRUCT_OFFSET(XPVFM, xpv_cur),
999 SVt_PVFM, TRUE, NONV, NOARENA,
1000 FIT_ARENA(20, sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur)) },
1002 /* XPVIO is 84 bytes, fits 48x */
1003 { sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1004 sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1005 + STRUCT_OFFSET(XPVIO, xpv_cur),
1006 SVt_PVIO, TRUE, NONV, HASARENA,
1007 FIT_ARENA(24, sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur)) },
1010 #define new_body_type(sv_type) \
1011 (void *)((char *)S_new_body(aTHX_ sv_type))
1013 #define del_body_type(p, sv_type) \
1014 del_body(p, &PL_body_roots[sv_type])
1017 #define new_body_allocated(sv_type) \
1018 (void *)((char *)S_new_body(aTHX_ sv_type) \
1019 - bodies_by_type[sv_type].offset)
1021 #define del_body_allocated(p, sv_type) \
1022 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1025 #define my_safemalloc(s) (void*)safemalloc(s)
1026 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1027 #define my_safefree(p) safefree((char*)p)
1031 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1032 #define del_XNV(p) my_safefree(p)
1034 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1035 #define del_XPVNV(p) my_safefree(p)
1037 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1038 #define del_XPVAV(p) my_safefree(p)
1040 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1041 #define del_XPVHV(p) my_safefree(p)
1043 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1044 #define del_XPVMG(p) my_safefree(p)
1046 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1047 #define del_XPVGV(p) my_safefree(p)
1051 #define new_XNV() new_body_type(SVt_NV)
1052 #define del_XNV(p) del_body_type(p, SVt_NV)
1054 #define new_XPVNV() new_body_type(SVt_PVNV)
1055 #define del_XPVNV(p) del_body_type(p, SVt_PVNV)
1057 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1058 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1060 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1061 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1063 #define new_XPVMG() new_body_type(SVt_PVMG)
1064 #define del_XPVMG(p) del_body_type(p, SVt_PVMG)
1066 #define new_XPVGV() new_body_type(SVt_PVGV)
1067 #define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1071 /* no arena for you! */
1073 #define new_NOARENA(details) \
1074 my_safemalloc((details)->body_size + (details)->offset)
1075 #define new_NOARENAZ(details) \
1076 my_safecalloc((details)->body_size + (details)->offset)
1079 S_more_bodies (pTHX_ const svtype sv_type)
1082 void ** const root = &PL_body_roots[sv_type];
1083 const struct body_details * const bdp = &bodies_by_type[sv_type];
1084 const size_t body_size = bdp->body_size;
1087 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1088 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1089 static bool done_sanity_check;
1091 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1092 * variables like done_sanity_check. */
1093 if (!done_sanity_check) {
1094 unsigned int i = SVt_LAST;
1096 done_sanity_check = TRUE;
1099 assert (bodies_by_type[i].type == i);
1103 assert(bdp->arena_size);
1105 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1107 end = start + arena_size - 2 * body_size;
1109 /* computed count doesnt reflect the 1st slot reservation */
1110 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1111 DEBUG_m(PerlIO_printf(Perl_debug_log,
1112 "arena %p end %p arena-size %d (from %d) type %d "
1114 (void*)start, (void*)end, (int)arena_size,
1115 (int)bdp->arena_size, sv_type, (int)body_size,
1116 (int)arena_size / (int)body_size));
1118 DEBUG_m(PerlIO_printf(Perl_debug_log,
1119 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1120 (void*)start, (void*)end,
1121 (int)bdp->arena_size, sv_type, (int)body_size,
1122 (int)bdp->arena_size / (int)body_size));
1124 *root = (void *)start;
1126 while (start <= end) {
1127 char * const next = start + body_size;
1128 *(void**) start = (void *)next;
1131 *(void **)start = 0;
1136 /* grab a new thing from the free list, allocating more if necessary.
1137 The inline version is used for speed in hot routines, and the
1138 function using it serves the rest (unless PURIFY).
1140 #define new_body_inline(xpv, sv_type) \
1142 void ** const r3wt = &PL_body_roots[sv_type]; \
1143 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1144 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1145 *(r3wt) = *(void**)(xpv); \
1151 S_new_body(pTHX_ const svtype sv_type)
1155 new_body_inline(xpv, sv_type);
1161 static const struct body_details fake_rv =
1162 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1165 =for apidoc sv_upgrade
1167 Upgrade an SV to a more complex form. Generally adds a new body type to the
1168 SV, then copies across as much information as possible from the old body.
1169 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1175 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1180 const svtype old_type = SvTYPE(sv);
1181 const struct body_details *new_type_details;
1182 const struct body_details *old_type_details
1183 = bodies_by_type + old_type;
1184 SV *referant = NULL;
1186 PERL_ARGS_ASSERT_SV_UPGRADE;
1188 if (old_type == new_type)
1191 /* This clause was purposefully added ahead of the early return above to
1192 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1193 inference by Nick I-S that it would fix other troublesome cases. See
1194 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1196 Given that shared hash key scalars are no longer PVIV, but PV, there is
1197 no longer need to unshare so as to free up the IVX slot for its proper
1198 purpose. So it's safe to move the early return earlier. */
1200 if (new_type != SVt_PV && SvIsCOW(sv)) {
1201 sv_force_normal_flags(sv, 0);
1204 old_body = SvANY(sv);
1206 /* Copying structures onto other structures that have been neatly zeroed
1207 has a subtle gotcha. Consider XPVMG
1209 +------+------+------+------+------+-------+-------+
1210 | NV | CUR | LEN | IV | MAGIC | STASH |
1211 +------+------+------+------+------+-------+-------+
1212 0 4 8 12 16 20 24 28
1214 where NVs are aligned to 8 bytes, so that sizeof that structure is
1215 actually 32 bytes long, with 4 bytes of padding at the end:
1217 +------+------+------+------+------+-------+-------+------+
1218 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1219 +------+------+------+------+------+-------+-------+------+
1220 0 4 8 12 16 20 24 28 32
1222 so what happens if you allocate memory for this structure:
1224 +------+------+------+------+------+-------+-------+------+------+...
1225 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1226 +------+------+------+------+------+-------+-------+------+------+...
1227 0 4 8 12 16 20 24 28 32 36
1229 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1230 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1231 started out as zero once, but it's quite possible that it isn't. So now,
1232 rather than a nicely zeroed GP, you have it pointing somewhere random.
1235 (In fact, GP ends up pointing at a previous GP structure, because the
1236 principle cause of the padding in XPVMG getting garbage is a copy of
1237 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1238 this happens to be moot because XPVGV has been re-ordered, with GP
1239 no longer after STASH)
1241 So we are careful and work out the size of used parts of all the
1249 referant = SvRV(sv);
1250 old_type_details = &fake_rv;
1251 if (new_type == SVt_NV)
1252 new_type = SVt_PVNV;
1254 if (new_type < SVt_PVIV) {
1255 new_type = (new_type == SVt_NV)
1256 ? SVt_PVNV : SVt_PVIV;
1261 if (new_type < SVt_PVNV) {
1262 new_type = SVt_PVNV;
1266 assert(new_type > SVt_PV);
1267 assert(SVt_IV < SVt_PV);
1268 assert(SVt_NV < SVt_PV);
1275 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1276 there's no way that it can be safely upgraded, because perl.c
1277 expects to Safefree(SvANY(PL_mess_sv)) */
1278 assert(sv != PL_mess_sv);
1279 /* This flag bit is used to mean other things in other scalar types.
1280 Given that it only has meaning inside the pad, it shouldn't be set
1281 on anything that can get upgraded. */
1282 assert(!SvPAD_TYPED(sv));
1285 if (old_type_details->cant_upgrade)
1286 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1287 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1290 if (old_type > new_type)
1291 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1292 (int)old_type, (int)new_type);
1294 new_type_details = bodies_by_type + new_type;
1296 SvFLAGS(sv) &= ~SVTYPEMASK;
1297 SvFLAGS(sv) |= new_type;
1299 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1300 the return statements above will have triggered. */
1301 assert (new_type != SVt_NULL);
1304 assert(old_type == SVt_NULL);
1305 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1309 assert(old_type == SVt_NULL);
1310 SvANY(sv) = new_XNV();
1315 assert(new_type_details->body_size);
1318 assert(new_type_details->arena);
1319 assert(new_type_details->arena_size);
1320 /* This points to the start of the allocated area. */
1321 new_body_inline(new_body, new_type);
1322 Zero(new_body, new_type_details->body_size, char);
1323 new_body = ((char *)new_body) - new_type_details->offset;
1325 /* We always allocated the full length item with PURIFY. To do this
1326 we fake things so that arena is false for all 16 types.. */
1327 new_body = new_NOARENAZ(new_type_details);
1329 SvANY(sv) = new_body;
1330 if (new_type == SVt_PVAV) {
1334 if (old_type_details->body_size) {
1337 /* It will have been zeroed when the new body was allocated.
1338 Lets not write to it, in case it confuses a write-back
1344 #ifndef NODEFAULT_SHAREKEYS
1345 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1347 HvMAX(sv) = 7; /* (start with 8 buckets) */
1348 if (old_type_details->body_size) {
1351 /* It will have been zeroed when the new body was allocated.
1352 Lets not write to it, in case it confuses a write-back
1357 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1358 The target created by newSVrv also is, and it can have magic.
1359 However, it never has SvPVX set.
1361 if (old_type == SVt_IV) {
1363 } else if (old_type >= SVt_PV) {
1364 assert(SvPVX_const(sv) == 0);
1367 if (old_type >= SVt_PVMG) {
1368 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1369 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1371 sv->sv_u.svu_array = NULL; /* or svu_hash */
1377 /* XXX Is this still needed? Was it ever needed? Surely as there is
1378 no route from NV to PVIV, NOK can never be true */
1379 assert(!SvNOKp(sv));
1391 assert(new_type_details->body_size);
1392 /* We always allocated the full length item with PURIFY. To do this
1393 we fake things so that arena is false for all 16 types.. */
1394 if(new_type_details->arena) {
1395 /* This points to the start of the allocated area. */
1396 new_body_inline(new_body, new_type);
1397 Zero(new_body, new_type_details->body_size, char);
1398 new_body = ((char *)new_body) - new_type_details->offset;
1400 new_body = new_NOARENAZ(new_type_details);
1402 SvANY(sv) = new_body;
1404 if (old_type_details->copy) {
1405 /* There is now the potential for an upgrade from something without
1406 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1407 int offset = old_type_details->offset;
1408 int length = old_type_details->copy;
1410 if (new_type_details->offset > old_type_details->offset) {
1411 const int difference
1412 = new_type_details->offset - old_type_details->offset;
1413 offset += difference;
1414 length -= difference;
1416 assert (length >= 0);
1418 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1422 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1423 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1424 * correct 0.0 for us. Otherwise, if the old body didn't have an
1425 * NV slot, but the new one does, then we need to initialise the
1426 * freshly created NV slot with whatever the correct bit pattern is
1428 if (old_type_details->zero_nv && !new_type_details->zero_nv
1429 && !isGV_with_GP(sv))
1433 if (new_type == SVt_PVIO) {
1434 IO * const io = MUTABLE_IO(sv);
1435 GV *iogv = gv_fetchpvs("FileHandle::", 0, SVt_PVHV);
1438 /* Clear the stashcache because a new IO could overrule a package
1440 hv_clear(PL_stashcache);
1442 /* unless exists($main::{FileHandle}) and
1443 defined(%main::FileHandle::) */
1444 if (!(iogv && GvHV(iogv) && HvARRAY(GvHV(iogv))))
1445 iogv = gv_fetchpvs("IO::Handle::", GV_ADD, SVt_PVHV);
1446 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1447 IoPAGE_LEN(sv) = 60;
1449 if (old_type < SVt_PV) {
1450 /* referant will be NULL unless the old type was SVt_IV emulating
1452 sv->sv_u.svu_rv = referant;
1456 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1457 (unsigned long)new_type);
1460 if (old_type_details->arena) {
1461 /* If there was an old body, then we need to free it.
1462 Note that there is an assumption that all bodies of types that
1463 can be upgraded came from arenas. Only the more complex non-
1464 upgradable types are allowed to be directly malloc()ed. */
1466 my_safefree(old_body);
1468 del_body((void*)((char*)old_body + old_type_details->offset),
1469 &PL_body_roots[old_type]);
1475 =for apidoc sv_backoff
1477 Remove any string offset. You should normally use the C<SvOOK_off> macro
1484 Perl_sv_backoff(pTHX_ register SV *const sv)
1487 const char * const s = SvPVX_const(sv);
1489 PERL_ARGS_ASSERT_SV_BACKOFF;
1490 PERL_UNUSED_CONTEXT;
1493 assert(SvTYPE(sv) != SVt_PVHV);
1494 assert(SvTYPE(sv) != SVt_PVAV);
1496 SvOOK_offset(sv, delta);
1498 SvLEN_set(sv, SvLEN(sv) + delta);
1499 SvPV_set(sv, SvPVX(sv) - delta);
1500 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1501 SvFLAGS(sv) &= ~SVf_OOK;
1508 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1509 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1510 Use the C<SvGROW> wrapper instead.
1516 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1520 PERL_ARGS_ASSERT_SV_GROW;
1522 if (PL_madskills && newlen >= 0x100000) {
1523 PerlIO_printf(Perl_debug_log,
1524 "Allocation too large: %"UVxf"\n", (UV)newlen);
1526 #ifdef HAS_64K_LIMIT
1527 if (newlen >= 0x10000) {
1528 PerlIO_printf(Perl_debug_log,
1529 "Allocation too large: %"UVxf"\n", (UV)newlen);
1532 #endif /* HAS_64K_LIMIT */
1535 if (SvTYPE(sv) < SVt_PV) {
1536 sv_upgrade(sv, SVt_PV);
1537 s = SvPVX_mutable(sv);
1539 else if (SvOOK(sv)) { /* pv is offset? */
1541 s = SvPVX_mutable(sv);
1542 if (newlen > SvLEN(sv))
1543 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1544 #ifdef HAS_64K_LIMIT
1545 if (newlen >= 0x10000)
1550 s = SvPVX_mutable(sv);
1552 if (newlen > SvLEN(sv)) { /* need more room? */
1553 #ifndef Perl_safesysmalloc_size
1554 newlen = PERL_STRLEN_ROUNDUP(newlen);
1556 if (SvLEN(sv) && s) {
1557 s = (char*)saferealloc(s, newlen);
1560 s = (char*)safemalloc(newlen);
1561 if (SvPVX_const(sv) && SvCUR(sv)) {
1562 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1566 #ifdef Perl_safesysmalloc_size
1567 /* Do this here, do it once, do it right, and then we will never get
1568 called back into sv_grow() unless there really is some growing
1570 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1572 SvLEN_set(sv, newlen);
1579 =for apidoc sv_setiv
1581 Copies an integer into the given SV, upgrading first if necessary.
1582 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1588 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1592 PERL_ARGS_ASSERT_SV_SETIV;
1594 SV_CHECK_THINKFIRST_COW_DROP(sv);
1595 switch (SvTYPE(sv)) {
1598 sv_upgrade(sv, SVt_IV);
1601 sv_upgrade(sv, SVt_PVIV);
1605 if (!isGV_with_GP(sv))
1612 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1616 (void)SvIOK_only(sv); /* validate number */
1622 =for apidoc sv_setiv_mg
1624 Like C<sv_setiv>, but also handles 'set' magic.
1630 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1632 PERL_ARGS_ASSERT_SV_SETIV_MG;
1639 =for apidoc sv_setuv
1641 Copies an unsigned integer into the given SV, upgrading first if necessary.
1642 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1648 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1650 PERL_ARGS_ASSERT_SV_SETUV;
1652 /* With these two if statements:
1653 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1656 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1658 If you wish to remove them, please benchmark to see what the effect is
1660 if (u <= (UV)IV_MAX) {
1661 sv_setiv(sv, (IV)u);
1670 =for apidoc sv_setuv_mg
1672 Like C<sv_setuv>, but also handles 'set' magic.
1678 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1680 PERL_ARGS_ASSERT_SV_SETUV_MG;
1687 =for apidoc sv_setnv
1689 Copies a double into the given SV, upgrading first if necessary.
1690 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1696 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV;
1702 SV_CHECK_THINKFIRST_COW_DROP(sv);
1703 switch (SvTYPE(sv)) {
1706 sv_upgrade(sv, SVt_NV);
1710 sv_upgrade(sv, SVt_PVNV);
1714 if (!isGV_with_GP(sv))
1721 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1726 (void)SvNOK_only(sv); /* validate number */
1731 =for apidoc sv_setnv_mg
1733 Like C<sv_setnv>, but also handles 'set' magic.
1739 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1741 PERL_ARGS_ASSERT_SV_SETNV_MG;
1747 /* Print an "isn't numeric" warning, using a cleaned-up,
1748 * printable version of the offending string
1752 S_not_a_number(pTHX_ SV *const sv)
1759 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1762 dsv = newSVpvs_flags("", SVs_TEMP);
1763 pv = sv_uni_display(dsv, sv, 10, 0);
1766 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1767 /* each *s can expand to 4 chars + "...\0",
1768 i.e. need room for 8 chars */
1770 const char *s = SvPVX_const(sv);
1771 const char * const end = s + SvCUR(sv);
1772 for ( ; s < end && d < limit; s++ ) {
1774 if (ch & 128 && !isPRINT_LC(ch)) {
1783 else if (ch == '\r') {
1787 else if (ch == '\f') {
1791 else if (ch == '\\') {
1795 else if (ch == '\0') {
1799 else if (isPRINT_LC(ch))
1816 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1817 "Argument \"%s\" isn't numeric in %s", pv,
1820 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1821 "Argument \"%s\" isn't numeric", pv);
1825 =for apidoc looks_like_number
1827 Test if the content of an SV looks like a number (or is a number).
1828 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1829 non-numeric warning), even if your atof() doesn't grok them.
1835 Perl_looks_like_number(pTHX_ SV *const sv)
1837 register const char *sbegin;
1840 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1843 sbegin = SvPVX_const(sv);
1846 else if (SvPOKp(sv))
1847 sbegin = SvPV_const(sv, len);
1849 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1850 return grok_number(sbegin, len, NULL);
1854 S_glob_2number(pTHX_ GV * const gv)
1856 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1857 SV *const buffer = sv_newmortal();
1859 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1861 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1864 gv_efullname3(buffer, gv, "*");
1865 SvFLAGS(gv) |= wasfake;
1867 /* We know that all GVs stringify to something that is not-a-number,
1868 so no need to test that. */
1869 if (ckWARN(WARN_NUMERIC))
1870 not_a_number(buffer);
1871 /* We just want something true to return, so that S_sv_2iuv_common
1872 can tail call us and return true. */
1876 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1877 until proven guilty, assume that things are not that bad... */
1882 As 64 bit platforms often have an NV that doesn't preserve all bits of
1883 an IV (an assumption perl has been based on to date) it becomes necessary
1884 to remove the assumption that the NV always carries enough precision to
1885 recreate the IV whenever needed, and that the NV is the canonical form.
1886 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1887 precision as a side effect of conversion (which would lead to insanity
1888 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1889 1) to distinguish between IV/UV/NV slots that have cached a valid
1890 conversion where precision was lost and IV/UV/NV slots that have a
1891 valid conversion which has lost no precision
1892 2) to ensure that if a numeric conversion to one form is requested that
1893 would lose precision, the precise conversion (or differently
1894 imprecise conversion) is also performed and cached, to prevent
1895 requests for different numeric formats on the same SV causing
1896 lossy conversion chains. (lossless conversion chains are perfectly
1901 SvIOKp is true if the IV slot contains a valid value
1902 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1903 SvNOKp is true if the NV slot contains a valid value
1904 SvNOK is true only if the NV value is accurate
1907 while converting from PV to NV, check to see if converting that NV to an
1908 IV(or UV) would lose accuracy over a direct conversion from PV to
1909 IV(or UV). If it would, cache both conversions, return NV, but mark
1910 SV as IOK NOKp (ie not NOK).
1912 While converting from PV to IV, check to see if converting that IV to an
1913 NV would lose accuracy over a direct conversion from PV to NV. If it
1914 would, cache both conversions, flag similarly.
1916 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1917 correctly because if IV & NV were set NV *always* overruled.
1918 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1919 changes - now IV and NV together means that the two are interchangeable:
1920 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1922 The benefit of this is that operations such as pp_add know that if
1923 SvIOK is true for both left and right operands, then integer addition
1924 can be used instead of floating point (for cases where the result won't
1925 overflow). Before, floating point was always used, which could lead to
1926 loss of precision compared with integer addition.
1928 * making IV and NV equal status should make maths accurate on 64 bit
1930 * may speed up maths somewhat if pp_add and friends start to use
1931 integers when possible instead of fp. (Hopefully the overhead in
1932 looking for SvIOK and checking for overflow will not outweigh the
1933 fp to integer speedup)
1934 * will slow down integer operations (callers of SvIV) on "inaccurate"
1935 values, as the change from SvIOK to SvIOKp will cause a call into
1936 sv_2iv each time rather than a macro access direct to the IV slot
1937 * should speed up number->string conversion on integers as IV is
1938 favoured when IV and NV are equally accurate
1940 ####################################################################
1941 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1942 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1943 On the other hand, SvUOK is true iff UV.
1944 ####################################################################
1946 Your mileage will vary depending your CPU's relative fp to integer
1950 #ifndef NV_PRESERVES_UV
1951 # define IS_NUMBER_UNDERFLOW_IV 1
1952 # define IS_NUMBER_UNDERFLOW_UV 2
1953 # define IS_NUMBER_IV_AND_UV 2
1954 # define IS_NUMBER_OVERFLOW_IV 4
1955 # define IS_NUMBER_OVERFLOW_UV 5
1957 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1959 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1961 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1969 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1971 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));
1972 if (SvNVX(sv) < (NV)IV_MIN) {
1973 (void)SvIOKp_on(sv);
1975 SvIV_set(sv, IV_MIN);
1976 return IS_NUMBER_UNDERFLOW_IV;
1978 if (SvNVX(sv) > (NV)UV_MAX) {
1979 (void)SvIOKp_on(sv);
1982 SvUV_set(sv, UV_MAX);
1983 return IS_NUMBER_OVERFLOW_UV;
1985 (void)SvIOKp_on(sv);
1987 /* Can't use strtol etc to convert this string. (See truth table in
1989 if (SvNVX(sv) <= (UV)IV_MAX) {
1990 SvIV_set(sv, I_V(SvNVX(sv)));
1991 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1992 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1994 /* Integer is imprecise. NOK, IOKp */
1996 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1999 SvUV_set(sv, U_V(SvNVX(sv)));
2000 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2001 if (SvUVX(sv) == UV_MAX) {
2002 /* As we know that NVs don't preserve UVs, UV_MAX cannot
2003 possibly be preserved by NV. Hence, it must be overflow.
2005 return IS_NUMBER_OVERFLOW_UV;
2007 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2009 /* Integer is imprecise. NOK, IOKp */
2011 return IS_NUMBER_OVERFLOW_IV;
2013 #endif /* !NV_PRESERVES_UV*/
2016 S_sv_2iuv_common(pTHX_ SV *const sv)
2020 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2023 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2024 * without also getting a cached IV/UV from it at the same time
2025 * (ie PV->NV conversion should detect loss of accuracy and cache
2026 * IV or UV at same time to avoid this. */
2027 /* IV-over-UV optimisation - choose to cache IV if possible */
2029 if (SvTYPE(sv) == SVt_NV)
2030 sv_upgrade(sv, SVt_PVNV);
2032 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2033 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2034 certainly cast into the IV range at IV_MAX, whereas the correct
2035 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2037 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2038 if (Perl_isnan(SvNVX(sv))) {
2044 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2045 SvIV_set(sv, I_V(SvNVX(sv)));
2046 if (SvNVX(sv) == (NV) SvIVX(sv)
2047 #ifndef NV_PRESERVES_UV
2048 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2049 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2050 /* Don't flag it as "accurately an integer" if the number
2051 came from a (by definition imprecise) NV operation, and
2052 we're outside the range of NV integer precision */
2056 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2058 /* scalar has trailing garbage, eg "42a" */
2060 DEBUG_c(PerlIO_printf(Perl_debug_log,
2061 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2067 /* IV not precise. No need to convert from PV, as NV
2068 conversion would already have cached IV if it detected
2069 that PV->IV would be better than PV->NV->IV
2070 flags already correct - don't set public IOK. */
2071 DEBUG_c(PerlIO_printf(Perl_debug_log,
2072 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2077 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2078 but the cast (NV)IV_MIN rounds to a the value less (more
2079 negative) than IV_MIN which happens to be equal to SvNVX ??
2080 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2081 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2082 (NV)UVX == NVX are both true, but the values differ. :-(
2083 Hopefully for 2s complement IV_MIN is something like
2084 0x8000000000000000 which will be exact. NWC */
2087 SvUV_set(sv, U_V(SvNVX(sv)));
2089 (SvNVX(sv) == (NV) SvUVX(sv))
2090 #ifndef NV_PRESERVES_UV
2091 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2092 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2093 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2094 /* Don't flag it as "accurately an integer" if the number
2095 came from a (by definition imprecise) NV operation, and
2096 we're outside the range of NV integer precision */
2102 DEBUG_c(PerlIO_printf(Perl_debug_log,
2103 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2109 else if (SvPOKp(sv) && SvLEN(sv)) {
2111 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2112 /* We want to avoid a possible problem when we cache an IV/ a UV which
2113 may be later translated to an NV, and the resulting NV is not
2114 the same as the direct translation of the initial string
2115 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2116 be careful to ensure that the value with the .456 is around if the
2117 NV value is requested in the future).
2119 This means that if we cache such an IV/a UV, we need to cache the
2120 NV as well. Moreover, we trade speed for space, and do not
2121 cache the NV if we are sure it's not needed.
2124 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2125 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2126 == IS_NUMBER_IN_UV) {
2127 /* It's definitely an integer, only upgrade to PVIV */
2128 if (SvTYPE(sv) < SVt_PVIV)
2129 sv_upgrade(sv, SVt_PVIV);
2131 } else if (SvTYPE(sv) < SVt_PVNV)
2132 sv_upgrade(sv, SVt_PVNV);
2134 /* If NVs preserve UVs then we only use the UV value if we know that
2135 we aren't going to call atof() below. If NVs don't preserve UVs
2136 then the value returned may have more precision than atof() will
2137 return, even though value isn't perfectly accurate. */
2138 if ((numtype & (IS_NUMBER_IN_UV
2139 #ifdef NV_PRESERVES_UV
2142 )) == IS_NUMBER_IN_UV) {
2143 /* This won't turn off the public IOK flag if it was set above */
2144 (void)SvIOKp_on(sv);
2146 if (!(numtype & IS_NUMBER_NEG)) {
2148 if (value <= (UV)IV_MAX) {
2149 SvIV_set(sv, (IV)value);
2151 /* it didn't overflow, and it was positive. */
2152 SvUV_set(sv, value);
2156 /* 2s complement assumption */
2157 if (value <= (UV)IV_MIN) {
2158 SvIV_set(sv, -(IV)value);
2160 /* Too negative for an IV. This is a double upgrade, but
2161 I'm assuming it will be rare. */
2162 if (SvTYPE(sv) < SVt_PVNV)
2163 sv_upgrade(sv, SVt_PVNV);
2167 SvNV_set(sv, -(NV)value);
2168 SvIV_set(sv, IV_MIN);
2172 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2173 will be in the previous block to set the IV slot, and the next
2174 block to set the NV slot. So no else here. */
2176 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2177 != IS_NUMBER_IN_UV) {
2178 /* It wasn't an (integer that doesn't overflow the UV). */
2179 SvNV_set(sv, Atof(SvPVX_const(sv)));
2181 if (! numtype && ckWARN(WARN_NUMERIC))
2184 #if defined(USE_LONG_DOUBLE)
2185 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2186 PTR2UV(sv), SvNVX(sv)));
2188 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2189 PTR2UV(sv), SvNVX(sv)));
2192 #ifdef NV_PRESERVES_UV
2193 (void)SvIOKp_on(sv);
2195 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2196 SvIV_set(sv, I_V(SvNVX(sv)));
2197 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2200 NOOP; /* Integer is imprecise. NOK, IOKp */
2202 /* UV will not work better than IV */
2204 if (SvNVX(sv) > (NV)UV_MAX) {
2206 /* Integer is inaccurate. NOK, IOKp, is UV */
2207 SvUV_set(sv, UV_MAX);
2209 SvUV_set(sv, U_V(SvNVX(sv)));
2210 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2211 NV preservse UV so can do correct comparison. */
2212 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2215 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2220 #else /* NV_PRESERVES_UV */
2221 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2222 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2223 /* The IV/UV slot will have been set from value returned by
2224 grok_number above. The NV slot has just been set using
2227 assert (SvIOKp(sv));
2229 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2230 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2231 /* Small enough to preserve all bits. */
2232 (void)SvIOKp_on(sv);
2234 SvIV_set(sv, I_V(SvNVX(sv)));
2235 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2237 /* Assumption: first non-preserved integer is < IV_MAX,
2238 this NV is in the preserved range, therefore: */
2239 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2241 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);
2245 0 0 already failed to read UV.
2246 0 1 already failed to read UV.
2247 1 0 you won't get here in this case. IV/UV
2248 slot set, public IOK, Atof() unneeded.
2249 1 1 already read UV.
2250 so there's no point in sv_2iuv_non_preserve() attempting
2251 to use atol, strtol, strtoul etc. */
2253 sv_2iuv_non_preserve (sv, numtype);
2255 sv_2iuv_non_preserve (sv);
2259 #endif /* NV_PRESERVES_UV */
2260 /* It might be more code efficient to go through the entire logic above
2261 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2262 gets complex and potentially buggy, so more programmer efficient
2263 to do it this way, by turning off the public flags: */
2265 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2269 if (isGV_with_GP(sv))
2270 return glob_2number(MUTABLE_GV(sv));
2272 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2273 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2276 if (SvTYPE(sv) < SVt_IV)
2277 /* Typically the caller expects that sv_any is not NULL now. */
2278 sv_upgrade(sv, SVt_IV);
2279 /* Return 0 from the caller. */
2286 =for apidoc sv_2iv_flags
2288 Return the integer value of an SV, doing any necessary string
2289 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2290 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2296 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2301 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2302 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2303 cache IVs just in case. In practice it seems that they never
2304 actually anywhere accessible by user Perl code, let alone get used
2305 in anything other than a string context. */
2306 if (flags & SV_GMAGIC)
2311 return I_V(SvNVX(sv));
2313 if (SvPOKp(sv) && SvLEN(sv)) {
2316 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2318 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2319 == IS_NUMBER_IN_UV) {
2320 /* It's definitely an integer */
2321 if (numtype & IS_NUMBER_NEG) {
2322 if (value < (UV)IV_MIN)
2325 if (value < (UV)IV_MAX)
2330 if (ckWARN(WARN_NUMERIC))
2333 return I_V(Atof(SvPVX_const(sv)));
2338 assert(SvTYPE(sv) >= SVt_PVMG);
2339 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2340 } else if (SvTHINKFIRST(sv)) {
2344 SV * const tmpstr=AMG_CALLun(sv,numer);
2345 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2346 return SvIV(tmpstr);
2349 return PTR2IV(SvRV(sv));
2352 sv_force_normal_flags(sv, 0);
2354 if (SvREADONLY(sv) && !SvOK(sv)) {
2355 if (ckWARN(WARN_UNINITIALIZED))
2361 if (S_sv_2iuv_common(aTHX_ sv))
2364 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2365 PTR2UV(sv),SvIVX(sv)));
2366 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2370 =for apidoc sv_2uv_flags
2372 Return the unsigned integer value of an SV, doing any necessary string
2373 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2374 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2380 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2385 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2386 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2387 cache IVs just in case. */
2388 if (flags & SV_GMAGIC)
2393 return U_V(SvNVX(sv));
2394 if (SvPOKp(sv) && SvLEN(sv)) {
2397 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2399 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2400 == IS_NUMBER_IN_UV) {
2401 /* It's definitely an integer */
2402 if (!(numtype & IS_NUMBER_NEG))
2406 if (ckWARN(WARN_NUMERIC))
2409 return U_V(Atof(SvPVX_const(sv)));
2414 assert(SvTYPE(sv) >= SVt_PVMG);
2415 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2416 } else if (SvTHINKFIRST(sv)) {
2420 SV *const tmpstr = AMG_CALLun(sv,numer);
2421 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2422 return SvUV(tmpstr);
2425 return PTR2UV(SvRV(sv));
2428 sv_force_normal_flags(sv, 0);
2430 if (SvREADONLY(sv) && !SvOK(sv)) {
2431 if (ckWARN(WARN_UNINITIALIZED))
2437 if (S_sv_2iuv_common(aTHX_ sv))
2441 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2442 PTR2UV(sv),SvUVX(sv)));
2443 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2449 Return the num value of an SV, doing any necessary string or integer
2450 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2457 Perl_sv_2nv(pTHX_ register SV *const sv)
2462 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2463 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2464 cache IVs just in case. */
2468 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2469 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2470 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2472 return Atof(SvPVX_const(sv));
2476 return (NV)SvUVX(sv);
2478 return (NV)SvIVX(sv);
2483 assert(SvTYPE(sv) >= SVt_PVMG);
2484 /* This falls through to the report_uninit near the end of the
2486 } else if (SvTHINKFIRST(sv)) {
2490 SV *const tmpstr = AMG_CALLun(sv,numer);
2491 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2492 return SvNV(tmpstr);
2495 return PTR2NV(SvRV(sv));
2498 sv_force_normal_flags(sv, 0);
2500 if (SvREADONLY(sv) && !SvOK(sv)) {
2501 if (ckWARN(WARN_UNINITIALIZED))
2506 if (SvTYPE(sv) < SVt_NV) {
2507 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2508 sv_upgrade(sv, SVt_NV);
2509 #ifdef USE_LONG_DOUBLE
2511 STORE_NUMERIC_LOCAL_SET_STANDARD();
2512 PerlIO_printf(Perl_debug_log,
2513 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2514 PTR2UV(sv), SvNVX(sv));
2515 RESTORE_NUMERIC_LOCAL();
2519 STORE_NUMERIC_LOCAL_SET_STANDARD();
2520 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2521 PTR2UV(sv), SvNVX(sv));
2522 RESTORE_NUMERIC_LOCAL();
2526 else if (SvTYPE(sv) < SVt_PVNV)
2527 sv_upgrade(sv, SVt_PVNV);
2532 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2533 #ifdef NV_PRESERVES_UV
2539 /* Only set the public NV OK flag if this NV preserves the IV */
2540 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2542 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2543 : (SvIVX(sv) == I_V(SvNVX(sv))))
2549 else if (SvPOKp(sv) && SvLEN(sv)) {
2551 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2552 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2554 #ifdef NV_PRESERVES_UV
2555 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2556 == IS_NUMBER_IN_UV) {
2557 /* It's definitely an integer */
2558 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2560 SvNV_set(sv, Atof(SvPVX_const(sv)));
2566 SvNV_set(sv, Atof(SvPVX_const(sv)));
2567 /* Only set the public NV OK flag if this NV preserves the value in
2568 the PV at least as well as an IV/UV would.
2569 Not sure how to do this 100% reliably. */
2570 /* if that shift count is out of range then Configure's test is
2571 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2573 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2574 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2575 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2576 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2577 /* Can't use strtol etc to convert this string, so don't try.
2578 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2581 /* value has been set. It may not be precise. */
2582 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2583 /* 2s complement assumption for (UV)IV_MIN */
2584 SvNOK_on(sv); /* Integer is too negative. */
2589 if (numtype & IS_NUMBER_NEG) {
2590 SvIV_set(sv, -(IV)value);
2591 } else if (value <= (UV)IV_MAX) {
2592 SvIV_set(sv, (IV)value);
2594 SvUV_set(sv, value);
2598 if (numtype & IS_NUMBER_NOT_INT) {
2599 /* I believe that even if the original PV had decimals,
2600 they are lost beyond the limit of the FP precision.
2601 However, neither is canonical, so both only get p
2602 flags. NWC, 2000/11/25 */
2603 /* Both already have p flags, so do nothing */
2605 const NV nv = SvNVX(sv);
2606 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2607 if (SvIVX(sv) == I_V(nv)) {
2610 /* It had no "." so it must be integer. */
2614 /* between IV_MAX and NV(UV_MAX).
2615 Could be slightly > UV_MAX */
2617 if (numtype & IS_NUMBER_NOT_INT) {
2618 /* UV and NV both imprecise. */
2620 const UV nv_as_uv = U_V(nv);
2622 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2631 /* It might be more code efficient to go through the entire logic above
2632 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2633 gets complex and potentially buggy, so more programmer efficient
2634 to do it this way, by turning off the public flags: */
2636 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2637 #endif /* NV_PRESERVES_UV */
2640 if (isGV_with_GP(sv)) {
2641 glob_2number(MUTABLE_GV(sv));
2645 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2647 assert (SvTYPE(sv) >= SVt_NV);
2648 /* Typically the caller expects that sv_any is not NULL now. */
2649 /* XXX Ilya implies that this is a bug in callers that assume this
2650 and ideally should be fixed. */
2653 #if defined(USE_LONG_DOUBLE)
2655 STORE_NUMERIC_LOCAL_SET_STANDARD();
2656 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2657 PTR2UV(sv), SvNVX(sv));
2658 RESTORE_NUMERIC_LOCAL();
2662 STORE_NUMERIC_LOCAL_SET_STANDARD();
2663 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2664 PTR2UV(sv), SvNVX(sv));
2665 RESTORE_NUMERIC_LOCAL();
2674 Return an SV with the numeric value of the source SV, doing any necessary
2675 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2676 access this function.
2682 Perl_sv_2num(pTHX_ register SV *const sv)
2684 PERL_ARGS_ASSERT_SV_2NUM;
2689 SV * const tmpsv = AMG_CALLun(sv,numer);
2690 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2691 return sv_2num(tmpsv);
2693 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2696 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2697 * UV as a string towards the end of buf, and return pointers to start and
2700 * We assume that buf is at least TYPE_CHARS(UV) long.
2704 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2706 char *ptr = buf + TYPE_CHARS(UV);
2707 char * const ebuf = ptr;
2710 PERL_ARGS_ASSERT_UIV_2BUF;
2722 *--ptr = '0' + (char)(uv % 10);
2731 =for apidoc sv_2pv_flags
2733 Returns a pointer to the string value of an SV, and sets *lp to its length.
2734 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2736 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2737 usually end up here too.
2743 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2753 if (SvGMAGICAL(sv)) {
2754 if (flags & SV_GMAGIC)
2759 if (flags & SV_MUTABLE_RETURN)
2760 return SvPVX_mutable(sv);
2761 if (flags & SV_CONST_RETURN)
2762 return (char *)SvPVX_const(sv);
2765 if (SvIOKp(sv) || SvNOKp(sv)) {
2766 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2771 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2772 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2774 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2781 #ifdef FIXNEGATIVEZERO
2782 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2788 SvUPGRADE(sv, SVt_PV);
2791 s = SvGROW_mutable(sv, len + 1);
2794 return (char*)memcpy(s, tbuf, len + 1);
2800 assert(SvTYPE(sv) >= SVt_PVMG);
2801 /* This falls through to the report_uninit near the end of the
2803 } else if (SvTHINKFIRST(sv)) {
2807 SV *const tmpstr = AMG_CALLun(sv,string);
2808 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2810 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2814 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2815 if (flags & SV_CONST_RETURN) {
2816 pv = (char *) SvPVX_const(tmpstr);
2818 pv = (flags & SV_MUTABLE_RETURN)
2819 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2822 *lp = SvCUR(tmpstr);
2824 pv = sv_2pv_flags(tmpstr, lp, flags);
2837 SV *const referent = SvRV(sv);
2841 retval = buffer = savepvn("NULLREF", len);
2842 } else if (SvTYPE(referent) == SVt_REGEXP) {
2843 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2848 /* If the regex is UTF-8 we want the containing scalar to
2849 have an UTF-8 flag too */
2855 if ((seen_evals = RX_SEEN_EVALS(re)))
2856 PL_reginterp_cnt += seen_evals;
2859 *lp = RX_WRAPLEN(re);
2861 return RX_WRAPPED(re);
2863 const char *const typestr = sv_reftype(referent, 0);
2864 const STRLEN typelen = strlen(typestr);
2865 UV addr = PTR2UV(referent);
2866 const char *stashname = NULL;
2867 STRLEN stashnamelen = 0; /* hush, gcc */
2868 const char *buffer_end;
2870 if (SvOBJECT(referent)) {
2871 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2874 stashname = HEK_KEY(name);
2875 stashnamelen = HEK_LEN(name);
2877 if (HEK_UTF8(name)) {
2883 stashname = "__ANON__";
2886 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2887 + 2 * sizeof(UV) + 2 /* )\0 */;
2889 len = typelen + 3 /* (0x */
2890 + 2 * sizeof(UV) + 2 /* )\0 */;
2893 Newx(buffer, len, char);
2894 buffer_end = retval = buffer + len;
2896 /* Working backwards */
2900 *--retval = PL_hexdigit[addr & 15];
2901 } while (addr >>= 4);
2907 memcpy(retval, typestr, typelen);
2911 retval -= stashnamelen;
2912 memcpy(retval, stashname, stashnamelen);
2914 /* retval may not neccesarily have reached the start of the
2916 assert (retval >= buffer);
2918 len = buffer_end - retval - 1; /* -1 for that \0 */
2926 if (SvREADONLY(sv) && !SvOK(sv)) {
2929 if (flags & SV_UNDEF_RETURNS_NULL)
2931 if (ckWARN(WARN_UNINITIALIZED))
2936 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2937 /* I'm assuming that if both IV and NV are equally valid then
2938 converting the IV is going to be more efficient */
2939 const U32 isUIOK = SvIsUV(sv);
2940 char buf[TYPE_CHARS(UV)];
2944 if (SvTYPE(sv) < SVt_PVIV)
2945 sv_upgrade(sv, SVt_PVIV);
2946 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2948 /* inlined from sv_setpvn */
2949 s = SvGROW_mutable(sv, len + 1);
2950 Move(ptr, s, len, char);
2954 else if (SvNOKp(sv)) {
2956 if (SvTYPE(sv) < SVt_PVNV)
2957 sv_upgrade(sv, SVt_PVNV);
2958 /* The +20 is pure guesswork. Configure test needed. --jhi */
2959 s = SvGROW_mutable(sv, NV_DIG + 20);
2960 /* some Xenix systems wipe out errno here */
2962 if (SvNVX(sv) == 0.0)
2963 my_strlcpy(s, "0", SvLEN(sv));
2967 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2970 #ifdef FIXNEGATIVEZERO
2971 if (*s == '-' && s[1] == '0' && !s[2]) {
2983 if (isGV_with_GP(sv)) {
2984 GV *const gv = MUTABLE_GV(sv);
2985 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2986 SV *const buffer = sv_newmortal();
2988 /* FAKE globs can get coerced, so need to turn this off temporarily
2991 gv_efullname3(buffer, gv, "*");
2992 SvFLAGS(gv) |= wasfake;
2994 assert(SvPOK(buffer));
2996 *lp = SvCUR(buffer);
2998 return SvPVX(buffer);
3003 if (flags & SV_UNDEF_RETURNS_NULL)
3005 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3007 if (SvTYPE(sv) < SVt_PV)
3008 /* Typically the caller expects that sv_any is not NULL now. */
3009 sv_upgrade(sv, SVt_PV);
3013 const STRLEN len = s - SvPVX_const(sv);
3019 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3020 PTR2UV(sv),SvPVX_const(sv)));
3021 if (flags & SV_CONST_RETURN)
3022 return (char *)SvPVX_const(sv);
3023 if (flags & SV_MUTABLE_RETURN)
3024 return SvPVX_mutable(sv);
3029 =for apidoc sv_copypv
3031 Copies a stringified representation of the source SV into the
3032 destination SV. Automatically performs any necessary mg_get and
3033 coercion of numeric values into strings. Guaranteed to preserve
3034 UTF8 flag even from overloaded objects. Similar in nature to
3035 sv_2pv[_flags] but operates directly on an SV instead of just the
3036 string. Mostly uses sv_2pv_flags to do its work, except when that
3037 would lose the UTF-8'ness of the PV.
3043 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3046 const char * const s = SvPV_const(ssv,len);
3048 PERL_ARGS_ASSERT_SV_COPYPV;
3050 sv_setpvn(dsv,s,len);
3058 =for apidoc sv_2pvbyte
3060 Return a pointer to the byte-encoded representation of the SV, and set *lp
3061 to its length. May cause the SV to be downgraded from UTF-8 as a
3064 Usually accessed via the C<SvPVbyte> macro.
3070 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3072 PERL_ARGS_ASSERT_SV_2PVBYTE;
3074 sv_utf8_downgrade(sv,0);
3075 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3079 =for apidoc sv_2pvutf8
3081 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3082 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3084 Usually accessed via the C<SvPVutf8> macro.
3090 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3092 PERL_ARGS_ASSERT_SV_2PVUTF8;
3094 sv_utf8_upgrade(sv);
3095 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3100 =for apidoc sv_2bool
3102 This function is only called on magical items, and is only used by
3103 sv_true() or its macro equivalent.
3109 Perl_sv_2bool(pTHX_ register SV *const sv)
3113 PERL_ARGS_ASSERT_SV_2BOOL;
3121 SV * const tmpsv = AMG_CALLun(sv,bool_);
3122 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3123 return (bool)SvTRUE(tmpsv);
3125 return SvRV(sv) != 0;
3128 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3130 (*sv->sv_u.svu_pv > '0' ||
3131 Xpvtmp->xpv_cur > 1 ||
3132 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3139 return SvIVX(sv) != 0;
3142 return SvNVX(sv) != 0.0;
3144 if (isGV_with_GP(sv))
3154 =for apidoc sv_utf8_upgrade
3156 Converts the PV of an SV to its UTF-8-encoded form.
3157 Forces the SV to string form if it is not already.
3158 Will C<mg_get> on C<sv> if appropriate.
3159 Always sets the SvUTF8 flag to avoid future validity checks even
3160 if the whole string is the same in UTF-8 as not.
3161 Returns the number of bytes in the converted string
3163 This is not as a general purpose byte encoding to Unicode interface:
3164 use the Encode extension for that.
3166 =for apidoc sv_utf8_upgrade_nomg
3168 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3170 =for apidoc sv_utf8_upgrade_flags
3172 Converts the PV of an SV to its UTF-8-encoded form.
3173 Forces the SV to string form if it is not already.
3174 Always sets the SvUTF8 flag to avoid future validity checks even
3175 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3176 will C<mg_get> on C<sv> if appropriate, else not.
3177 Returns the number of bytes in the converted string
3178 C<sv_utf8_upgrade> and
3179 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3181 This is not as a general purpose byte encoding to Unicode interface:
3182 use the Encode extension for that.
3186 The grow version is currently not externally documented. It adds a parameter,
3187 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3188 have free after it upon return. This allows the caller to reserve extra space
3189 that it intends to fill, to avoid extra grows.
3191 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3192 which can be used to tell this function to not first check to see if there are
3193 any characters that are different in UTF-8 (variant characters) which would
3194 force it to allocate a new string to sv, but to assume there are. Typically
3195 this flag is used by a routine that has already parsed the string to find that
3196 there are such characters, and passes this information on so that the work
3197 doesn't have to be repeated.
3199 (One might think that the calling routine could pass in the position of the
3200 first such variant, so it wouldn't have to be found again. But that is not the
3201 case, because typically when the caller is likely to use this flag, it won't be
3202 calling this routine unless it finds something that won't fit into a byte.
3203 Otherwise it tries to not upgrade and just use bytes. But some things that
3204 do fit into a byte are variants in utf8, and the caller may not have been
3205 keeping track of these.)
3207 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3208 isn't guaranteed due to having other routines do the work in some input cases,
3209 or if the input is already flagged as being in utf8.
3211 The speed of this could perhaps be improved for many cases if someone wanted to
3212 write a fast function that counts the number of variant characters in a string,
3213 especially if it could return the position of the first one.
3218 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3222 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3224 if (sv == &PL_sv_undef)
3228 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3229 (void) sv_2pv_flags(sv,&len, flags);
3231 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3235 (void) SvPV_force(sv,len);
3240 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3245 sv_force_normal_flags(sv, 0);
3248 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3249 sv_recode_to_utf8(sv, PL_encoding);
3250 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3254 if (SvCUR(sv) > 0) { /* Assume Latin-1/EBCDIC */
3255 /* This function could be much more efficient if we
3256 * had a FLAG in SVs to signal if there are any variant
3257 * chars in the PV. Given that there isn't such a flag
3258 * make the loop as fast as possible (although there are certainly ways
3259 * to speed this up, eg. through vectorization) */
3260 U8 * s = (U8 *) SvPVX_const(sv);
3261 U8 * e = (U8 *) SvEND(sv);
3263 STRLEN two_byte_count = 0;
3265 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3267 /* See if really will need to convert to utf8. We mustn't rely on our
3268 * incoming SV being well formed and having a trailing '\0', as certain
3269 * code in pp_formline can send us partially built SVs. */
3273 if (NATIVE_IS_INVARIANT(ch)) continue;
3275 t--; /* t already incremented; re-point to first variant */
3280 /* utf8 conversion not needed because all are invariants. Mark as
3281 * UTF-8 even if no variant - saves scanning loop */
3287 /* Here, the string should be converted to utf8, either because of an
3288 * input flag (two_byte_count = 0), or because a character that
3289 * requires 2 bytes was found (two_byte_count = 1). t points either to
3290 * the beginning of the string (if we didn't examine anything), or to
3291 * the first variant. In either case, everything from s to t - 1 will
3292 * occupy only 1 byte each on output.
3294 * There are two main ways to convert. One is to create a new string
3295 * and go through the input starting from the beginning, appending each
3296 * converted value onto the new string as we go along. It's probably
3297 * best to allocate enough space in the string for the worst possible
3298 * case rather than possibly running out of space and having to
3299 * reallocate and then copy what we've done so far. Since everything
3300 * from s to t - 1 is invariant, the destination can be initialized
3301 * with these using a fast memory copy
3303 * The other way is to figure out exactly how big the string should be
3304 * by parsing the entire input. Then you don't have to make it big
3305 * enough to handle the worst possible case, and more importantly, if
3306 * the string you already have is large enough, you don't have to
3307 * allocate a new string, you can copy the last character in the input
3308 * string to the final position(s) that will be occupied by the
3309 * converted string and go backwards, stopping at t, since everything
3310 * before that is invariant.
3312 * There are advantages and disadvantages to each method.
3314 * In the first method, we can allocate a new string, do the memory
3315 * copy from the s to t - 1, and then proceed through the rest of the
3316 * string byte-by-byte.
3318 * In the second method, we proceed through the rest of the input
3319 * string just calculating how big the converted string will be. Then
3320 * there are two cases:
3321 * 1) if the string has enough extra space to handle the converted
3322 * value. We go backwards through the string, converting until we
3323 * get to the position we are at now, and then stop. If this
3324 * position is far enough along in the string, this method is
3325 * faster than the other method. If the memory copy were the same
3326 * speed as the byte-by-byte loop, that position would be about
3327 * half-way, as at the half-way mark, parsing to the end and back
3328 * is one complete string's parse, the same amount as starting
3329 * over and going all the way through. Actually, it would be
3330 * somewhat less than half-way, as it's faster to just count bytes
3331 * than to also copy, and we don't have the overhead of allocating
3332 * a new string, changing the scalar to use it, and freeing the
3333 * existing one. But if the memory copy is fast, the break-even
3334 * point is somewhere after half way. The counting loop could be
3335 * sped up by vectorization, etc, to move the break-even point
3336 * further towards the beginning.
3337 * 2) if the string doesn't have enough space to handle the converted
3338 * value. A new string will have to be allocated, and one might
3339 * as well, given that, start from the beginning doing the first
3340 * method. We've spent extra time parsing the string and in
3341 * exchange all we've gotten is that we know precisely how big to
3342 * make the new one. Perl is more optimized for time than space,
3343 * so this case is a loser.
3344 * So what I've decided to do is not use the 2nd method unless it is
3345 * guaranteed that a new string won't have to be allocated, assuming
3346 * the worst case. I also decided not to put any more conditions on it
3347 * than this, for now. It seems likely that, since the worst case is
3348 * twice as big as the unknown portion of the string (plus 1), we won't
3349 * be guaranteed enough space, causing us to go to the first method,
3350 * unless the string is short, or the first variant character is near
3351 * the end of it. In either of these cases, it seems best to use the
3352 * 2nd method. The only circumstance I can think of where this would
3353 * be really slower is if the string had once had much more data in it
3354 * than it does now, but there is still a substantial amount in it */
3357 STRLEN invariant_head = t - s;
3358 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3359 if (SvLEN(sv) < size) {
3361 /* Here, have decided to allocate a new string */
3366 Newx(dst, size, U8);
3368 /* If no known invariants at the beginning of the input string,
3369 * set so starts from there. Otherwise, can use memory copy to
3370 * get up to where we are now, and then start from here */
3372 if (invariant_head <= 0) {
3375 Copy(s, dst, invariant_head, char);
3376 d = dst + invariant_head;
3380 const UV uv = NATIVE8_TO_UNI(*t++);
3381 if (UNI_IS_INVARIANT(uv))
3382 *d++ = (U8)UNI_TO_NATIVE(uv);
3384 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3385 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3389 SvPV_free(sv); /* No longer using pre-existing string */
3390 SvPV_set(sv, (char*)dst);
3391 SvCUR_set(sv, d - dst);
3392 SvLEN_set(sv, size);
3395 /* Here, have decided to get the exact size of the string.
3396 * Currently this happens only when we know that there is
3397 * guaranteed enough space to fit the converted string, so
3398 * don't have to worry about growing. If two_byte_count is 0,
3399 * then t points to the first byte of the string which hasn't
3400 * been examined yet. Otherwise two_byte_count is 1, and t
3401 * points to the first byte in the string that will expand to
3402 * two. Depending on this, start examining at t or 1 after t.
3405 U8 *d = t + two_byte_count;
3408 /* Count up the remaining bytes that expand to two */
3411 const U8 chr = *d++;
3412 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3415 /* The string will expand by just the number of bytes that
3416 * occupy two positions. But we are one afterwards because of
3417 * the increment just above. This is the place to put the
3418 * trailing NUL, and to set the length before we decrement */
3420 d += two_byte_count;
3421 SvCUR_set(sv, d - s);
3425 /* Having decremented d, it points to the position to put the
3426 * very last byte of the expanded string. Go backwards through
3427 * the string, copying and expanding as we go, stopping when we
3428 * get to the part that is invariant the rest of the way down */
3432 const U8 ch = NATIVE8_TO_UNI(*e--);
3433 if (UNI_IS_INVARIANT(ch)) {
3434 *d-- = UNI_TO_NATIVE(ch);
3436 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3437 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3444 /* Mark as UTF-8 even if no variant - saves scanning loop */
3450 =for apidoc sv_utf8_downgrade
3452 Attempts to convert the PV of an SV from characters to bytes.
3453 If the PV contains a character that cannot fit
3454 in a byte, this conversion will fail;
3455 in this case, either returns false or, if C<fail_ok> is not
3458 This is not as a general purpose Unicode to byte encoding interface:
3459 use the Encode extension for that.
3465 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3469 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3471 if (SvPOKp(sv) && SvUTF8(sv)) {
3477 sv_force_normal_flags(sv, 0);
3479 s = (U8 *) SvPV(sv, len);
3480 if (!utf8_to_bytes(s, &len)) {
3485 Perl_croak(aTHX_ "Wide character in %s",
3488 Perl_croak(aTHX_ "Wide character");
3499 =for apidoc sv_utf8_encode
3501 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3502 flag off so that it looks like octets again.
3508 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3510 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3513 sv_force_normal_flags(sv, 0);
3515 if (SvREADONLY(sv)) {
3516 Perl_croak(aTHX_ "%s", PL_no_modify);
3518 (void) sv_utf8_upgrade(sv);
3523 =for apidoc sv_utf8_decode
3525 If the PV of the SV is an octet sequence in UTF-8
3526 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3527 so that it looks like a character. If the PV contains only single-byte
3528 characters, the C<SvUTF8> flag stays being off.
3529 Scans PV for validity and returns false if the PV is invalid UTF-8.
3535 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3537 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3543 /* The octets may have got themselves encoded - get them back as
3546 if (!sv_utf8_downgrade(sv, TRUE))
3549 /* it is actually just a matter of turning the utf8 flag on, but
3550 * we want to make sure everything inside is valid utf8 first.
3552 c = (const U8 *) SvPVX_const(sv);
3553 if (!is_utf8_string(c, SvCUR(sv)+1))
3555 e = (const U8 *) SvEND(sv);
3558 if (!UTF8_IS_INVARIANT(ch)) {
3568 =for apidoc sv_setsv
3570 Copies the contents of the source SV C<ssv> into the destination SV
3571 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3572 function if the source SV needs to be reused. Does not handle 'set' magic.
3573 Loosely speaking, it performs a copy-by-value, obliterating any previous
3574 content of the destination.
3576 You probably want to use one of the assortment of wrappers, such as
3577 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3578 C<SvSetMagicSV_nosteal>.
3580 =for apidoc sv_setsv_flags
3582 Copies the contents of the source SV C<ssv> into the destination SV
3583 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3584 function if the source SV needs to be reused. Does not handle 'set' magic.
3585 Loosely speaking, it performs a copy-by-value, obliterating any previous
3586 content of the destination.
3587 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3588 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3589 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3590 and C<sv_setsv_nomg> are implemented in terms of this function.
3592 You probably want to use one of the assortment of wrappers, such as
3593 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3594 C<SvSetMagicSV_nosteal>.
3596 This is the primary function for copying scalars, and most other
3597 copy-ish functions and macros use this underneath.
3603 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3605 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3607 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3609 if (dtype != SVt_PVGV) {
3610 const char * const name = GvNAME(sstr);
3611 const STRLEN len = GvNAMELEN(sstr);
3613 if (dtype >= SVt_PV) {
3619 SvUPGRADE(dstr, SVt_PVGV);
3620 (void)SvOK_off(dstr);
3621 /* FIXME - why are we doing this, then turning it off and on again
3623 isGV_with_GP_on(dstr);
3625 GvSTASH(dstr) = GvSTASH(sstr);
3627 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3628 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3629 SvFAKE_on(dstr); /* can coerce to non-glob */
3632 if(GvGP(MUTABLE_GV(sstr))) {
3633 /* If source has method cache entry, clear it */
3635 SvREFCNT_dec(GvCV(sstr));
3639 /* If source has a real method, then a method is
3641 else if(GvCV((const GV *)sstr)) {
3646 /* If dest already had a real method, that's a change as well */
3647 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3651 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3654 gp_free(MUTABLE_GV(dstr));
3655 isGV_with_GP_off(dstr);
3656 (void)SvOK_off(dstr);
3657 isGV_with_GP_on(dstr);
3658 GvINTRO_off(dstr); /* one-shot flag */
3659 GvGP(dstr) = gp_ref(GvGP(sstr));
3660 if (SvTAINTED(sstr))
3662 if (GvIMPORTED(dstr) != GVf_IMPORTED
3663 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3665 GvIMPORTED_on(dstr);
3668 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3669 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3674 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3676 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3678 const int intro = GvINTRO(dstr);
3681 const U32 stype = SvTYPE(sref);
3682 bool mro_changes = FALSE;
3684 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3687 GvINTRO_off(dstr); /* one-shot flag */
3688 GvLINE(dstr) = CopLINE(PL_curcop);
3689 GvEGV(dstr) = MUTABLE_GV(dstr);
3694 location = (SV **) &GvCV(dstr);
3695 import_flag = GVf_IMPORTED_CV;
3698 location = (SV **) &GvHV(dstr);
3699 import_flag = GVf_IMPORTED_HV;
3702 location = (SV **) &GvAV(dstr);
3703 if (strEQ(GvNAME((GV*)dstr), "ISA"))
3705 import_flag = GVf_IMPORTED_AV;
3708 location = (SV **) &GvIOp(dstr);
3711 location = (SV **) &GvFORM(dstr);
3714 location = &GvSV(dstr);
3715 import_flag = GVf_IMPORTED_SV;
3718 if (stype == SVt_PVCV) {
3719 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3720 if (GvCVGEN(dstr)) {
3721 SvREFCNT_dec(GvCV(dstr));
3723 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3726 SAVEGENERICSV(*location);
3730 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3731 CV* const cv = MUTABLE_CV(*location);
3733 if (!GvCVGEN((const GV *)dstr) &&
3734 (CvROOT(cv) || CvXSUB(cv)))
3736 /* Redefining a sub - warning is mandatory if
3737 it was a const and its value changed. */
3738 if (CvCONST(cv) && CvCONST((const CV *)sref)
3740 == cv_const_sv((const CV *)sref)) {
3742 /* They are 2 constant subroutines generated from
3743 the same constant. This probably means that
3744 they are really the "same" proxy subroutine
3745 instantiated in 2 places. Most likely this is
3746 when a constant is exported twice. Don't warn.
3749 else if (ckWARN(WARN_REDEFINE)
3751 && (!CvCONST((const CV *)sref)
3752 || sv_cmp(cv_const_sv(cv),
3753 cv_const_sv((const CV *)
3755 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3758 ? "Constant subroutine %s::%s redefined"
3759 : "Subroutine %s::%s redefined"),
3760 HvNAME_get(GvSTASH((const GV *)dstr)),
3761 GvENAME(MUTABLE_GV(dstr)));
3765 cv_ckproto_len(cv, (const GV *)dstr,
3766 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3767 SvPOK(sref) ? SvCUR(sref) : 0);
3769 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3770 GvASSUMECV_on(dstr);
3771 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3774 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3775 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3776 GvFLAGS(dstr) |= import_flag;
3781 if (SvTAINTED(sstr))
3783 if (mro_changes) mro_isa_changed_in(GvSTASH(dstr));
3788 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3791 register U32 sflags;
3793 register svtype stype;
3795 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3800 if (SvIS_FREED(dstr)) {
3801 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3802 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3804 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3806 sstr = &PL_sv_undef;
3807 if (SvIS_FREED(sstr)) {
3808 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3809 (void*)sstr, (void*)dstr);
3811 stype = SvTYPE(sstr);
3812 dtype = SvTYPE(dstr);
3814 (void)SvAMAGIC_off(dstr);
3817 /* need to nuke the magic */
3821 /* There's a lot of redundancy below but we're going for speed here */
3826 if (dtype != SVt_PVGV) {
3827 (void)SvOK_off(dstr);
3835 sv_upgrade(dstr, SVt_IV);
3839 sv_upgrade(dstr, SVt_PVIV);
3842 goto end_of_first_switch;
3844 (void)SvIOK_only(dstr);
3845 SvIV_set(dstr, SvIVX(sstr));
3848 /* SvTAINTED can only be true if the SV has taint magic, which in
3849 turn means that the SV type is PVMG (or greater). This is the
3850 case statement for SVt_IV, so this cannot be true (whatever gcov
3852 assert(!SvTAINTED(sstr));
3857 if (dtype < SVt_PV && dtype != SVt_IV)
3858 sv_upgrade(dstr, SVt_IV);
3866 sv_upgrade(dstr, SVt_NV);
3870 sv_upgrade(dstr, SVt_PVNV);
3873 goto end_of_first_switch;
3875 SvNV_set(dstr, SvNVX(sstr));
3876 (void)SvNOK_only(dstr);
3877 /* SvTAINTED can only be true if the SV has taint magic, which in
3878 turn means that the SV type is PVMG (or greater). This is the
3879 case statement for SVt_NV, so this cannot be true (whatever gcov
3881 assert(!SvTAINTED(sstr));
3887 #ifdef PERL_OLD_COPY_ON_WRITE
3888 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3889 if (dtype < SVt_PVIV)
3890 sv_upgrade(dstr, SVt_PVIV);
3898 sv_upgrade(dstr, SVt_PV);
3901 if (dtype < SVt_PVIV)
3902 sv_upgrade(dstr, SVt_PVIV);
3905 if (dtype < SVt_PVNV)
3906 sv_upgrade(dstr, SVt_PVNV);
3910 const char * const type = sv_reftype(sstr,0);
3912 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3914 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3918 /* case SVt_BIND: */
3921 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3922 glob_assign_glob(dstr, sstr, dtype);
3925 /* SvVALID means that this PVGV is playing at being an FBM. */
3929 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3931 if (SvTYPE(sstr) != stype) {
3932 stype = SvTYPE(sstr);
3933 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3934 glob_assign_glob(dstr, sstr, dtype);
3939 if (stype == SVt_PVLV)
3940 SvUPGRADE(dstr, SVt_PVNV);
3942 SvUPGRADE(dstr, (svtype)stype);
3944 end_of_first_switch:
3946 /* dstr may have been upgraded. */
3947 dtype = SvTYPE(dstr);
3948 sflags = SvFLAGS(sstr);
3950 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3951 /* Assigning to a subroutine sets the prototype. */
3954 const char *const ptr = SvPV_const(sstr, len);
3956 SvGROW(dstr, len + 1);
3957 Copy(ptr, SvPVX(dstr), len + 1, char);
3958 SvCUR_set(dstr, len);
3960 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3964 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3965 const char * const type = sv_reftype(dstr,0);
3967 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3969 Perl_croak(aTHX_ "Cannot copy to %s", type);
3970 } else if (sflags & SVf_ROK) {
3971 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3972 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3975 if (GvIMPORTED(dstr) != GVf_IMPORTED
3976 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3978 GvIMPORTED_on(dstr);
3983 glob_assign_glob(dstr, sstr, dtype);
3987 if (dtype >= SVt_PV) {
3988 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3989 glob_assign_ref(dstr, sstr);
3992 if (SvPVX_const(dstr)) {
3998 (void)SvOK_off(dstr);
3999 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4000 SvFLAGS(dstr) |= sflags & SVf_ROK;
4001 assert(!(sflags & SVp_NOK));
4002 assert(!(sflags & SVp_IOK));
4003 assert(!(sflags & SVf_NOK));
4004 assert(!(sflags & SVf_IOK));
4006 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4007 if (!(sflags & SVf_OK)) {
4008 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4009 "Undefined value assigned to typeglob");
4012 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4013 if (dstr != (const SV *)gv) {
4015 gp_free(MUTABLE_GV(dstr));
4016 GvGP(dstr) = gp_ref(GvGP(gv));
4020 else if (sflags & SVp_POK) {
4024 * Check to see if we can just swipe the string. If so, it's a
4025 * possible small lose on short strings, but a big win on long ones.
4026 * It might even be a win on short strings if SvPVX_const(dstr)
4027 * has to be allocated and SvPVX_const(sstr) has to be freed.
4028 * Likewise if we can set up COW rather than doing an actual copy, we
4029 * drop to the else clause, as the swipe code and the COW setup code
4030 * have much in common.
4033 /* Whichever path we take through the next code, we want this true,
4034 and doing it now facilitates the COW check. */
4035 (void)SvPOK_only(dstr);
4038 /* If we're already COW then this clause is not true, and if COW
4039 is allowed then we drop down to the else and make dest COW
4040 with us. If caller hasn't said that we're allowed to COW
4041 shared hash keys then we don't do the COW setup, even if the
4042 source scalar is a shared hash key scalar. */
4043 (((flags & SV_COW_SHARED_HASH_KEYS)
4044 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4045 : 1 /* If making a COW copy is forbidden then the behaviour we
4046 desire is as if the source SV isn't actually already
4047 COW, even if it is. So we act as if the source flags
4048 are not COW, rather than actually testing them. */
4050 #ifndef PERL_OLD_COPY_ON_WRITE
4051 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4052 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4053 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4054 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4055 but in turn, it's somewhat dead code, never expected to go
4056 live, but more kept as a placeholder on how to do it better
4057 in a newer implementation. */
4058 /* If we are COW and dstr is a suitable target then we drop down
4059 into the else and make dest a COW of us. */
4060 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4065 (sflags & SVs_TEMP) && /* slated for free anyway? */
4066 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4067 (!(flags & SV_NOSTEAL)) &&
4068 /* and we're allowed to steal temps */
4069 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4070 SvLEN(sstr) && /* and really is a string */
4071 /* and won't be needed again, potentially */
4072 !(PL_op && PL_op->op_type == OP_AASSIGN))
4073 #ifdef PERL_OLD_COPY_ON_WRITE
4074 && ((flags & SV_COW_SHARED_HASH_KEYS)
4075 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4076 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4077 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4081 /* Failed the swipe test, and it's not a shared hash key either.
4082 Have to copy the string. */
4083 STRLEN len = SvCUR(sstr);
4084 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4085 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4086 SvCUR_set(dstr, len);
4087 *SvEND(dstr) = '\0';
4089 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4091 /* Either it's a shared hash key, or it's suitable for
4092 copy-on-write or we can swipe the string. */
4094 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4098 #ifdef PERL_OLD_COPY_ON_WRITE
4100 if ((sflags & (SVf_FAKE | SVf_READONLY))
4101 != (SVf_FAKE | SVf_READONLY)) {
4102 SvREADONLY_on(sstr);
4104 /* Make the source SV into a loop of 1.
4105 (about to become 2) */
4106 SV_COW_NEXT_SV_SET(sstr, sstr);
4110 /* Initial code is common. */
4111 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4116 /* making another shared SV. */
4117 STRLEN cur = SvCUR(sstr);
4118 STRLEN len = SvLEN(sstr);
4119 #ifdef PERL_OLD_COPY_ON_WRITE
4121 assert (SvTYPE(dstr) >= SVt_PVIV);
4122 /* SvIsCOW_normal */
4123 /* splice us in between source and next-after-source. */
4124 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4125 SV_COW_NEXT_SV_SET(sstr, dstr);
4126 SvPV_set(dstr, SvPVX_mutable(sstr));
4130 /* SvIsCOW_shared_hash */
4131 DEBUG_C(PerlIO_printf(Perl_debug_log,
4132 "Copy on write: Sharing hash\n"));
4134 assert (SvTYPE(dstr) >= SVt_PV);
4136 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4138 SvLEN_set(dstr, len);
4139 SvCUR_set(dstr, cur);
4140 SvREADONLY_on(dstr);
4144 { /* Passes the swipe test. */
4145 SvPV_set(dstr, SvPVX_mutable(sstr));
4146 SvLEN_set(dstr, SvLEN(sstr));
4147 SvCUR_set(dstr, SvCUR(sstr));
4150 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4151 SvPV_set(sstr, NULL);
4157 if (sflags & SVp_NOK) {
4158 SvNV_set(dstr, SvNVX(sstr));
4160 if (sflags & SVp_IOK) {
4161 SvIV_set(dstr, SvIVX(sstr));
4162 /* Must do this otherwise some other overloaded use of 0x80000000
4163 gets confused. I guess SVpbm_VALID */
4164 if (sflags & SVf_IVisUV)
4167 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4169 const MAGIC * const smg = SvVSTRING_mg(sstr);
4171 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4172 smg->mg_ptr, smg->mg_len);
4173 SvRMAGICAL_on(dstr);
4177 else if (sflags & (SVp_IOK|SVp_NOK)) {
4178 (void)SvOK_off(dstr);
4179 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4180 if (sflags & SVp_IOK) {
4181 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4182 SvIV_set(dstr, SvIVX(sstr));
4184 if (sflags & SVp_NOK) {
4185 SvNV_set(dstr, SvNVX(sstr));
4189 if (isGV_with_GP(sstr)) {
4190 /* This stringification rule for globs is spread in 3 places.
4191 This feels bad. FIXME. */
4192 const U32 wasfake = sflags & SVf_FAKE;
4194 /* FAKE globs can get coerced, so need to turn this off
4195 temporarily if it is on. */
4197 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4198 SvFLAGS(sstr) |= wasfake;
4201 (void)SvOK_off(dstr);
4203 if (SvTAINTED(sstr))
4208 =for apidoc sv_setsv_mg
4210 Like C<sv_setsv>, but also handles 'set' magic.
4216 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4218 PERL_ARGS_ASSERT_SV_SETSV_MG;
4220 sv_setsv(dstr,sstr);
4224 #ifdef PERL_OLD_COPY_ON_WRITE
4226 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4228 STRLEN cur = SvCUR(sstr);
4229 STRLEN len = SvLEN(sstr);
4230 register char *new_pv;
4232 PERL_ARGS_ASSERT_SV_SETSV_COW;
4235 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4236 (void*)sstr, (void*)dstr);
4243 if (SvTHINKFIRST(dstr))
4244 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4245 else if (SvPVX_const(dstr))
4246 Safefree(SvPVX_const(dstr));
4250 SvUPGRADE(dstr, SVt_PVIV);
4252 assert (SvPOK(sstr));
4253 assert (SvPOKp(sstr));
4254 assert (!SvIOK(sstr));
4255 assert (!SvIOKp(sstr));
4256 assert (!SvNOK(sstr));
4257 assert (!SvNOKp(sstr));
4259 if (SvIsCOW(sstr)) {
4261 if (SvLEN(sstr) == 0) {
4262 /* source is a COW shared hash key. */
4263 DEBUG_C(PerlIO_printf(Perl_debug_log,
4264 "Fast copy on write: Sharing hash\n"));
4265 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4268 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4270 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4271 SvUPGRADE(sstr, SVt_PVIV);
4272 SvREADONLY_on(sstr);
4274 DEBUG_C(PerlIO_printf(Perl_debug_log,
4275 "Fast copy on write: Converting sstr to COW\n"));
4276 SV_COW_NEXT_SV_SET(dstr, sstr);
4278 SV_COW_NEXT_SV_SET(sstr, dstr);
4279 new_pv = SvPVX_mutable(sstr);
4282 SvPV_set(dstr, new_pv);
4283 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4286 SvLEN_set(dstr, len);
4287 SvCUR_set(dstr, cur);
4296 =for apidoc sv_setpvn
4298 Copies a string into an SV. The C<len> parameter indicates the number of
4299 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4300 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4306 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4309 register char *dptr;
4311 PERL_ARGS_ASSERT_SV_SETPVN;
4313 SV_CHECK_THINKFIRST_COW_DROP(sv);
4319 /* len is STRLEN which is unsigned, need to copy to signed */
4322 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4324 SvUPGRADE(sv, SVt_PV);
4326 dptr = SvGROW(sv, len + 1);
4327 Move(ptr,dptr,len,char);
4330 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4335 =for apidoc sv_setpvn_mg
4337 Like C<sv_setpvn>, but also handles 'set' magic.
4343 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4345 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4347 sv_setpvn(sv,ptr,len);
4352 =for apidoc sv_setpv
4354 Copies a string into an SV. The string must be null-terminated. Does not
4355 handle 'set' magic. See C<sv_setpv_mg>.
4361 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4364 register STRLEN len;
4366 PERL_ARGS_ASSERT_SV_SETPV;
4368 SV_CHECK_THINKFIRST_COW_DROP(sv);
4374 SvUPGRADE(sv, SVt_PV);
4376 SvGROW(sv, len + 1);
4377 Move(ptr,SvPVX(sv),len+1,char);
4379 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4384 =for apidoc sv_setpv_mg
4386 Like C<sv_setpv>, but also handles 'set' magic.
4392 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4394 PERL_ARGS_ASSERT_SV_SETPV_MG;
4401 =for apidoc sv_usepvn_flags
4403 Tells an SV to use C<ptr> to find its string value. Normally the
4404 string is stored inside the SV but sv_usepvn allows the SV to use an
4405 outside string. The C<ptr> should point to memory that was allocated
4406 by C<malloc>. The string length, C<len>, must be supplied. By default
4407 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4408 so that pointer should not be freed or used by the programmer after
4409 giving it to sv_usepvn, and neither should any pointers from "behind"
4410 that pointer (e.g. ptr + 1) be used.
4412 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4413 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4414 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4415 C<len>, and already meets the requirements for storing in C<SvPVX>)
4421 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4426 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4428 SV_CHECK_THINKFIRST_COW_DROP(sv);
4429 SvUPGRADE(sv, SVt_PV);
4432 if (flags & SV_SMAGIC)
4436 if (SvPVX_const(sv))
4440 if (flags & SV_HAS_TRAILING_NUL)
4441 assert(ptr[len] == '\0');
4444 allocate = (flags & SV_HAS_TRAILING_NUL)
4446 #ifdef Perl_safesysmalloc_size
4449 PERL_STRLEN_ROUNDUP(len + 1);
4451 if (flags & SV_HAS_TRAILING_NUL) {
4452 /* It's long enough - do nothing.
4453 Specfically Perl_newCONSTSUB is relying on this. */
4456 /* Force a move to shake out bugs in callers. */
4457 char *new_ptr = (char*)safemalloc(allocate);
4458 Copy(ptr, new_ptr, len, char);
4459 PoisonFree(ptr,len,char);
4463 ptr = (char*) saferealloc (ptr, allocate);
4466 #ifdef Perl_safesysmalloc_size
4467 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4469 SvLEN_set(sv, allocate);
4473 if (!(flags & SV_HAS_TRAILING_NUL)) {
4476 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4478 if (flags & SV_SMAGIC)
4482 #ifdef PERL_OLD_COPY_ON_WRITE
4483 /* Need to do this *after* making the SV normal, as we need the buffer
4484 pointer to remain valid until after we've copied it. If we let go too early,
4485 another thread could invalidate it by unsharing last of the same hash key
4486 (which it can do by means other than releasing copy-on-write Svs)
4487 or by changing the other copy-on-write SVs in the loop. */
4489 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4491 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4493 { /* this SV was SvIsCOW_normal(sv) */
4494 /* we need to find the SV pointing to us. */
4495 SV *current = SV_COW_NEXT_SV(after);
4497 if (current == sv) {
4498 /* The SV we point to points back to us (there were only two of us
4500 Hence other SV is no longer copy on write either. */
4502 SvREADONLY_off(after);
4504 /* We need to follow the pointers around the loop. */
4506 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4509 /* don't loop forever if the structure is bust, and we have
4510 a pointer into a closed loop. */
4511 assert (current != after);
4512 assert (SvPVX_const(current) == pvx);
4514 /* Make the SV before us point to the SV after us. */
4515 SV_COW_NEXT_SV_SET(current, after);
4521 =for apidoc sv_force_normal_flags
4523 Undo various types of fakery on an SV: if the PV is a shared string, make
4524 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4525 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4526 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4527 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4528 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4529 set to some other value.) In addition, the C<flags> parameter gets passed to
4530 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4531 with flags set to 0.
4537 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4541 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4543 #ifdef PERL_OLD_COPY_ON_WRITE
4544 if (SvREADONLY(sv)) {
4546 const char * const pvx = SvPVX_const(sv);
4547 const STRLEN len = SvLEN(sv);
4548 const STRLEN cur = SvCUR(sv);
4549 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4550 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4551 we'll fail an assertion. */
4552 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4555 PerlIO_printf(Perl_debug_log,
4556 "Copy on write: Force normal %ld\n",
4562 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4565 if (flags & SV_COW_DROP_PV) {
4566 /* OK, so we don't need to copy our buffer. */
4569 SvGROW(sv, cur + 1);
4570 Move(pvx,SvPVX(sv),cur,char);
4575 sv_release_COW(sv, pvx, next);
4577 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4583 else if (IN_PERL_RUNTIME)
4584 Perl_croak(aTHX_ "%s", PL_no_modify);
4587 if (SvREADONLY(sv)) {
4589 const char * const pvx = SvPVX_const(sv);
4590 const STRLEN len = SvCUR(sv);
4595 SvGROW(sv, len + 1);
4596 Move(pvx,SvPVX(sv),len,char);
4598 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4600 else if (IN_PERL_RUNTIME)
4601 Perl_croak(aTHX_ "%s", PL_no_modify);
4605 sv_unref_flags(sv, flags);
4606 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4613 Efficient removal of characters from the beginning of the string buffer.
4614 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4615 the string buffer. The C<ptr> becomes the first character of the adjusted
4616 string. Uses the "OOK hack".
4617 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4618 refer to the same chunk of data.
4624 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4630 const U8 *real_start;
4634 PERL_ARGS_ASSERT_SV_CHOP;
4636 if (!ptr || !SvPOKp(sv))
4638 delta = ptr - SvPVX_const(sv);
4640 /* Nothing to do. */
4643 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4644 nothing uses the value of ptr any more. */
4645 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4646 if (ptr <= SvPVX_const(sv))
4647 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4648 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4649 SV_CHECK_THINKFIRST(sv);
4650 if (delta > max_delta)
4651 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4652 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4653 SvPVX_const(sv) + max_delta);
4656 if (!SvLEN(sv)) { /* make copy of shared string */
4657 const char *pvx = SvPVX_const(sv);
4658 const STRLEN len = SvCUR(sv);
4659 SvGROW(sv, len + 1);
4660 Move(pvx,SvPVX(sv),len,char);
4663 SvFLAGS(sv) |= SVf_OOK;
4666 SvOOK_offset(sv, old_delta);
4668 SvLEN_set(sv, SvLEN(sv) - delta);
4669 SvCUR_set(sv, SvCUR(sv) - delta);
4670 SvPV_set(sv, SvPVX(sv) + delta);
4672 p = (U8 *)SvPVX_const(sv);
4677 real_start = p - delta;
4681 if (delta < 0x100) {
4685 p -= sizeof(STRLEN);
4686 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4690 /* Fill the preceding buffer with sentinals to verify that no-one is
4692 while (p > real_start) {
4700 =for apidoc sv_catpvn
4702 Concatenates the string onto the end of the string which is in the SV. The
4703 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4704 status set, then the bytes appended should be valid UTF-8.
4705 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4707 =for apidoc sv_catpvn_flags
4709 Concatenates the string onto the end of the string which is in the SV. The
4710 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4711 status set, then the bytes appended should be valid UTF-8.
4712 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4713 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4714 in terms of this function.
4720 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4724 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4726 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4728 SvGROW(dsv, dlen + slen + 1);
4730 sstr = SvPVX_const(dsv);
4731 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4732 SvCUR_set(dsv, SvCUR(dsv) + slen);
4734 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4736 if (flags & SV_SMAGIC)
4741 =for apidoc sv_catsv
4743 Concatenates the string from SV C<ssv> onto the end of the string in
4744 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4745 not 'set' magic. See C<sv_catsv_mg>.
4747 =for apidoc sv_catsv_flags
4749 Concatenates the string from SV C<ssv> onto the end of the string in
4750 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4751 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4752 and C<sv_catsv_nomg> are implemented in terms of this function.
4757 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4761 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4765 const char *spv = SvPV_const(ssv, slen);
4767 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4768 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4769 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4770 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4771 dsv->sv_flags doesn't have that bit set.
4772 Andy Dougherty 12 Oct 2001
4774 const I32 sutf8 = DO_UTF8(ssv);
4777 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4779 dutf8 = DO_UTF8(dsv);
4781 if (dutf8 != sutf8) {
4783 /* Not modifying source SV, so taking a temporary copy. */
4784 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4786 sv_utf8_upgrade(csv);
4787 spv = SvPV_const(csv, slen);
4790 /* Leave enough space for the cat that's about to happen */
4791 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4793 sv_catpvn_nomg(dsv, spv, slen);
4796 if (flags & SV_SMAGIC)
4801 =for apidoc sv_catpv
4803 Concatenates the string onto the end of the string which is in the SV.
4804 If the SV has the UTF-8 status set, then the bytes appended should be
4805 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4810 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4813 register STRLEN len;
4817 PERL_ARGS_ASSERT_SV_CATPV;
4821 junk = SvPV_force(sv, tlen);
4823 SvGROW(sv, tlen + len + 1);
4825 ptr = SvPVX_const(sv);
4826 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4827 SvCUR_set(sv, SvCUR(sv) + len);
4828 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4833 =for apidoc sv_catpv_mg
4835 Like C<sv_catpv>, but also handles 'set' magic.
4841 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4843 PERL_ARGS_ASSERT_SV_CATPV_MG;
4852 Creates a new SV. A non-zero C<len> parameter indicates the number of
4853 bytes of preallocated string space the SV should have. An extra byte for a
4854 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4855 space is allocated.) The reference count for the new SV is set to 1.
4857 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4858 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4859 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4860 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4861 modules supporting older perls.
4867 Perl_newSV(pTHX_ const STRLEN len)
4874 sv_upgrade(sv, SVt_PV);
4875 SvGROW(sv, len + 1);
4880 =for apidoc sv_magicext
4882 Adds magic to an SV, upgrading it if necessary. Applies the
4883 supplied vtable and returns a pointer to the magic added.
4885 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4886 In particular, you can add magic to SvREADONLY SVs, and add more than
4887 one instance of the same 'how'.
4889 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4890 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4891 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4892 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4894 (This is now used as a subroutine by C<sv_magic>.)
4899 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4900 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4905 PERL_ARGS_ASSERT_SV_MAGICEXT;
4907 SvUPGRADE(sv, SVt_PVMG);
4908 Newxz(mg, 1, MAGIC);
4909 mg->mg_moremagic = SvMAGIC(sv);
4910 SvMAGIC_set(sv, mg);
4912 /* Sometimes a magic contains a reference loop, where the sv and
4913 object refer to each other. To prevent a reference loop that
4914 would prevent such objects being freed, we look for such loops
4915 and if we find one we avoid incrementing the object refcount.
4917 Note we cannot do this to avoid self-tie loops as intervening RV must
4918 have its REFCNT incremented to keep it in existence.
4921 if (!obj || obj == sv ||
4922 how == PERL_MAGIC_arylen ||
4923 how == PERL_MAGIC_symtab ||
4924 (SvTYPE(obj) == SVt_PVGV &&
4925 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4926 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4927 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4932 mg->mg_obj = SvREFCNT_inc_simple(obj);
4933 mg->mg_flags |= MGf_REFCOUNTED;
4936 /* Normal self-ties simply pass a null object, and instead of
4937 using mg_obj directly, use the SvTIED_obj macro to produce a
4938 new RV as needed. For glob "self-ties", we are tieing the PVIO
4939 with an RV obj pointing to the glob containing the PVIO. In
4940 this case, to avoid a reference loop, we need to weaken the
4944 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4945 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4951 mg->mg_len = namlen;
4954 mg->mg_ptr = savepvn(name, namlen);
4955 else if (namlen == HEf_SVKEY) {
4956 /* Yes, this is casting away const. This is only for the case of
4957 HEf_SVKEY. I think we need to document this abberation of the
4958 constness of the API, rather than making name non-const, as
4959 that change propagating outwards a long way. */
4960 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4962 mg->mg_ptr = (char *) name;
4964 mg->mg_virtual = (MGVTBL *) vtable;
4968 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4973 =for apidoc sv_magic
4975 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4976 then adds a new magic item of type C<how> to the head of the magic list.
4978 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4979 handling of the C<name> and C<namlen> arguments.
4981 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4982 to add more than one instance of the same 'how'.
4988 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4989 const char *const name, const I32 namlen)
4992 const MGVTBL *vtable;
4995 PERL_ARGS_ASSERT_SV_MAGIC;
4997 #ifdef PERL_OLD_COPY_ON_WRITE
4999 sv_force_normal_flags(sv, 0);
5001 if (SvREADONLY(sv)) {
5003 /* its okay to attach magic to shared strings; the subsequent
5004 * upgrade to PVMG will unshare the string */
5005 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5008 && how != PERL_MAGIC_regex_global
5009 && how != PERL_MAGIC_bm
5010 && how != PERL_MAGIC_fm
5011 && how != PERL_MAGIC_sv
5012 && how != PERL_MAGIC_backref
5015 Perl_croak(aTHX_ "%s", PL_no_modify);
5018 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5019 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5020 /* sv_magic() refuses to add a magic of the same 'how' as an
5023 if (how == PERL_MAGIC_taint) {
5025 /* Any scalar which already had taint magic on which someone
5026 (erroneously?) did SvIOK_on() or similar will now be
5027 incorrectly sporting public "OK" flags. */
5028 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5036 vtable = &PL_vtbl_sv;
5038 case PERL_MAGIC_overload:
5039 vtable = &PL_vtbl_amagic;
5041 case PERL_MAGIC_overload_elem:
5042 vtable = &PL_vtbl_amagicelem;
5044 case PERL_MAGIC_overload_table:
5045 vtable = &PL_vtbl_ovrld;
5048 vtable = &PL_vtbl_bm;
5050 case PERL_MAGIC_regdata:
5051 vtable = &PL_vtbl_regdata;
5053 case PERL_MAGIC_regdatum:
5054 vtable = &PL_vtbl_regdatum;
5056 case PERL_MAGIC_env:
5057 vtable = &PL_vtbl_env;
5060 vtable = &PL_vtbl_fm;
5062 case PERL_MAGIC_envelem:
5063 vtable = &PL_vtbl_envelem;
5065 case PERL_MAGIC_regex_global:
5066 vtable = &PL_vtbl_mglob;
5068 case PERL_MAGIC_isa:
5069 vtable = &PL_vtbl_isa;
5071 case PERL_MAGIC_isaelem:
5072 vtable = &PL_vtbl_isaelem;
5074 case PERL_MAGIC_nkeys:
5075 vtable = &PL_vtbl_nkeys;
5077 case PERL_MAGIC_dbfile:
5080 case PERL_MAGIC_dbline:
5081 vtable = &PL_vtbl_dbline;
5083 #ifdef USE_LOCALE_COLLATE
5084 case PERL_MAGIC_collxfrm:
5085 vtable = &PL_vtbl_collxfrm;
5087 #endif /* USE_LOCALE_COLLATE */
5088 case PERL_MAGIC_tied:
5089 vtable = &PL_vtbl_pack;
5091 case PERL_MAGIC_tiedelem:
5092 case PERL_MAGIC_tiedscalar:
5093 vtable = &PL_vtbl_packelem;
5096 vtable = &PL_vtbl_regexp;
5098 case PERL_MAGIC_sig:
5099 vtable = &PL_vtbl_sig;
5101 case PERL_MAGIC_sigelem:
5102 vtable = &PL_vtbl_sigelem;
5104 case PERL_MAGIC_taint:
5105 vtable = &PL_vtbl_taint;
5107 case PERL_MAGIC_uvar:
5108 vtable = &PL_vtbl_uvar;
5110 case PERL_MAGIC_vec:
5111 vtable = &PL_vtbl_vec;
5113 case PERL_MAGIC_arylen_p:
5114 case PERL_MAGIC_rhash:
5115 case PERL_MAGIC_symtab:
5116 case PERL_MAGIC_vstring:
5119 case PERL_MAGIC_utf8:
5120 vtable = &PL_vtbl_utf8;
5122 case PERL_MAGIC_substr:
5123 vtable = &PL_vtbl_substr;
5125 case PERL_MAGIC_defelem:
5126 vtable = &PL_vtbl_defelem;
5128 case PERL_MAGIC_arylen:
5129 vtable = &PL_vtbl_arylen;
5131 case PERL_MAGIC_pos:
5132 vtable = &PL_vtbl_pos;
5134 case PERL_MAGIC_backref:
5135 vtable = &PL_vtbl_backref;
5137 case PERL_MAGIC_hintselem:
5138 vtable = &PL_vtbl_hintselem;
5140 case PERL_MAGIC_hints:
5141 vtable = &PL_vtbl_hints;
5143 case PERL_MAGIC_ext:
5144 /* Reserved for use by extensions not perl internals. */
5145 /* Useful for attaching extension internal data to perl vars. */
5146 /* Note that multiple extensions may clash if magical scalars */
5147 /* etc holding private data from one are passed to another. */
5151 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5154 /* Rest of work is done else where */
5155 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5158 case PERL_MAGIC_taint:
5161 case PERL_MAGIC_ext:
5162 case PERL_MAGIC_dbfile:
5169 =for apidoc sv_unmagic
5171 Removes all magic of type C<type> from an SV.
5177 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5182 PERL_ARGS_ASSERT_SV_UNMAGIC;
5184 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5186 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5187 for (mg = *mgp; mg; mg = *mgp) {
5188 if (mg->mg_type == type) {
5189 const MGVTBL* const vtbl = mg->mg_virtual;
5190 *mgp = mg->mg_moremagic;
5191 if (vtbl && vtbl->svt_free)
5192 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5193 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5195 Safefree(mg->mg_ptr);
5196 else if (mg->mg_len == HEf_SVKEY)
5197 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5198 else if (mg->mg_type == PERL_MAGIC_utf8)
5199 Safefree(mg->mg_ptr);
5201 if (mg->mg_flags & MGf_REFCOUNTED)
5202 SvREFCNT_dec(mg->mg_obj);
5206 mgp = &mg->mg_moremagic;
5210 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5211 SvMAGIC_set(sv, NULL);
5218 =for apidoc sv_rvweaken
5220 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5221 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5222 push a back-reference to this RV onto the array of backreferences
5223 associated with that magic. If the RV is magical, set magic will be
5224 called after the RV is cleared.
5230 Perl_sv_rvweaken(pTHX_ SV *const sv)
5234 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5236 if (!SvOK(sv)) /* let undefs pass */
5239 Perl_croak(aTHX_ "Can't weaken a nonreference");
5240 else if (SvWEAKREF(sv)) {
5241 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5245 Perl_sv_add_backref(aTHX_ tsv, sv);
5251 /* Give tsv backref magic if it hasn't already got it, then push a
5252 * back-reference to sv onto the array associated with the backref magic.
5255 /* A discussion about the backreferences array and its refcount:
5257 * The AV holding the backreferences is pointed to either as the mg_obj of
5258 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5259 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5260 * have the standard magic instead.) The array is created with a refcount
5261 * of 2. This means that if during global destruction the array gets
5262 * picked on first to have its refcount decremented by the random zapper,
5263 * it won't actually be freed, meaning it's still theere for when its
5264 * parent gets freed.
5265 * When the parent SV is freed, in the case of magic, the magic is freed,
5266 * Perl_magic_killbackrefs is called which decrements one refcount, then
5267 * mg_obj is freed which kills the second count.
5268 * In the vase of a HV being freed, one ref is removed by
5269 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5274 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5279 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5281 if (SvTYPE(tsv) == SVt_PVHV) {
5282 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5286 /* There is no AV in the offical place - try a fixup. */
5287 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5290 /* Aha. They've got it stowed in magic. Bring it back. */
5291 av = MUTABLE_AV(mg->mg_obj);
5292 /* Stop mg_free decreasing the refernce count. */
5294 /* Stop mg_free even calling the destructor, given that
5295 there's no AV to free up. */
5297 sv_unmagic(tsv, PERL_MAGIC_backref);
5301 SvREFCNT_inc_simple_void(av); /* see discussion above */
5306 const MAGIC *const mg
5307 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5309 av = MUTABLE_AV(mg->mg_obj);
5313 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5314 /* av now has a refcnt of 2; see discussion above */
5317 if (AvFILLp(av) >= AvMAX(av)) {
5318 av_extend(av, AvFILLp(av)+1);
5320 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5323 /* delete a back-reference to ourselves from the backref magic associated
5324 * with the SV we point to.
5328 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5335 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5337 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5338 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5339 /* We mustn't attempt to "fix up" the hash here by moving the
5340 backreference array back to the hv_aux structure, as that is stored
5341 in the main HvARRAY(), and hfreentries assumes that no-one
5342 reallocates HvARRAY() while it is running. */
5345 const MAGIC *const mg
5346 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5348 av = MUTABLE_AV(mg->mg_obj);
5352 Perl_croak(aTHX_ "panic: del_backref");
5354 assert(!SvIS_FREED(av));
5357 /* We shouldn't be in here more than once, but for paranoia reasons lets
5359 for (i = AvFILLp(av); i >= 0; i--) {
5361 const SSize_t fill = AvFILLp(av);
5363 /* We weren't the last entry.
5364 An unordered list has this property that you can take the
5365 last element off the end to fill the hole, and it's still
5366 an unordered list :-)
5371 AvFILLp(av) = fill - 1;
5377 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5379 SV **svp = AvARRAY(av);
5381 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5382 PERL_UNUSED_ARG(sv);
5384 assert(!svp || !SvIS_FREED(av));
5386 SV *const *const last = svp + AvFILLp(av);
5388 while (svp <= last) {
5390 SV *const referrer = *svp;
5391 if (SvWEAKREF(referrer)) {
5392 /* XXX Should we check that it hasn't changed? */
5393 SvRV_set(referrer, 0);
5395 SvWEAKREF_off(referrer);
5396 SvSETMAGIC(referrer);
5397 } else if (SvTYPE(referrer) == SVt_PVGV ||
5398 SvTYPE(referrer) == SVt_PVLV) {
5399 /* You lookin' at me? */
5400 assert(GvSTASH(referrer));
5401 assert(GvSTASH(referrer) == (const HV *)sv);
5402 GvSTASH(referrer) = 0;
5405 "panic: magic_killbackrefs (flags=%"UVxf")",
5406 (UV)SvFLAGS(referrer));
5414 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5419 =for apidoc sv_insert
5421 Inserts a string at the specified offset/length within the SV. Similar to
5422 the Perl substr() function. Handles get magic.
5424 =for apidoc sv_insert_flags
5426 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5432 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5437 register char *midend;
5438 register char *bigend;
5442 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5445 Perl_croak(aTHX_ "Can't modify non-existent substring");
5446 SvPV_force_flags(bigstr, curlen, flags);
5447 (void)SvPOK_only_UTF8(bigstr);
5448 if (offset + len > curlen) {
5449 SvGROW(bigstr, offset+len+1);
5450 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5451 SvCUR_set(bigstr, offset+len);
5455 i = littlelen - len;
5456 if (i > 0) { /* string might grow */
5457 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5458 mid = big + offset + len;
5459 midend = bigend = big + SvCUR(bigstr);
5462 while (midend > mid) /* shove everything down */
5463 *--bigend = *--midend;
5464 Move(little,big+offset,littlelen,char);
5465 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5470 Move(little,SvPVX(bigstr)+offset,len,char);
5475 big = SvPVX(bigstr);
5478 bigend = big + SvCUR(bigstr);
5480 if (midend > bigend)
5481 Perl_croak(aTHX_ "panic: sv_insert");
5483 if (mid - big > bigend - midend) { /* faster to shorten from end */
5485 Move(little, mid, littlelen,char);
5488 i = bigend - midend;
5490 Move(midend, mid, i,char);
5494 SvCUR_set(bigstr, mid - big);
5496 else if ((i = mid - big)) { /* faster from front */
5497 midend -= littlelen;
5499 Move(big, midend - i, i, char);
5500 sv_chop(bigstr,midend-i);
5502 Move(little, mid, littlelen,char);
5504 else if (littlelen) {
5505 midend -= littlelen;
5506 sv_chop(bigstr,midend);
5507 Move(little,midend,littlelen,char);
5510 sv_chop(bigstr,midend);
5516 =for apidoc sv_replace
5518 Make the first argument a copy of the second, then delete the original.
5519 The target SV physically takes over ownership of the body of the source SV
5520 and inherits its flags; however, the target keeps any magic it owns,
5521 and any magic in the source is discarded.
5522 Note that this is a rather specialist SV copying operation; most of the
5523 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5529 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5532 const U32 refcnt = SvREFCNT(sv);
5534 PERL_ARGS_ASSERT_SV_REPLACE;
5536 SV_CHECK_THINKFIRST_COW_DROP(sv);
5537 if (SvREFCNT(nsv) != 1) {
5538 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5539 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5541 if (SvMAGICAL(sv)) {
5545 sv_upgrade(nsv, SVt_PVMG);
5546 SvMAGIC_set(nsv, SvMAGIC(sv));
5547 SvFLAGS(nsv) |= SvMAGICAL(sv);
5549 SvMAGIC_set(sv, NULL);
5553 assert(!SvREFCNT(sv));
5554 #ifdef DEBUG_LEAKING_SCALARS
5555 sv->sv_flags = nsv->sv_flags;
5556 sv->sv_any = nsv->sv_any;
5557 sv->sv_refcnt = nsv->sv_refcnt;
5558 sv->sv_u = nsv->sv_u;
5560 StructCopy(nsv,sv,SV);
5562 if(SvTYPE(sv) == SVt_IV) {
5564 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5568 #ifdef PERL_OLD_COPY_ON_WRITE
5569 if (SvIsCOW_normal(nsv)) {
5570 /* We need to follow the pointers around the loop to make the
5571 previous SV point to sv, rather than nsv. */
5574 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5577 assert(SvPVX_const(current) == SvPVX_const(nsv));
5579 /* Make the SV before us point to the SV after us. */
5581 PerlIO_printf(Perl_debug_log, "previous is\n");
5583 PerlIO_printf(Perl_debug_log,
5584 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5585 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5587 SV_COW_NEXT_SV_SET(current, sv);
5590 SvREFCNT(sv) = refcnt;
5591 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5597 =for apidoc sv_clear
5599 Clear an SV: call any destructors, free up any memory used by the body,
5600 and free the body itself. The SV's head is I<not> freed, although
5601 its type is set to all 1's so that it won't inadvertently be assumed
5602 to be live during global destruction etc.
5603 This function should only be called when REFCNT is zero. Most of the time
5604 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5611 Perl_sv_clear(pTHX_ register SV *const sv)
5614 const U32 type = SvTYPE(sv);
5615 const struct body_details *const sv_type_details
5616 = bodies_by_type + type;
5619 PERL_ARGS_ASSERT_SV_CLEAR;
5620 assert(SvREFCNT(sv) == 0);
5621 assert(SvTYPE(sv) != SVTYPEMASK);
5623 if (type <= SVt_IV) {
5624 /* See the comment in sv.h about the collusion between this early
5625 return and the overloading of the NULL and IV slots in the size
5628 SV * const target = SvRV(sv);
5630 sv_del_backref(target, sv);
5632 SvREFCNT_dec(target);
5634 SvFLAGS(sv) &= SVf_BREAK;
5635 SvFLAGS(sv) |= SVTYPEMASK;
5640 if (PL_defstash && /* Still have a symbol table? */
5647 stash = SvSTASH(sv);
5648 destructor = StashHANDLER(stash,DESTROY);
5650 /* A constant subroutine can have no side effects, so
5651 don't bother calling it. */
5652 && !CvCONST(destructor)
5653 /* Don't bother calling an empty destructor */
5654 && (CvISXSUB(destructor)
5655 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5657 SV* const tmpref = newRV(sv);
5658 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5660 PUSHSTACKi(PERLSI_DESTROY);
5665 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5671 if(SvREFCNT(tmpref) < 2) {
5672 /* tmpref is not kept alive! */
5674 SvRV_set(tmpref, NULL);
5677 SvREFCNT_dec(tmpref);
5679 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5683 if (PL_in_clean_objs)
5684 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5686 /* DESTROY gave object new lease on life */
5692 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5693 SvOBJECT_off(sv); /* Curse the object. */
5694 if (type != SVt_PVIO)
5695 --PL_sv_objcount; /* XXX Might want something more general */
5698 if (type >= SVt_PVMG) {
5699 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5700 SvREFCNT_dec(SvOURSTASH(sv));
5701 } else if (SvMAGIC(sv))
5703 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5704 SvREFCNT_dec(SvSTASH(sv));
5707 /* case SVt_BIND: */
5710 IoIFP(sv) != PerlIO_stdin() &&
5711 IoIFP(sv) != PerlIO_stdout() &&
5712 IoIFP(sv) != PerlIO_stderr())
5714 io_close(MUTABLE_IO(sv), FALSE);
5716 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5717 PerlDir_close(IoDIRP(sv));
5718 IoDIRP(sv) = (DIR*)NULL;
5719 Safefree(IoTOP_NAME(sv));
5720 Safefree(IoFMT_NAME(sv));
5721 Safefree(IoBOTTOM_NAME(sv));
5724 /* FIXME for plugins */
5725 pregfree2((REGEXP*) sv);
5729 cv_undef(MUTABLE_CV(sv));
5732 if (PL_last_swash_hv == (const HV *)sv) {
5733 PL_last_swash_hv = NULL;
5735 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5736 hv_undef(MUTABLE_HV(sv));
5739 if (PL_comppad == MUTABLE_AV(sv)) {
5743 av_undef(MUTABLE_AV(sv));
5746 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5747 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5748 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5749 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5751 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5752 SvREFCNT_dec(LvTARG(sv));
5754 if (isGV_with_GP(sv)) {
5755 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5756 && HvNAME_get(stash))
5757 mro_method_changed_in(stash);
5758 gp_free(MUTABLE_GV(sv));
5760 unshare_hek(GvNAME_HEK(sv));
5761 /* If we're in a stash, we don't own a reference to it. However it does
5762 have a back reference to us, which needs to be cleared. */
5763 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5764 sv_del_backref(MUTABLE_SV(stash), sv);
5766 /* FIXME. There are probably more unreferenced pointers to SVs in the
5767 interpreter struct that we should check and tidy in a similar
5769 if ((const GV *)sv == PL_last_in_gv)
5770 PL_last_in_gv = NULL;
5776 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5779 SvOOK_offset(sv, offset);
5780 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5781 /* Don't even bother with turning off the OOK flag. */
5784 SV * const target = SvRV(sv);
5786 sv_del_backref(target, sv);
5788 SvREFCNT_dec(target);
5790 #ifdef PERL_OLD_COPY_ON_WRITE
5791 else if (SvPVX_const(sv)) {
5794 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5798 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5800 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5804 } else if (SvLEN(sv)) {
5805 Safefree(SvPVX_const(sv));
5809 else if (SvPVX_const(sv) && SvLEN(sv))
5810 Safefree(SvPVX_mutable(sv));
5811 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5812 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5821 SvFLAGS(sv) &= SVf_BREAK;
5822 SvFLAGS(sv) |= SVTYPEMASK;
5824 if (sv_type_details->arena) {
5825 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5826 &PL_body_roots[type]);
5828 else if (sv_type_details->body_size) {
5829 my_safefree(SvANY(sv));
5834 =for apidoc sv_newref
5836 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5843 Perl_sv_newref(pTHX_ SV *const sv)
5845 PERL_UNUSED_CONTEXT;
5854 Decrement an SV's reference count, and if it drops to zero, call
5855 C<sv_clear> to invoke destructors and free up any memory used by
5856 the body; finally, deallocate the SV's head itself.
5857 Normally called via a wrapper macro C<SvREFCNT_dec>.
5863 Perl_sv_free(pTHX_ SV *const sv)
5868 if (SvREFCNT(sv) == 0) {
5869 if (SvFLAGS(sv) & SVf_BREAK)
5870 /* this SV's refcnt has been artificially decremented to
5871 * trigger cleanup */
5873 if (PL_in_clean_all) /* All is fair */
5875 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5876 /* make sure SvREFCNT(sv)==0 happens very seldom */
5877 SvREFCNT(sv) = (~(U32)0)/2;
5880 if (ckWARN_d(WARN_INTERNAL)) {
5881 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5882 Perl_dump_sv_child(aTHX_ sv);
5884 #ifdef DEBUG_LEAKING_SCALARS
5887 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5888 if (PL_warnhook == PERL_WARNHOOK_FATAL
5889 || ckDEAD(packWARN(WARN_INTERNAL))) {
5890 /* Don't let Perl_warner cause us to escape our fate: */
5894 /* This may not return: */
5895 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5896 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5897 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5900 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5905 if (--(SvREFCNT(sv)) > 0)
5907 Perl_sv_free2(aTHX_ sv);
5911 Perl_sv_free2(pTHX_ SV *const sv)
5915 PERL_ARGS_ASSERT_SV_FREE2;
5919 if (ckWARN_d(WARN_DEBUGGING))
5920 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
5921 "Attempt to free temp prematurely: SV 0x%"UVxf
5922 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5926 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5927 /* make sure SvREFCNT(sv)==0 happens very seldom */
5928 SvREFCNT(sv) = (~(U32)0)/2;
5939 Returns the length of the string in the SV. Handles magic and type
5940 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5946 Perl_sv_len(pTHX_ register SV *const sv)
5954 len = mg_length(sv);
5956 (void)SvPV_const(sv, len);
5961 =for apidoc sv_len_utf8
5963 Returns the number of characters in the string in an SV, counting wide
5964 UTF-8 bytes as a single character. Handles magic and type coercion.
5970 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5971 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5972 * (Note that the mg_len is not the length of the mg_ptr field.
5973 * This allows the cache to store the character length of the string without
5974 * needing to malloc() extra storage to attach to the mg_ptr.)
5979 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5985 return mg_length(sv);
5989 const U8 *s = (U8*)SvPV_const(sv, len);
5993 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5995 if (mg && mg->mg_len != -1) {
5997 if (PL_utf8cache < 0) {
5998 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6000 /* Need to turn the assertions off otherwise we may
6001 recurse infinitely while printing error messages.
6003 SAVEI8(PL_utf8cache);
6005 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6006 " real %"UVuf" for %"SVf,
6007 (UV) ulen, (UV) real, SVfARG(sv));
6012 ulen = Perl_utf8_length(aTHX_ s, s + len);
6013 if (!SvREADONLY(sv)) {
6015 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6016 &PL_vtbl_utf8, 0, 0);
6024 return Perl_utf8_length(aTHX_ s, s + len);
6028 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6031 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6034 const U8 *s = start;
6036 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6038 while (s < send && uoffset--)
6041 /* This is the existing behaviour. Possibly it should be a croak, as
6042 it's actually a bounds error */
6048 /* Given the length of the string in both bytes and UTF-8 characters, decide
6049 whether to walk forwards or backwards to find the byte corresponding to
6050 the passed in UTF-8 offset. */
6052 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6053 const STRLEN uoffset, const STRLEN uend)
6055 STRLEN backw = uend - uoffset;
6057 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6059 if (uoffset < 2 * backw) {
6060 /* The assumption is that going forwards is twice the speed of going
6061 forward (that's where the 2 * backw comes from).
6062 (The real figure of course depends on the UTF-8 data.) */
6063 return sv_pos_u2b_forwards(start, send, uoffset);
6068 while (UTF8_IS_CONTINUATION(*send))
6071 return send - start;
6074 /* For the string representation of the given scalar, find the byte
6075 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6076 give another position in the string, *before* the sought offset, which
6077 (which is always true, as 0, 0 is a valid pair of positions), which should
6078 help reduce the amount of linear searching.
6079 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6080 will be used to reduce the amount of linear searching. The cache will be
6081 created if necessary, and the found value offered to it for update. */
6083 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6084 const U8 *const send, const STRLEN uoffset,
6085 STRLEN uoffset0, STRLEN boffset0)
6087 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6090 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6092 assert (uoffset >= uoffset0);
6094 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6095 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6096 if ((*mgp)->mg_ptr) {
6097 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6098 if (cache[0] == uoffset) {
6099 /* An exact match. */
6102 if (cache[2] == uoffset) {
6103 /* An exact match. */
6107 if (cache[0] < uoffset) {
6108 /* The cache already knows part of the way. */
6109 if (cache[0] > uoffset0) {
6110 /* The cache knows more than the passed in pair */
6111 uoffset0 = cache[0];
6112 boffset0 = cache[1];
6114 if ((*mgp)->mg_len != -1) {
6115 /* And we know the end too. */
6117 + sv_pos_u2b_midway(start + boffset0, send,
6119 (*mgp)->mg_len - uoffset0);
6122 + sv_pos_u2b_forwards(start + boffset0,
6123 send, uoffset - uoffset0);
6126 else if (cache[2] < uoffset) {
6127 /* We're between the two cache entries. */
6128 if (cache[2] > uoffset0) {
6129 /* and the cache knows more than the passed in pair */
6130 uoffset0 = cache[2];
6131 boffset0 = cache[3];
6135 + sv_pos_u2b_midway(start + boffset0,
6138 cache[0] - uoffset0);
6141 + sv_pos_u2b_midway(start + boffset0,
6144 cache[2] - uoffset0);
6148 else if ((*mgp)->mg_len != -1) {
6149 /* If we can take advantage of a passed in offset, do so. */
6150 /* In fact, offset0 is either 0, or less than offset, so don't
6151 need to worry about the other possibility. */
6153 + sv_pos_u2b_midway(start + boffset0, send,
6155 (*mgp)->mg_len - uoffset0);
6160 if (!found || PL_utf8cache < 0) {
6161 const STRLEN real_boffset
6162 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6163 send, uoffset - uoffset0);
6165 if (found && PL_utf8cache < 0) {
6166 if (real_boffset != boffset) {
6167 /* Need to turn the assertions off otherwise we may recurse
6168 infinitely while printing error messages. */
6169 SAVEI8(PL_utf8cache);
6171 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6172 " real %"UVuf" for %"SVf,
6173 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6176 boffset = real_boffset;
6180 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6186 =for apidoc sv_pos_u2b
6188 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6189 the start of the string, to a count of the equivalent number of bytes; if
6190 lenp is non-zero, it does the same to lenp, but this time starting from
6191 the offset, rather than from the start of the string. Handles magic and
6198 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6199 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6200 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6205 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6210 PERL_ARGS_ASSERT_SV_POS_U2B;
6215 start = (U8*)SvPV_const(sv, len);
6217 STRLEN uoffset = (STRLEN) *offsetp;
6218 const U8 * const send = start + len;
6220 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6223 *offsetp = (I32) boffset;
6226 /* Convert the relative offset to absolute. */
6227 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6228 const STRLEN boffset2
6229 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6230 uoffset, boffset) - boffset;
6244 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6245 byte length pairing. The (byte) length of the total SV is passed in too,
6246 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6247 may not have updated SvCUR, so we can't rely on reading it directly.
6249 The proffered utf8/byte length pairing isn't used if the cache already has
6250 two pairs, and swapping either for the proffered pair would increase the
6251 RMS of the intervals between known byte offsets.
6253 The cache itself consists of 4 STRLEN values
6254 0: larger UTF-8 offset
6255 1: corresponding byte offset
6256 2: smaller UTF-8 offset
6257 3: corresponding byte offset
6259 Unused cache pairs have the value 0, 0.
6260 Keeping the cache "backwards" means that the invariant of
6261 cache[0] >= cache[2] is maintained even with empty slots, which means that
6262 the code that uses it doesn't need to worry if only 1 entry has actually
6263 been set to non-zero. It also makes the "position beyond the end of the
6264 cache" logic much simpler, as the first slot is always the one to start
6268 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6269 const STRLEN utf8, const STRLEN blen)
6273 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6279 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6281 (*mgp)->mg_len = -1;
6285 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6286 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6287 (*mgp)->mg_ptr = (char *) cache;
6291 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6292 /* SvPOKp() because it's possible that sv has string overloading, and
6293 therefore is a reference, hence SvPVX() is actually a pointer.
6294 This cures the (very real) symptoms of RT 69422, but I'm not actually
6295 sure whether we should even be caching the results of UTF-8
6296 operations on overloading, given that nothing stops overloading
6297 returning a different value every time it's called. */
6298 const U8 *start = (const U8 *) SvPVX_const(sv);
6299 const STRLEN realutf8 = utf8_length(start, start + byte);
6301 if (realutf8 != utf8) {
6302 /* Need to turn the assertions off otherwise we may recurse
6303 infinitely while printing error messages. */
6304 SAVEI8(PL_utf8cache);
6306 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6307 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6311 /* Cache is held with the later position first, to simplify the code
6312 that deals with unbounded ends. */
6314 ASSERT_UTF8_CACHE(cache);
6315 if (cache[1] == 0) {
6316 /* Cache is totally empty */
6319 } else if (cache[3] == 0) {
6320 if (byte > cache[1]) {
6321 /* New one is larger, so goes first. */
6322 cache[2] = cache[0];
6323 cache[3] = cache[1];
6331 #define THREEWAY_SQUARE(a,b,c,d) \
6332 ((float)((d) - (c))) * ((float)((d) - (c))) \
6333 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6334 + ((float)((b) - (a))) * ((float)((b) - (a)))
6336 /* Cache has 2 slots in use, and we know three potential pairs.
6337 Keep the two that give the lowest RMS distance. Do the
6338 calcualation in bytes simply because we always know the byte
6339 length. squareroot has the same ordering as the positive value,
6340 so don't bother with the actual square root. */
6341 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6342 if (byte > cache[1]) {
6343 /* New position is after the existing pair of pairs. */
6344 const float keep_earlier
6345 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6346 const float keep_later
6347 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6349 if (keep_later < keep_earlier) {
6350 if (keep_later < existing) {
6351 cache[2] = cache[0];
6352 cache[3] = cache[1];
6358 if (keep_earlier < existing) {
6364 else if (byte > cache[3]) {
6365 /* New position is between the existing pair of pairs. */
6366 const float keep_earlier
6367 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6368 const float keep_later
6369 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6371 if (keep_later < keep_earlier) {
6372 if (keep_later < existing) {
6378 if (keep_earlier < existing) {
6385 /* New position is before the existing pair of pairs. */
6386 const float keep_earlier
6387 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6388 const float keep_later
6389 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6391 if (keep_later < keep_earlier) {
6392 if (keep_later < existing) {
6398 if (keep_earlier < existing) {
6399 cache[0] = cache[2];
6400 cache[1] = cache[3];
6407 ASSERT_UTF8_CACHE(cache);
6410 /* We already know all of the way, now we may be able to walk back. The same
6411 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6412 backward is half the speed of walking forward. */
6414 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6415 const U8 *end, STRLEN endu)
6417 const STRLEN forw = target - s;
6418 STRLEN backw = end - target;
6420 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6422 if (forw < 2 * backw) {
6423 return utf8_length(s, target);
6426 while (end > target) {
6428 while (UTF8_IS_CONTINUATION(*end)) {
6437 =for apidoc sv_pos_b2u
6439 Converts the value pointed to by offsetp from a count of bytes from the
6440 start of the string, to a count of the equivalent number of UTF-8 chars.
6441 Handles magic and type coercion.
6447 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6448 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6453 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6456 const STRLEN byte = *offsetp;
6457 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6463 PERL_ARGS_ASSERT_SV_POS_B2U;
6468 s = (const U8*)SvPV_const(sv, blen);
6471 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6475 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6476 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6478 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6479 if (cache[1] == byte) {
6480 /* An exact match. */
6481 *offsetp = cache[0];
6484 if (cache[3] == byte) {
6485 /* An exact match. */
6486 *offsetp = cache[2];
6490 if (cache[1] < byte) {
6491 /* We already know part of the way. */
6492 if (mg->mg_len != -1) {
6493 /* Actually, we know the end too. */
6495 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6496 s + blen, mg->mg_len - cache[0]);
6498 len = cache[0] + utf8_length(s + cache[1], send);
6501 else if (cache[3] < byte) {
6502 /* We're between the two cached pairs, so we do the calculation
6503 offset by the byte/utf-8 positions for the earlier pair,
6504 then add the utf-8 characters from the string start to
6506 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6507 s + cache[1], cache[0] - cache[2])
6511 else { /* cache[3] > byte */
6512 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6516 ASSERT_UTF8_CACHE(cache);
6518 } else if (mg->mg_len != -1) {
6519 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6523 if (!found || PL_utf8cache < 0) {
6524 const STRLEN real_len = utf8_length(s, send);
6526 if (found && PL_utf8cache < 0) {
6527 if (len != real_len) {
6528 /* Need to turn the assertions off otherwise we may recurse
6529 infinitely while printing error messages. */
6530 SAVEI8(PL_utf8cache);
6532 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6533 " real %"UVuf" for %"SVf,
6534 (UV) len, (UV) real_len, SVfARG(sv));
6542 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6548 Returns a boolean indicating whether the strings in the two SVs are
6549 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6550 coerce its args to strings if necessary.
6556 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6565 SV* svrecode = NULL;
6572 /* if pv1 and pv2 are the same, second SvPV_const call may
6573 * invalidate pv1, so we may need to make a copy */
6574 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6575 pv1 = SvPV_const(sv1, cur1);
6576 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6578 pv1 = SvPV_const(sv1, cur1);
6586 pv2 = SvPV_const(sv2, cur2);
6588 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6589 /* Differing utf8ness.
6590 * Do not UTF8size the comparands as a side-effect. */
6593 svrecode = newSVpvn(pv2, cur2);
6594 sv_recode_to_utf8(svrecode, PL_encoding);
6595 pv2 = SvPV_const(svrecode, cur2);
6598 svrecode = newSVpvn(pv1, cur1);
6599 sv_recode_to_utf8(svrecode, PL_encoding);
6600 pv1 = SvPV_const(svrecode, cur1);
6602 /* Now both are in UTF-8. */
6604 SvREFCNT_dec(svrecode);
6609 bool is_utf8 = TRUE;
6612 /* sv1 is the UTF-8 one,
6613 * if is equal it must be downgrade-able */
6614 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6620 /* sv2 is the UTF-8 one,
6621 * if is equal it must be downgrade-able */
6622 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6628 /* Downgrade not possible - cannot be eq */
6636 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6638 SvREFCNT_dec(svrecode);
6648 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6649 string in C<sv1> is less than, equal to, or greater than the string in
6650 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6651 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6657 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6661 const char *pv1, *pv2;
6664 SV *svrecode = NULL;
6671 pv1 = SvPV_const(sv1, cur1);
6678 pv2 = SvPV_const(sv2, cur2);
6680 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6681 /* Differing utf8ness.
6682 * Do not UTF8size the comparands as a side-effect. */
6685 svrecode = newSVpvn(pv2, cur2);
6686 sv_recode_to_utf8(svrecode, PL_encoding);
6687 pv2 = SvPV_const(svrecode, cur2);
6690 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6695 svrecode = newSVpvn(pv1, cur1);
6696 sv_recode_to_utf8(svrecode, PL_encoding);
6697 pv1 = SvPV_const(svrecode, cur1);
6700 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6706 cmp = cur2 ? -1 : 0;
6710 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6713 cmp = retval < 0 ? -1 : 1;
6714 } else if (cur1 == cur2) {
6717 cmp = cur1 < cur2 ? -1 : 1;
6721 SvREFCNT_dec(svrecode);
6729 =for apidoc sv_cmp_locale
6731 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6732 'use bytes' aware, handles get magic, and will coerce its args to strings
6733 if necessary. See also C<sv_cmp>.
6739 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6742 #ifdef USE_LOCALE_COLLATE
6748 if (PL_collation_standard)
6752 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6754 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6756 if (!pv1 || !len1) {
6767 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6770 return retval < 0 ? -1 : 1;
6773 * When the result of collation is equality, that doesn't mean
6774 * that there are no differences -- some locales exclude some
6775 * characters from consideration. So to avoid false equalities,
6776 * we use the raw string as a tiebreaker.
6782 #endif /* USE_LOCALE_COLLATE */
6784 return sv_cmp(sv1, sv2);
6788 #ifdef USE_LOCALE_COLLATE
6791 =for apidoc sv_collxfrm
6793 Add Collate Transform magic to an SV if it doesn't already have it.
6795 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6796 scalar data of the variable, but transformed to such a format that a normal
6797 memory comparison can be used to compare the data according to the locale
6804 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6809 PERL_ARGS_ASSERT_SV_COLLXFRM;
6811 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6812 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6818 Safefree(mg->mg_ptr);
6819 s = SvPV_const(sv, len);
6820 if ((xf = mem_collxfrm(s, len, &xlen))) {
6822 #ifdef PERL_OLD_COPY_ON_WRITE
6824 sv_force_normal_flags(sv, 0);
6826 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6840 if (mg && mg->mg_ptr) {
6842 return mg->mg_ptr + sizeof(PL_collation_ix);
6850 #endif /* USE_LOCALE_COLLATE */
6855 Get a line from the filehandle and store it into the SV, optionally
6856 appending to the currently-stored string.
6862 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6867 register STDCHAR rslast;
6868 register STDCHAR *bp;
6873 PERL_ARGS_ASSERT_SV_GETS;
6875 if (SvTHINKFIRST(sv))
6876 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6877 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6879 However, perlbench says it's slower, because the existing swipe code
6880 is faster than copy on write.
6881 Swings and roundabouts. */
6882 SvUPGRADE(sv, SVt_PV);
6887 if (PerlIO_isutf8(fp)) {
6889 sv_utf8_upgrade_nomg(sv);
6890 sv_pos_u2b(sv,&append,0);
6892 } else if (SvUTF8(sv)) {
6893 SV * const tsv = newSV(0);
6894 sv_gets(tsv, fp, 0);
6895 sv_utf8_upgrade_nomg(tsv);
6896 SvCUR_set(sv,append);
6899 goto return_string_or_null;
6904 if (PerlIO_isutf8(fp))
6907 if (IN_PERL_COMPILETIME) {
6908 /* we always read code in line mode */
6912 else if (RsSNARF(PL_rs)) {
6913 /* If it is a regular disk file use size from stat() as estimate
6914 of amount we are going to read -- may result in mallocing
6915 more memory than we really need if the layers below reduce
6916 the size we read (e.g. CRLF or a gzip layer).
6919 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6920 const Off_t offset = PerlIO_tell(fp);
6921 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6922 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6928 else if (RsRECORD(PL_rs)) {
6936 /* Grab the size of the record we're getting */
6937 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6938 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6941 /* VMS wants read instead of fread, because fread doesn't respect */
6942 /* RMS record boundaries. This is not necessarily a good thing to be */
6943 /* doing, but we've got no other real choice - except avoid stdio
6944 as implementation - perhaps write a :vms layer ?
6946 fd = PerlIO_fileno(fp);
6947 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6948 bytesread = PerlIO_read(fp, buffer, recsize);
6951 bytesread = PerlLIO_read(fd, buffer, recsize);
6954 bytesread = PerlIO_read(fp, buffer, recsize);
6958 SvCUR_set(sv, bytesread + append);
6959 buffer[bytesread] = '\0';
6960 goto return_string_or_null;
6962 else if (RsPARA(PL_rs)) {
6968 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6969 if (PerlIO_isutf8(fp)) {
6970 rsptr = SvPVutf8(PL_rs, rslen);
6973 if (SvUTF8(PL_rs)) {
6974 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6975 Perl_croak(aTHX_ "Wide character in $/");
6978 rsptr = SvPV_const(PL_rs, rslen);
6982 rslast = rslen ? rsptr[rslen - 1] : '\0';
6984 if (rspara) { /* have to do this both before and after */
6985 do { /* to make sure file boundaries work right */
6988 i = PerlIO_getc(fp);
6992 PerlIO_ungetc(fp,i);
6998 /* See if we know enough about I/O mechanism to cheat it ! */
7000 /* This used to be #ifdef test - it is made run-time test for ease
7001 of abstracting out stdio interface. One call should be cheap
7002 enough here - and may even be a macro allowing compile
7006 if (PerlIO_fast_gets(fp)) {
7009 * We're going to steal some values from the stdio struct
7010 * and put EVERYTHING in the innermost loop into registers.
7012 register STDCHAR *ptr;
7016 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7017 /* An ungetc()d char is handled separately from the regular
7018 * buffer, so we getc() it back out and stuff it in the buffer.
7020 i = PerlIO_getc(fp);
7021 if (i == EOF) return 0;
7022 *(--((*fp)->_ptr)) = (unsigned char) i;
7026 /* Here is some breathtakingly efficient cheating */
7028 cnt = PerlIO_get_cnt(fp); /* get count into register */
7029 /* make sure we have the room */
7030 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7031 /* Not room for all of it
7032 if we are looking for a separator and room for some
7034 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7035 /* just process what we have room for */
7036 shortbuffered = cnt - SvLEN(sv) + append + 1;
7037 cnt -= shortbuffered;
7041 /* remember that cnt can be negative */
7042 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7047 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7048 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7049 DEBUG_P(PerlIO_printf(Perl_debug_log,
7050 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7051 DEBUG_P(PerlIO_printf(Perl_debug_log,
7052 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7053 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7054 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7059 while (cnt > 0) { /* this | eat */
7061 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7062 goto thats_all_folks; /* screams | sed :-) */
7066 Copy(ptr, bp, cnt, char); /* this | eat */
7067 bp += cnt; /* screams | dust */
7068 ptr += cnt; /* louder | sed :-) */
7073 if (shortbuffered) { /* oh well, must extend */
7074 cnt = shortbuffered;
7076 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7078 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7079 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7083 DEBUG_P(PerlIO_printf(Perl_debug_log,
7084 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7085 PTR2UV(ptr),(long)cnt));
7086 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7088 DEBUG_P(PerlIO_printf(Perl_debug_log,
7089 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7090 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7091 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7093 /* This used to call 'filbuf' in stdio form, but as that behaves like
7094 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7095 another abstraction. */
7096 i = PerlIO_getc(fp); /* get more characters */
7098 DEBUG_P(PerlIO_printf(Perl_debug_log,
7099 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7100 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7101 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7103 cnt = PerlIO_get_cnt(fp);
7104 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7105 DEBUG_P(PerlIO_printf(Perl_debug_log,
7106 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7108 if (i == EOF) /* all done for ever? */
7109 goto thats_really_all_folks;
7111 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7113 SvGROW(sv, bpx + cnt + 2);
7114 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7116 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7118 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7119 goto thats_all_folks;
7123 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7124 memNE((char*)bp - rslen, rsptr, rslen))
7125 goto screamer; /* go back to the fray */
7126 thats_really_all_folks:
7128 cnt += shortbuffered;
7129 DEBUG_P(PerlIO_printf(Perl_debug_log,
7130 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7131 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7132 DEBUG_P(PerlIO_printf(Perl_debug_log,
7133 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7134 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7135 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7137 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7138 DEBUG_P(PerlIO_printf(Perl_debug_log,
7139 "Screamer: done, len=%ld, string=|%.*s|\n",
7140 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7144 /*The big, slow, and stupid way. */
7145 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7146 STDCHAR *buf = NULL;
7147 Newx(buf, 8192, STDCHAR);
7155 register const STDCHAR * const bpe = buf + sizeof(buf);
7157 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7158 ; /* keep reading */
7162 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7163 /* Accomodate broken VAXC compiler, which applies U8 cast to
7164 * both args of ?: operator, causing EOF to change into 255
7167 i = (U8)buf[cnt - 1];
7173 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7175 sv_catpvn(sv, (char *) buf, cnt);
7177 sv_setpvn(sv, (char *) buf, cnt);
7179 if (i != EOF && /* joy */
7181 SvCUR(sv) < rslen ||
7182 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7186 * If we're reading from a TTY and we get a short read,
7187 * indicating that the user hit his EOF character, we need
7188 * to notice it now, because if we try to read from the TTY
7189 * again, the EOF condition will disappear.
7191 * The comparison of cnt to sizeof(buf) is an optimization
7192 * that prevents unnecessary calls to feof().
7196 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7200 #ifdef USE_HEAP_INSTEAD_OF_STACK
7205 if (rspara) { /* have to do this both before and after */
7206 while (i != EOF) { /* to make sure file boundaries work right */
7207 i = PerlIO_getc(fp);
7209 PerlIO_ungetc(fp,i);
7215 return_string_or_null:
7216 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7222 Auto-increment of the value in the SV, doing string to numeric conversion
7223 if necessary. Handles 'get' magic.
7229 Perl_sv_inc(pTHX_ register SV *const sv)
7238 if (SvTHINKFIRST(sv)) {
7240 sv_force_normal_flags(sv, 0);
7241 if (SvREADONLY(sv)) {
7242 if (IN_PERL_RUNTIME)
7243 Perl_croak(aTHX_ "%s", PL_no_modify);
7247 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7249 i = PTR2IV(SvRV(sv));
7254 flags = SvFLAGS(sv);
7255 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7256 /* It's (privately or publicly) a float, but not tested as an
7257 integer, so test it to see. */
7259 flags = SvFLAGS(sv);
7261 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7262 /* It's publicly an integer, or privately an integer-not-float */
7263 #ifdef PERL_PRESERVE_IVUV
7267 if (SvUVX(sv) == UV_MAX)
7268 sv_setnv(sv, UV_MAX_P1);
7270 (void)SvIOK_only_UV(sv);
7271 SvUV_set(sv, SvUVX(sv) + 1);
7273 if (SvIVX(sv) == IV_MAX)
7274 sv_setuv(sv, (UV)IV_MAX + 1);
7276 (void)SvIOK_only(sv);
7277 SvIV_set(sv, SvIVX(sv) + 1);
7282 if (flags & SVp_NOK) {
7283 const NV was = SvNVX(sv);
7284 if (NV_OVERFLOWS_INTEGERS_AT &&
7285 was >= NV_OVERFLOWS_INTEGERS_AT) {
7286 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7287 "Lost precision when incrementing %" NVff " by 1",
7290 (void)SvNOK_only(sv);
7291 SvNV_set(sv, was + 1.0);
7295 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7296 if ((flags & SVTYPEMASK) < SVt_PVIV)
7297 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7298 (void)SvIOK_only(sv);
7303 while (isALPHA(*d)) d++;
7304 while (isDIGIT(*d)) d++;
7305 if (d < SvEND(sv)) {
7306 #ifdef PERL_PRESERVE_IVUV
7307 /* Got to punt this as an integer if needs be, but we don't issue
7308 warnings. Probably ought to make the sv_iv_please() that does
7309 the conversion if possible, and silently. */
7310 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7311 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7312 /* Need to try really hard to see if it's an integer.
7313 9.22337203685478e+18 is an integer.
7314 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7315 so $a="9.22337203685478e+18"; $a+0; $a++
7316 needs to be the same as $a="9.22337203685478e+18"; $a++
7323 /* sv_2iv *should* have made this an NV */
7324 if (flags & SVp_NOK) {
7325 (void)SvNOK_only(sv);
7326 SvNV_set(sv, SvNVX(sv) + 1.0);
7329 /* I don't think we can get here. Maybe I should assert this
7330 And if we do get here I suspect that sv_setnv will croak. NWC
7332 #if defined(USE_LONG_DOUBLE)
7333 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",
7334 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7336 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7337 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7340 #endif /* PERL_PRESERVE_IVUV */
7341 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7345 while (d >= SvPVX_const(sv)) {
7353 /* MKS: The original code here died if letters weren't consecutive.
7354 * at least it didn't have to worry about non-C locales. The
7355 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7356 * arranged in order (although not consecutively) and that only
7357 * [A-Za-z] are accepted by isALPHA in the C locale.
7359 if (*d != 'z' && *d != 'Z') {
7360 do { ++*d; } while (!isALPHA(*d));
7363 *(d--) -= 'z' - 'a';
7368 *(d--) -= 'z' - 'a' + 1;
7372 /* oh,oh, the number grew */
7373 SvGROW(sv, SvCUR(sv) + 2);
7374 SvCUR_set(sv, SvCUR(sv) + 1);
7375 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7386 Auto-decrement of the value in the SV, doing string to numeric conversion
7387 if necessary. Handles 'get' magic.
7393 Perl_sv_dec(pTHX_ register SV *const sv)
7401 if (SvTHINKFIRST(sv)) {
7403 sv_force_normal_flags(sv, 0);
7404 if (SvREADONLY(sv)) {
7405 if (IN_PERL_RUNTIME)
7406 Perl_croak(aTHX_ "%s", PL_no_modify);
7410 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7412 i = PTR2IV(SvRV(sv));
7417 /* Unlike sv_inc we don't have to worry about string-never-numbers
7418 and keeping them magic. But we mustn't warn on punting */
7419 flags = SvFLAGS(sv);
7420 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7421 /* It's publicly an integer, or privately an integer-not-float */
7422 #ifdef PERL_PRESERVE_IVUV
7426 if (SvUVX(sv) == 0) {
7427 (void)SvIOK_only(sv);
7431 (void)SvIOK_only_UV(sv);
7432 SvUV_set(sv, SvUVX(sv) - 1);
7435 if (SvIVX(sv) == IV_MIN) {
7436 sv_setnv(sv, (NV)IV_MIN);
7440 (void)SvIOK_only(sv);
7441 SvIV_set(sv, SvIVX(sv) - 1);
7446 if (flags & SVp_NOK) {
7449 const NV was = SvNVX(sv);
7450 if (NV_OVERFLOWS_INTEGERS_AT &&
7451 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7452 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7453 "Lost precision when decrementing %" NVff " by 1",
7456 (void)SvNOK_only(sv);
7457 SvNV_set(sv, was - 1.0);
7461 if (!(flags & SVp_POK)) {
7462 if ((flags & SVTYPEMASK) < SVt_PVIV)
7463 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7465 (void)SvIOK_only(sv);
7468 #ifdef PERL_PRESERVE_IVUV
7470 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7471 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7472 /* Need to try really hard to see if it's an integer.
7473 9.22337203685478e+18 is an integer.
7474 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7475 so $a="9.22337203685478e+18"; $a+0; $a--
7476 needs to be the same as $a="9.22337203685478e+18"; $a--
7483 /* sv_2iv *should* have made this an NV */
7484 if (flags & SVp_NOK) {
7485 (void)SvNOK_only(sv);
7486 SvNV_set(sv, SvNVX(sv) - 1.0);
7489 /* I don't think we can get here. Maybe I should assert this
7490 And if we do get here I suspect that sv_setnv will croak. NWC
7492 #if defined(USE_LONG_DOUBLE)
7493 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",
7494 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7496 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7497 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7501 #endif /* PERL_PRESERVE_IVUV */
7502 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7505 /* this define is used to eliminate a chunk of duplicated but shared logic
7506 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7507 * used anywhere but here - yves
7509 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7512 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7516 =for apidoc sv_mortalcopy
7518 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7519 The new SV is marked as mortal. It will be destroyed "soon", either by an
7520 explicit call to FREETMPS, or by an implicit call at places such as
7521 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7526 /* Make a string that will exist for the duration of the expression
7527 * evaluation. Actually, it may have to last longer than that, but
7528 * hopefully we won't free it until it has been assigned to a
7529 * permanent location. */
7532 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7538 sv_setsv(sv,oldstr);
7539 PUSH_EXTEND_MORTAL__SV_C(sv);
7545 =for apidoc sv_newmortal
7547 Creates a new null SV which is mortal. The reference count of the SV is
7548 set to 1. It will be destroyed "soon", either by an explicit call to
7549 FREETMPS, or by an implicit call at places such as statement boundaries.
7550 See also C<sv_mortalcopy> and C<sv_2mortal>.
7556 Perl_sv_newmortal(pTHX)
7562 SvFLAGS(sv) = SVs_TEMP;
7563 PUSH_EXTEND_MORTAL__SV_C(sv);
7569 =for apidoc newSVpvn_flags
7571 Creates a new SV and copies a string into it. The reference count for the
7572 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7573 string. You are responsible for ensuring that the source string is at least
7574 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7575 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7576 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7577 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7578 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7580 #define newSVpvn_utf8(s, len, u) \
7581 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7587 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7592 /* All the flags we don't support must be zero.
7593 And we're new code so I'm going to assert this from the start. */
7594 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7596 sv_setpvn(sv,s,len);
7598 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7599 * and do what it does outselves here.
7600 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7601 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7602 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7603 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7606 SvFLAGS(sv) |= flags;
7608 if(flags & SVs_TEMP){
7609 PUSH_EXTEND_MORTAL__SV_C(sv);
7616 =for apidoc sv_2mortal
7618 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7619 by an explicit call to FREETMPS, or by an implicit call at places such as
7620 statement boundaries. SvTEMP() is turned on which means that the SV's
7621 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7622 and C<sv_mortalcopy>.
7628 Perl_sv_2mortal(pTHX_ register SV *const sv)
7633 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7635 PUSH_EXTEND_MORTAL__SV_C(sv);
7643 Creates a new SV and copies a string into it. The reference count for the
7644 SV is set to 1. If C<len> is zero, Perl will compute the length using
7645 strlen(). For efficiency, consider using C<newSVpvn> instead.
7651 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7657 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7662 =for apidoc newSVpvn
7664 Creates a new SV and copies a string into it. The reference count for the
7665 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7666 string. You are responsible for ensuring that the source string is at least
7667 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7673 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7679 sv_setpvn(sv,s,len);
7684 =for apidoc newSVhek
7686 Creates a new SV from the hash key structure. It will generate scalars that
7687 point to the shared string table where possible. Returns a new (undefined)
7688 SV if the hek is NULL.
7694 Perl_newSVhek(pTHX_ const HEK *const hek)
7704 if (HEK_LEN(hek) == HEf_SVKEY) {
7705 return newSVsv(*(SV**)HEK_KEY(hek));
7707 const int flags = HEK_FLAGS(hek);
7708 if (flags & HVhek_WASUTF8) {
7710 Andreas would like keys he put in as utf8 to come back as utf8
7712 STRLEN utf8_len = HEK_LEN(hek);
7713 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7714 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7717 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7719 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7720 /* We don't have a pointer to the hv, so we have to replicate the
7721 flag into every HEK. This hv is using custom a hasing
7722 algorithm. Hence we can't return a shared string scalar, as
7723 that would contain the (wrong) hash value, and might get passed
7724 into an hv routine with a regular hash.
7725 Similarly, a hash that isn't using shared hash keys has to have
7726 the flag in every key so that we know not to try to call
7727 share_hek_kek on it. */
7729 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7734 /* This will be overwhelminly the most common case. */
7736 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7737 more efficient than sharepvn(). */
7741 sv_upgrade(sv, SVt_PV);
7742 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7743 SvCUR_set(sv, HEK_LEN(hek));
7756 =for apidoc newSVpvn_share
7758 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7759 table. If the string does not already exist in the table, it is created
7760 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7761 value is used; otherwise the hash is computed. The string's hash can be later
7762 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7763 that as the string table is used for shared hash keys these strings will have
7764 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7770 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7774 bool is_utf8 = FALSE;
7775 const char *const orig_src = src;
7778 STRLEN tmplen = -len;
7780 /* See the note in hv.c:hv_fetch() --jhi */
7781 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7785 PERL_HASH(hash, src, len);
7787 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7788 changes here, update it there too. */
7789 sv_upgrade(sv, SVt_PV);
7790 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7798 if (src != orig_src)
7804 #if defined(PERL_IMPLICIT_CONTEXT)
7806 /* pTHX_ magic can't cope with varargs, so this is a no-context
7807 * version of the main function, (which may itself be aliased to us).
7808 * Don't access this version directly.
7812 Perl_newSVpvf_nocontext(const char *const pat, ...)
7818 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7820 va_start(args, pat);
7821 sv = vnewSVpvf(pat, &args);
7828 =for apidoc newSVpvf
7830 Creates a new SV and initializes it with the string formatted like
7837 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7842 PERL_ARGS_ASSERT_NEWSVPVF;
7844 va_start(args, pat);
7845 sv = vnewSVpvf(pat, &args);
7850 /* backend for newSVpvf() and newSVpvf_nocontext() */
7853 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7858 PERL_ARGS_ASSERT_VNEWSVPVF;
7861 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7868 Creates a new SV and copies a floating point value into it.
7869 The reference count for the SV is set to 1.
7875 Perl_newSVnv(pTHX_ const NV n)
7888 Creates a new SV and copies an integer into it. The reference count for the
7895 Perl_newSViv(pTHX_ const IV i)
7908 Creates a new SV and copies an unsigned integer into it.
7909 The reference count for the SV is set to 1.
7915 Perl_newSVuv(pTHX_ const UV u)
7926 =for apidoc newSV_type
7928 Creates a new SV, of the type specified. The reference count for the new SV
7935 Perl_newSV_type(pTHX_ const svtype type)
7940 sv_upgrade(sv, type);
7945 =for apidoc newRV_noinc
7947 Creates an RV wrapper for an SV. The reference count for the original
7948 SV is B<not> incremented.
7954 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7957 register SV *sv = newSV_type(SVt_IV);
7959 PERL_ARGS_ASSERT_NEWRV_NOINC;
7962 SvRV_set(sv, tmpRef);
7967 /* newRV_inc is the official function name to use now.
7968 * newRV_inc is in fact #defined to newRV in sv.h
7972 Perl_newRV(pTHX_ SV *const sv)
7976 PERL_ARGS_ASSERT_NEWRV;
7978 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7984 Creates a new SV which is an exact duplicate of the original SV.
7991 Perl_newSVsv(pTHX_ register SV *const old)
7998 if (SvTYPE(old) == SVTYPEMASK) {
7999 if (ckWARN_d(WARN_INTERNAL))
8000 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8004 /* SV_GMAGIC is the default for sv_setv()
8005 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8006 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8007 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8012 =for apidoc sv_reset
8014 Underlying implementation for the C<reset> Perl function.
8015 Note that the perl-level function is vaguely deprecated.
8021 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8024 char todo[PERL_UCHAR_MAX+1];
8026 PERL_ARGS_ASSERT_SV_RESET;
8031 if (!*s) { /* reset ?? searches */
8032 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8034 const U32 count = mg->mg_len / sizeof(PMOP**);
8035 PMOP **pmp = (PMOP**) mg->mg_ptr;
8036 PMOP *const *const end = pmp + count;
8040 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8042 (*pmp)->op_pmflags &= ~PMf_USED;
8050 /* reset variables */
8052 if (!HvARRAY(stash))
8055 Zero(todo, 256, char);
8058 I32 i = (unsigned char)*s;
8062 max = (unsigned char)*s++;
8063 for ( ; i <= max; i++) {
8066 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8068 for (entry = HvARRAY(stash)[i];
8070 entry = HeNEXT(entry))
8075 if (!todo[(U8)*HeKEY(entry)])
8077 gv = MUTABLE_GV(HeVAL(entry));
8080 if (SvTHINKFIRST(sv)) {
8081 if (!SvREADONLY(sv) && SvROK(sv))
8083 /* XXX Is this continue a bug? Why should THINKFIRST
8084 exempt us from resetting arrays and hashes? */
8088 if (SvTYPE(sv) >= SVt_PV) {
8090 if (SvPVX_const(sv) != NULL)
8098 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8100 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8103 # if defined(USE_ENVIRON_ARRAY)
8106 # endif /* USE_ENVIRON_ARRAY */
8117 Using various gambits, try to get an IO from an SV: the IO slot if its a
8118 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8119 named after the PV if we're a string.
8125 Perl_sv_2io(pTHX_ SV *const sv)
8130 PERL_ARGS_ASSERT_SV_2IO;
8132 switch (SvTYPE(sv)) {
8134 io = MUTABLE_IO(sv);
8137 if (isGV_with_GP(sv)) {
8138 gv = MUTABLE_GV(sv);
8141 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8147 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8149 return sv_2io(SvRV(sv));
8150 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8156 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8165 Using various gambits, try to get a CV from an SV; in addition, try if
8166 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8167 The flags in C<lref> are passed to gv_fetchsv.
8173 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8179 PERL_ARGS_ASSERT_SV_2CV;
8186 switch (SvTYPE(sv)) {
8190 return MUTABLE_CV(sv);
8197 if (isGV_with_GP(sv)) {
8198 gv = MUTABLE_GV(sv);
8207 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8209 tryAMAGICunDEREF(to_cv);
8212 if (SvTYPE(sv) == SVt_PVCV) {
8213 cv = MUTABLE_CV(sv);
8218 else if(isGV_with_GP(sv))
8219 gv = MUTABLE_GV(sv);
8221 Perl_croak(aTHX_ "Not a subroutine reference");
8223 else if (isGV_with_GP(sv)) {
8225 gv = MUTABLE_GV(sv);
8228 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8234 /* Some flags to gv_fetchsv mean don't really create the GV */
8235 if (!isGV_with_GP(gv)) {
8241 if (lref && !GvCVu(gv)) {
8245 gv_efullname3(tmpsv, gv, NULL);
8246 /* XXX this is probably not what they think they're getting.
8247 * It has the same effect as "sub name;", i.e. just a forward
8249 newSUB(start_subparse(FALSE, 0),
8250 newSVOP(OP_CONST, 0, tmpsv),
8254 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8255 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8264 Returns true if the SV has a true value by Perl's rules.
8265 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8266 instead use an in-line version.
8272 Perl_sv_true(pTHX_ register SV *const sv)
8277 register const XPV* const tXpv = (XPV*)SvANY(sv);
8279 (tXpv->xpv_cur > 1 ||
8280 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8287 return SvIVX(sv) != 0;
8290 return SvNVX(sv) != 0.0;
8292 return sv_2bool(sv);
8298 =for apidoc sv_pvn_force
8300 Get a sensible string out of the SV somehow.
8301 A private implementation of the C<SvPV_force> macro for compilers which
8302 can't cope with complex macro expressions. Always use the macro instead.
8304 =for apidoc sv_pvn_force_flags
8306 Get a sensible string out of the SV somehow.
8307 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8308 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8309 implemented in terms of this function.
8310 You normally want to use the various wrapper macros instead: see
8311 C<SvPV_force> and C<SvPV_force_nomg>
8317 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8321 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8323 if (SvTHINKFIRST(sv) && !SvROK(sv))
8324 sv_force_normal_flags(sv, 0);
8334 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8335 const char * const ref = sv_reftype(sv,0);
8337 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8338 ref, OP_NAME(PL_op));
8340 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8342 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8343 || isGV_with_GP(sv))
8344 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8346 s = sv_2pv_flags(sv, &len, flags);
8350 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8353 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8354 SvGROW(sv, len + 1);
8355 Move(s,SvPVX(sv),len,char);
8357 SvPVX(sv)[len] = '\0';
8360 SvPOK_on(sv); /* validate pointer */
8362 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8363 PTR2UV(sv),SvPVX_const(sv)));
8366 return SvPVX_mutable(sv);
8370 =for apidoc sv_pvbyten_force
8372 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8378 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8380 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8382 sv_pvn_force(sv,lp);
8383 sv_utf8_downgrade(sv,0);
8389 =for apidoc sv_pvutf8n_force
8391 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8397 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8399 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8401 sv_pvn_force(sv,lp);
8402 sv_utf8_upgrade(sv);
8408 =for apidoc sv_reftype
8410 Returns a string describing what the SV is a reference to.
8416 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8418 PERL_ARGS_ASSERT_SV_REFTYPE;
8420 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8421 inside return suggests a const propagation bug in g++. */
8422 if (ob && SvOBJECT(sv)) {
8423 char * const name = HvNAME_get(SvSTASH(sv));
8424 return name ? name : (char *) "__ANON__";
8427 switch (SvTYPE(sv)) {
8442 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8443 /* tied lvalues should appear to be
8444 * scalars for backwards compatitbility */
8445 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8446 ? "SCALAR" : "LVALUE");
8447 case SVt_PVAV: return "ARRAY";
8448 case SVt_PVHV: return "HASH";
8449 case SVt_PVCV: return "CODE";
8450 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8451 ? "GLOB" : "SCALAR");
8452 case SVt_PVFM: return "FORMAT";
8453 case SVt_PVIO: return "IO";
8454 case SVt_BIND: return "BIND";
8455 case SVt_REGEXP: return "REGEXP";
8456 default: return "UNKNOWN";
8462 =for apidoc sv_isobject
8464 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8465 object. If the SV is not an RV, or if the object is not blessed, then this
8472 Perl_sv_isobject(pTHX_ SV *sv)
8488 Returns a boolean indicating whether the SV is blessed into the specified
8489 class. This does not check for subtypes; use C<sv_derived_from> to verify
8490 an inheritance relationship.
8496 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8500 PERL_ARGS_ASSERT_SV_ISA;
8510 hvname = HvNAME_get(SvSTASH(sv));
8514 return strEQ(hvname, name);
8520 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8521 it will be upgraded to one. If C<classname> is non-null then the new SV will
8522 be blessed in the specified package. The new SV is returned and its
8523 reference count is 1.
8529 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8534 PERL_ARGS_ASSERT_NEWSVRV;
8538 SV_CHECK_THINKFIRST_COW_DROP(rv);
8539 (void)SvAMAGIC_off(rv);
8541 if (SvTYPE(rv) >= SVt_PVMG) {
8542 const U32 refcnt = SvREFCNT(rv);
8546 SvREFCNT(rv) = refcnt;
8548 sv_upgrade(rv, SVt_IV);
8549 } else if (SvROK(rv)) {
8550 SvREFCNT_dec(SvRV(rv));
8552 prepare_SV_for_RV(rv);
8560 HV* const stash = gv_stashpv(classname, GV_ADD);
8561 (void)sv_bless(rv, stash);
8567 =for apidoc sv_setref_pv
8569 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8570 argument will be upgraded to an RV. That RV will be modified to point to
8571 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8572 into the SV. The C<classname> argument indicates the package for the
8573 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8574 will have a reference count of 1, and the RV will be returned.
8576 Do not use with other Perl types such as HV, AV, SV, CV, because those
8577 objects will become corrupted by the pointer copy process.
8579 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8585 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8589 PERL_ARGS_ASSERT_SV_SETREF_PV;
8592 sv_setsv(rv, &PL_sv_undef);
8596 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8601 =for apidoc sv_setref_iv
8603 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8604 argument will be upgraded to an RV. That RV will be modified to point to
8605 the new SV. The C<classname> argument indicates the package for the
8606 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8607 will have a reference count of 1, and the RV will be returned.
8613 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8615 PERL_ARGS_ASSERT_SV_SETREF_IV;
8617 sv_setiv(newSVrv(rv,classname), iv);
8622 =for apidoc sv_setref_uv
8624 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8625 argument will be upgraded to an RV. That RV will be modified to point to
8626 the new SV. The C<classname> argument indicates the package for the
8627 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8628 will have a reference count of 1, and the RV will be returned.
8634 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8636 PERL_ARGS_ASSERT_SV_SETREF_UV;
8638 sv_setuv(newSVrv(rv,classname), uv);
8643 =for apidoc sv_setref_nv
8645 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8646 argument will be upgraded to an RV. That RV will be modified to point to
8647 the new SV. The C<classname> argument indicates the package for the
8648 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8649 will have a reference count of 1, and the RV will be returned.
8655 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8657 PERL_ARGS_ASSERT_SV_SETREF_NV;
8659 sv_setnv(newSVrv(rv,classname), nv);
8664 =for apidoc sv_setref_pvn
8666 Copies a string into a new SV, optionally blessing the SV. The length of the
8667 string must be specified with C<n>. The C<rv> argument will be upgraded to
8668 an RV. That RV will be modified to point to the new SV. The C<classname>
8669 argument indicates the package for the blessing. Set C<classname> to
8670 C<NULL> to avoid the blessing. The new SV will have a reference count
8671 of 1, and the RV will be returned.
8673 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8679 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8680 const char *const pv, const STRLEN n)
8682 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8684 sv_setpvn(newSVrv(rv,classname), pv, n);
8689 =for apidoc sv_bless
8691 Blesses an SV into a specified package. The SV must be an RV. The package
8692 must be designated by its stash (see C<gv_stashpv()>). The reference count
8693 of the SV is unaffected.
8699 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8704 PERL_ARGS_ASSERT_SV_BLESS;
8707 Perl_croak(aTHX_ "Can't bless non-reference value");
8709 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8710 if (SvIsCOW(tmpRef))
8711 sv_force_normal_flags(tmpRef, 0);
8712 if (SvREADONLY(tmpRef))
8713 Perl_croak(aTHX_ "%s", PL_no_modify);
8714 if (SvOBJECT(tmpRef)) {
8715 if (SvTYPE(tmpRef) != SVt_PVIO)
8717 SvREFCNT_dec(SvSTASH(tmpRef));
8720 SvOBJECT_on(tmpRef);
8721 if (SvTYPE(tmpRef) != SVt_PVIO)
8723 SvUPGRADE(tmpRef, SVt_PVMG);
8724 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8729 (void)SvAMAGIC_off(sv);
8731 if(SvSMAGICAL(tmpRef))
8732 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8740 /* Downgrades a PVGV to a PVMG.
8744 S_sv_unglob(pTHX_ SV *const sv)
8749 SV * const temp = sv_newmortal();
8751 PERL_ARGS_ASSERT_SV_UNGLOB;
8753 assert(SvTYPE(sv) == SVt_PVGV);
8755 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8758 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8759 && HvNAME_get(stash))
8760 mro_method_changed_in(stash);
8761 gp_free(MUTABLE_GV(sv));
8764 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8768 if (GvNAME_HEK(sv)) {
8769 unshare_hek(GvNAME_HEK(sv));
8771 isGV_with_GP_off(sv);
8773 /* need to keep SvANY(sv) in the right arena */
8774 xpvmg = new_XPVMG();
8775 StructCopy(SvANY(sv), xpvmg, XPVMG);
8776 del_XPVGV(SvANY(sv));
8779 SvFLAGS(sv) &= ~SVTYPEMASK;
8780 SvFLAGS(sv) |= SVt_PVMG;
8782 /* Intentionally not calling any local SET magic, as this isn't so much a
8783 set operation as merely an internal storage change. */
8784 sv_setsv_flags(sv, temp, 0);
8788 =for apidoc sv_unref_flags
8790 Unsets the RV status of the SV, and decrements the reference count of
8791 whatever was being referenced by the RV. This can almost be thought of
8792 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8793 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8794 (otherwise the decrementing is conditional on the reference count being
8795 different from one or the reference being a readonly SV).
8802 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8804 SV* const target = SvRV(ref);
8806 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8808 if (SvWEAKREF(ref)) {
8809 sv_del_backref(target, ref);
8811 SvRV_set(ref, NULL);
8814 SvRV_set(ref, NULL);
8816 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8817 assigned to as BEGIN {$a = \"Foo"} will fail. */
8818 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8819 SvREFCNT_dec(target);
8820 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8821 sv_2mortal(target); /* Schedule for freeing later */
8825 =for apidoc sv_untaint
8827 Untaint an SV. Use C<SvTAINTED_off> instead.
8832 Perl_sv_untaint(pTHX_ SV *const sv)
8834 PERL_ARGS_ASSERT_SV_UNTAINT;
8836 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8837 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8844 =for apidoc sv_tainted
8846 Test an SV for taintedness. Use C<SvTAINTED> instead.
8851 Perl_sv_tainted(pTHX_ SV *const sv)
8853 PERL_ARGS_ASSERT_SV_TAINTED;
8855 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8856 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8857 if (mg && (mg->mg_len & 1) )
8864 =for apidoc sv_setpviv
8866 Copies an integer into the given SV, also updating its string value.
8867 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8873 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8875 char buf[TYPE_CHARS(UV)];
8877 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8879 PERL_ARGS_ASSERT_SV_SETPVIV;
8881 sv_setpvn(sv, ptr, ebuf - ptr);
8885 =for apidoc sv_setpviv_mg
8887 Like C<sv_setpviv>, but also handles 'set' magic.
8893 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8895 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8901 #if defined(PERL_IMPLICIT_CONTEXT)
8903 /* pTHX_ magic can't cope with varargs, so this is a no-context
8904 * version of the main function, (which may itself be aliased to us).
8905 * Don't access this version directly.
8909 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8914 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8916 va_start(args, pat);
8917 sv_vsetpvf(sv, pat, &args);
8921 /* pTHX_ magic can't cope with varargs, so this is a no-context
8922 * version of the main function, (which may itself be aliased to us).
8923 * Don't access this version directly.
8927 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8932 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8934 va_start(args, pat);
8935 sv_vsetpvf_mg(sv, pat, &args);
8941 =for apidoc sv_setpvf
8943 Works like C<sv_catpvf> but copies the text into the SV instead of
8944 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8950 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8954 PERL_ARGS_ASSERT_SV_SETPVF;
8956 va_start(args, pat);
8957 sv_vsetpvf(sv, pat, &args);
8962 =for apidoc sv_vsetpvf
8964 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8965 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8967 Usually used via its frontend C<sv_setpvf>.
8973 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8975 PERL_ARGS_ASSERT_SV_VSETPVF;
8977 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8981 =for apidoc sv_setpvf_mg
8983 Like C<sv_setpvf>, but also handles 'set' magic.
8989 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8993 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8995 va_start(args, pat);
8996 sv_vsetpvf_mg(sv, pat, &args);
9001 =for apidoc sv_vsetpvf_mg
9003 Like C<sv_vsetpvf>, but also handles 'set' magic.
9005 Usually used via its frontend C<sv_setpvf_mg>.
9011 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9013 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9015 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9019 #if defined(PERL_IMPLICIT_CONTEXT)
9021 /* pTHX_ magic can't cope with varargs, so this is a no-context
9022 * version of the main function, (which may itself be aliased to us).
9023 * Don't access this version directly.
9027 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9032 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9034 va_start(args, pat);
9035 sv_vcatpvf(sv, pat, &args);
9039 /* pTHX_ magic can't cope with varargs, so this is a no-context
9040 * version of the main function, (which may itself be aliased to us).
9041 * Don't access this version directly.
9045 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9050 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9052 va_start(args, pat);
9053 sv_vcatpvf_mg(sv, pat, &args);
9059 =for apidoc sv_catpvf
9061 Processes its arguments like C<sprintf> and appends the formatted
9062 output to an SV. If the appended data contains "wide" characters
9063 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9064 and characters >255 formatted with %c), the original SV might get
9065 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9066 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9067 valid UTF-8; if the original SV was bytes, the pattern should be too.
9072 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9076 PERL_ARGS_ASSERT_SV_CATPVF;
9078 va_start(args, pat);
9079 sv_vcatpvf(sv, pat, &args);
9084 =for apidoc sv_vcatpvf
9086 Processes its arguments like C<vsprintf> and appends the formatted output
9087 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9089 Usually used via its frontend C<sv_catpvf>.
9095 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9097 PERL_ARGS_ASSERT_SV_VCATPVF;
9099 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9103 =for apidoc sv_catpvf_mg
9105 Like C<sv_catpvf>, but also handles 'set' magic.
9111 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9115 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9117 va_start(args, pat);
9118 sv_vcatpvf_mg(sv, pat, &args);
9123 =for apidoc sv_vcatpvf_mg
9125 Like C<sv_vcatpvf>, but also handles 'set' magic.
9127 Usually used via its frontend C<sv_catpvf_mg>.
9133 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9135 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9137 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9142 =for apidoc sv_vsetpvfn
9144 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9147 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9153 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9154 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9156 PERL_ARGS_ASSERT_SV_VSETPVFN;
9159 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9163 S_expect_number(pTHX_ char **const pattern)
9168 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9170 switch (**pattern) {
9171 case '1': case '2': case '3':
9172 case '4': case '5': case '6':
9173 case '7': case '8': case '9':
9174 var = *(*pattern)++ - '0';
9175 while (isDIGIT(**pattern)) {
9176 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9178 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9186 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9188 const int neg = nv < 0;
9191 PERL_ARGS_ASSERT_F0CONVERT;
9199 if (uv & 1 && uv == nv)
9200 uv--; /* Round to even */
9202 const unsigned dig = uv % 10;
9215 =for apidoc sv_vcatpvfn
9217 Processes its arguments like C<vsprintf> and appends the formatted output
9218 to an SV. Uses an array of SVs if the C style variable argument list is
9219 missing (NULL). When running with taint checks enabled, indicates via
9220 C<maybe_tainted> if results are untrustworthy (often due to the use of
9223 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9229 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9230 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9231 vec_utf8 = DO_UTF8(vecsv);
9233 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9236 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9237 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9245 static const char nullstr[] = "(null)";
9247 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9248 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9250 /* Times 4: a decimal digit takes more than 3 binary digits.
9251 * NV_DIG: mantissa takes than many decimal digits.
9252 * Plus 32: Playing safe. */
9253 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9254 /* large enough for "%#.#f" --chip */
9255 /* what about long double NVs? --jhi */
9257 PERL_ARGS_ASSERT_SV_VCATPVFN;
9258 PERL_UNUSED_ARG(maybe_tainted);
9260 /* no matter what, this is a string now */
9261 (void)SvPV_force(sv, origlen);
9263 /* special-case "", "%s", and "%-p" (SVf - see below) */
9266 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9268 const char * const s = va_arg(*args, char*);
9269 sv_catpv(sv, s ? s : nullstr);
9271 else if (svix < svmax) {
9272 sv_catsv(sv, *svargs);
9276 if (args && patlen == 3 && pat[0] == '%' &&
9277 pat[1] == '-' && pat[2] == 'p') {
9278 argsv = MUTABLE_SV(va_arg(*args, void*));
9279 sv_catsv(sv, argsv);
9283 #ifndef USE_LONG_DOUBLE
9284 /* special-case "%.<number>[gf]" */
9285 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9286 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9287 unsigned digits = 0;
9291 while (*pp >= '0' && *pp <= '9')
9292 digits = 10 * digits + (*pp++ - '0');
9293 if (pp - pat == (int)patlen - 1) {
9301 /* Add check for digits != 0 because it seems that some
9302 gconverts are buggy in this case, and we don't yet have
9303 a Configure test for this. */
9304 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9305 /* 0, point, slack */
9306 Gconvert(nv, (int)digits, 0, ebuf);
9308 if (*ebuf) /* May return an empty string for digits==0 */
9311 } else if (!digits) {
9314 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9315 sv_catpvn(sv, p, l);
9321 #endif /* !USE_LONG_DOUBLE */
9323 if (!args && svix < svmax && DO_UTF8(*svargs))
9326 patend = (char*)pat + patlen;
9327 for (p = (char*)pat; p < patend; p = q) {
9330 bool vectorize = FALSE;
9331 bool vectorarg = FALSE;
9332 bool vec_utf8 = FALSE;
9338 bool has_precis = FALSE;
9340 const I32 osvix = svix;
9341 bool is_utf8 = FALSE; /* is this item utf8? */
9342 #ifdef HAS_LDBL_SPRINTF_BUG
9343 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9344 with sfio - Allen <allens@cpan.org> */
9345 bool fix_ldbl_sprintf_bug = FALSE;
9349 U8 utf8buf[UTF8_MAXBYTES+1];
9350 STRLEN esignlen = 0;
9352 const char *eptr = NULL;
9353 const char *fmtstart;
9356 const U8 *vecstr = NULL;
9363 /* we need a long double target in case HAS_LONG_DOUBLE but
9366 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9374 const char *dotstr = ".";
9375 STRLEN dotstrlen = 1;
9376 I32 efix = 0; /* explicit format parameter index */
9377 I32 ewix = 0; /* explicit width index */
9378 I32 epix = 0; /* explicit precision index */
9379 I32 evix = 0; /* explicit vector index */
9380 bool asterisk = FALSE;
9382 /* echo everything up to the next format specification */
9383 for (q = p; q < patend && *q != '%'; ++q) ;
9385 if (has_utf8 && !pat_utf8)
9386 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9388 sv_catpvn(sv, p, q - p);
9397 We allow format specification elements in this order:
9398 \d+\$ explicit format parameter index
9400 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9401 0 flag (as above): repeated to allow "v02"
9402 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9403 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9405 [%bcdefginopsuxDFOUX] format (mandatory)
9410 As of perl5.9.3, printf format checking is on by default.
9411 Internally, perl uses %p formats to provide an escape to
9412 some extended formatting. This block deals with those
9413 extensions: if it does not match, (char*)q is reset and
9414 the normal format processing code is used.
9416 Currently defined extensions are:
9417 %p include pointer address (standard)
9418 %-p (SVf) include an SV (previously %_)
9419 %-<num>p include an SV with precision <num>
9420 %<num>p reserved for future extensions
9422 Robin Barker 2005-07-14
9424 %1p (VDf) removed. RMB 2007-10-19
9431 n = expect_number(&q);
9438 argsv = MUTABLE_SV(va_arg(*args, void*));
9439 eptr = SvPV_const(argsv, elen);
9445 if (ckWARN_d(WARN_INTERNAL))
9446 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9447 "internal %%<num>p might conflict with future printf extensions");
9453 if ( (width = expect_number(&q)) ) {
9468 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9497 if ( (ewix = expect_number(&q)) )
9506 if ((vectorarg = asterisk)) {
9519 width = expect_number(&q);
9525 vecsv = va_arg(*args, SV*);
9527 vecsv = (evix > 0 && evix <= svmax)
9528 ? svargs[evix-1] : &PL_sv_undef;
9530 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9532 dotstr = SvPV_const(vecsv, dotstrlen);
9533 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9534 bad with tied or overloaded values that return UTF8. */
9537 else if (has_utf8) {
9538 vecsv = sv_mortalcopy(vecsv);
9539 sv_utf8_upgrade(vecsv);
9540 dotstr = SvPV_const(vecsv, dotstrlen);
9547 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9548 vecsv = svargs[efix ? efix-1 : svix++];
9549 vecstr = (U8*)SvPV_const(vecsv,veclen);
9550 vec_utf8 = DO_UTF8(vecsv);
9552 /* if this is a version object, we need to convert
9553 * back into v-string notation and then let the
9554 * vectorize happen normally
9556 if (sv_derived_from(vecsv, "version")) {
9557 char *version = savesvpv(vecsv);
9558 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9559 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9560 "vector argument not supported with alpha versions");
9563 vecsv = sv_newmortal();
9564 scan_vstring(version, version + veclen, vecsv);
9565 vecstr = (U8*)SvPV_const(vecsv, veclen);
9566 vec_utf8 = DO_UTF8(vecsv);
9578 i = va_arg(*args, int);
9580 i = (ewix ? ewix <= svmax : svix < svmax) ?
9581 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9583 width = (i < 0) ? -i : i;
9593 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9595 /* XXX: todo, support specified precision parameter */
9599 i = va_arg(*args, int);
9601 i = (ewix ? ewix <= svmax : svix < svmax)
9602 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9604 has_precis = !(i < 0);
9609 precis = precis * 10 + (*q++ - '0');
9618 case 'I': /* Ix, I32x, and I64x */
9620 if (q[1] == '6' && q[2] == '4') {
9626 if (q[1] == '3' && q[2] == '2') {
9636 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9647 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9648 if (*(q + 1) == 'l') { /* lld, llf */
9674 if (!vectorize && !args) {
9676 const I32 i = efix-1;
9677 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9679 argsv = (svix >= 0 && svix < svmax)
9680 ? svargs[svix++] : &PL_sv_undef;
9691 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9693 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9695 eptr = (char*)utf8buf;
9696 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9710 eptr = va_arg(*args, char*);
9712 elen = strlen(eptr);
9714 eptr = (char *)nullstr;
9715 elen = sizeof nullstr - 1;
9719 eptr = SvPV_const(argsv, elen);
9720 if (DO_UTF8(argsv)) {
9721 STRLEN old_precis = precis;
9722 if (has_precis && precis < elen) {
9723 STRLEN ulen = sv_len_utf8(argsv);
9724 I32 p = precis > ulen ? ulen : precis;
9725 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9728 if (width) { /* fudge width (can't fudge elen) */
9729 if (has_precis && precis < elen)
9730 width += precis - old_precis;
9732 width += elen - sv_len_utf8(argsv);
9739 if (has_precis && precis < elen)
9746 if (alt || vectorize)
9748 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9769 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9778 esignbuf[esignlen++] = plus;
9782 case 'h': iv = (short)va_arg(*args, int); break;
9783 case 'l': iv = va_arg(*args, long); break;
9784 case 'V': iv = va_arg(*args, IV); break;
9785 default: iv = va_arg(*args, int); break;
9788 iv = va_arg(*args, Quad_t); break;
9795 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9797 case 'h': iv = (short)tiv; break;
9798 case 'l': iv = (long)tiv; break;
9800 default: iv = tiv; break;
9803 iv = (Quad_t)tiv; break;
9809 if ( !vectorize ) /* we already set uv above */
9814 esignbuf[esignlen++] = plus;
9818 esignbuf[esignlen++] = '-';
9862 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9873 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9874 case 'l': uv = va_arg(*args, unsigned long); break;
9875 case 'V': uv = va_arg(*args, UV); break;
9876 default: uv = va_arg(*args, unsigned); break;
9879 uv = va_arg(*args, Uquad_t); break;
9886 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9888 case 'h': uv = (unsigned short)tuv; break;
9889 case 'l': uv = (unsigned long)tuv; break;
9891 default: uv = tuv; break;
9894 uv = (Uquad_t)tuv; break;
9903 char *ptr = ebuf + sizeof ebuf;
9904 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9910 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9916 esignbuf[esignlen++] = '0';
9917 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9925 if (alt && *ptr != '0')
9934 esignbuf[esignlen++] = '0';
9935 esignbuf[esignlen++] = c;
9938 default: /* it had better be ten or less */
9942 } while (uv /= base);
9945 elen = (ebuf + sizeof ebuf) - ptr;
9949 zeros = precis - elen;
9950 else if (precis == 0 && elen == 1 && *eptr == '0'
9951 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9954 /* a precision nullifies the 0 flag. */
9961 /* FLOATING POINT */
9964 c = 'f'; /* maybe %F isn't supported here */
9972 /* This is evil, but floating point is even more evil */
9974 /* for SV-style calling, we can only get NV
9975 for C-style calling, we assume %f is double;
9976 for simplicity we allow any of %Lf, %llf, %qf for long double
9980 #if defined(USE_LONG_DOUBLE)
9984 /* [perl #20339] - we should accept and ignore %lf rather than die */
9988 #if defined(USE_LONG_DOUBLE)
9989 intsize = args ? 0 : 'q';
9993 #if defined(HAS_LONG_DOUBLE)
10002 /* now we need (long double) if intsize == 'q', else (double) */
10004 #if LONG_DOUBLESIZE > DOUBLESIZE
10006 va_arg(*args, long double) :
10007 va_arg(*args, double)
10009 va_arg(*args, double)
10014 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10015 else. frexp() has some unspecified behaviour for those three */
10016 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10018 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10019 will cast our (long double) to (double) */
10020 (void)Perl_frexp(nv, &i);
10021 if (i == PERL_INT_MIN)
10022 Perl_die(aTHX_ "panic: frexp");
10024 need = BIT_DIGITS(i);
10026 need += has_precis ? precis : 6; /* known default */
10031 #ifdef HAS_LDBL_SPRINTF_BUG
10032 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10033 with sfio - Allen <allens@cpan.org> */
10036 # define MY_DBL_MAX DBL_MAX
10037 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10038 # if DOUBLESIZE >= 8
10039 # define MY_DBL_MAX 1.7976931348623157E+308L
10041 # define MY_DBL_MAX 3.40282347E+38L
10045 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10046 # define MY_DBL_MAX_BUG 1L
10048 # define MY_DBL_MAX_BUG MY_DBL_MAX
10052 # define MY_DBL_MIN DBL_MIN
10053 # else /* XXX guessing! -Allen */
10054 # if DOUBLESIZE >= 8
10055 # define MY_DBL_MIN 2.2250738585072014E-308L
10057 # define MY_DBL_MIN 1.17549435E-38L
10061 if ((intsize == 'q') && (c == 'f') &&
10062 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10063 (need < DBL_DIG)) {
10064 /* it's going to be short enough that
10065 * long double precision is not needed */
10067 if ((nv <= 0L) && (nv >= -0L))
10068 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10070 /* would use Perl_fp_class as a double-check but not
10071 * functional on IRIX - see perl.h comments */
10073 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10074 /* It's within the range that a double can represent */
10075 #if defined(DBL_MAX) && !defined(DBL_MIN)
10076 if ((nv >= ((long double)1/DBL_MAX)) ||
10077 (nv <= (-(long double)1/DBL_MAX)))
10079 fix_ldbl_sprintf_bug = TRUE;
10082 if (fix_ldbl_sprintf_bug == TRUE) {
10092 # undef MY_DBL_MAX_BUG
10095 #endif /* HAS_LDBL_SPRINTF_BUG */
10097 need += 20; /* fudge factor */
10098 if (PL_efloatsize < need) {
10099 Safefree(PL_efloatbuf);
10100 PL_efloatsize = need + 20; /* more fudge */
10101 Newx(PL_efloatbuf, PL_efloatsize, char);
10102 PL_efloatbuf[0] = '\0';
10105 if ( !(width || left || plus || alt) && fill != '0'
10106 && has_precis && intsize != 'q' ) { /* Shortcuts */
10107 /* See earlier comment about buggy Gconvert when digits,
10109 if ( c == 'g' && precis) {
10110 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10111 /* May return an empty string for digits==0 */
10112 if (*PL_efloatbuf) {
10113 elen = strlen(PL_efloatbuf);
10114 goto float_converted;
10116 } else if ( c == 'f' && !precis) {
10117 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10122 char *ptr = ebuf + sizeof ebuf;
10125 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10126 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10127 if (intsize == 'q') {
10128 /* Copy the one or more characters in a long double
10129 * format before the 'base' ([efgEFG]) character to
10130 * the format string. */
10131 static char const prifldbl[] = PERL_PRIfldbl;
10132 char const *p = prifldbl + sizeof(prifldbl) - 3;
10133 while (p >= prifldbl) { *--ptr = *p--; }
10138 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10143 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10155 /* No taint. Otherwise we are in the strange situation
10156 * where printf() taints but print($float) doesn't.
10158 #if defined(HAS_LONG_DOUBLE)
10159 elen = ((intsize == 'q')
10160 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10161 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10163 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10167 eptr = PL_efloatbuf;
10175 i = SvCUR(sv) - origlen;
10178 case 'h': *(va_arg(*args, short*)) = i; break;
10179 default: *(va_arg(*args, int*)) = i; break;
10180 case 'l': *(va_arg(*args, long*)) = i; break;
10181 case 'V': *(va_arg(*args, IV*)) = i; break;
10184 *(va_arg(*args, Quad_t*)) = i; break;
10191 sv_setuv_mg(argsv, (UV)i);
10192 continue; /* not "break" */
10199 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10200 && ckWARN(WARN_PRINTF))
10202 SV * const msg = sv_newmortal();
10203 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10204 (PL_op->op_type == OP_PRTF) ? "" : "s");
10205 if (fmtstart < patend) {
10206 const char * const fmtend = q < patend ? q : patend;
10208 sv_catpvs(msg, "\"%");
10209 for (f = fmtstart; f < fmtend; f++) {
10211 sv_catpvn(msg, f, 1);
10213 Perl_sv_catpvf(aTHX_ msg,
10214 "\\%03"UVof, (UV)*f & 0xFF);
10217 sv_catpvs(msg, "\"");
10219 sv_catpvs(msg, "end of string");
10221 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10224 /* output mangled stuff ... */
10230 /* ... right here, because formatting flags should not apply */
10231 SvGROW(sv, SvCUR(sv) + elen + 1);
10233 Copy(eptr, p, elen, char);
10236 SvCUR_set(sv, p - SvPVX_const(sv));
10238 continue; /* not "break" */
10241 if (is_utf8 != has_utf8) {
10244 sv_utf8_upgrade(sv);
10247 const STRLEN old_elen = elen;
10248 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10249 sv_utf8_upgrade(nsv);
10250 eptr = SvPVX_const(nsv);
10253 if (width) { /* fudge width (can't fudge elen) */
10254 width += elen - old_elen;
10260 have = esignlen + zeros + elen;
10262 Perl_croak_nocontext("%s", PL_memory_wrap);
10264 need = (have > width ? have : width);
10267 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10268 Perl_croak_nocontext("%s", PL_memory_wrap);
10269 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10271 if (esignlen && fill == '0') {
10273 for (i = 0; i < (int)esignlen; i++)
10274 *p++ = esignbuf[i];
10276 if (gap && !left) {
10277 memset(p, fill, gap);
10280 if (esignlen && fill != '0') {
10282 for (i = 0; i < (int)esignlen; i++)
10283 *p++ = esignbuf[i];
10287 for (i = zeros; i; i--)
10291 Copy(eptr, p, elen, char);
10295 memset(p, ' ', gap);
10300 Copy(dotstr, p, dotstrlen, char);
10304 vectorize = FALSE; /* done iterating over vecstr */
10311 SvCUR_set(sv, p - SvPVX_const(sv));
10319 /* =========================================================================
10321 =head1 Cloning an interpreter
10323 All the macros and functions in this section are for the private use of
10324 the main function, perl_clone().
10326 The foo_dup() functions make an exact copy of an existing foo thingy.
10327 During the course of a cloning, a hash table is used to map old addresses
10328 to new addresses. The table is created and manipulated with the
10329 ptr_table_* functions.
10333 * =========================================================================*/
10336 #if defined(USE_ITHREADS)
10338 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10339 #ifndef GpREFCNT_inc
10340 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10344 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10345 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10346 If this changes, please unmerge ss_dup.
10347 Likewise, sv_dup_inc_multiple() relies on this fact. */
10348 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10349 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10350 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10351 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10352 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10353 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10354 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10355 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10356 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10357 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10358 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10359 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10360 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10361 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10363 /* clone a parser */
10366 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10370 PERL_ARGS_ASSERT_PARSER_DUP;
10375 /* look for it in the table first */
10376 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10380 /* create anew and remember what it is */
10381 Newxz(parser, 1, yy_parser);
10382 ptr_table_store(PL_ptr_table, proto, parser);
10384 parser->yyerrstatus = 0;
10385 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10387 /* XXX these not yet duped */
10388 parser->old_parser = NULL;
10389 parser->stack = NULL;
10391 parser->stack_size = 0;
10392 /* XXX parser->stack->state = 0; */
10394 /* XXX eventually, just Copy() most of the parser struct ? */
10396 parser->lex_brackets = proto->lex_brackets;
10397 parser->lex_casemods = proto->lex_casemods;
10398 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10399 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10400 parser->lex_casestack = savepvn(proto->lex_casestack,
10401 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10402 parser->lex_defer = proto->lex_defer;
10403 parser->lex_dojoin = proto->lex_dojoin;
10404 parser->lex_expect = proto->lex_expect;
10405 parser->lex_formbrack = proto->lex_formbrack;
10406 parser->lex_inpat = proto->lex_inpat;
10407 parser->lex_inwhat = proto->lex_inwhat;
10408 parser->lex_op = proto->lex_op;
10409 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10410 parser->lex_starts = proto->lex_starts;
10411 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10412 parser->multi_close = proto->multi_close;
10413 parser->multi_open = proto->multi_open;
10414 parser->multi_start = proto->multi_start;
10415 parser->multi_end = proto->multi_end;
10416 parser->pending_ident = proto->pending_ident;
10417 parser->preambled = proto->preambled;
10418 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10419 parser->linestr = sv_dup_inc(proto->linestr, param);
10420 parser->expect = proto->expect;
10421 parser->copline = proto->copline;
10422 parser->last_lop_op = proto->last_lop_op;
10423 parser->lex_state = proto->lex_state;
10424 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10425 /* rsfp_filters entries have fake IoDIRP() */
10426 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10427 parser->in_my = proto->in_my;
10428 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10429 parser->error_count = proto->error_count;
10432 parser->linestr = sv_dup_inc(proto->linestr, param);
10435 char * const ols = SvPVX(proto->linestr);
10436 char * const ls = SvPVX(parser->linestr);
10438 parser->bufptr = ls + (proto->bufptr >= ols ?
10439 proto->bufptr - ols : 0);
10440 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10441 proto->oldbufptr - ols : 0);
10442 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10443 proto->oldoldbufptr - ols : 0);
10444 parser->linestart = ls + (proto->linestart >= ols ?
10445 proto->linestart - ols : 0);
10446 parser->last_uni = ls + (proto->last_uni >= ols ?
10447 proto->last_uni - ols : 0);
10448 parser->last_lop = ls + (proto->last_lop >= ols ?
10449 proto->last_lop - ols : 0);
10451 parser->bufend = ls + SvCUR(parser->linestr);
10454 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10458 parser->endwhite = proto->endwhite;
10459 parser->faketokens = proto->faketokens;
10460 parser->lasttoke = proto->lasttoke;
10461 parser->nextwhite = proto->nextwhite;
10462 parser->realtokenstart = proto->realtokenstart;
10463 parser->skipwhite = proto->skipwhite;
10464 parser->thisclose = proto->thisclose;
10465 parser->thismad = proto->thismad;
10466 parser->thisopen = proto->thisopen;
10467 parser->thisstuff = proto->thisstuff;
10468 parser->thistoken = proto->thistoken;
10469 parser->thiswhite = proto->thiswhite;
10471 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10472 parser->curforce = proto->curforce;
10474 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10475 Copy(proto->nexttype, parser->nexttype, 5, I32);
10476 parser->nexttoke = proto->nexttoke;
10479 /* XXX should clone saved_curcop here, but we aren't passed
10480 * proto_perl; so do it in perl_clone_using instead */
10486 /* duplicate a file handle */
10489 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10493 PERL_ARGS_ASSERT_FP_DUP;
10494 PERL_UNUSED_ARG(type);
10497 return (PerlIO*)NULL;
10499 /* look for it in the table first */
10500 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10504 /* create anew and remember what it is */
10505 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10506 ptr_table_store(PL_ptr_table, fp, ret);
10510 /* duplicate a directory handle */
10513 Perl_dirp_dup(pTHX_ DIR *const dp)
10515 PERL_UNUSED_CONTEXT;
10522 /* duplicate a typeglob */
10525 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10529 PERL_ARGS_ASSERT_GP_DUP;
10533 /* look for it in the table first */
10534 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10538 /* create anew and remember what it is */
10540 ptr_table_store(PL_ptr_table, gp, ret);
10543 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10544 on Newxz() to do this for us. */
10545 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10546 ret->gp_io = io_dup_inc(gp->gp_io, param);
10547 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10548 ret->gp_av = av_dup_inc(gp->gp_av, param);
10549 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10550 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10551 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10552 ret->gp_cvgen = gp->gp_cvgen;
10553 ret->gp_line = gp->gp_line;
10554 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10558 /* duplicate a chain of magic */
10561 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10563 MAGIC *mgret = NULL;
10564 MAGIC **mgprev_p = &mgret;
10566 PERL_ARGS_ASSERT_MG_DUP;
10568 for (; mg; mg = mg->mg_moremagic) {
10570 Newx(nmg, 1, MAGIC);
10572 mgprev_p = &(nmg->mg_moremagic);
10574 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10575 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10576 from the original commit adding Perl_mg_dup() - revision 4538.
10577 Similarly there is the annotation "XXX random ptr?" next to the
10578 assignment to nmg->mg_ptr. */
10581 /* FIXME for plugins
10582 if (nmg->mg_type == PERL_MAGIC_qr) {
10583 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10587 if(nmg->mg_type == PERL_MAGIC_backref) {
10588 /* The backref AV has its reference count deliberately bumped by
10591 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10594 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10595 ? sv_dup_inc(nmg->mg_obj, param)
10596 : sv_dup(nmg->mg_obj, param);
10599 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10600 if (nmg->mg_len > 0) {
10601 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10602 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10603 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10605 AMT * const namtp = (AMT*)nmg->mg_ptr;
10606 sv_dup_inc_multiple((SV**)(namtp->table),
10607 (SV**)(namtp->table), NofAMmeth, param);
10610 else if (nmg->mg_len == HEf_SVKEY)
10611 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10613 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10614 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10620 #endif /* USE_ITHREADS */
10622 /* create a new pointer-mapping table */
10625 Perl_ptr_table_new(pTHX)
10628 PERL_UNUSED_CONTEXT;
10630 Newx(tbl, 1, PTR_TBL_t);
10631 tbl->tbl_max = 511;
10632 tbl->tbl_items = 0;
10633 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10637 #define PTR_TABLE_HASH(ptr) \
10638 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10641 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10642 following define) and at call to new_body_inline made below in
10643 Perl_ptr_table_store()
10646 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10648 /* map an existing pointer using a table */
10650 STATIC PTR_TBL_ENT_t *
10651 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10653 PTR_TBL_ENT_t *tblent;
10654 const UV hash = PTR_TABLE_HASH(sv);
10656 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10658 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10659 for (; tblent; tblent = tblent->next) {
10660 if (tblent->oldval == sv)
10667 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10669 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10671 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10672 PERL_UNUSED_CONTEXT;
10674 return tblent ? tblent->newval : NULL;
10677 /* add a new entry to a pointer-mapping table */
10680 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10682 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10684 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10685 PERL_UNUSED_CONTEXT;
10688 tblent->newval = newsv;
10690 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10692 new_body_inline(tblent, PTE_SVSLOT);
10694 tblent->oldval = oldsv;
10695 tblent->newval = newsv;
10696 tblent->next = tbl->tbl_ary[entry];
10697 tbl->tbl_ary[entry] = tblent;
10699 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10700 ptr_table_split(tbl);
10704 /* double the hash bucket size of an existing ptr table */
10707 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10709 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10710 const UV oldsize = tbl->tbl_max + 1;
10711 UV newsize = oldsize * 2;
10714 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10715 PERL_UNUSED_CONTEXT;
10717 Renew(ary, newsize, PTR_TBL_ENT_t*);
10718 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10719 tbl->tbl_max = --newsize;
10720 tbl->tbl_ary = ary;
10721 for (i=0; i < oldsize; i++, ary++) {
10722 PTR_TBL_ENT_t **curentp, **entp, *ent;
10725 curentp = ary + oldsize;
10726 for (entp = ary, ent = *ary; ent; ent = *entp) {
10727 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10729 ent->next = *curentp;
10739 /* remove all the entries from a ptr table */
10742 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10744 if (tbl && tbl->tbl_items) {
10745 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10746 UV riter = tbl->tbl_max;
10749 PTR_TBL_ENT_t *entry = array[riter];
10752 PTR_TBL_ENT_t * const oentry = entry;
10753 entry = entry->next;
10758 tbl->tbl_items = 0;
10762 /* clear and free a ptr table */
10765 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10770 ptr_table_clear(tbl);
10771 Safefree(tbl->tbl_ary);
10775 #if defined(USE_ITHREADS)
10778 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10780 PERL_ARGS_ASSERT_RVPV_DUP;
10783 SvRV_set(dstr, SvWEAKREF(sstr)
10784 ? sv_dup(SvRV_const(sstr), param)
10785 : sv_dup_inc(SvRV_const(sstr), param));
10788 else if (SvPVX_const(sstr)) {
10789 /* Has something there */
10791 /* Normal PV - clone whole allocated space */
10792 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10793 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10794 /* Not that normal - actually sstr is copy on write.
10795 But we are a true, independant SV, so: */
10796 SvREADONLY_off(dstr);
10801 /* Special case - not normally malloced for some reason */
10802 if (isGV_with_GP(sstr)) {
10803 /* Don't need to do anything here. */
10805 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10806 /* A "shared" PV - clone it as "shared" PV */
10808 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10812 /* Some other special case - random pointer */
10813 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10818 /* Copy the NULL */
10819 SvPV_set(dstr, NULL);
10823 /* duplicate a list of SVs. source and dest may point to the same memory. */
10825 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10826 SSize_t items, CLONE_PARAMS *const param)
10828 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10830 while (items-- > 0) {
10831 *dest++ = sv_dup_inc(*source++, param);
10837 /* duplicate an SV of any type (including AV, HV etc) */
10840 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10845 PERL_ARGS_ASSERT_SV_DUP;
10849 if (SvTYPE(sstr) == SVTYPEMASK) {
10850 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10855 /* look for it in the table first */
10856 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10860 if(param->flags & CLONEf_JOIN_IN) {
10861 /** We are joining here so we don't want do clone
10862 something that is bad **/
10863 if (SvTYPE(sstr) == SVt_PVHV) {
10864 const HEK * const hvname = HvNAME_HEK(sstr);
10866 /** don't clone stashes if they already exist **/
10867 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10871 /* create anew and remember what it is */
10874 #ifdef DEBUG_LEAKING_SCALARS
10875 dstr->sv_debug_optype = sstr->sv_debug_optype;
10876 dstr->sv_debug_line = sstr->sv_debug_line;
10877 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10878 dstr->sv_debug_cloned = 1;
10879 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10882 ptr_table_store(PL_ptr_table, sstr, dstr);
10885 SvFLAGS(dstr) = SvFLAGS(sstr);
10886 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10887 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10890 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10891 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10892 (void*)PL_watch_pvx, SvPVX_const(sstr));
10895 /* don't clone objects whose class has asked us not to */
10896 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10901 switch (SvTYPE(sstr)) {
10903 SvANY(dstr) = NULL;
10906 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10908 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10910 SvIV_set(dstr, SvIVX(sstr));
10914 SvANY(dstr) = new_XNV();
10915 SvNV_set(dstr, SvNVX(sstr));
10917 /* case SVt_BIND: */
10920 /* These are all the types that need complex bodies allocating. */
10922 const svtype sv_type = SvTYPE(sstr);
10923 const struct body_details *const sv_type_details
10924 = bodies_by_type + sv_type;
10928 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10943 assert(sv_type_details->body_size);
10944 if (sv_type_details->arena) {
10945 new_body_inline(new_body, sv_type);
10947 = (void*)((char*)new_body - sv_type_details->offset);
10949 new_body = new_NOARENA(sv_type_details);
10953 SvANY(dstr) = new_body;
10956 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10957 ((char*)SvANY(dstr)) + sv_type_details->offset,
10958 sv_type_details->copy, char);
10960 Copy(((char*)SvANY(sstr)),
10961 ((char*)SvANY(dstr)),
10962 sv_type_details->body_size + sv_type_details->offset, char);
10965 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10966 && !isGV_with_GP(dstr))
10967 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10969 /* The Copy above means that all the source (unduplicated) pointers
10970 are now in the destination. We can check the flags and the
10971 pointers in either, but it's possible that there's less cache
10972 missing by always going for the destination.
10973 FIXME - instrument and check that assumption */
10974 if (sv_type >= SVt_PVMG) {
10975 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10976 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10977 } else if (SvMAGIC(dstr))
10978 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10980 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10983 /* The cast silences a GCC warning about unhandled types. */
10984 switch ((int)sv_type) {
10994 /* FIXME for plugins */
10995 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10998 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10999 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11000 LvTARG(dstr) = dstr;
11001 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11002 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11004 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11006 if(isGV_with_GP(sstr)) {
11007 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11008 /* Don't call sv_add_backref here as it's going to be
11009 created as part of the magic cloning of the symbol
11011 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11012 at the point of this comment. */
11013 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11014 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11015 (void)GpREFCNT_inc(GvGP(dstr));
11017 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11020 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11021 if (IoOFP(dstr) == IoIFP(sstr))
11022 IoOFP(dstr) = IoIFP(dstr);
11024 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11025 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11026 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11027 /* I have no idea why fake dirp (rsfps)
11028 should be treated differently but otherwise
11029 we end up with leaks -- sky*/
11030 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11031 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11032 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11034 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11035 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11036 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11037 if (IoDIRP(dstr)) {
11038 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11041 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11044 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11045 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11046 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11049 /* avoid cloning an empty array */
11050 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11051 SV **dst_ary, **src_ary;
11052 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11054 src_ary = AvARRAY((const AV *)sstr);
11055 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11056 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11057 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11058 AvALLOC((const AV *)dstr) = dst_ary;
11059 if (AvREAL((const AV *)sstr)) {
11060 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11064 while (items-- > 0)
11065 *dst_ary++ = sv_dup(*src_ary++, param);
11067 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11068 while (items-- > 0) {
11069 *dst_ary++ = &PL_sv_undef;
11073 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11074 AvALLOC((const AV *)dstr) = (SV**)NULL;
11075 AvMAX( (const AV *)dstr) = -1;
11076 AvFILLp((const AV *)dstr) = -1;
11080 if (HvARRAY((const HV *)sstr)) {
11082 const bool sharekeys = !!HvSHAREKEYS(sstr);
11083 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11084 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11086 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11087 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11089 HvARRAY(dstr) = (HE**)darray;
11090 while (i <= sxhv->xhv_max) {
11091 const HE * const source = HvARRAY(sstr)[i];
11092 HvARRAY(dstr)[i] = source
11093 ? he_dup(source, sharekeys, param) : 0;
11098 const struct xpvhv_aux * const saux = HvAUX(sstr);
11099 struct xpvhv_aux * const daux = HvAUX(dstr);
11100 /* This flag isn't copied. */
11101 /* SvOOK_on(hv) attacks the IV flags. */
11102 SvFLAGS(dstr) |= SVf_OOK;
11104 hvname = saux->xhv_name;
11105 daux->xhv_name = hek_dup(hvname, param);
11107 daux->xhv_riter = saux->xhv_riter;
11108 daux->xhv_eiter = saux->xhv_eiter
11109 ? he_dup(saux->xhv_eiter,
11110 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11111 /* backref array needs refcnt=2; see sv_add_backref */
11112 daux->xhv_backreferences =
11113 saux->xhv_backreferences
11114 ? MUTABLE_AV(SvREFCNT_inc(
11115 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11118 daux->xhv_mro_meta = saux->xhv_mro_meta
11119 ? mro_meta_dup(saux->xhv_mro_meta, param)
11122 /* Record stashes for possible cloning in Perl_clone(). */
11124 av_push(param->stashes, dstr);
11128 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11131 if (!(param->flags & CLONEf_COPY_STACKS)) {
11135 /* NOTE: not refcounted */
11136 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11138 if (!CvISXSUB(dstr))
11139 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11141 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11142 CvXSUBANY(dstr).any_ptr =
11143 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11145 /* don't dup if copying back - CvGV isn't refcounted, so the
11146 * duped GV may never be freed. A bit of a hack! DAPM */
11147 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11148 NULL : gv_dup(CvGV(dstr), param) ;
11149 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11151 CvWEAKOUTSIDE(sstr)
11152 ? cv_dup( CvOUTSIDE(dstr), param)
11153 : cv_dup_inc(CvOUTSIDE(dstr), param);
11154 if (!CvISXSUB(dstr))
11155 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11161 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11167 /* duplicate a context */
11170 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11172 PERL_CONTEXT *ncxs;
11174 PERL_ARGS_ASSERT_CX_DUP;
11177 return (PERL_CONTEXT*)NULL;
11179 /* look for it in the table first */
11180 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11184 /* create anew and remember what it is */
11185 Newx(ncxs, max + 1, PERL_CONTEXT);
11186 ptr_table_store(PL_ptr_table, cxs, ncxs);
11187 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11190 PERL_CONTEXT * const ncx = &ncxs[ix];
11191 if (CxTYPE(ncx) == CXt_SUBST) {
11192 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11195 switch (CxTYPE(ncx)) {
11197 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11198 ? cv_dup_inc(ncx->blk_sub.cv, param)
11199 : cv_dup(ncx->blk_sub.cv,param));
11200 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11201 ? av_dup_inc(ncx->blk_sub.argarray,
11204 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11206 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11207 ncx->blk_sub.oldcomppad);
11210 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11212 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11214 case CXt_LOOP_LAZYSV:
11215 ncx->blk_loop.state_u.lazysv.end
11216 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11217 /* We are taking advantage of av_dup_inc and sv_dup_inc
11218 actually being the same function, and order equivalance of
11220 We can assert the later [but only at run time :-(] */
11221 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11222 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11224 ncx->blk_loop.state_u.ary.ary
11225 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11226 case CXt_LOOP_LAZYIV:
11227 case CXt_LOOP_PLAIN:
11228 if (CxPADLOOP(ncx)) {
11229 ncx->blk_loop.oldcomppad
11230 = (PAD*)ptr_table_fetch(PL_ptr_table,
11231 ncx->blk_loop.oldcomppad);
11233 ncx->blk_loop.oldcomppad
11234 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11239 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11240 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11241 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11254 /* duplicate a stack info structure */
11257 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11261 PERL_ARGS_ASSERT_SI_DUP;
11264 return (PERL_SI*)NULL;
11266 /* look for it in the table first */
11267 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11271 /* create anew and remember what it is */
11272 Newxz(nsi, 1, PERL_SI);
11273 ptr_table_store(PL_ptr_table, si, nsi);
11275 nsi->si_stack = av_dup_inc(si->si_stack, param);
11276 nsi->si_cxix = si->si_cxix;
11277 nsi->si_cxmax = si->si_cxmax;
11278 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11279 nsi->si_type = si->si_type;
11280 nsi->si_prev = si_dup(si->si_prev, param);
11281 nsi->si_next = si_dup(si->si_next, param);
11282 nsi->si_markoff = si->si_markoff;
11287 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11288 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11289 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11290 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11291 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11292 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11293 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11294 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11295 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11296 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11297 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11298 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11299 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11300 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11303 #define pv_dup_inc(p) SAVEPV(p)
11304 #define pv_dup(p) SAVEPV(p)
11305 #define svp_dup_inc(p,pp) any_dup(p,pp)
11307 /* map any object to the new equivent - either something in the
11308 * ptr table, or something in the interpreter structure
11312 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11316 PERL_ARGS_ASSERT_ANY_DUP;
11319 return (void*)NULL;
11321 /* look for it in the table first */
11322 ret = ptr_table_fetch(PL_ptr_table, v);
11326 /* see if it is part of the interpreter structure */
11327 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11328 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11336 /* duplicate the save stack */
11339 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11342 ANY * const ss = proto_perl->Isavestack;
11343 const I32 max = proto_perl->Isavestack_max;
11344 I32 ix = proto_perl->Isavestack_ix;
11357 void (*dptr) (void*);
11358 void (*dxptr) (pTHX_ void*);
11360 PERL_ARGS_ASSERT_SS_DUP;
11362 Newxz(nss, max, ANY);
11365 const I32 type = POPINT(ss,ix);
11366 TOPINT(nss,ix) = type;
11368 case SAVEt_HELEM: /* hash element */
11369 sv = (const SV *)POPPTR(ss,ix);
11370 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11372 case SAVEt_ITEM: /* normal string */
11373 case SAVEt_SV: /* scalar reference */
11374 sv = (const SV *)POPPTR(ss,ix);
11375 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11378 case SAVEt_MORTALIZESV:
11379 sv = (const SV *)POPPTR(ss,ix);
11380 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11382 case SAVEt_SHARED_PVREF: /* char* in shared space */
11383 c = (char*)POPPTR(ss,ix);
11384 TOPPTR(nss,ix) = savesharedpv(c);
11385 ptr = POPPTR(ss,ix);
11386 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11388 case SAVEt_GENERIC_SVREF: /* generic sv */
11389 case SAVEt_SVREF: /* scalar reference */
11390 sv = (const SV *)POPPTR(ss,ix);
11391 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11392 ptr = POPPTR(ss,ix);
11393 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11395 case SAVEt_HV: /* hash reference */
11396 case SAVEt_AV: /* array reference */
11397 sv = (const SV *) POPPTR(ss,ix);
11398 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11400 case SAVEt_COMPPAD:
11402 sv = (const SV *) POPPTR(ss,ix);
11403 TOPPTR(nss,ix) = sv_dup(sv, param);
11405 case SAVEt_INT: /* int reference */
11406 ptr = POPPTR(ss,ix);
11407 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11408 intval = (int)POPINT(ss,ix);
11409 TOPINT(nss,ix) = intval;
11411 case SAVEt_LONG: /* long reference */
11412 ptr = POPPTR(ss,ix);
11413 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11415 case SAVEt_CLEARSV:
11416 longval = (long)POPLONG(ss,ix);
11417 TOPLONG(nss,ix) = longval;
11419 case SAVEt_I32: /* I32 reference */
11420 case SAVEt_I16: /* I16 reference */
11421 case SAVEt_I8: /* I8 reference */
11422 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11423 ptr = POPPTR(ss,ix);
11424 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11426 TOPINT(nss,ix) = i;
11428 case SAVEt_IV: /* IV reference */
11429 ptr = POPPTR(ss,ix);
11430 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11432 TOPIV(nss,ix) = iv;
11434 case SAVEt_HPTR: /* HV* reference */
11435 case SAVEt_APTR: /* AV* reference */
11436 case SAVEt_SPTR: /* SV* reference */
11437 ptr = POPPTR(ss,ix);
11438 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11439 sv = (const SV *)POPPTR(ss,ix);
11440 TOPPTR(nss,ix) = sv_dup(sv, param);
11442 case SAVEt_VPTR: /* random* reference */
11443 ptr = POPPTR(ss,ix);
11444 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11445 ptr = POPPTR(ss,ix);
11446 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11448 case SAVEt_GENERIC_PVREF: /* generic char* */
11449 case SAVEt_PPTR: /* char* reference */
11450 ptr = POPPTR(ss,ix);
11451 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11452 c = (char*)POPPTR(ss,ix);
11453 TOPPTR(nss,ix) = pv_dup(c);
11455 case SAVEt_GP: /* scalar reference */
11456 gp = (GP*)POPPTR(ss,ix);
11457 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11458 (void)GpREFCNT_inc(gp);
11459 gv = (const GV *)POPPTR(ss,ix);
11460 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11463 ptr = POPPTR(ss,ix);
11464 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11465 /* these are assumed to be refcounted properly */
11467 switch (((OP*)ptr)->op_type) {
11469 case OP_LEAVESUBLV:
11473 case OP_LEAVEWRITE:
11474 TOPPTR(nss,ix) = ptr;
11477 (void) OpREFCNT_inc(o);
11481 TOPPTR(nss,ix) = NULL;
11486 TOPPTR(nss,ix) = NULL;
11489 hv = (const HV *)POPPTR(ss,ix);
11490 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11492 TOPINT(nss,ix) = i;
11495 c = (char*)POPPTR(ss,ix);
11496 TOPPTR(nss,ix) = pv_dup_inc(c);
11498 case SAVEt_STACK_POS: /* Position on Perl stack */
11500 TOPINT(nss,ix) = i;
11502 case SAVEt_DESTRUCTOR:
11503 ptr = POPPTR(ss,ix);
11504 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11505 dptr = POPDPTR(ss,ix);
11506 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11507 any_dup(FPTR2DPTR(void *, dptr),
11510 case SAVEt_DESTRUCTOR_X:
11511 ptr = POPPTR(ss,ix);
11512 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11513 dxptr = POPDXPTR(ss,ix);
11514 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11515 any_dup(FPTR2DPTR(void *, dxptr),
11518 case SAVEt_REGCONTEXT:
11521 TOPINT(nss,ix) = i;
11524 case SAVEt_AELEM: /* array element */
11525 sv = (const SV *)POPPTR(ss,ix);
11526 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11528 TOPINT(nss,ix) = i;
11529 av = (const AV *)POPPTR(ss,ix);
11530 TOPPTR(nss,ix) = av_dup_inc(av, param);
11533 ptr = POPPTR(ss,ix);
11534 TOPPTR(nss,ix) = ptr;
11537 ptr = POPPTR(ss,ix);
11540 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11541 HINTS_REFCNT_UNLOCK;
11543 TOPPTR(nss,ix) = ptr;
11545 TOPINT(nss,ix) = i;
11546 if (i & HINT_LOCALIZE_HH) {
11547 hv = (const HV *)POPPTR(ss,ix);
11548 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11551 case SAVEt_PADSV_AND_MORTALIZE:
11552 longval = (long)POPLONG(ss,ix);
11553 TOPLONG(nss,ix) = longval;
11554 ptr = POPPTR(ss,ix);
11555 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11556 sv = (const SV *)POPPTR(ss,ix);
11557 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11560 ptr = POPPTR(ss,ix);
11561 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11562 longval = (long)POPBOOL(ss,ix);
11563 TOPBOOL(nss,ix) = (bool)longval;
11565 case SAVEt_SET_SVFLAGS:
11567 TOPINT(nss,ix) = i;
11569 TOPINT(nss,ix) = i;
11570 sv = (const SV *)POPPTR(ss,ix);
11571 TOPPTR(nss,ix) = sv_dup(sv, param);
11573 case SAVEt_RE_STATE:
11575 const struct re_save_state *const old_state
11576 = (struct re_save_state *)
11577 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11578 struct re_save_state *const new_state
11579 = (struct re_save_state *)
11580 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11582 Copy(old_state, new_state, 1, struct re_save_state);
11583 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11585 new_state->re_state_bostr
11586 = pv_dup(old_state->re_state_bostr);
11587 new_state->re_state_reginput
11588 = pv_dup(old_state->re_state_reginput);
11589 new_state->re_state_regeol
11590 = pv_dup(old_state->re_state_regeol);
11591 new_state->re_state_regoffs
11592 = (regexp_paren_pair*)
11593 any_dup(old_state->re_state_regoffs, proto_perl);
11594 new_state->re_state_reglastparen
11595 = (U32*) any_dup(old_state->re_state_reglastparen,
11597 new_state->re_state_reglastcloseparen
11598 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11600 /* XXX This just has to be broken. The old save_re_context
11601 code did SAVEGENERICPV(PL_reg_start_tmp);
11602 PL_reg_start_tmp is char **.
11603 Look above to what the dup code does for
11604 SAVEt_GENERIC_PVREF
11605 It can never have worked.
11606 So this is merely a faithful copy of the exiting bug: */
11607 new_state->re_state_reg_start_tmp
11608 = (char **) pv_dup((char *)
11609 old_state->re_state_reg_start_tmp);
11610 /* I assume that it only ever "worked" because no-one called
11611 (pseudo)fork while the regexp engine had re-entered itself.
11613 #ifdef PERL_OLD_COPY_ON_WRITE
11614 new_state->re_state_nrs
11615 = sv_dup(old_state->re_state_nrs, param);
11617 new_state->re_state_reg_magic
11618 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11620 new_state->re_state_reg_oldcurpm
11621 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11623 new_state->re_state_reg_curpm
11624 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11626 new_state->re_state_reg_oldsaved
11627 = pv_dup(old_state->re_state_reg_oldsaved);
11628 new_state->re_state_reg_poscache
11629 = pv_dup(old_state->re_state_reg_poscache);
11630 new_state->re_state_reg_starttry
11631 = pv_dup(old_state->re_state_reg_starttry);
11634 case SAVEt_COMPILE_WARNINGS:
11635 ptr = POPPTR(ss,ix);
11636 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11639 ptr = POPPTR(ss,ix);
11640 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11644 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11652 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11653 * flag to the result. This is done for each stash before cloning starts,
11654 * so we know which stashes want their objects cloned */
11657 do_mark_cloneable_stash(pTHX_ SV *const sv)
11659 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11661 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11662 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11663 if (cloner && GvCV(cloner)) {
11670 mXPUSHs(newSVhek(hvname));
11672 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11679 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11687 =for apidoc perl_clone
11689 Create and return a new interpreter by cloning the current one.
11691 perl_clone takes these flags as parameters:
11693 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11694 without it we only clone the data and zero the stacks,
11695 with it we copy the stacks and the new perl interpreter is
11696 ready to run at the exact same point as the previous one.
11697 The pseudo-fork code uses COPY_STACKS while the
11698 threads->create doesn't.
11700 CLONEf_KEEP_PTR_TABLE
11701 perl_clone keeps a ptr_table with the pointer of the old
11702 variable as a key and the new variable as a value,
11703 this allows it to check if something has been cloned and not
11704 clone it again but rather just use the value and increase the
11705 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11706 the ptr_table using the function
11707 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11708 reason to keep it around is if you want to dup some of your own
11709 variable who are outside the graph perl scans, example of this
11710 code is in threads.xs create
11713 This is a win32 thing, it is ignored on unix, it tells perls
11714 win32host code (which is c++) to clone itself, this is needed on
11715 win32 if you want to run two threads at the same time,
11716 if you just want to do some stuff in a separate perl interpreter
11717 and then throw it away and return to the original one,
11718 you don't need to do anything.
11723 /* XXX the above needs expanding by someone who actually understands it ! */
11724 EXTERN_C PerlInterpreter *
11725 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11728 perl_clone(PerlInterpreter *proto_perl, UV flags)
11731 #ifdef PERL_IMPLICIT_SYS
11733 PERL_ARGS_ASSERT_PERL_CLONE;
11735 /* perlhost.h so we need to call into it
11736 to clone the host, CPerlHost should have a c interface, sky */
11738 if (flags & CLONEf_CLONE_HOST) {
11739 return perl_clone_host(proto_perl,flags);
11741 return perl_clone_using(proto_perl, flags,
11743 proto_perl->IMemShared,
11744 proto_perl->IMemParse,
11746 proto_perl->IStdIO,
11750 proto_perl->IProc);
11754 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11755 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11756 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11757 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11758 struct IPerlDir* ipD, struct IPerlSock* ipS,
11759 struct IPerlProc* ipP)
11761 /* XXX many of the string copies here can be optimized if they're
11762 * constants; they need to be allocated as common memory and just
11763 * their pointers copied. */
11766 CLONE_PARAMS clone_params;
11767 CLONE_PARAMS* const param = &clone_params;
11769 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11771 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11773 /* for each stash, determine whether its objects should be cloned */
11774 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11775 PERL_SET_THX(my_perl);
11778 PoisonNew(my_perl, 1, PerlInterpreter);
11784 PL_savestack_ix = 0;
11785 PL_savestack_max = -1;
11786 PL_sig_pending = 0;
11788 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11789 # else /* !DEBUGGING */
11790 Zero(my_perl, 1, PerlInterpreter);
11791 # endif /* DEBUGGING */
11793 /* host pointers */
11795 PL_MemShared = ipMS;
11796 PL_MemParse = ipMP;
11803 #else /* !PERL_IMPLICIT_SYS */
11805 CLONE_PARAMS clone_params;
11806 CLONE_PARAMS* param = &clone_params;
11807 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11809 PERL_ARGS_ASSERT_PERL_CLONE;
11811 /* for each stash, determine whether its objects should be cloned */
11812 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11813 PERL_SET_THX(my_perl);
11816 PoisonNew(my_perl, 1, PerlInterpreter);
11822 PL_savestack_ix = 0;
11823 PL_savestack_max = -1;
11824 PL_sig_pending = 0;
11826 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11827 # else /* !DEBUGGING */
11828 Zero(my_perl, 1, PerlInterpreter);
11829 # endif /* DEBUGGING */
11830 #endif /* PERL_IMPLICIT_SYS */
11831 param->flags = flags;
11832 param->proto_perl = proto_perl;
11834 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11836 PL_body_arenas = NULL;
11837 Zero(&PL_body_roots, 1, PL_body_roots);
11839 PL_nice_chunk = NULL;
11840 PL_nice_chunk_size = 0;
11842 PL_sv_objcount = 0;
11844 PL_sv_arenaroot = NULL;
11846 PL_debug = proto_perl->Idebug;
11848 PL_hash_seed = proto_perl->Ihash_seed;
11849 PL_rehash_seed = proto_perl->Irehash_seed;
11851 #ifdef USE_REENTRANT_API
11852 /* XXX: things like -Dm will segfault here in perlio, but doing
11853 * PERL_SET_CONTEXT(proto_perl);
11854 * breaks too many other things
11856 Perl_reentrant_init(aTHX);
11859 /* create SV map for pointer relocation */
11860 PL_ptr_table = ptr_table_new();
11862 /* initialize these special pointers as early as possible */
11863 SvANY(&PL_sv_undef) = NULL;
11864 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11865 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11866 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11868 SvANY(&PL_sv_no) = new_XPVNV();
11869 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11870 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11871 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11872 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11873 SvCUR_set(&PL_sv_no, 0);
11874 SvLEN_set(&PL_sv_no, 1);
11875 SvIV_set(&PL_sv_no, 0);
11876 SvNV_set(&PL_sv_no, 0);
11877 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11879 SvANY(&PL_sv_yes) = new_XPVNV();
11880 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11881 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11882 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11883 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11884 SvCUR_set(&PL_sv_yes, 1);
11885 SvLEN_set(&PL_sv_yes, 2);
11886 SvIV_set(&PL_sv_yes, 1);
11887 SvNV_set(&PL_sv_yes, 1);
11888 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11890 /* create (a non-shared!) shared string table */
11891 PL_strtab = newHV();
11892 HvSHAREKEYS_off(PL_strtab);
11893 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11894 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11896 PL_compiling = proto_perl->Icompiling;
11898 /* These two PVs will be free'd special way so must set them same way op.c does */
11899 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11900 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11902 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11903 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11905 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11906 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11907 if (PL_compiling.cop_hints_hash) {
11909 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11910 HINTS_REFCNT_UNLOCK;
11912 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11913 #ifdef PERL_DEBUG_READONLY_OPS
11918 /* pseudo environmental stuff */
11919 PL_origargc = proto_perl->Iorigargc;
11920 PL_origargv = proto_perl->Iorigargv;
11922 param->stashes = newAV(); /* Setup array of objects to call clone on */
11924 /* Set tainting stuff before PerlIO_debug can possibly get called */
11925 PL_tainting = proto_perl->Itainting;
11926 PL_taint_warn = proto_perl->Itaint_warn;
11928 #ifdef PERLIO_LAYERS
11929 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11930 PerlIO_clone(aTHX_ proto_perl, param);
11933 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11934 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11935 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11936 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11937 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11938 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11941 PL_minus_c = proto_perl->Iminus_c;
11942 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11943 PL_localpatches = proto_perl->Ilocalpatches;
11944 PL_splitstr = proto_perl->Isplitstr;
11945 PL_minus_n = proto_perl->Iminus_n;
11946 PL_minus_p = proto_perl->Iminus_p;
11947 PL_minus_l = proto_perl->Iminus_l;
11948 PL_minus_a = proto_perl->Iminus_a;
11949 PL_minus_E = proto_perl->Iminus_E;
11950 PL_minus_F = proto_perl->Iminus_F;
11951 PL_doswitches = proto_perl->Idoswitches;
11952 PL_dowarn = proto_perl->Idowarn;
11953 PL_doextract = proto_perl->Idoextract;
11954 PL_sawampersand = proto_perl->Isawampersand;
11955 PL_unsafe = proto_perl->Iunsafe;
11956 PL_inplace = SAVEPV(proto_perl->Iinplace);
11957 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11958 PL_perldb = proto_perl->Iperldb;
11959 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11960 PL_exit_flags = proto_perl->Iexit_flags;
11962 /* magical thingies */
11963 /* XXX time(&PL_basetime) when asked for? */
11964 PL_basetime = proto_perl->Ibasetime;
11965 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11967 PL_maxsysfd = proto_perl->Imaxsysfd;
11968 PL_statusvalue = proto_perl->Istatusvalue;
11970 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11972 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11974 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11976 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11977 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11978 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11981 /* RE engine related */
11982 Zero(&PL_reg_state, 1, struct re_save_state);
11983 PL_reginterp_cnt = 0;
11984 PL_regmatch_slab = NULL;
11986 /* Clone the regex array */
11987 /* ORANGE FIXME for plugins, probably in the SV dup code.
11988 newSViv(PTR2IV(CALLREGDUPE(
11989 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11991 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11992 PL_regex_pad = AvARRAY(PL_regex_padav);
11994 /* shortcuts to various I/O objects */
11995 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11996 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11997 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11998 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11999 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12000 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12001 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12003 /* shortcuts to regexp stuff */
12004 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12006 /* shortcuts to misc objects */
12007 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12009 /* shortcuts to debugging objects */
12010 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12011 PL_DBline = gv_dup(proto_perl->IDBline, param);
12012 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12013 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12014 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12015 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12016 PL_dbargs = av_dup(proto_perl->Idbargs, param);
12018 /* symbol tables */
12019 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12020 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12021 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12022 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12023 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12025 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12026 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12027 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12028 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12029 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12030 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12031 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12032 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12034 PL_sub_generation = proto_perl->Isub_generation;
12035 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12037 /* funky return mechanisms */
12038 PL_forkprocess = proto_perl->Iforkprocess;
12040 /* subprocess state */
12041 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12043 /* internal state */
12044 PL_maxo = proto_perl->Imaxo;
12045 if (proto_perl->Iop_mask)
12046 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12049 /* PL_asserting = proto_perl->Iasserting; */
12051 /* current interpreter roots */
12052 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12054 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12056 PL_main_start = proto_perl->Imain_start;
12057 PL_eval_root = proto_perl->Ieval_root;
12058 PL_eval_start = proto_perl->Ieval_start;
12060 /* runtime control stuff */
12061 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12063 PL_filemode = proto_perl->Ifilemode;
12064 PL_lastfd = proto_perl->Ilastfd;
12065 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12068 PL_gensym = proto_perl->Igensym;
12069 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12070 PL_laststatval = proto_perl->Ilaststatval;
12071 PL_laststype = proto_perl->Ilaststype;
12074 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12076 /* interpreter atexit processing */
12077 PL_exitlistlen = proto_perl->Iexitlistlen;
12078 if (PL_exitlistlen) {
12079 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12080 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12083 PL_exitlist = (PerlExitListEntry*)NULL;
12085 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12086 if (PL_my_cxt_size) {
12087 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12088 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12089 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12090 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12091 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12095 PL_my_cxt_list = (void**)NULL;
12096 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12097 PL_my_cxt_keys = (const char**)NULL;
12100 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12101 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12102 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12104 PL_profiledata = NULL;
12106 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12108 PAD_CLONE_VARS(proto_perl, param);
12110 #ifdef HAVE_INTERP_INTERN
12111 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12114 /* more statics moved here */
12115 PL_generation = proto_perl->Igeneration;
12116 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12118 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12119 PL_in_clean_all = proto_perl->Iin_clean_all;
12121 PL_uid = proto_perl->Iuid;
12122 PL_euid = proto_perl->Ieuid;
12123 PL_gid = proto_perl->Igid;
12124 PL_egid = proto_perl->Iegid;
12125 PL_nomemok = proto_perl->Inomemok;
12126 PL_an = proto_perl->Ian;
12127 PL_evalseq = proto_perl->Ievalseq;
12128 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12129 PL_origalen = proto_perl->Iorigalen;
12130 #ifdef PERL_USES_PL_PIDSTATUS
12131 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12133 PL_osname = SAVEPV(proto_perl->Iosname);
12134 PL_sighandlerp = proto_perl->Isighandlerp;
12136 PL_runops = proto_perl->Irunops;
12138 PL_parser = parser_dup(proto_perl->Iparser, param);
12140 /* XXX this only works if the saved cop has already been cloned */
12141 if (proto_perl->Iparser) {
12142 PL_parser->saved_curcop = (COP*)any_dup(
12143 proto_perl->Iparser->saved_curcop,
12147 PL_subline = proto_perl->Isubline;
12148 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12151 PL_cryptseen = proto_perl->Icryptseen;
12154 PL_hints = proto_perl->Ihints;
12156 PL_amagic_generation = proto_perl->Iamagic_generation;
12158 #ifdef USE_LOCALE_COLLATE
12159 PL_collation_ix = proto_perl->Icollation_ix;
12160 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12161 PL_collation_standard = proto_perl->Icollation_standard;
12162 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12163 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12164 #endif /* USE_LOCALE_COLLATE */
12166 #ifdef USE_LOCALE_NUMERIC
12167 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12168 PL_numeric_standard = proto_perl->Inumeric_standard;
12169 PL_numeric_local = proto_perl->Inumeric_local;
12170 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12171 #endif /* !USE_LOCALE_NUMERIC */
12173 /* utf8 character classes */
12174 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12175 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12176 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12177 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12178 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12179 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12180 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12181 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12182 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12183 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12184 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12185 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12186 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12187 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12188 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12189 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12190 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12191 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12192 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12194 /* Did the locale setup indicate UTF-8? */
12195 PL_utf8locale = proto_perl->Iutf8locale;
12196 /* Unicode features (see perlrun/-C) */
12197 PL_unicode = proto_perl->Iunicode;
12199 /* Pre-5.8 signals control */
12200 PL_signals = proto_perl->Isignals;
12202 /* times() ticks per second */
12203 PL_clocktick = proto_perl->Iclocktick;
12205 /* Recursion stopper for PerlIO_find_layer */
12206 PL_in_load_module = proto_perl->Iin_load_module;
12208 /* sort() routine */
12209 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12211 /* Not really needed/useful since the reenrant_retint is "volatile",
12212 * but do it for consistency's sake. */
12213 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12215 /* Hooks to shared SVs and locks. */
12216 PL_sharehook = proto_perl->Isharehook;
12217 PL_lockhook = proto_perl->Ilockhook;
12218 PL_unlockhook = proto_perl->Iunlockhook;
12219 PL_threadhook = proto_perl->Ithreadhook;
12220 PL_destroyhook = proto_perl->Idestroyhook;
12222 #ifdef THREADS_HAVE_PIDS
12223 PL_ppid = proto_perl->Ippid;
12227 PL_last_swash_hv = NULL; /* reinits on demand */
12228 PL_last_swash_klen = 0;
12229 PL_last_swash_key[0]= '\0';
12230 PL_last_swash_tmps = (U8*)NULL;
12231 PL_last_swash_slen = 0;
12233 PL_glob_index = proto_perl->Iglob_index;
12234 PL_srand_called = proto_perl->Isrand_called;
12236 if (proto_perl->Ipsig_pend) {
12237 Newxz(PL_psig_pend, SIG_SIZE, int);
12240 PL_psig_pend = (int*)NULL;
12243 if (proto_perl->Ipsig_name) {
12244 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12245 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12247 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12250 PL_psig_ptr = (SV**)NULL;
12251 PL_psig_name = (SV**)NULL;
12254 /* intrpvar.h stuff */
12256 if (flags & CLONEf_COPY_STACKS) {
12257 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12258 PL_tmps_ix = proto_perl->Itmps_ix;
12259 PL_tmps_max = proto_perl->Itmps_max;
12260 PL_tmps_floor = proto_perl->Itmps_floor;
12261 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12262 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack, PL_tmps_ix,
12265 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12266 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12267 Newxz(PL_markstack, i, I32);
12268 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12269 - proto_perl->Imarkstack);
12270 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12271 - proto_perl->Imarkstack);
12272 Copy(proto_perl->Imarkstack, PL_markstack,
12273 PL_markstack_ptr - PL_markstack + 1, I32);
12275 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12276 * NOTE: unlike the others! */
12277 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12278 PL_scopestack_max = proto_perl->Iscopestack_max;
12279 Newxz(PL_scopestack, PL_scopestack_max, I32);
12280 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12282 /* NOTE: si_dup() looks at PL_markstack */
12283 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12285 /* PL_curstack = PL_curstackinfo->si_stack; */
12286 PL_curstack = av_dup(proto_perl->Icurstack, param);
12287 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12289 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12290 PL_stack_base = AvARRAY(PL_curstack);
12291 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12292 - proto_perl->Istack_base);
12293 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12295 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12296 * NOTE: unlike the others! */
12297 PL_savestack_ix = proto_perl->Isavestack_ix;
12298 PL_savestack_max = proto_perl->Isavestack_max;
12299 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12300 PL_savestack = ss_dup(proto_perl, param);
12304 ENTER; /* perl_destruct() wants to LEAVE; */
12306 /* although we're not duplicating the tmps stack, we should still
12307 * add entries for any SVs on the tmps stack that got cloned by a
12308 * non-refcount means (eg a temp in @_); otherwise they will be
12311 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12312 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12313 proto_perl->Itmps_stack[i]));
12314 if (nsv && !SvREFCNT(nsv)) {
12315 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
12320 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12321 PL_top_env = &PL_start_env;
12323 PL_op = proto_perl->Iop;
12326 PL_Xpv = (XPV*)NULL;
12327 my_perl->Ina = proto_perl->Ina;
12329 PL_statbuf = proto_perl->Istatbuf;
12330 PL_statcache = proto_perl->Istatcache;
12331 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12332 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12334 PL_timesbuf = proto_perl->Itimesbuf;
12337 PL_tainted = proto_perl->Itainted;
12338 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12339 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12340 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12341 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12342 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12343 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12344 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12345 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12347 PL_restartop = proto_perl->Irestartop;
12348 PL_in_eval = proto_perl->Iin_eval;
12349 PL_delaymagic = proto_perl->Idelaymagic;
12350 PL_dirty = proto_perl->Idirty;
12351 PL_localizing = proto_perl->Ilocalizing;
12353 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12354 PL_hv_fetch_ent_mh = NULL;
12355 PL_modcount = proto_perl->Imodcount;
12356 PL_lastgotoprobe = NULL;
12357 PL_dumpindent = proto_perl->Idumpindent;
12359 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12360 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12361 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12362 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12363 PL_efloatbuf = NULL; /* reinits on demand */
12364 PL_efloatsize = 0; /* reinits on demand */
12368 PL_screamfirst = NULL;
12369 PL_screamnext = NULL;
12370 PL_maxscream = -1; /* reinits on demand */
12371 PL_lastscream = NULL;
12374 PL_regdummy = proto_perl->Iregdummy;
12375 PL_colorset = 0; /* reinits PL_colors[] */
12376 /*PL_colors[6] = {0,0,0,0,0,0};*/
12380 /* Pluggable optimizer */
12381 PL_peepp = proto_perl->Ipeepp;
12382 /* op_free() hook */
12383 PL_opfreehook = proto_perl->Iopfreehook;
12385 PL_stashcache = newHV();
12387 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12388 proto_perl->Iwatchaddr);
12389 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12390 if (PL_debug && PL_watchaddr) {
12391 PerlIO_printf(Perl_debug_log,
12392 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12393 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12394 PTR2UV(PL_watchok));
12397 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12399 /* Call the ->CLONE method, if it exists, for each of the stashes
12400 identified by sv_dup() above.
12402 while(av_len(param->stashes) != -1) {
12403 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12404 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12405 if (cloner && GvCV(cloner)) {
12410 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12412 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12418 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12419 ptr_table_free(PL_ptr_table);
12420 PL_ptr_table = NULL;
12424 SvREFCNT_dec(param->stashes);
12426 /* orphaned? eg threads->new inside BEGIN or use */
12427 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12428 SvREFCNT_inc_simple_void(PL_compcv);
12429 SAVEFREESV(PL_compcv);
12435 #endif /* USE_ITHREADS */
12438 =head1 Unicode Support
12440 =for apidoc sv_recode_to_utf8
12442 The encoding is assumed to be an Encode object, on entry the PV
12443 of the sv is assumed to be octets in that encoding, and the sv
12444 will be converted into Unicode (and UTF-8).
12446 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12447 is not a reference, nothing is done to the sv. If the encoding is not
12448 an C<Encode::XS> Encoding object, bad things will happen.
12449 (See F<lib/encoding.pm> and L<Encode>).
12451 The PV of the sv is returned.
12456 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12460 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12462 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12476 Passing sv_yes is wrong - it needs to be or'ed set of constants
12477 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12478 remove converted chars from source.
12480 Both will default the value - let them.
12482 XPUSHs(&PL_sv_yes);
12485 call_method("decode", G_SCALAR);
12489 s = SvPV_const(uni, len);
12490 if (s != SvPVX_const(sv)) {
12491 SvGROW(sv, len + 1);
12492 Move(s, SvPVX(sv), len + 1, char);
12493 SvCUR_set(sv, len);
12500 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12504 =for apidoc sv_cat_decode
12506 The encoding is assumed to be an Encode object, the PV of the ssv is
12507 assumed to be octets in that encoding and decoding the input starts
12508 from the position which (PV + *offset) pointed to. The dsv will be
12509 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12510 when the string tstr appears in decoding output or the input ends on
12511 the PV of the ssv. The value which the offset points will be modified
12512 to the last input position on the ssv.
12514 Returns TRUE if the terminator was found, else returns FALSE.
12519 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12520 SV *ssv, int *offset, char *tstr, int tlen)
12525 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12527 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12538 offsv = newSViv(*offset);
12540 mXPUSHp(tstr, tlen);
12542 call_method("cat_decode", G_SCALAR);
12544 ret = SvTRUE(TOPs);
12545 *offset = SvIV(offsv);
12551 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12556 /* ---------------------------------------------------------------------
12558 * support functions for report_uninit()
12561 /* the maxiumum size of array or hash where we will scan looking
12562 * for the undefined element that triggered the warning */
12564 #define FUV_MAX_SEARCH_SIZE 1000
12566 /* Look for an entry in the hash whose value has the same SV as val;
12567 * If so, return a mortal copy of the key. */
12570 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12573 register HE **array;
12576 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12578 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12579 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12582 array = HvARRAY(hv);
12584 for (i=HvMAX(hv); i>0; i--) {
12585 register HE *entry;
12586 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12587 if (HeVAL(entry) != val)
12589 if ( HeVAL(entry) == &PL_sv_undef ||
12590 HeVAL(entry) == &PL_sv_placeholder)
12594 if (HeKLEN(entry) == HEf_SVKEY)
12595 return sv_mortalcopy(HeKEY_sv(entry));
12596 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12602 /* Look for an entry in the array whose value has the same SV as val;
12603 * If so, return the index, otherwise return -1. */
12606 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12610 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12612 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12613 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12616 if (val != &PL_sv_undef) {
12617 SV ** const svp = AvARRAY(av);
12620 for (i=AvFILLp(av); i>=0; i--)
12627 /* S_varname(): return the name of a variable, optionally with a subscript.
12628 * If gv is non-zero, use the name of that global, along with gvtype (one
12629 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12630 * targ. Depending on the value of the subscript_type flag, return:
12633 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12634 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12635 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12636 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12639 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12640 const SV *const keyname, I32 aindex, int subscript_type)
12643 SV * const name = sv_newmortal();
12646 buffer[0] = gvtype;
12649 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12651 gv_fullname4(name, gv, buffer, 0);
12653 if ((unsigned int)SvPVX(name)[1] <= 26) {
12655 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12657 /* Swap the 1 unprintable control character for the 2 byte pretty
12658 version - ie substr($name, 1, 1) = $buffer; */
12659 sv_insert(name, 1, 1, buffer, 2);
12663 CV * const cv = find_runcv(NULL);
12667 if (!cv || !CvPADLIST(cv))
12669 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12670 sv = *av_fetch(av, targ, FALSE);
12671 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12674 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12675 SV * const sv = newSV(0);
12676 *SvPVX(name) = '$';
12677 Perl_sv_catpvf(aTHX_ name, "{%s}",
12678 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12681 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12682 *SvPVX(name) = '$';
12683 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12685 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12686 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12687 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12695 =for apidoc find_uninit_var
12697 Find the name of the undefined variable (if any) that caused the operator o
12698 to issue a "Use of uninitialized value" warning.
12699 If match is true, only return a name if it's value matches uninit_sv.
12700 So roughly speaking, if a unary operator (such as OP_COS) generates a
12701 warning, then following the direct child of the op may yield an
12702 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12703 other hand, with OP_ADD there are two branches to follow, so we only print
12704 the variable name if we get an exact match.
12706 The name is returned as a mortal SV.
12708 Assumes that PL_op is the op that originally triggered the error, and that
12709 PL_comppad/PL_curpad points to the currently executing pad.
12715 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12721 const OP *o, *o2, *kid;
12723 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12724 uninit_sv == &PL_sv_placeholder)))
12727 switch (obase->op_type) {
12734 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12735 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12738 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12740 if (pad) { /* @lex, %lex */
12741 sv = PAD_SVl(obase->op_targ);
12745 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12746 /* @global, %global */
12747 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12750 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12752 else /* @{expr}, %{expr} */
12753 return find_uninit_var(cUNOPx(obase)->op_first,
12757 /* attempt to find a match within the aggregate */
12759 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12761 subscript_type = FUV_SUBSCRIPT_HASH;
12764 index = find_array_subscript((const AV *)sv, uninit_sv);
12766 subscript_type = FUV_SUBSCRIPT_ARRAY;
12769 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12772 return varname(gv, hash ? '%' : '@', obase->op_targ,
12773 keysv, index, subscript_type);
12777 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12779 return varname(NULL, '$', obase->op_targ,
12780 NULL, 0, FUV_SUBSCRIPT_NONE);
12783 gv = cGVOPx_gv(obase);
12784 if (!gv || (match && GvSV(gv) != uninit_sv))
12786 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12789 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12792 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12793 if (!av || SvRMAGICAL(av))
12795 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12796 if (!svp || *svp != uninit_sv)
12799 return varname(NULL, '$', obase->op_targ,
12800 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12803 gv = cGVOPx_gv(obase);
12808 AV *const av = GvAV(gv);
12809 if (!av || SvRMAGICAL(av))
12811 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12812 if (!svp || *svp != uninit_sv)
12815 return varname(gv, '$', 0,
12816 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12821 o = cUNOPx(obase)->op_first;
12822 if (!o || o->op_type != OP_NULL ||
12823 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12825 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12829 if (PL_op == obase)
12830 /* $a[uninit_expr] or $h{uninit_expr} */
12831 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12834 o = cBINOPx(obase)->op_first;
12835 kid = cBINOPx(obase)->op_last;
12837 /* get the av or hv, and optionally the gv */
12839 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12840 sv = PAD_SV(o->op_targ);
12842 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12843 && cUNOPo->op_first->op_type == OP_GV)
12845 gv = cGVOPx_gv(cUNOPo->op_first);
12849 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12854 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12855 /* index is constant */
12859 if (obase->op_type == OP_HELEM) {
12860 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12861 if (!he || HeVAL(he) != uninit_sv)
12865 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12866 if (!svp || *svp != uninit_sv)
12870 if (obase->op_type == OP_HELEM)
12871 return varname(gv, '%', o->op_targ,
12872 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12874 return varname(gv, '@', o->op_targ, NULL,
12875 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12878 /* index is an expression;
12879 * attempt to find a match within the aggregate */
12880 if (obase->op_type == OP_HELEM) {
12881 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12883 return varname(gv, '%', o->op_targ,
12884 keysv, 0, FUV_SUBSCRIPT_HASH);
12888 = find_array_subscript((const AV *)sv, uninit_sv);
12890 return varname(gv, '@', o->op_targ,
12891 NULL, index, FUV_SUBSCRIPT_ARRAY);
12896 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12898 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12903 /* only examine RHS */
12904 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12907 o = cUNOPx(obase)->op_first;
12908 if (o->op_type == OP_PUSHMARK)
12911 if (!o->op_sibling) {
12912 /* one-arg version of open is highly magical */
12914 if (o->op_type == OP_GV) { /* open FOO; */
12916 if (match && GvSV(gv) != uninit_sv)
12918 return varname(gv, '$', 0,
12919 NULL, 0, FUV_SUBSCRIPT_NONE);
12921 /* other possibilities not handled are:
12922 * open $x; or open my $x; should return '${*$x}'
12923 * open expr; should return '$'.expr ideally
12929 /* ops where $_ may be an implicit arg */
12933 if ( !(obase->op_flags & OPf_STACKED)) {
12934 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12935 ? PAD_SVl(obase->op_targ)
12938 sv = sv_newmortal();
12939 sv_setpvs(sv, "$_");
12948 match = 1; /* print etc can return undef on defined args */
12949 /* skip filehandle as it can't produce 'undef' warning */
12950 o = cUNOPx(obase)->op_first;
12951 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12952 o = o->op_sibling->op_sibling;
12956 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12958 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12960 /* the following ops are capable of returning PL_sv_undef even for
12961 * defined arg(s) */
12980 case OP_GETPEERNAME:
13028 case OP_SMARTMATCH:
13037 /* XXX tmp hack: these two may call an XS sub, and currently
13038 XS subs don't have a SUB entry on the context stack, so CV and
13039 pad determination goes wrong, and BAD things happen. So, just
13040 don't try to determine the value under those circumstances.
13041 Need a better fix at dome point. DAPM 11/2007 */
13047 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13048 if (gv && GvSV(gv) == uninit_sv)
13049 return newSVpvs_flags("$.", SVs_TEMP);
13054 /* def-ness of rval pos() is independent of the def-ness of its arg */
13055 if ( !(obase->op_flags & OPf_MOD))
13060 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13061 return newSVpvs_flags("${$/}", SVs_TEMP);
13066 if (!(obase->op_flags & OPf_KIDS))
13068 o = cUNOPx(obase)->op_first;
13074 /* if all except one arg are constant, or have no side-effects,
13075 * or are optimized away, then it's unambiguous */
13077 for (kid=o; kid; kid = kid->op_sibling) {
13079 const OPCODE type = kid->op_type;
13080 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13081 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13082 || (type == OP_PUSHMARK)
13086 if (o2) { /* more than one found */
13093 return find_uninit_var(o2, uninit_sv, match);
13095 /* scan all args */
13097 sv = find_uninit_var(o, uninit_sv, 1);
13109 =for apidoc report_uninit
13111 Print appropriate "Use of uninitialized variable" warning
13117 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13121 SV* varname = NULL;
13123 varname = find_uninit_var(PL_op, uninit_sv,0);
13125 sv_insert(varname, 0, 0, " ", 1);
13127 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13128 varname ? SvPV_nolen_const(varname) : "",
13129 " in ", OP_DESC(PL_op));
13132 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13138 * c-indentation-style: bsd
13139 * c-basic-offset: 4
13140 * indent-tabs-mode: t
13143 * ex: set ts=8 sts=4 sw=4 noet:
projects / p5sagit/p5-mst-13.2.git / blob